I Part 1

1. Macroeconomics for the
© The McGrawHill
Introducing Advanced
Empirics about Prosperity
long run and for the short
Companies. 2005
Macroeconomics
and Growth
run
SorensenWhitta Jacobsen:
Basic Theory and
10
Chapter
Macroeconomics for
the long run and for
the short run
W
hat is the subject matter of macroeconomics? What methods and simpli(ying
assumptions do macroeconomists use in their efforts to explruiain how the
economy works? These are the Issues addressed In thls Introductory chapter.
The chapter explains the philosophy underlying our course in macroeconomics. In
particular, it explains why we have divided the main body of the text intou two Books. one
concerned with the long run, and another one dealing with the short run. Reading this
chapter will help you to understand the links between the following chapters and to see
where the material in each individual chapter fitsiyui into the big scheme. In addition, this
chapter presents some important concepts and elements of macroeconomic models which
we will use repeatedly in later chapters.
We start by discussing how to define macroeconomics. We then go on to explain why
it is useful to develop separate macroeconomic theories lor the long run and for the short
run. Finally, we end the chapter by summing up the dillerent assumptions underlying
macroeconomic models for the short run versus macro models for the long run.
1.1
Wh at is macroeconomics?
It seems natural for a macroeconomic textbook to start by giving a general definition of
macroeconomics. Yet you are not going to see a perfectly clear deflnition here; we simply
do not think that there is one. In fact. economists have even found it hard to oller a clearcut dellnition of the general science of economics.
The economist and Nobel Prize wim1er Gary Becker once said that 'economics is the
study of the allocation of scarce resources to satisfY competing ends'. 1 This certainly says
something essential about what (mainly micro) economics is. However, some parts of
macroeconomics are concerned '.vith situations where resources are not scarce. because the
available supplies of labour and capital are not fully utilized. These important realworld
situations would not fall within the realm of economics according to the above definition.
1. The idea behind this definition goes back to the nineteent h century, and the first scholar to give a comprehensive
statement of it was Lionel Robbins in his famous Essay on the Nature and Significance of Economic Science,
London, Macmillan, First edition, 1932.
1
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
2
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
INTRO DUC ING ADVANCED MACRO ECONOM ICS
Given the difficulties of providing a brief and accurate definition of economics, one
should not be surprised that a perfectly clear subdivision into micro and macroeconomics
is also elusive. It is sometimes said that macroeconomics is that part of economics which
is concerned with the economy as a whole. This suggests that microeconomics focuses
only on the small elements of the economy such as the single agent or the market for a
particular product. Although much of macroeconomics is concerned with the economy in
the large, this distinction between macro and micro is inaccurate. Important parts of
macroeconomics are not (directly) concerned with the whole economy, but rather with
understanding particular markets such as the labour market or the credit market. And a
large and important part of microeconomics, general equilibrium theory, is concerned
with the interaction among markets. that is. with the economy as a whole.
We therefore think that the best characterization of macroeconomics is one that
simply states the main questions asked in this branch of economics.
A definition of macroeconomics by subject
What creates growth in aggregate output and income per capita in the long run? And
what causes the fluctuations in economic activity that we observe in the short run? These
are the basic questions in macroeconomics. At the risk of oversimplilying. we may therefore say that macroeconomics is the study of eco11omic growth and lwsiness cycles. As we shall
see later on, to explain the movements in total output we must also understand the movements in total consumption, investment and the rate of unemployment, as well as the
interaction of these real variables with nominal variables such as the general level of
wages, prices, nominal interest rates, foreign exchange rates, etc. Hence macroeconomics
also includes the study of these variables.
A definition of macroeconomics by method
What we have offered above is a definition of macroeconomics by subject: macroeconomics is defined by the issues studied by macroeconomists. A strict 'empiricist'
version of this definition. which also involves the choice of method, is to say that macroecmwmics is concerned with explaining observed time series for eco11omic variables like GDP,
col!sumption, investme11t, prices a11d wages, the rate of unemployme11t, etc. This reflects the
view that a scientillc discipline should be dellned in terms of the data it seeks to explain.
We do not agree completely with this view. We do think there are some purely
theoretical scientific contributions that should be considered as part of the body of macroeconomics. However, like the rest of the macroeconomics profession , we are heavily
influenced by the empiricist view. To secure the link between theory and the real world,
theories should be evaluated by holding them up against the facts, and new theory should
ideally be justified and accompanied by illustrative empirical material. You will see that
our book very much reflects this approach. 2
2. According to this view it is important that one is familiar w ith the definitions of the main statistic al series on macro·
economic variables. We do not include the definitions in this book, since we assume that you know the basic
concepts of national income accounting and the standard measures of inflation and unemployment, etc., from an
earlier course in macroeconomics.
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
1 MACRO ECONOMI CS FO R THE LO NG RUN AN D FO R T HE SH ORT RUN
3
Why do macroeconomists aggregate?
The variables entering into macroeconomic models are typically aggregate variables
covering the economy as a w hole. For instance, in macroeconomics we often describe the
entire production side of the economy as if a single commodity were produced by the use
of just two dillerent inputs, capital and labour, both onedimensional variables which can
be represented by a number. By contrast, microeconomics studies disaggregated models in
which it is typically not 'allowed' to aggregate the production of. say, oranges and apples
into the production of fruit.
The aggregation undertaken in macroeconomic models raises obvious problems.
Take the concept oft he aggregate capital stock, for example. Capital is defined as produced
means of production, so it includes, lor example, buildings as well as computers. How
should the quantities of these two capital goods be added into one number representing
their combined productivity? In practice, the aggregate real capital stock is measured by
multiplying the quantities of the dillerent capital good~ by their respective prices in some
base year, and then adding up the values of the stocks of buildings, computers. etc., calculated at the fixed baseyear prices. This would seem to be a sensible way of measuring the
quantity of aggregate capital input provided the relative prices of th e different capital
goods remain reasonably constant over time. But we know that over the past decades the
relative price of computers has decreased enormously, and at the same time computers
have become tremendously more productive. Using relative prices to obtain an aggregate
measure of capital representing the productivity of produced inputs is therefore a dubious
procedure.
Nevertheless, the assumption that capital input as well as total output can be
represented by single numbers is standard in macroeconomics. How can we defend.
for instance, letting the production of myriads of dillerent goods and services being
represented by one number called 'aggregate output'? There are several lines of defence:
1. Over time the outputs of a lot of goods and services including capital goods do in fact
tend to move in the same direction. Given that the production volumes of most
industries tend to be positively correlated. Hence it seems defensible to use concepts
lil<e aggregate output or aggregate investment, eYen if we do not have a method of
measuring these aggregates which is fully correct in all circumstances.
2. The economy is such a complex mechanism that we cannot hope to explain and
describe it all in detail. To understand at least some of the economic regularities
observed, we have to make strong simplifications by abstracting from many details.
Aggregation of variables is one convenient way of simpli(ying.
3. Whether a certain h ighly simplifying assumption is useful or not is ultimately an
empirical question. If a model built on strong simplifications yields predictions which
accord with observed movements in some economic variables, then that model seems
useful for understanding (some of) the determinants of those variables. Presumably, it
will then also be usefu l for evaluating the eflects of economic policies intended to affect
the variables considered.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
4
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
© The McGrawHill
1. Macroeconomics for the
long run and for the short
run
Companies. 2005
INTRO DUCI NG ADVANCED MACRO ECONOM ICS
Economist Robert Solow (of whom you will hear a Jot more in this course) has put it
the following way:
All theory d epends on assumptions wh ich are not quite true. That is what makes it theory. The
art of successful theo rizing is to make the inevitable simplifying assumptions in such a way
that the final results are not very sensitive. 3
The long run versus the short run
As we have noted, macroeconomics seeks answers to the questions 'what creates growth
in GDP per capita in the long run?' and 'what creates fluctuations in GDP in the short
run?' It also tries to answer some related questions like 'what explains the level oflongrun
unemployment?', and 'what explains the shortrun variations in unemployment?'
Because these questions relate to dillerent time horizons, we have organized the main
body of this text in two Books, where the first is concerned with the longrun questions
above. and the second is concerned with the shortrun questions.
We lind both parts of macroeconomics very important, but one can argue that the
issues addressed in longrun macroeconomics are the most important ones. Consider a
poor country like Uganda whose GDP per capita is only about 3 per cent of GDP per capita
in the United States. A typical policy issue in longrun macroeconomics is: h ow could a
country like Uganda initiate a growth process that would gradually take it up to, say, 20
per cent of the US level? A typical policy issue in shortrun macroeconomics is: what could
a government in, say, the UK, Sweden, Denmark, or the Netherlands do to avoid an
increase in the rate of unemployment from 5 to 8 per cent which would otherwise follow
after a negattve shock to the economy? We do find the latter question very important,
but for anyone concerned with the longrun welfare of human beings, the first type of
question seems more essential.
1.2
Longrun and shortrun economic phenomena
................................................................................................................................................................................
The distinction bea.veen longrun and shortnm macroeconomics is first and fundamp,ntnlly n ilistinr.ti nn hPtWP.P11 thp, p /·rtmnmtma
WI'
want tn wrrfprstmrd. 'T'his IPnils to n
distinction between the jimdmnentnl characteristics of the models we use, and of the policies
we analyse. Below we will motivate these statements and explain the nature of the
difference between longrun and shortrun macroeconomic models.
Let us start by looking at some data. In Figs 1.1 and 1.2 we have chosen to focus on
two countries, the USA and Denmark, since these two nations can be seen as polar cases:
the USA is a large and relatively closed economy with a fairly small public sector (by
European standards), whereas Denmark is a small and relatively open economy with a
large public sector. We want to illustrate that, despite these differences, some important
qualitative features of the data are shared between the two economies. Similar figures
drawn lor countries like the UK, Sweden, the Netherlands, and for practically any
developed market economy would show the same qualitative features.
3. Robert M. Solow, 'A Contribution to the Theory of Economic Growth', Quarterly Journal of Economics, 70, 1956,
pp. 6594.
I Part 1  Basic Theory and
1. Macroeconomics for the
© The McGrawHill
Introducing Advanced
Empirics about Prosperity
long run and for the short
Companies. 2005
Macroeconomics
and Growth
run
SorensenWhitta Jacobsen:
1 MACROECONOMICS FOR THE LONG RUN AND FOR THE SHORT RUN
5
140
135
125
~
120
110
100
/
r:2d!_
115
105
~
USA
130
/
~
/
95
1873 1883 1893 1903 1913 1923 1933 1943 1953 1963 1973 1983 1993
145
140
135
Denmark
130
125
A/
120
~v
115
110
105
100
/
/
~
~
/
~
95
1873 1883 1893 1903 1913 1923
1933 1943 1953 1963 1973 1983 1993
Figure 1.1: Logarithm of real GOP in the USA and Denmark, 1873 1995
Note: Logarithm of GOP in 1873 normalized to 100.
Source: Angus Maddison,
The World Economy: Historical Statistics, Davelopment Centre S tudies, OECD, 2003.
In Fig. 1.1 the natural logarithms of the rumual real GDPs of the USA and Denmark
are drawn up for the period from 1873 to 1995, with the value in 1873 indexed to 100 lor
both countries. Figure 1.2 shows the average annual rates of unemployment lor the two
countries during the past century. In both ligures the actual data are represented by the
most solidly drawn zigzagged curves.
The ligures also include curves which are much more smooth. These are meant to
express the tre11ds ofthe relevant series. In Chapter 14 you will learn about a technique for
constructing such a trend (that technique has in fact been used here). For now let us just
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
6
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
INTROD UCI NG ADVANCED MACROECONOMICS
30,.
USA
25 +·~~
0 +,,~rrr~~~,~~~
1903 1911 1919 1927 1935 1943 1951 1959 1967 1975 1983 1991 1999
18.,
16 +~1
Denmark
14 +1~ 1
3
0+............,,J
1906 191 3 1920 1927 1934 1941 1948 1955 1962 1969 1976 1983 1990 1997
Figure 1.2: Annual unemployment rate in the USA and Denmark in the twentieth century
Source: RB. Mitchell, International Historical Statistics, Macmillan, 1998; US Bureau of labor Statistics; S.A.
Hansen, rlJkonorrisk Vaekst i Danmark, Bind 2, 191483, Akademisk Forlag, 1984; Statistisk Tiarsoversigt, Statistics
Denmark.
view the trend curves intuitively as the curves which a man in the street would draw with
a pencil if he were asked to illustrate the 'underlying movement" or the 'underlying
gravity level'.
In any event, the figures suggest that one way to interpret the movements of each of
the series is to view them as being made up of two components: a trend componerrt representing the overall evolution and captured by the smooth curves. and a cyclical component
representing the yeartoyear fluctuations and captured by the shifting vertical deviations
between the actual data curve and the trend curve.
I Part 1 
1. Macroeconomics for the
© The McGrawHill
Introducing Advanced
Empirics about Prosperity
long run and for the short
Companies. 2005
Macroeconomics
and Growth
run
SorensenWhittaJacobsen:
Basic Theory and
1 MACROECONOMICS FOR THE LONG RUN AND FOR T HE SHORT RUN
10
7
You may flnd that the cyclical components in Fig. 1.1 do not seem that great. To
highlight the fluctuations (around the longrun trend), we have drawn Fig. 1.3 which
shows the annual rate of change in real GDP. and the annual absolute change in the rate
of unemployment for the USA and Denmark for the shorter period 19802003. The
annual rate of change in GDP varies from sometimes higher than 6 per cent. dm>\ll1 to
sometimes around  2 per cent. and the rate of unemployment occasionally increases by
up to two percentage points in just one year, and a lso sometimes drops by two points in
one year. These cyclical movements are of considerable size.
Macroeconomics for the long run is about understanding the trends in series like
those just shm>\ll1, representing the longrun growth in GOP and the longnm or socalled
structural unemployment. respectively. Macroeconomics lor the short run is about
understanding the annual or quarterly .fluctuations in. lor instance. the GOP and the rate
of unemployment. Note how each part of macroeconomic theory corresponds to each of
the main questions asked in our deHnition of macroeconomics.
Most economists believe that an understanding of the trend requires a different type
of explanation than an understanding of the fluctuations. The dillerent macroeconomic
models are formal expressions of these different explanations. The fundamental assumptions of the models for the long run and for the short run therefore diller, and hence the
models themselves become dillerent. 4 But why should there be diilerent macroeconomic
theories for understanding the trends and for explaining the fluctuations? We will now
consider this question in more detail.
Macroeconomic theory for the short run
................................................................................................................................................................................
It may be illuminating to start with an example. Consider the Danish economy in 1993
and 1994 as illustrated in Fig. 1.3. In 1993 Danish real GOP was completely stagnant,
but then in 1994 it suddenly increased by about 5.5 per cent! This dramatic shift ailected
the rate of unemployment which dropped by almost two percentage points.
Exogenous shocks
A standard macroeconomic explanation for this shortrun fluctuation would start by
saying that in 1994 the Danish economy was hit by a positive exogerwus shock. that is, by
a sudden event which is best seen as coming from outside the (Danish) economic system.
Such an event could be either a supply shock like a sudden increase in the productivity of
resources, or a demand shock like a sudden rise in domestic consumption or investment
rooted. for example, in more optimistic expectations concerning the future, or in a more
expansionary llscal or monetary policy, or in a sudden increase in the demand for Danish
exports. Danish llscal policy was in fact relaxed in 199394. and the real interest rate
4. Not all economists think that different assumptions are necessarily needed to explain short·run and long·run macro·
economic phenomena. Some believe that one model can be used fer understanding both growth and fluctuations,
since they do not see the f luctuations as coming (only) from exogenous shocks, but (also) from forces w ithin the
economic system itself. This hypothesis of endogenous business cydes is too advanced for the present course.
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
8
I Part 1 
Basic Theory and
Empirics about Prosperity
and Growth
© The McGrawHill
Companies. 2005
1. Macroeconomics for the
long run and for the short
run
INTROD UC ING ADVANCED MACROECONOMICS
Annual relative change In real GOP
USA
8
6
c
4
tf
2
~
u~
0
2
n
I
n
....
r1
u
nnn
....
~
LJ_]....J_I
II
1990
1985
1980
1995
2000
Annu al absolute change In the rate of unemployment
USA
UJ
c
·o
a.
Q)
C!)
1.5
c"'
 0.5
0
 1.5
Q)
Q;
a.
~rlrr
0.5
 2.5
~ ~
1980
1985
nnr
UU''
Ll
1990
....
L..l''
1995
nn;: :;
2000
Annual relative change In real GOP
Denmark
6
5
4
3
fl_
2
1
0
1
2
3
flJlj l
.J=u
:u~

1985
1980
2
f
~ ITII
u~=
 0.5
1
 1.5
2
1990
flnJ Ln.nJl
1995
U
2000
Annual absolute change In the rate of unemployment
Denmark
1.5 f0
H1,
1980
1985
n11n nn11
1990
I
.I
I [1
r~l
~l l u
1995
LJ
2000
Figure 1.3: Annual relative change in GOP and absolute change in the rate of unemployment
Note: The origins of t he columns indicating relative changes in GOP are the average annual growth rates of GOP,
which are 2.97 and 1.74 for the USA and Denmark, respectively. The actual annual growth rates are the distances
from zero t o the :opjbottom.
Source: O ECD, Economic O u11ook Database.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
1 MACRO ECONOMI CS FO R THE LO NG RUN AN D FO R T HE SH ORT RUN
9
fell at that time. Moreover. the Danish export markets in Europe started to grow more
quickly in 1994. Thus we may assume that the 1994 shock was mainly a positive demand
shock.
To say that the sudden increase in Danish production was caused by an exogenous
shock, basically an unexplained event. is not exactly deep. But note two things. First,
economies are hit all the time by events which are best considered as exogenous from the
viewpoint of economic theory. For example. economists should not be concerned with
explaining fluctuations in harvests due to shifting weather conditions, and probably they
should not try to explain all sudden changes in the moods of consumers and investors.
Focusing on the small economy of Denmark. a sudden increase in the demand for exports
due to events in foreign countries should also be considered exogenous. Second. the occurrence of the shock is not the end of the story. A shock may be what initiated the change in
economic activity, but it cannot itself explain all the subsequent economic reactions of
households and firms. There is more explanation to do.
Nominal rigidities
For an increase in the aggregate demand for goods and services to lead to an increase in
production . as observed in Denmark in 1994, it must have been prolltable lor Danish
flrms to increase their supply of output to accommodate the increase in demand. In the
short run the capital stock is more or less given, so production can only increase if more
labour is used. As more workers come to utilize the given capital stock, the marginal
productivity of additional hours worked is likely to decline. If the marginal productivity of
labour is indeed falling, it seems plausible that frrms will only want to expand employment
if the real wage falls. The real wage rate is dellned as W/P. where W is the money wage
rate and P is the price level. Presumably the higher demand for goods and services will
lead to some increase in P. If the nominal wage rate W is rigid in the short run. this rise in
prices will indeed cause a fall in the real wage, inducing firms to hire more labour to
increase their supply of output.
Thus another key ingredient in our explanation of the fluctuation is the assumption
of some short run nominal rigidi ty, in this case a rigid money wage. The assumption th at
nominal wage rates are fixed for a certain period of time is quite realistic. Firms and
workers do not renegotiate their wage rates every day or every month. because negotiation is a timeconsuming process involving the risk of unpleasant and costly industrial
conflict.
But experience also indicates that the nominal prices of most goods and services are
only adjusted with certain time intervals. In an interesting empirical analysis of newsstand prices of American magazines. economist Stephen G. Cecchetti found that under an
average general inflation of 4 per cent per year magazine prices were only changed every
6 years on the average. This means that on average the real price of a magazine is eroded
by about 2 5 per cent by inflation before the nominal price is changed. 5
5. Stephen G. Cecchetti: 'The Frequency of Price Adjustment: A S tudy of the Newsst and Prices of Magazines', Journal
of Econometrics, 31 , 1986, pp. 255274.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
10
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
INTRODUCING ADVANCED MAC ROECONOM ICS
The rigidity of nominal magazine prices is not just a special case. In an empirical
analysis of price rigidity. economist Alan Blinder asked a sample of business managers:
'How often do the prices of your most important products change in a typical year?' About
50 per cent of the managers responded that they only changed their prices once or less
than once a year. 6 Explaining why (most) firms do not immediately adjust their prices in
response to changes in demand and cost is an intriguing issue to which we return in
Section 5.
We argued above that a rigid nominal wage in association with a flexible. upward
adjusting nominal price could create the fall in real wages that would make it proiHable for
llrms to supply more output in response to a positive demand shock. If a fall in the real
wage is the typical reason why llrms want to increase their output in reaction to a positive
demand shock. we should expect to observe a negative relationship between output and
real wages. However, output often increases without a simultaneous decrease in real
wages. It is therefore important to ask if an increase in the demand for output can induce
llrms to increase their supply even if all nominal prices are llxed in the short run (so the
real wage does not fall)? The answer is 'yes', provided that prices are above marginal costs
before the demand shock hits the economy. In practice most markets are characterized by
imperfect competition where llrms have some monopoly power enabling them to charge
prices which are indeed above marginal costs. In that case they '>\Till be able to increase
their total prollt by increasing their output in response to an increase in demand, even if
they have to keep their prices temporarily tlxed.
The basic point is that shortrun nominal price or wage rigidity can explain why an
exogenous increase in nominal aggregate demand leads to a shortrun increase in real
output and employment. If nominal prices are flxed. all of the increase in demand '>\Till be
reflected in a rise in real output, because imperfectly competitive llrms will be happy to
increase their supply as long as their (fixed) prices remain above marginal costs. And even
if prices increase in response to higher demand. a rigid nominal wage means that the price
hike will drive down the real wage which in tum will stimulate employment and output.
In practice, both nominal wages and nominal prices are rigid in the short run, although to
different degrees in diflerent markets.
Expectational errors
So far our explanation for the shortrun fluctuation in Danish output has relied on two
ingredients: exogenous shocks and shortrun nominal rigidities. But there is a third ingredient which is essential for a full understanding of the fluctuation: expectational errors. To
see this. consider the Danish boom in 1994 and suppose that there was in fact some fall in
the Danish real wage, as some prices increased in response to growing demand and
nominal wages Jagged behind. Faced with falling real wages, why were Danish workers
nevertheless willing to increase their supply of labour, thereby enabling firms to expand
employment and output? One possible answer is that trade unions in the heavily unionized Danish labour market had pushed real wages above the marginal disutility of work so
6. Alan S. B linde·: 'On S ticky Prices: Academic Theories Meet the Real World', in N.G. Mankiw, ed.,
Chicago, University of Chicago Press, 1994.
Monetary Policy,
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
1 MACRO ECON O M IC S FO R THE LONG RU N AND FOR TH E SHO RT RUN
11
that some workers were involuntarily unemployed prior to the demand boom. By dellnition. a worker who is involuntarily unemployed is willing to take a job even if the real
wage falls below its current level (provided it does not fall too much).
But this hypothesis only begs the question why trade unions faced with growing
labour demand would allow the real wage to fall? If unions were bargaining in 1993 to
achieve a certain real wage w for 1994, why would they suddenly accept that the real
wage fell below the target for 19 94 at a time when labour demand was increasing? The
most plausible answer is that the fall in real wages was uninte11ded by unions. If unions
h ad perfectly anticipated the positive 1994 demand shock and its ellect on the price
level, they would h ave bargained for a higher 1994 money wage rate to secure their
target real wage w, that is. they would have demanded a money wage satisfying
W = P·w.
However, since (most) wages had to be set before the occurrence of the shock. and
since the shock was 11ot perfectly foreseen, the negotiated money wage had to be based on
the expected price level P' which did not include the full inllationary effect of the shock:
W = P' · w < P · w. When the shock hit and prices increased above their expected level,
unions were locked into their nominal wage contracts. and given the employer's right to
hire more workers at the negotiated money wage, unions had to allow their members to
supply additional labour even though the realized real wage. W/P . turned out to be lower
than the target real wage, W/P'' = OJ .
This example illustrates our point that business lluctuations typically involve expectational errors, in this case errors made by workers (trade unions). In the case where some
prices as well as money wages are rigid in the short run, an unanticipated shock will also
cause some firms to err in their expectations. When firms preset their prices for a certain
period. they will base their pricing decisions on their expected costs which will be influenced by the expected general price level P''. When the unanticipated shock hits the
economy. some firms (call them Group 1) will be just about to adjust their prices and will
be able to account for the inflationary cost eflect of the shock. But many other firms
(Group 2) wh ich have recently reset their prices will choose to maintain their existing
prices for a while, even though the increase in the prices charged by Group 1 1lrms drives
the costs of Group 2 flrms above the previously expected level. As long as the shock does
not push marginal costs above the preset prices, even Group 2 llrms will want to expand
their output to accommodate the unexpected rise in demand.
Macroeconomics for the short run: summing up
We may sum up the points of this section as follows: mo.croeco11omic theory for the short ru11.
intended to explain the economic fluctuations from year to year or from quarter to
quarter, typically includes the following three modelling features:
1. exogenous shocks, i.e., sudden abrupt influences on the economy coming from changes
in preferences, technology, or economic policies.
2. shortrun nominal rigidity, i.e., some period after the occurrence of a shock during
which some prices and/or wages are sticky.
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
12
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
INTRODUCING ADVANCED MAC ROECONOM ICS
3. expectational errors, i.e .. a period after the occurrence of a shock during which some
prices are different from what was expected before the shock. 7
1.4
M;~g~g~~.~~~.~!~..~9.:~~EY...f.?.~.~.~.~. ~~.~J~.E~.~.............................................................
While exogenous shocks. temporarily sticky wages or prices and erroneous expectations
are required for an understanding of the changes from year to year in unemployment and
GDP. most economists think that these features are best disregarded when we try to
explain the 'gravity level' of the rate of unemployment and the trendwise gradual growth
of GDP over long periods. The smooth trend curves in Figs 1.1 and 1.2 could not possibly
reflect a succession of random shocks over the more than 100 years considered. By definition, shocks have to go in opposite directions from time to time. If technology improves
constantly each year to imply a 2 per cent increase in GDP per head, then this annual shift
in technology should not be considered as a shock. but rather as a foreseeable gradual
movement. Moreover. although nominal wages and prices may be sticky in the short run.
they do adjust in the longer run. In macroeconomics for the long run we therefore
abstract ti·om the three features which define shortrun macroeconomics.
Longrun modelling: the basic assumptions
In other words. macroeconomic theory for the long run, intended to explain the trendwise
movements of main economic variables around which the yeartoyear fluctuations
occur, portrays the economy as if exogenous shocks do not occur, i.e., the economic fundamentals like preferences and technology develop in a smooth and foreseeable way over
time; prices are ji.tl/y adjusted in all periods in accordance with the economy's full longrun
price flexibility ; and expectations are correct all the time.
You should carefully note the 'as if in this definition. If the assumptions of macroeconomics for the short run give the right picture. then in (almost) every year the
economy will be reacting to shocks, with prices still not fully adjusted and expectational
errors still prevailing. This is because new shocks occur all the time. Nevertheless, certain
phenomena may be better understood by considering the economy as if shocks did not
occur,
pric~s w~r~
always fully
aujust~d.
ami
~xpc::ctalious
always correcl. Among such
phenomena we include the longrun growth performance and the longrun gravity level
of unemployment. Thus macroeconomic models lor the longrun describe the underlying
longrun equilibrium towards which the economy is gravitating. even though recurring
7. Some short·run macroeconomic theories do not assume price rigidity, e.g. the theory of Milton Friedman present ed in
'The Role of Monetary Policy', American Economic Review, 58, 1968, pp. 1 17, and related more formal contri·
butions such as Robert E. Lucas, Jr: 'Expectations and the Neutrality of Money', Journal of Economic Theory, 4 , 1972,
pp. 103 124, or the real business cycle theory explained in Sect ion 4 of Chapt er 19. A main point in these t heories is
that fluct uations can be understood w ithin a framework t hat only assumes shocks and possibly expectational errors.
Hence one cannot conclude from the mere observation of fluctuations t hat wages or prices are rigid: fluctuations are
theoretically compatible with all w ages and prices being fully adjust ed all the time, as we shall see in Chapters 18 and
19. In fact, a main point of real business cycle theory is that one and t he same class of economic models can be used
for understanding both longrun growth issues and shortrun issues of fluctuat ions. The models in this class are much
like t he growth models to be dealt with in subsequent chapters. The only rele·1ant element to add to the growth models
to make them also models of fluctuations is, according to this view, exogenous supply shocks. We find this theo·
retically interesting, but today it is widely accepted that short·run price and wage rigidities are indeed important for
understanding the economy's short·run reactions to shocks and t hat demand shocks are important.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
1. Macroeconomics for the
long run and for the short
run
Empirics about Prosperity
and Growth
© The McGrawHill
Companies. 2005
1 MACRO ECON O M IC S FO R THE LONG RU N AND FOR TH E SHO RT RUN
13
shocks and the timeconsuming adjustment to these shocks imply that the economy is
never exactly in this longrun equilibrium.
To put it another way. in explaining how economic growth in Denmark could change
from zero per cent in 1993 to 5.5 per cent in 1994. and unemployment at the same time
be signiftcantly reduced, we have to allow lor exogenous shocks, price rigidities and
expectational errors. In an explanation of the average annual GDP growth of 3.4 per cent
over the period from 19 50 to 1998, or of the average rate of unemployment of 5.3 per
cent. it L~ better to abstract from these three complicating features to focus attention on the
fundamental forces governing the longrun evolution of the economy.
Real rigidities and natural rate theory
The assumption in longrun macroeconomics that all wages and prices are tully adjusted in
all periods means that there are no nominal rigidities in the long term. But there may well be
permanent real rigidities preventing real prices and wages from adjusting to the values
which would prevail under perfect competition. The economy's longnm wage and price
flexibility can be different from the flexibility assumed in the traditional model of perfect competition. We will now go through a simplified example to illustrate what we mean. For concreteness, we will show how real rigidities may arise from the market power of trade unions.
but we emphasi7..e that even if unions are weak or nonexistent, there are other mechanisms
which may cause significant real rigidities, as our analysis in Chapter 12 '.viii make clear.
Consider an economy divided into many different sectors. each represented by a fum
producing a dillerentiated product, and each having an industry trade union which controls the supply oflabour to firms in the sector. Suppose lor simplicity that labour is the only
variable factor of production and that it takes one unit of labour to produce one unit of
output in each sector. The average and marginal cost of production for the individual fum
will then be equal to the wage rate for its industry. It is well known from microeconomics
that a prolltmaximizing firm facing a downwardsloping demand curve '.vith a constant
price elasticity of demand will set its price as a constant markup over its marginal cost. If
the average nominal wage level is W, the average price level P may thus be specilled as:
P = m1'W,
mP> 1,
(1)
where mPis the markup factor of the representative llrm.
Now consider the trade union monopolizing the supply of labour to a particular
sector i. To simplify. suppose the union is so strong that it can ellectively dictate the
nominal wage rate Wi to be paid by the firm representing sector i. The trade union sets
the wage rate with the purpose of maximizing the objective function Q , where:
Q=
w )
(j" v Li'
(2)
Here Li is the level of employment in sector i, IA.';/P is the real wage for union members
employed in firm i, and v is the real income which members are able to earn elsewhere if
they fail to find a job in sector i. The variable v is sometimes referred to as the workers'
'outside option' and '.viii be specilled below. Since {11.1JP  v is the rent (surplus) which the
in dividual union member earns by being employed in firm i rather than having to look
for a job elsewhere. the magnitude Q is the total rent that union mem bers gain from
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
14
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
© The McGrawHill
1. Macroeconomics for the
long run and for the short
run
Companies. 2005
INTROD UC ING ADVAN C ED MA CROEC O NOMI C S
employment in firm i. In Chapter 13 we shall explain in detail why it is reasonable to
assume that the trade union wants to maximize this total rent; lor the moment we ask you
to accept this plausible assumption.
When setting the wage rate Wi, the union must take into account that a higher wage
will induce the firm to hire fewer workers. If the union charges an average wage rate so
that W/W = l, the firm in sector i will h ave the same cost level as its competitors in other
sectors and will experience an average volume of sales. The union may then expect to gain
a proportional share of total employment. so employment in sector i will be L/n. where Lis
aggregate employment and n is the number of diflerent sectors. But if the union for sector
i raises its relative wage rate, W/W. its employer \>Viii have to raise his relative output
price, leading to a lower volume of sales and employment in the sector. Noting from (1)
that W = P/mP, we may thus specify the demand for labour in sector i as:
(] > 1.
( 3)
where the parameter a measures the numerical wage elasticity oflabour demand. Since
the trade union for sector i is small relative to the rest of the economy, it takes the outside
option v and the general price level Pas given. By setting the nominal wage rate, Wi, the
union thus implicitly chooses its real wage. ~/P. Using (3) to eliminate Li from (2). and
differentiating the resulting expression for Q with respect to Hl;. we obtain a necessary
(first order) condition for the optimality of the wage rate set by the union. Since the logfunction is increasing we may as well difl"erentiate the log of Q, which is convenient here:
din Q
1
1
1
 =     a  =0
dW
w,_ v P · W
r
W;
p
r
\V
p=m u.
(f
mw => 1.
(4)
a 1
We see that the union will set the real wage as a constant markup over the outside
option. Let us take a closer look at this option. If the unemployment rate is u, the representative union member has a probability 1  u of finding another job if he has to look for
work outside sector i. In that case he may expect to earn the average real wage W /P.There
is also a probability u that the union member \>Viii remain unemployed if he does not obtain
a job in sector i. In that situation the worker will receive a real unemployment benefit, b,
assumed to be lower than the average real wage. The outside option. defmed as the real
income a worker may expect to earn outside sector i, may thus be specified as:
w
u = (1  11)  + ub.
(5)
p
From (4) and ( 5) we may derive a simple theory of longrun unemployment which
will help to clarify the concept of real rigidities. Inserting (5) into (4 ), and denoting the real
wage rates inside and outside firm i by wi and w, respectively, we get:
wi = m"'[(l  u)w + ub].
w
IV= .
p
(6)
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
1 MACRO ECON O M IC S FO R THE LONG RUN AND FOR TH E SHO RT RUN
15
Assuming that the union markup factor mwis the same across sectors, it follows from (6)
that trade unions will charge the same wage rate in all sectors. Hence we may set W; = w
in ( 6) and solve for w to get:
W=
m'"u
1  mw(1  u)
b=
1
b
 l ..!.'
1 _ m"'m"'
u
(7)
where we assume u > (rn w 1)/m"" for a meaningful expression. Equation (7) is the
economy's wage cun1e showing how the real wage claimed by unions depends on the
unemployment rate and the rate of unemployment benefit: under the condition just
stated, the real wage claim falls if unemployment goes up.
In a macroeconomic equilibrium the real wage claimed by unions must be consistent
with the real wage implicitly oflered by firms through their price setting. According to (1),
firms implicitly offer the real wage IV= vV/P = 1/m". Inserting this into (7) and solving for
u, we get the longrzm equilibrium unernployment rate, denoted by U:
11 =  ,w =
1  mPb
m"' 1
m w(1  m~'b)
.
You may check for yourself that this equilibrium unemployment rate will be positive but
smaller than one provided that mwm"b < 1 or, equivalently. b/IV < 1/rn"'. In other words,
as long as the replacement ratio b/ w ollered by the system of unemployment insurance is
not 'too· high, we get a meaningful solution for the equilibrium unemployment rate. (Our
assumption mwmPb < 1 implies IV> b as well as u > (m w 1)/m "" as assumed above.
which you should check.)
Note that a positive unemployment rate emerges even though we did not assume any
nominal rigidities: the nominal variables Wand P can be fi:'eely scaled up or down in the
equations above.
The unemployment problem is rooted in the market distortions reflected in the
markup factors m" and m w. If there were no unions and labour markets were perfectly
competitive, workers could not drive wages above their outside option. According to our
simplified model equilibrium unemployment would then be zero, as you can see by setting
m w = 1 In ( 8). But unions do exist and do have market power. g In the situation where
mw> 1, the unemployment generated by union monopoly power will be ~:acerbated by
imperfect competition in product markets. This is because weaker product market competition implies a higher value of the price markup. mt'. and according to (8) this 11Vill
increase the value of u: when firms drive down the level of real wages by pushing prices
above marginal costs, the amount of involuntary unemployment must rise to bring union
wage claims down in line with the real wages implicitly otlered by firms through their
monopolistic price setting. We also see from (8) that a higher level of unemployment
benefits will raise the equilibrium unemployment rate when unions have market
power.
8. You may ask w hy rational workers would want to form trade unions if unions create involuntary unemployment
Chapter 13 w ill offer an answer to this.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
16
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
1. Macroeconomics for the
long run and for the short
run
© The McGrawHill
Companies. 2005
INTROD UC ING ADVAN C ED MA CROEC O NOM IC S
When economists speak of'real rigidities' , they often mean that market imperfections
permanently distort the real prices and real wages claimed by Hrms and workers away
from the competitive relative (real) prices which would ensure full resource utilization.
Even in the long run, when all nominal wages and prices have had time to ft1lly adjust to
their desired levels, these structural market distortions will persist, leaving the economy in
a state of permanent unemployment. Our analysis suggests that the degree of real rigidity
in the economy may be measured by the size of the parameters m ••, m Pand b which contribute to equilibrium unemployment. Alternatively, we might simply measure the degree
of real rigidity by the size of the equilibrium unemployment rate, since we have seen that
IT = 0 in the benchmark case of perfect competition.
According to established tradition in macroeconomics, the longrun equilibrium
unemployment rate implied by the economy's real rigidities is called the rwtural rate. This
is the rate of unemployment emerging when all relative prices have fully adjusted in
accordance with the economy's longrun wage and price flexibility. In line with this tradition, we will use the term 'natural rate of resource utilization' to denote the rate at which
factors of production are utilized in longrun equilibrium when the economy has
exhausted its potential for price adjustment. By dellnition, there are no real rigidities if
perfect competition prevails, since real (relative) prices in a competitive economy adjust
until the demand for each resource equals the full supply ofthat resource. Th is just means
that the natural rate of resource utilization is 100 per cent. If real rigidities do exist, the
natural rate of resource utilization will be less than 100 per cent.
We have suggested that the degree of real rigidity may be measured by the level of the
natural unemployment rate. We should add that economists sometimes Hnd it fruit!ul to
work with a slightly dillerent concept of real rigidity. In this alternative definition . the
degree of real rigidity is measured by the responsiveness of real wages and real prices to a
short run cha~~ge in the unemployment rate away from the natural rate. If this responsiveness is low, the degree of real rigidity is said to be high . In Section 5 we \<\Till adopt this alternative notion of real rigidity which \<\Till turn out to be useful for explaining shortrun
nominal rigidities. At any rate, real rigidity in a broad sense refers to the fact that the
economy's degree of relative price flexibility is less than it would be in an ideal world of
perfect competition.
The crucial role of the supply side in longrun modelling
As indicated by the simple example in equation (8) . the natural rate of employment. 1  fi,
is given from the supply side of the economy, since the parameters m••, mP and b are
characteristics of the economy's supply side structures. This has an important implication: in macroeconomic models for the long run, where wages and prices are assumed
to be fully adjusted in all periods, output is determined solely from the supply side. In any
given period there is a certain labour force L, and a certain (predetermined) capital stock
K. Output is then completely constrained from the supply side, since it cannot exceed
the volume which can be produced by means of the labour input (1  u)L and the
predetermined capital input K.
By contrast, in shortrun macroeconomic models nominal wage and price rigidities
and/or expectational errors may cause employment to deviate from its natural rate.
Hence we do not have the simple supply side determination of employment just described.
One then just describes the economy in the end period. Static versus dynamic models A macroeconomic model for the long run can be a singleperiod static model. employment is also intluenced by the aggregate dema11d for goods and services. but sometimes it is useful to focus on a single period which is an 'end period' in the sense that no new shocks have occurred for a long time and the economy has Hnished all its adjustments.Basic Theory and Empirics about Prosperity and Growth 1. This may seem surprising. Macroeconomics for the long run and for the short run © The McGrawHill Companies. and typically they also describe the evolution of other important stock variables such as the labour force. the stock of natural resources. . dynamic models for the long run describe the process of capital accumulation explicitly. etc. Solow who made fundamental contribut ions to growt h theory. The purpose is to characterize the equilibrium towards which the economy is tending in the long run. for analysing the eflects of economic policies. these models have the shortcoming that they do not contain utility functions and therefore do not allow an explicit evaluation of the welfare consequences of alternative economic policies. are called Solow models. without complicating the theory with an explicit analysis of the dynamic process which takes the economy to the longrun equilibrium. and simply assume that it is 'good' to have high per capita income or consumption. We leave these more complicated models lor a later macroeconomics course. 9. These models. Some of Solow's contribut ions are the subjects of Chapters 3 and 5 below. In Solow models one will just have to look at the policy implications lor economic variables like GDP or the level of consumption. in longrun macroeconomic models employment always corresponds to the natural rate.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . However. They ha\'e t his name after the Nobel Prize Laureate Robert M. is the amount of capital available in period t. Thus. 0 < (5 < 1. always contain the following dynamic link between the current flow of investment and the increase in the stock of capital: (9) where K. In many models it is simply assumed that households always save an exogenous fractions of total income so that in any period t we have S1 = sY1• The growth models using this assumption. and c~ is the rate at which capital depreciates. In a closed economy the level of gross investment I1 must equal gross savings S1• Por a closed economy the capital accumulation equation therefore becomes: (10) Thus savings play a central role in growth models. Solow models are well suited and widely used lor understanding many growth issues. also called growth models. By contrast.9 There are more advanced growth models in which savings are derived from maximization of utility functions under appropriate budget constraints. whereas in shortrun models employment is determined also from the demand side and lluctuates around the natural rate. 2005 1 MACROECONOMICS FO R THE LONG RU N AND FOR THE SHORT RUN 17 Instead. I 1 is gross investment in period t. The models of longrun structural unemployment presented in Chapters 12 and 13 will be of this nature.
. Gregory Mankiw: 'Small Menu Costs and Large Business Cycles: A Macroeconomic Model of Monopoly'. like the costs to a restaurant of having to print a new menu card..... = (11) 10.. 57.. For convenience... Later on. where the replacement rate c < 1 is a constant.. Similarly.... . wage setters may have to spend some resources estimating the exact magnitude of the change if they want to be sure that they set the optimal wage rate. union leaders may have to spend some time sitting down at the bargaining table with the employer to write a new wage contract..... and hence they are usually neglected in economic models. then the privately optimal behaviour may be to leave the wage rate unchanged to avoid the cost of wage adjustment. A seminal contribution to the menu cost theory of nominal rigidity was made in the highly readable article by N. As we shaU demonstrate in this section.... For instance......menu costs'.... we \>\Till show that when target real wages are not very responsive to changes in the level of economic activity............... Moreover. pp.. Hrms may incur small costs in resetting their nominal prices..(1 .. We noted earlier that some form of nominal rigidity typically plays an important part in the explanation of shortrun business fluctuations.. Specillcally.. the menu cost theory was elaborated by Mankiw and several other so·called New Keynesians. we now assume that the system of unemployment insurance offers a fixed compensation relative to the average wage level... If this loss is very small..... Quarterly Journal of Economics. Macroeconomics for the long run and for the short run © The McGrawHill Companies... For example... 100...... 183203..Basic Theory and Empirics about Prosperity and Growth 1..... 1985.c)u)w.................... 2005 INTROD UC ING ADVAN C ED MA CROEC O NOMI C S 1.... when the demand for labour changes... 10 Explaining nominal wage stickiness We start by considering the circumstances in which nominal wage rigidity may be the outcome of optimizing economic behaviour...... 529539.......... It is not unreasonable to assume that there are some small costs of adjusting nominal wage rates... they may have to print new catalogues or spend other resources communicating the new prices to the market... Recalling that w W(P... pp. we have a potential explanation for the nominal rigidities which seem to exist in the real world..... including Laurence Ball and Da·1id Romer: 'Real Rigidities and the Non· Neutrality of Money'. But if we can show that even very small menu costs may be sufficient to induce economic agents not to adjust nominal wages or prices in response to even considerable shocks................. 1990.... Review of Economic Studies..... The costs of adjusting nominal prices or wages are often referred to as ... In most cases such costs are likely to be quite small.... Our strategy is to measure the loss incurred by individual wage setters if they do not adjust their nominal wage rate in reaction to a change in the unemployment rate of a realistic magnitude...5 Explaining nominal rigidities .. even very small costs of nominal wage or price adj ustment may be sullkient to induce rational agents to abstain from adjusting nominal wages or prices in the short run in response to fluctuations in economic activity. the outside option (5) may then be \vritten as a function of the unemployment rate: = v = v(u) = [1 . the phenomenon of real rigidity discussed in the previous section may help us understand how shortrun nominal rigidity may be compatible with rational economic behaviour...........SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 18 I Part 1 .......... We base our analysis on the simple model of wage formation developed in the previous section.......... b cw..
. it must satis(y the ilrstorder condition. ' ' 1 (1 2Q(w. it is convenient to take a secondorder Taylor approximation of Q (l\'. = m'•v(u) given by (12). since we know from (1) that w will remain constant at the level1/rn1' as long as nominal prices are not rigid.'. we get the optimal real wage which the union would choose in the absence of any costs of wage adjustment: = W.jP. implying L = 1 . To estimate how high C will ha ve to be to prevent wage adjustment.Jacobsen: Introducing Advanced Macroeconomics I Part 1 .. If it chooses not to change its wage rate. . (1 3) We can now analyse whether the union in a speciflc sector would actually want to adjust its own real wage W..w. ' (1 5) ' Note that since W. (w.• u). so nominal wage rigidity would be absent.) . and in that case its nominal wage will be rigid. Q (w.. Hence we may approximate the utility loss in (14) by: (16) . the union will suffer the following utility loss. Macroeconomics for the long run and for the short run Empirics about Prosperity and Growth Companies.Whitta. leaving it at t he initial level IV . Suppose that the union in sector i has initially set a real wage rate"' ..: = Oin optimum _ (1Q (w..c)u] w. we may write the objective function for the union organizing the workers in sector i as: L. calcula ted around the new optimum wage rate w. Hence the union will want to maintain the real wage rate IV. . V. u) + (w... this would initiate an economywide adjustment of nominal wage rates. relative to the constant average real wage if the unemployment rate were to change. In that case th e w1ion would want to change its nominal wage rate l\~. th e utility loss in (11) is directly comparable to the fiXed real cost of nominal wage adjust ment. excluding adjustment costs: (14) Since the union's objective function (and hence UL) is specified in terms of real income.) ". the union would then want to set the real wage W. .. Inserting (11) into (4) and remembering that w.Sorensen.(1 ... (12) Without loss of generality.lt) _ Q (w. if UL < C. u) = (w. And if all unions had an incentive to adjust their own real wage relative to the market real wage. = mwv(u) = mw[1 . = 0 .) 2 _ . since this is the only way it can aflect its real wage rate.. to avoid the costs of wage adjustment. even though real wage rates would stay constant due to the markup price setting of tirms. we may ch oose our units of measurement such that the total labour force is normalized to unity. 2005 1 MACRO ECON O M IC S FO R THE LONG RUN AND FOR TH E SHO RT RUN 19 We do not state explicitly that the outside option v is also a function of the average economywide real wage w.. denoted by C. u) "" Q (w.Basic Theory and © The McGrawHill 1. (1Q /i1w.. = mwv(ft) corresponding to some initial level of unemployment it. Suppose further that an exogenous shock drives unemployment to a new and higher levelu. UL.11.w) ' ' dW. .v(u))( 1 : 11 )(m1' w.. In the absence of any costs of nominal wage adjustment. u) + 2 (dw. is the new optinwl wage rate (in the absence of adjustment costs). Inserting this into (3) a nd substituting the resulting expression into (2) along with (11).
= w.( 1 . dividing by w1L..2} Inserting m " = r1j(a. Since W. it follows that: t:i'."" 0.? {~)[(m•w).1) plus the fact that (4) implies (w.w.2~ .) 2 ' ' v)/w. we get: =a(mP)(~)(m•w)·!•• •>.broadly interpreted to include the costs and inconveniences of re\vriting wage contracts plus the resource costs of reestimating the optimal wage.(1 . (1 .C ~ ")(mPt.W."" (17) Before the shock to unemployment.1] n =am•f ~ u)(m•w) !"' '>[(a + 1)(w~ "). we lind that: 11 = (rr1)(' v. suppose that the replacement rate cis equal to 0 . implying a lm·ge 40 per cent increase in unemployment.v)/w.c)u · Not surprisingly.02. First we note that when b = cw = cfmP.L. .05 and c = 0. = w.L.1) and (w1 (ilw. 2005 INTRODUCING ADVANCED MAC ROECONOM ICS The second derivative in (16) may be calculated from (13 ).c)(u . the costs of wage adjustment . assuming a set of parameter values which generate a realistic initial unemployment rate. Armed with Eqs (17) and (18).v.002 . 5 (not unrealistic for the average worker in many OECD countries) and that the economy is initially in long run equilibrium at an unemployment rate equal to 5 per cent.v)(m"w' ·•'J ~ n iJ2 Q P =amP(~)[mP(a + 1)(w1 . Macroeconomics for the long run and for the short run Empirics about Prosperity and Growth © The McGrawHill Companies.)a (where the expression lor w.) . and expressing UL as a fraction of the union's total wage bill w.. one llnds that UL/w. optimally in (18) 1 . and hence the larger the fall in the value of the outside option of union members.it = 0. In other words. Inserting these parameter values into (17) and (18).2(mPw) .ii.amP(w.= 1/a into the latter expression. suppose that the increase in the unemployment rate is two percentage points. we end up with (17). ")(m"w1)~.only have to amount to about onefifth of 1 per cent of rr the wage bill to make it suboptimal for the individual union to reduce its nominal wage 11. u . is also given by (12). we can calculate how high the costs of nominal wage adjustment will need to be to prevent such adjustment. the larger is the union's utility loss from not adjusting the wage rate.L. = 1/C1. From (13) we have (}Q = Ow.v)(mPw)~. u . and using the definition of m" again.L. 2 (} Q (ow. Doing this.Basic Theory and 1.ii) 1:i1.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 20 I Part 1 . = W. the expression for the natural unemployment rate given in (8) simplilles to u= mw1 1 m"'(1 c)  (19) a(1 c) As an example. using the dellnition m"' e1/(e1 .. it follows that = 40. follows from (3)). 5 into (19). Finally. we see from (17) and (18) that the larger the rise in unemployment. Inserting it = u = 0 .2 . the union had set its wage rate accordance with (12) . mw n ' Substituting :his expression into (16). 2 UL W. .
Jacobsen: Introducing Advanced Macroeconomics I Part 1 .SorensenWhitta. and as a consequence the general nominal wage level will be rigid. there may still be nominal price rigidity. On the other hand. . the degree of real wage rigidity is indeed quite high. and ifthe nominal wage rate is rigid.111. Explaining nominal price rigidity The menu cost theory may also explain the rigidity of nominal prices. Thus the theory outlined here may indeed be an interesting candidate for explaining nominal wage rigidity. Hence (IV. because a change in the unemployment rate only h as a small impact on the value of the outside option lor the average worker. even very small menu costs may cause the individual trade union's nominal wage claim to be rigid in the short run.Basic Theory and Empirics about Prosperity and Growth 1.)/w. But even if there is 110 nominal wage rigidity. Obviously. despite the sharp rise in unemployment. One can construct . The link between real and nominal rigidity Notice how nominal and real rigidities are intimately linked: the term (1~'. the weaker is the desire for real wage adjustment in response to fluctuations in unemployment.)/w. If we assume zero costs of nominal wage adjustment so that nominal wages are fully flexible.0 5 per cent of the wage bill to generate nominal wage rigidity. In short. if firms set their prices in accordance with our markup price setting equation P = m1' W. is large. The smaller this magnitude. To sum up. if (l:\'. so the higher is the degree of nominal wage rigidity. Of course the results from such a simplilled model should not be taken too literally..may be seen as an inverse measure of the degree of real wage rigidity. a relatively h igh degree of real wage rigidity seems to be a necessary condition lor nominal wage rigidity. 2005 1 MACRO ECON O M IC S FO R THE LONG RU N AND FOR TH E SHO RT RUN 21 claim. as one would expect if real wage rigidity is important in practice. it follows that nominal prices will also be rigid. . and the empirical evidence presented in Chapter 14 suggests that the correlation between real wages and economic activity is quite weak whereas the correlation between output and employment is very strong. according to (17). Macroeconomics for the long run and for the short run © The McGrawHill Companies. However. reflecting a low degree of real rigidity.w .)/w. And the greater the degree of real wage rigidity. appearing in (1 7) measures how much the union would like to change its real wage rate if adjustment were costless. and vice versa. Consequently ftrms will only experience moderate fluctuations in their real wage costs over the business cycle. our parameter values imply that the 'menu costs' of wage adjustment only have to make up a tiny 0 . the smaller are the menu costs needed to prevent 11ominal wage adjustment. In our trade union model of wage setting and unemployment. the labour market model set up above still implies considerable real wage rigidity for reasonable parameter values. so the higher is the degree of real wage rigidity . small menu costs are unlikely to prevent nominal wage adjustment.w. given plausible parameter values. other theories of imperfect labour markets such as the theory of 'elllciency wages' presented in Chapter 12 also tend to imply fairly rigid real wages. Adjustment costs of such small magnitude certainly do not seem unlikely. If the rise in the unemployment rate is only 1 percentage point (still a substantial 20 per cent increase in unemployment). This means that wage setters (unions) 1<\Till not want to change nominal and real wage rates very much when economic activity ch anges. despite the fact that wage adjustment is costless.
For exactly the same reasons. the social cost of nominal wage rigidity may be substantial. given the assumption underlying our price setting equation that the production function is linear in labour input. Hence nominal price rigidity is a rellection of real price rigidity .Basic Theory and Empirics about Prosperity and Growth 1. Since the individual llrm takes the prices set by all other Hrms as given. Finally.u) = 0. Similarly. it has little incentive to adjust its real price. total GOP ()I) after the shock to unemployment is Y = 1 . just as nominal wage rigidity arises from real wage rigidity.2 per cent of the wage bill. so if unemployment were to rise by one percentage point less. The reason why the individual union estimates such a small gain from adjusting its own nominal wage is that it takes all other wages and prices as given.93 ~ 0. they would collectively drive down the rate of inflation (via the pricesetting equation P = mP W') . Macroeconomics for the long run and for the short run © The McGrawHill Companies. Thus. With a linear production function. Final observations on nominal rigidities It should be emphasized that although the cost to the individual union (the private cost) of not adjusting its wage rate is likely to be small.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 22 I Part 1 . Over time as shocks accumulate. the social cost of nominal price rigidity is likely to be many times as great as the private cost perceived by individual price setters.93. and hence an even smaller percentage of total GDP.07 = 0. 12 Against this signillcant income gain must be set the small menu cost of wage adjustment which is only 0. Because the outside option is rather insensitive to changes in unemployment. and as we shall see in Book Two. this will tend to increase real aggregate demand. output would rise by dY/ Y = du/ (1 . its social cost is many times as large as the perceived private cost of nonadjustment of nominal wages. This parallel between the lack of incentives lor wage and price adjustment may be explained as follows: in our model of union wage setting the outside option is the union's real opportunity cost of 'selling' an extra job to the owner of firm i through wage moderation (since the outside option is the real income a union member could have earned elsewhere in the labour market). is also a decision to change its real (relative) price.0. if the Hrm's real labour cost of producing and selling an extra unit of output is insensitive to the business cycle. P/P. . Take our Hrst numerical example above: if a coordinated downward adjustment of wages and prices could secure an expansion of aggregate demand which would cut the two percentage point rise in unemployment in half. In this way it might be possible to limit the fall in output and employment signiHcantly.0 11 relative to the baseline scenario where unemployment rises by two percentage points. and hence it will olten abstain from adjusting its nominal price in order to save menu costs. total output and real income would rise by more than 1 percentage point relative to the baseline scenario where no nominal adjustment takes place. But if all unions could be induced to cut their nominal wage rates simultaneously in response to a drop in aggregate employment. we should stress that the menu cost theory outlined here is only intended to explain shortrun nominal rigidities.01/0. agents who keep 12. 2005 INTRODUCING ADVANCED MAC ROECONOM ICS plausible examples in which this rigidity in the real production costs ofllrms will generate nominal price rigidity even when the menu costs of price changes are very small. a decision by firm ito change its nominal price P. the union loses little by not adjusting its wage rate. since nominal wage rigidity causes a substantial loss of output.u = 1 . This is why the problem of nominal rigidity should be taken seriously.
The division of macroeconomics described above leads to a parallel division of economic policies into lo11grun policies aimed at promoting growth and longrun prosperity and at reducing longrun unemployment.. and slwrtrun policies aimed at mitigating economic fluctuations and their harmful consequences coming..... the division of macroeconomics suggests more than this categorization of policies according to their aims.......... etc.... . for example.I Part 1 . A shortrun policy to mitigate business cycles... such as improving the level and composi1ion of work skills through education and training to attain a better match between the skills possessed by workers and the skills demanded by employers. our analysis suggests that such a policy must try to increase the degree of competition in labour or product markets (by lowering the values of mw and m1' ) and/or reduce the generosity of the system of unemployment insurance (lowering b).... and hence changes in nominal demand will allect real economic variables. can be a monetary or fiscal policy of demand management................ For instance................... rrzP and b which reflect the structural characteristics of labour and product markets.... as we explained earlier........... The reason is that in the short run prices and expectations are not fully adjusted.. A main motivation for studying economic phenomena is the need to improve the basis lor economic policy advice.... 13.......... policies aimed at promoting longrun growth and prosperity must all"ect one or several structural characteristics of the economy such as the longrun propensities to save and invest...Jacobsen: 1 MACROECONOMICS FOR THE LONG RUN AND FOR THE SHORT RUN 23 their wages or prices constant are likely to be pushed ever further away from their optimal wages and prices....... recall that longrun macroeconomics assumes full nominal wage and price adjustment. In richer models of t he labour market there are other ways of reducing the nat ural unemployment rat e. the natural rate of unemployment is also referred to as the 'stmctural' unemployment rate.. In short. The basic and dillerent assumptions underlying the two parts of macroeconomic theory have consequences lor the channels through which longrun and shortrun policies can allect the economy. to engage in education and R&D...... policies lor the long run must be structural policies. this implies that longrun unemployment is determined exclusively by the parameters m'•.... Such a policy can affect the rate of employment in the short run even if it does not influence the basic structures and incentives in the economy..... However.6 Longrun versus shortrun economic policies ... For example....... Because it is rooted in the basic structural features of the economy.. 1 3 In a similar way....... on the other hand............. 2005 Macroeconomics and Growth run SorensenWhitta........Basic Theory and 1........ .................... It follows that a policy intended to reduce longrun unemployment can only be successful if it allects the economy's basic structures.... from sudden increases in unemployment. Specifically. we want to understand the sources of economic growth because this could be helpful in designing a growthpromoting policy lor a lowincome country trying to escape poverty . As we saw from Eq...... Macroeconomics for the © The McGrawHill Introducing Advanced Empirics about Prosperity long run and for the short Companies.. Eventually they '>\Till therefore find it optimal to pay the menu cost to realign their relative price position: in the long run wages and prices do adjust. (8).... 1.
.............. 6..... Understanding these ideas is important for a good understanding of this book....... 5. we say that real rigidities prevail............. The economy's longrun equilibrium is the combination of relative prices and quantities which would emerge in a general equilibrium where wages and prices have had time to adjust fully to past shocks and where no further shocks have occurred over a sufficiently long period...including the labour market. the fundamentals of the economy such as preferences and technology evolve smoothly and predictably.. and expectations are correct all the time...7 Summary .Basic Theory and Empirics about Prosperity and Growth 1..Jacobsen: •~traducing Advanced lacroeconomics 24 I Part 1 .. When this is the case. and these may cause the rate of resource utilization to deviate from the natural rate for periods of sufficient length to be of interest..... This chapter has explained the ideas behind the structuring of our course in macroeconomics.......... indicating that most markets......... Hence real wages and prices do not adjust sufficiently to prevent permanent underutilization of resources................ Individual agents adjust their nominal wages or prices w ith the purpose of changing their real wages or prices..... 3.. Thus nominal rigidities prevail in the short run....... In practice empirical studies often find that prices are above marginal costs. Real rigidities imply that individual agents do not w ish to reduce their real (relative) wages and prices very much in response to unemployment or excess capacity. Macroeconomics for the long run and for the short run © The McGrawHill Companies..... shortrun nominal wage and price rig idities can be privately optimal even if the menu costs of nominal wage and price ..... taking the key elements in a different sequence which w ill help to clarify the distinction between macroeconomics for the long run and macroeconomics for the short run: 1.....SorensenWhitta..... although not one that ensures equality between pricetaking supplies and demands... If the longrur equilibrium is the outcome of perfect competition..... Macroeconomics for the long run aims at explaining the trends in main economic time series and the effects of structural economic policies....are characterized by imperfect competition..... In longrun macroeconomics the economy is analysed as if relative prices are fully adjusted to their longrun equilibrium values in each period.. as the level of output that can be produced when available resources are utilized at their natural rates............ Imperfect competition typically means that the natural rate is less than 100 per cent......... 2005 INTRODUCING ADVANCED MACROECONOM ICS 1.. When the degree of real rigidity is strong. and prices have adjusted to equate these supplies and demands market by market. Still we can think of long run relative prices as being determined by an underlying general equilibrium system............ Because real rigidities imply that agents do not want to change their real prices very much in response to changes in economic activity.. Empirical evidence shows that most money prices and money wages are only adjusted w ith certain time intervals even under considerable inflation.. 7...... One implication is that in longrun macroeconomics aggregate output is determined from the supply side alone.. The natural rate is the rate of resource utilization emerging when relative prices have fully adjusted to their longrun equilibrium values. 2. all economic agents have taken wages and prices as given and found their optimal pricetaking supplies and demands.............. Shortrun nominal wage and price rigidities mean that some money wages and/ or prices are fixed over a certain period.. real rig idities also tend to generate nominal rigidities.. 4. Let us therefore summarize our main concepts and arguments..
Fighting structural unemployment Try to think of specific polic ies wh ich might reduce the natural rate of unemployment. 'real rigidities'. Do you find these assumptions reasonable? Justify your answers. Exercise 4.SorensenWhitta. and the 'natural rate of resource utilization'. and expectational errors are fundamental for understanding shortrun fluctuations. Try to g ive concrete examples of typical shortrun polic ies and typical long run policies. Do you find the arguments reasonable? Justify your answers. Exercise 3. Explain why you th ink that these particular policies might reduce longrun unemployment. shortrun nominal wage and price rigidities. At the same time the social cost of nominal rig idity can be many times as large as the perceived private cost for the individual agent. Macroeconomic policies for the short run and for the long run Explain the d ifference between structural policies and demand management polic ies. the actual rate of resource utilization can deviate fro m the natural rate. Economists believe that exogenous shocks (sudden unpredictable changes in factors such as business confidence. Shortrun nominal wage and price rigid ities may imply that some relative prices are also fixed in the short run. Because of nominal rigidities and expectational errors. Menu costs and real and nominal rigidities Explain and discuss the type of costs included in the concept of 'menu costs'. In the short run aggregate demand is therefore just as important for economic activity as aggregate supply. preferences and technology).8 Exercises Exercise 1. Macroeconomics for the long run and for the short run © The McGrawHill Companies. Shortrun versus longrun macroeconomics Explain the different assumptions underlying macroeconomics for the short run and macroeconomics for the long run.Jacobsen: Introducing Advanced Macroeconomics I Part 1  Basic Theory and Empirics about Prosperity and Growth 1. Explain why and in what sense socalled real rigid ities are necessary to generate nominal wage and price rigidity. 9. real wages are fixed. if both nominal wages and nominal prices are fixed. Exercise 5. For instance. Exercise 2. Explain the concepts of 'nominal rig idities'. 8. . Shortrun nominal rigidities can explain why it takes time for relative (real) prices to adjust to their longrun equilibrium levels. 2005 1 MACROECONOMICS FOR THE LONG RUN AND FOR THE SHORT RUN 25 adjustment are very small. Aggregation in macroeconomics Discuss why macroeconomists typically work w ith aggregate variables and explain the arguments they use to defend this procedure. Macroeconomics for the short run seeks to explain the fluctuations in main economic time series around their trends and the effects of stabilization polic ies. 1.
Basic Theory and I 2... ..: .... ...tu .oA I Part 1. o ...... " .Jacobsen: l n t ... Longrun Unemployment and Structural Economic Policy 27 . _...... ..._ #0 . .... . Some facts about r..:.. .......:.. inn 6 Auttn .. ...eI SorensenWhitta... : ... I © The McGrawHill ·~ Book The Long Run Economic Growth....................
2005 .Basic Theory and Empirics about Prosperity and Growth 2.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Part 1 . Some facts about prosperity and growth © The McGrawHill Companies.
#0 .~~.......Basic Theory and I 2....?...~..~p~~.... ..Jacobsen: I Part 1.........~~.....P....~~..?...~?.........~~........ :.~.?. ...~.?.~~.... o ... .~.....~ 4 Wealth accumulation and capital mobility: the Solow ~?...~....... ~?...~..~...~Y.. inn 6Auttn.....~........~E~ry.. :.<?~. I SorensenWhitta.~~...~~~~ ..... Some facts about l nt ..:...~~.. ............~~!~.. ?..<?J?:.....?....f...:..~...: ...~. ".tu .~?._ 3 .~~~~~~~~.f....~.~~..P.~ .................g.P..~?. .........~.~.... "' I © The McGrawHill Part Basic Theory and Empirics about Prosperity and Growth 2 ...~~... 29 ...~...~..~g.......~.~~~....oA  r..~J?:4...~.....~.&E?..?... ...
Basic Theory and Empirics about Prosperity and Growth 2. 2005 .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Part 1 . Some facts about prosperity and growth © The McGrawHill Companies.
the nation will be better or worse supplied with all the necessaries and conveniences for which it has occasion. bears a greater or smaller proportion to the nu mber of those who are to consume it. i. Clearly. and becomes either investment or an export surplus. However. and which consists always either in the immediate produce of that labour. a book which greatly advanced economics as a scientific discipline. but what is not consumed is saved. and in so far as consumption is a good proxy for the economic wellbeing of people. Already in the title Adam Smith stated what he considered to be the most important issue in economics: what is it that makes a nation prosperous. it is the annual/eve/ of GDP per person that is of importance. not as an end in itself. an increase in average GDP per person is to the benefit of everybody. Consumption is thus always rooted in production and income. the annual increase in GDP per person. Nevertheless. According therefore as this produce. Smith says that the average level of prosperity in a country can be measured by the country's GDP or income per person. or what is purchased with it. to the benefit of its citizens? Smith was clear from the very beginning about how prosperity should be measured: A The annual labour of every nation is the fund which originally supplies it with all the necessaries and conveniences of life which it annually consumes. (Opening remarks of 'Introduction and Plan of the Work' of The Wealth of Nations. This was the title Adam Smith gave his famous book published in 1776. but because the way a country can reach a higher level of income is through n process oJgrowth. or in what is purchased with that produce from other nations. 31 .prosperity and growth 11 Inquiry into the Nature and Causes of the Wealth of Nations. We are interested in growth.e. in Book One we will be much concerned with economic growth. According to Adam Smith's view. 2005 Chapter Some facts about . In modem economic language. 1 776). Some facts about prosperity and growth © The McGrawHill Companies. annual GDP or income per person is a relevant measure of prosperity. not only the avemge income per person but also the distrilJUtion of income across persons is of interest.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1  Basic Theory and Empirics about Prosperity and Growth 2. In both uses it adds to the national wealth and thereby becomes a source of future consumption. for a given degree of income inequality. It is not necessarily all of the annual income that is consumed during the year.
This implied that around the year 2000 the level of GDP per worker in Botswana was more than 10 times greater than in Nigeria and Uganda. Today Hong Kong is one of the world's rich areas. until 2000. The figure shows GOP per worker (G OP per person in the labour force).3% of US Figure 2. Both Jamaica and Hong Kong had a GDP per worker of around 20 per cent of that in the USA in 1960.1. levels of income per person. no: G OP per capita (G OP per person in the population) for reasons to be explained. The differences in the conditions of life caused by these different growth experiences are not ditncult to imagine. were all at about the same level of GDP per worker. From that time until2000.SorensenWhitta.1 gives some clear illustrations of the importance of gradual growth for reaching higher. that is one of the world's richest countries. a GDP per worker at or below 3 per cent of the US level. Nigeria and Uganda.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill 2. Venezuela has fallen to 28 per cent of the US level. average annual growth in Italy was 2. Botswana. Botswana had reached a level of 3 6 per cent ofUS income per capita. In 1960 Venezuela was one of the world's richest countries with a GDP per worker of 83 per cent of the US level. Over the subsequent 40 years. Today it is Italy. but in Venezuela it was negative. Jamaica grew at an average rate of0. and this level was very low by international standards. while Jamaica is relatively poor at around 11 per cent of the US level. Some facts about prosperity and growth Companies. with a GDP per worker at 84 per cent of the US level.1: The importance of growth in GOP per worker for the level of GOP per worker • Figure for 1999 1. or falling to lower. but rather than having. while Hong Kong grew at an annual rate of 5. However. 25000 36%of us• .5 per cent.4 per cent. 0 0 'iii Q) 0::: 5000 Year 2. not Venezuela. while the growth rates of Nigeria and Uganda were . . By 2000 Botswana had not become extremely rich. 2005 PART 1 : B AS IC TH EO RY AND EMPIRI CS A BOUT PRO S PERITY AND GROWTH Figure 2. while Italy was considerably behind at 55 per cent of the US level. respectively.2 per cent and 1.('? ~ 20000 "0 CD 0> 0> ~ 15000 ~ 0 $ ~ 10000 a. GDP per worker in Botswana grew at an impressive average rate of about 6 per cent per year. 1 Around 19 60.Jacobsen: Introducing Advanced Macroeconomics 32 I Part 1 .9 per cent.2 per cent a year. three African nations. as Nigeria and Uganda. with a GDP per worker of about 80 per cent of the US level.
probably the most important one. is a process of growth over a succession of years.~.. 1955 1960 1965 1970 1975 1980 Year 1985 1990 1995 84% of US 60000 ""' (/) ~ 50000 0 "0 (!) <» <» 2000 40000 .. We start in this chapter by stating and discussing some empirical regularities concerning prosperity and growth. . 2005 2. is: what creates growth? Growth theory and growth empirics are fascinating subjects in economics. 0. 0 20000 ~ ~ a:: 10000 1955 \ 28o/o of US 55% of US 1960 1965 1970 1975 1980 1985 1990 1995 2000 Year Figure 2.. or to achieve prosperity. 1 What allows a country to escape from poverty. These 'stylized facts' constitute important knowledge in themselves. They will also be used later on as yardsticks: the growth theory to be presented \.2%) 0 +..viii be held up against the empirical facts.. Some facts about prosperity and growth 2 SOME FAC TS ABO UT PROSPERITY AND GROWTH 33 60000 (/) ~ 50000 0 "0 (!) ~ 40000 ~ ~ ~ 30000 ~ ~ 20000 19%ofUS ~ 11o/o of US ~ a:: 10000 Jamaica (0.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1  Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies. Therefore one of the most important questions in economics. Q..1: (continued) Source: Penn World Table 6..: Q) ~ 0 30000 3 Q.
...... (If the productivity in the formal and the in formal sector are identical. Egypt and Pakistan are only 13 per cent and 6 percent as rich as the US in the year 2000. the purchasingpoweradjusted exchange rate is one dollar for 100 pesos..... The GDP of each country is the value of the 'olllcial' marketed production.. Dividing the ofllcial GDP by the size of the ofllcial labour Ioree can be a way of correcting for this bias... Table 2. since the olllcial labour force does not include those working only in the informal sector....... from an earlier course in macroeconomics..1 shows that it may indeed make a difference whether one considers GDP per capita or per worker... the value of the marketed production.... say US dollars..... Some facts about prosperity and growth © The McGrawHill Companies...... measured by GDP per capita.... total consumption......... Instead one should use a rate of conversion that reflects purchasing power.veen a highly developed country such as the US. A second issue is whether GDP per person should mean GDP per capita........ We will therefore only discuss a lew aspects of measuring economic welfare by GDP per person.. where the GDP is only divided by the size of the labour Ioree.....000 pesos in the less developed country..... First. One should use this computed exchange rate to convert the GDP of the less developed country into a purchasingpoweradjusted GDP in US dollars....Basic Theory and Empirics about Prosperity and Growth 2. Current exch ange rates are not appropriate for converting GDPs if one is interested in standards of living........... and a somewhat less developed country with relatively fewer transactions in the olllcial market economy and relatively more sell~sulllciency.. one will probably underestimate the prosperity of the country in a comparison with the highly developed country....... etc. we will use GDP per person as a proxy for living standards...... Agreeing with Adam Smith that the ratio of the annual 'produce' to 'the number of those who are to consume it' is a relevant measure of the average standard of living in a nation...... which is the population size times the labour force participation rate? The less developed country will typically have a larger informal sector and relatively more people living from nonmarketed home production.. using GDP per worker as our measure of the standard of living makes at least some correction lor crosscountry dillerences in the degree of participation in the formal economy....... and the purchasingpoweradjusted exch ange rate could then be defined as the relation between the two total costs. For this purpose we need an exchange rate that converts amounts denominated in the currency of the less developed country..... The total cost of a relevant bundle of goods representing the 'necessaries and conveniences' of life should be computed in the two countries. respectively.. the correction should be perfect. 2005 PART 1 : B AS IC TH EO RY AND EMPIRI CS A BOUT PROSPERITY AND GROWTH 2..... For instance.... the peso say... You are assumed to know about statistics such as GDP.. Suppose we would like to compare the standard ofliving bea... If one divides the olllcial GDP of the country.... or whether it should mean GDP per worker. into amounts in dollars... If the bundle has a total cost of 1000 dollars in the US.............. and a cost of 100. It may be misleading to use the olllcial prevailing exchange rate between the peso and the dollar since prevailing exchange rates are volatile............... It will rather be a reflection of some change in expectations........... total investment.. Measured ..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 34 I Part 1 ........ A 10 per cent increase in the peso relative to dollars from one month to the next will typically not reflect that the less developed country has become 10 per cent richer compared to the US in one month......1 Measuring the wealth of a nation ..) Generally... where the GDP is divided by the size of the total population.. a comparison between the two countries requires that their GDPs be measured in the same currency. by the size of the entire population..
3. and lor most other countries 1960.06 1. Denmark is richer than the Netherlands. it can still be argued that GDP per worker is a more appropriate measure of economic standards ofliving than GDP per capita.381 4. average labour productivity is a key variable.765 7.htm. this is another reason lor studying GDP per worker.80 0.41 0. come from a database called the Penn World Table (PWT).293 26. 2005 2 SOM E FA C TS AB O UT PRO SPERIT Y AN D GROW TH 35 Table 2.01 0.47 0. 2 The statistics used lor international comparison in this book are of the purchasingpoweradjusted type.675 24. Nevertheless.1: GOP per capita and per worker (2000) Absolute.52 0. 1. reflecting the higher participation rate in Denmark.60 0. Since we want a close correspondence bet\<veen our empirical and theoretical variables. from which most growth data in this book are taken. and from other relevant databases.64 0.71 0. the US.2000.econ.023 0.70 0. 3 The PWT does not cover all countries in the world.50 0.453 44.71 0.52 0. An extract from the PWT 6. but also for differences in annual working time and unemployment rates. they are still not rich compared to the US. GDP per worker is closer to a productivity measure th an GDP per capita.74 0. which is an attempt to create internationally comparable statistics relevant lor growth issues for the countries of the world lor a considerable period (for many developed countries 19 50. For the examples oflreland vs. but according to GDP per worker it is a bit poorer.752 45.2000). Data \>Viii.81 0. can be found in Table A at the end of Book One. Lack of good inter· nationally comparable statistics for annual work hours prevents this. According to GDP per capita. The PWT can be found on the internet at http:l/pwt. The dillerence is a reflection of the much higher participation rate in the US.190 26. and we will mainly use GDP per worker as a proxy for the standard of living and for labour productivity .11 Source: Penn World Table 6. because home production and leisure should also count.69 1.313 23. the dHTerences in participation rates are probably not caused by diflerences in the relative sizes of the unofllcial sectors. the most current version was PWT 6.008 64.29 Relative to US GOP per capita GOP per worker 1.230 56.79 0. and Denmark vs.1.79 0.edJ/Downloads/index. now at 21 per cent and 11 per cent of th e US level. 1 by GDP per worker. In the growth models presented in the subsequent chapters. .73 0.737 52.635 22.42 0.608 24.649 65.30 0. to a large extent. Ireland was 79 per cent as rich as the US in 2000. Th is reflects the low participation rates in Egypt and Pakistan. Also note that according to GDP per capita. and productivity has a closer correspondence to GDP per worker than to GDP per capita.779 38. the Netherlands.67 0. but rather by 'true' differences in labour force participation .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . but they are considerably less poor. respectively. 1996 dollars Country USA Denmark Japan Netherlands Belgium Sweden UK Ireland Egypt Pakistan GOP per capita GOP per worker Implicit participation rate 33. By this argument one should perhaps go all the way to G OP per wori< hour to allow not only for differences in participation rates. At the time of w riting (2004). Ireland was just as rich as the US.781 23.00 0. Among the exceptions 2.52 0.21 0. but measuring by GDP per worker.184 2.053 13. Some facts about prosperity and growth © The McGrawHill Companies.upenn.13 0.537 50.Basic Theory and Empirics about Prosperity and Growth 2.00 0.88 0.
... For instance... but is there any sign that the world income distribution has become more equal over. The curves are based on average GDP per worker and assume that nll people living in a specific country............. 2005 2.... y ) on a Lorenz curve indicates that the fraction x of the people in the world with the smallest incomes earns the fraction y of all the income in the world.. 100 £ 80 "F 0 0 c 60 '0 ~ ~ ... We have already given some illustrations of the enormous di!Jerences between rich and poor countries.. the last 40 years? We can investigate this by means of socalled Lorenz curves................... The world income distribution Some countries are poor and some are rich.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies.... have an income equal to the average per worker GDP recorded in the PWT. "'E U: ~ 20 / 0 0 20 40 60 80 Fraction of world population (%) Figure 2..... not just all people in the labour force....... the Lorenz curve for 1960 shows that in that year........ A point (x......2 shows two Lorenz curves for the world.......2: Lorenz curves for the world.......... the 60 per cent of the world population that lived in the poorest countries earned around 15 per cent of the world's income..... and about moving in and out of these categories.......... Figure 2... Some facts about prosperity and growth PART 1 : B AS IC TH EO RY AND EMPIRI CS A BOUT PRO S PERITY AND GROWTH are nations that were formerly part of the Soviet Union and other formerly planned economies ofEastern Europe. say. in the country and for the year in question.. In 1998 the corresponding llgure was around 20 per cent.... 1960 and 1998......................... and of the fact that some countries have managed to develop from poor to not so poor through a good growth performance..2 The rich and the poor.. 0 40 c 0 1998: Gini 0.. 104 countries Source: Penn World Table 6.......1 100 ......SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 36 I Part 1 . 2........... one lor 1960 and one lor 1998.........52 ....... the growing and the declining ............. We now go into more detail about the issues of poverty and prosperity...
The Lorenz curves have been constructed from such personal incomes and population sizes (you will be asked to do a similar construction yourself in Exercise 2). called the Gini coefficient. reach conclusions similar to the one appearing from our more crude analysis. Because of the particular construction of the Lorenz curves based on GDP per worker. and it has pretty much stayed like that in relative terms over the last 40 years. in particular China. the OECD and the rest of middle·income countries diverge away from it. is a number between zero and one. respectively. If the Lorenz curve had been completely unchanged from 1960 to 1998 it would mean that the percentage change in 4. the data lor the three heavily populated counties mentioned above are not the best. the Lorenz curve and the associated Gini coelllcient illustrate the income distribution in relative terms.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1  Basic Theory and Empirics about Prosperity and Growth 2. and the closer it is to one the more ina:ruality there is. as a fraction of the entire area below the diagonal. 2.4 Stylized fa ct 1 Some countries are rich and some are poor. However. for instance. Two features should be noted. If. the personal income distribution in China has become more unequal over the period considered. but also India and Pakistan (as illustrated in Table 2. . Absolute equality corresponds to a Lorenz curve identical to the 45° line. they should be comparable to each other. This measure. there is some tendency towards a more equal world income distribution. Our investigation of the world income distribution is a bit crude. or diagonal. then China. is an aggregate measure of inequality.3 below). First. these Gini coefllcients are not comparable to traditional coetllcients computed lor the income distribution of a specific country. but in the construction we have completely neglected the influence of inequality within countries by attributing the same income to all persons in each country. he also points out that things have worsened considerably in most of Africa. The Gini coelllcients for 19 60 and 199 8 are shown in Fig. for instance.2 comes from the relatively good growth performances of some heavily populated countries. 59 and 0 . and not at the bottom. India. The procedure may seem a bit odd. and they are 0 . There are. However. and warns: 'If Africa does not st art growing. Between the two years the world income distribution does seem to have become more equal. Furthermore. The further the Lorenz curve is below the diagonal. but in this way we can base the Lorenz curves on data for average GDP per worker. but not lllltcl! at the bottom. 2005 2 SOME FAC TS ABO UT PROSPERITY AND GROWTH 37 We thus attribute to each individual in a country an artificial income equal to the recorded average GDP per worker. It is interesting to note that more thorough studies of the evolution in the world income distribution.2 gives a distorted impression. the more unequal is the income distribution. and global inequality w ill rise. Sala+Martin emphasizes more strongly than we have the improve· ment in the world income distribution (achieved between 1980 and 1998).2 will exaggerate the degree to which the per~~ona/ world income distribution has become more equal. Fig. growth of the African continent should be the priority of anyone concerned w ith global income inequality' . However. both Colombia University and NBER worl<ing papers. This does not necessarily mean that Fig. 2. 2. two studies by economist Xavier Sala+ Martin: 'The Disturbing 'Rise" of G lobal Income Inequality' and 'The World Distribution of Income (estimated from Individual Country Distributions)'. the differences are enormous.52. Investigating the Lorenz curves and the Gini coelllcients reveals that there were enormous income dillerences in the world both in 1960 and in 1998. but not much.2. Most of the improvement indicated by Fig. Some facts about prosperity and growth © The McGrawHill Companies. Therefore the area between the diagonal and the Lorenz curve. If anything. which account lor inequality within countries. 2. Thus.
9 per cent._ 1 "'In y. For instance. 2005 PART 1 : B AS IC TH EO RY AND EMPIRI CS A BOUT PROS PERITY AND GROWTH income per capita in the world's poorest country would have been the same as in the richest country. 5 Table 2.SorensenWhitta. so y. Over the same 3 8 years countries such as Mozambique and Niger moved the other way. which all had a GDP per worker of less than 4 per cent of the US level. 5 per cent. it is approximately equal to the variable's absolute change in (natural) logs. The rate of change (zx)/x= zfx. Some countries are among the 15 poorest both in 1960 and 1998. On the other hand. Their GDP per worker fell from being around 9 per cent of the US level down to less than 3. since. the slope of the log function is one. Indeed. the function In v goes like the function v ._1. Prosperity and poverty.. The fact that the 10 percent of people living in the poorest countries earned only around 1.Basic Theory and Introducing Advanced Empirics about Prosperity Macroeconomics and Growth 38 2...Jacobsen: I Part 1 .. The average annual growth rates between 1960 and 1998 reported in Table 2. in 1960 China was among the poorest w ith a level of GDP per worker at 4. for example.2 have been calculated as (In y 98 . We use here a general property that will be used throughout the book: if the relative change in a Yariable is not too large.l. and for the 15 richest countries in both years (still only including countries for which data are available from the PWT). the other countries that took their place among the poorest had sulllciently bad growth experiences to ensure that the group of poorest countries did not increase their share of the world's income. but other countries have left the group of the poorest. by having very bad growth performances./y. If the relative change we are interested in is a growth rate in.2 illustrates again the enormous di1Ierences between the rich and the poor. By 1998 China had advanced to 8.2. and incomes generally increase. Hence. Other countries that moved out of the bottom 15 are. bringing China safely out of the group of the 15 poorest nations in 1998.In y 60 )/38. into the group of the poorest. as long as a number v is not too far from one._ 1 is close to one. corresponding more or less to a doubling over the full period. the world average income per worker (computed '>\lith population sizes as weights) increased from 19 60 to 1999 by an average annual rate ofl ..I is approximately equal to In zln X= In zfx. 3 per cent of th at in the US.g. . locally around (I. is a good approximation.ln Yr. Second. This was achieved by an average annual growth rate in GDP per worker of 3. However. but also a fair amount of income mobility.y. then the absolute differences between the poorest and the richest increase. Some facts about © The McGrawHill prosperity and growth Companies. 0).8 per cent.9 per cent of the US level. In v "' v. growth and decline The movement in and out of the groups of the poorest and of the most prosperous countries is illustrated in Table 2. The table states some statistics for the countries that were among the 15 poorest in terms of GDP per worker in 1960 and in 1998.I . The fact that the Lorenz curve is relatively unchanged at the bottom means that the poorest countries have also enjoyed most of this increase. where y 1 is GDP per worker in year t. e. In absolute terms the poorest country would be much less poor by 1998 than in 1960. There were some countries that did move out of the group of the poorest countries by having high growth (compared to the world). and d In rJ/dv = 1/ " • so measured at "= I. 5._ 1)/y. if the world income distribution is unchanged in relative terms. Lesotho and Romania. and then (y.5 per cent of world income in 1960 as well as in 1998 does not mean that it is impossible to escape from poverty. G DP per worker. then typically it is not too far from zero. (y. it is not exactly the same countries that are the poorest (or the richest) in 1960 and 1998.y 1 _ 1)/y. The latterfollows because In 1 = 0.
4 0.6 1 .1 3. 1 At the r ich end there was also a substantial amount of mobility.1 3.0 4.0 61 .1 3.0 87.9 3.1 2.5 31.3 5.0 5.2 79.9 81.0 Real GDP per worker re lative to USA % % Average annual growth rate 1960.1 2.6 Richest in 1960 1960 1998 S witzerland New Zealand USA Canada Luxembourg Austra lia Netherlands Venezuela Denmark S weden Norway United Kingdom Ic eland Belg ium Argentina 105.4 2.0 1. moved into the top 15 .0 3.6 3.3 87.1 1.0 3.8 2.8 9.8 1.9 5.0 92.1 2.7 1.4 82.0 0. .0.8 70.9 91 .6 1.3 0.2 79.1 73. Hong Kong and Ireland.6 0. starting at levels of GDP per worker relative to the US of 19 per cent and 43 per cent.3 3.8 88.2 76. Venezuela and New Zealand.8 75.5 0.7 3.0 100.2 reveals anoth er interesting.8 61.1 2.6 3.5 5.2 3.5 9.6 1.5 2.8 4.1998 Poorest in 1998 1960 1998 % Tanzania Burundi Ethiopia GuineaBis sau Rwanda Niger Malawi Mali Uganda Madagascar Burkina Faso Mozambique Central African Republic Nigeria Gambia 2.2 85.7 8.2: The world's prosperity 'Top 15' and 'Bottom 15'.6 2.9 92.2 4.8 1.7 4.3 1.9 0.1 66.9 1.7 1.5 101 . Source: Penn World Table 6.5 1.3 18. Most spectacular drop outs of the top 15 are Argentina.0 50.3 8.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 .5 153.0 1.9 79.4 79.0 0.5 1.1998 % % 1960 1998 % 2.5 54.1 3.5 153.1 0.8 Richest in 1998 1960 1998 Luxembourg USA Ireland Norway Belgiu m Italy Australia Netherlands Canada Austria Hong Kong Denmark France Switze rlan d Finland 91.0 55.8 2. th ere are substantial differences in income per worker.2 85.9 2. fact: even among the world's 15 r ich est countries in 1998. Table 2.2 14. 2005 2 SOM E FA C TS AB O UT PRO SPERIT Y AN D GROW TH 39 Table 2.7 1.3 2.2 1. and perh aps surprising.4 79.9 5.7 2.0 1.1 87.5 3.6 5.1 69.2 5.0 43.5 3.9 6.8 45.8 0.8 1.5 1.8 77.3 88.3 88.6 Not e: Countries are chosen from a sample of 104 count ries for which dat a were available. the US level.Basic Theory and Empirics about Prosperity and Growth I © The McGrawHill 2.3 2.3 66. 7 1.3 0.0 100.6 2. to above.4 82.4 1.0 1.8 1.8 3.4 1.9 0.4 79.3 2.2 1.8 6. 1960 and 1998 Poorest in 1960 Average Real GDP per ann ual worker relative g rowth rate to USA 1960.4 3.9 7.7 70.4 3. ranging from 73 per cent of.2 0.5 3.0 1.9 68.2 8.8 88. Most impressively.5 100.8 3.8 75.0 70.8 2.6 11 .0 79.6 1.1 0.9 1.4 3.5 2.0 100.0 91 .6 5.2 69.9 8 1.2 5.2 83.6 3.9 0.7 1.5 0.6 Tanzania GuineaBissau Malawi Burundi Republic of Congo Eth iopia Uganda Burkina Faso Lesotho China Nepal Rwanda Gambia Kenya Romania 0.0 69.1 3.0 4.6 2.2 3.2 4.9 3.9 79.7 . and ended at levels 79 per cent and 91 per cent.2 59. respectively.6 105.6 2.9 2.9 2. respectively. Some facts about prosperity and growth Companies.2 2.5 3.
4 7.7 0.4 1.4 4.1 153.7 0.5 16.6 5.4 4.3: World growth 'bottom 20' and 'top 20'.0 4. 19601998 20 slowest growing Central African Republic Nicaragua Niger Mali Venezuela Zimbabwe Madagascar Mozambique Chad Senegal Nigeria Comoros Mauritania Rwanda Bolivia Peru Cameroon Jamaica Ethiopia Togo Real GDP per worker relative to USA Average annual growth rate 1960.2 3.0 28.2 14.7 91 .2 25. Singapore is missing for lack of 1998 data.5 14.3 5.8 3.3 0.9 5.4 3.3 % 20 fastest growing Taiwan Botswana Hong Kong Korea Thailand Japan Romania Malaysia Ireland China Mauritius Seychelles Barbados India Portugal Spain Pakistan Republic of Congo Luxembourg Syria Note: Sample of 104 countries as for Table 2.1 83.4 8.8 3.6 53.9 4.3 62.2 0.5 3.3 3.1 5.8 11 .9 4.5 33.4 6.5 38.9 2.2 8.2 16.0 31 .8 7.9 14.3 68.0 52.2 0.2 43.1 2.6 3.5 22.6 43.4 2.7 0. Source: Penn Wortd Table 6.7 7.6 10.1 .2 3.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 40 I Part 1.3 19.0 8.8 3.2 12.1 6.3 0.4 2.9 12.0 8.3 79.8 20.Basic Theory and Empirics about Prosperity and Growth I © The McGrawHill 2.4 0.2 3.9 45.6 3.5 11.8 3.1 0.4 40.2 3.1998 % 1960 1998 1.6 0.9 6.7 11 . 2005 PAR T 1: BAS IC THEORY AND EMPIRICS ABOU T PROSPERITY AND GROWTH Table 2.5 5.5 12.4 0.4 1. Some facts about prosperity and growth Companies.4 11.4 5.0 0.7 64.2 18.3 0.2 52.4 3.0 0.1 12.3 6.4 0.2 34.1 9.2 1.9 29.3 7.1 6.2 4.4 3.3 21.2.2 3. but would otherwise be at the top of the top 20.2 3.0 20.3 11.6 3.6 91.2 4.7 9.1 3.0 27.0 0.3 9.9 30.1 3.7 22.8 3.
but cer· tainly not all. Penguin Press. It is not surprising to find the socalled 'tiger economies' of East Asia in the top 20: Taiwan. You may perhaps think that this is a trivial fact: there is nothing inevitable about grm. Altogether. Likewise.ving countries in the bottom 20): see for instance Table A at the end of Book One. Table 2. Allen Lane.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . It is perhaps more surprising (if you had not read the previous sections) to find the African country Botswana near the very top. as the title suggests. lor instance). on average around or above 4 per cent per year. In 1968 the Nobel Prize Laureate Gunnar Myrdal published the book (in three volumes.3 per cent (the fastest grm.Basic Theory and Empirics about Prosperity and Growth 2. The countries at the top oft he top 20 are often called 'growth miracles' . but there may be grm. that over the preceding 38 years they grew much more slowly. In fact . India. Hong Kong and South Korea (Singapore could be included. and vice versa. Other SouthEast Asian countries such as Thailand and i\'lalaysia also grew quickly over the same period. subSaharan Africa or the Middle East for instance. Hong Kong.3 shows the bottom 20 and the top 20 lor growth in GDP per worker between 1960 and 1998. a country cnn move from being relatively poor to being relatively riel!. . it illustrates once again that the light against poverty is not a hopeless one. 6 All the growth miracles of SouthEast Asia experienced breaks in their growth rates from lower to higher values somewhere around the early 19 60s. There are no growthpreventing areas in the world. the countries in the bottom 20 can be called 'growth disasters'. and in particular in South·East Asia. Consider the tiger economies. it is of interest to look directly at growth performances. Hidden in the data is another interesting fact. but does not appear in Table 2. Taiwan. Some facts about prosperity and growth © The McGrawHill Companies. tuming. We know. average growth rates between 19 60 and 1998 were actually negative. 6. will never be able to grow out of poverty.3 because oflack of data for 1998)which all grew very quickly between 1960 and 1998. our insights can be summarized as follows: Stylized fact 2 Growth mtes vary substantially between countries. and also Japan. At the time of publication several of these countries. Asian Drama: An Inquiry into the Poverty of Nations. and by the process ofgrowing or declining quickly. 2005 2 SOM E FA CTS AB OUT PRO SPERIT Y AN D GROW TH 41 Since a country can move from poor to rich through a process of growth. Similar growth breaks have occurred lor several other countries in the growth top 20. covering more than 2000 pages). The book was. Pakistan and Syria on the list. China. with average annual GDP growth rates per worker above 5 per cent. South Korea. We stress this because sometimes in the public debate it is taken for granted that countries in some parts of the world. and to lind countries like the Republic of Congo.fmm a high rate to a low one or vice versa. much concerned with why countries in South Asia. However. or from being relatively rich to being relatively poor. one cannot point to a continent or part of the world in which the break in growth from low to high should be impossible. London. Note that lor many of these countries.vth. The data simply do not support such an idea. however. The countries of the world that are neither in the growth top 20 nor the bottom 20 have had annual growth rates between 3 per cent (the slowest growing countries in top 20) and 0. Stylized fact 3 Growth crm brenk in a country. had started t o grow quickly. were not able to escape from poverty by a rapid growth process. in the 1960s some economists worried about the prospects for countries in South Asia.vthpreventing policies or circumstances (such as civil war. In particular.
...ITA .. Convergence and Welfare: What the LongRun Data Show'.. Figures like Fig. 8.... Baumol in a famous article from 1986...... 2......... 1950..... 2005 2.. American Economic Review......... suggesting some of the 'laws' discussed below.......lny........ Stylized fact 3 raises again the perhaps most important question in economics: how can a country create a positive growth break? What kinds of policies can do that? Growth theory and empirics.. Absolute convergence Figure 2.. . as presented in the coming chapters........3 plots the log of annual GDP per worker for the period 1950 to 2000 lor a selection of OECD member countries in 1998.....IRL . 9. can contribute some possible answers.... BaLmol..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 42 I Part 1 .............. Since lny..... Figure 2............. constant growth corresponds to a straight line......1085....3 (drawn lor even longer periods.....Basic Theory and © The McGrawHill Companies. in a Hgure showing In y... _ 1 is (approximately) the rate of growth in y .3: Convergence of GOP per worker among selected countries.......... W illiam J...5 Ol 0 ' FIN .... a.. 76.............1 7.3 Con vergence .: USA ~ Q) ~ 0 :: 10 Q...USA BEL DNK ········NOR FRA 9 E ~ ...5 "'0..... if it were true... pp.... against timet. 1986....SGP ..... 7 Such ligures open up the fascinating possibility that the 11.2000 Source: Penn World Table 6... up to 100 years and more) were presented by William J........... '5 "0 CD 0> 0> 11 10... 'Productivity G rowth. An interesting idea in economics would...~ 1950 1955 1960 1965 1970 1975 Year 1980 1985 1990 1995 2000 Figure 2........5 0 0 0 .. A glance at the Hgure suggests that all the countries tend to approach one and the same 'growth path' . Some facts about prosperity and growth Empirics about Prosperity and Growth PART 1: BASIC THEORY AND EMPIRICS ABOU T PROSPERITY AND GROWTH Most of all.... HKG 8+.... 2 .... imply that poverty should disappear by itself.......3 suggests a tendency for countries to approach one and the same upward sloping straight line...... perhaps more or less a line corresponding to the USA.. a....5 ...... 1072.'te 0. and the slope of the line is the growth rate....
focuses on the club of relatively rich OECD members in 2000.4 seem to support the hy pothesis of absolute convergence. Figure 2. and {3 1 is a positive parameter) should then be expected to lit well: In y~. A main reason why this hypothesis is so fascinating is the implication that poverty should disappear by itself in the long run.J lnyo.4. It is worth thinking about what this would imply for the relevance of. However. so that all countries converge on the same level of income per worker. It could be argued that the speed by which poverty disappears is so slow. other t hings being equal. It would not make such policies irrelevant altogether. should be negatively correlated wit h GDP per worker in year 0 . The line has the particular formula: (2) Figures 2. and denoting country i's GDP per worker in year t by y~. The countries in Fig. say. The country points in Fig. Tn the long nm GDP per worker (or per capita) converges to one and the same growth path in all countries. in diflerent ways. o.Basic Theory and Empirics about Prosperity and Growth 2. .3 and 2. 8 In other words. Some facts about prosperity and growth © The McGrawHill Companies. However. Indexing cow1tries by i. It cannel be concluded the other way round that a negative associatiJn between initial G DP per worker and subse· quent growth is (necessarily) an indication of convergence! See the exercise on Galton's Fallacy at the end of the chapter. if poverty disappears by itself. {3 0 is a constant.4 are nicely located along a decreasing straight line. ordinary least squares (OLS) estimation.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . they are not. selected countries which had relatively high and relatively similar GDPs per worker in 2000.. say. The hypothesis of absolute convergence implies that countries with relatively low levels of GDP per worker in an initial year will grow relatively fa st after that initial year. but they only present data for some countries. 2. lor instance. The line that has been drawn in the figure is a best flt to the points according to a standard statistical method. (1) Figure 2. .ln y. foreign aid policies. it certainly makes foreign aid less necessary than if some countries are caught in poverty traps. wh ich is described in the flgure legend and in the statistical appendix to this book. 2.4 plots the average annual growth rate of GDP per worker from 19 51 to 2000 against the log of GDP per worker in 19 51 lor all OECD member countries in 2 000 lor which data were available from the PWT. that foreign a id is currently needed. Both ligures have.4 that had relatively low GDPs per worker around 19 SO must h ave gro\<Vn relatively fast to be able to join the rich club in 8.) T _ {3 = {3 . This would not be so much of a problem if those countries were randomly selected and representative. This possibility is sufficiently intriguing to have a name of its m<Vn: The Hypothesis of Absolute Convergence. 2005 2 SOM E FA CTS AB OUT PRO SPERIT Y AN D GROW TH 43 diflerences with respect to output and income per person between the countries of the world automatically vanish in the long run. average growth in GDP per worker from year 0 to year T. the following equation (where the lefthand side is the approximate average annual growth rate in GDP per worker. and the current poverty problems so severe.
24 OECO countries 1 Note: The countries behind the country codes can be found in Table A The straight line is estimat ed by ordinary least squares (OLS).01 > <( 0. Convergence. 1988.05 0.5 10 10.025 0.5 plots the average growth rate of GDP per worker from 1960 to 2000 against the log of GDP per worker in 1960 lor all countries fuliHling the criteria mentioned.14. The slope of the line of best IH has even become positive.SorensenWhitta. 9 To avoid the sample selection bias problem one should try to get a more representative sample.Basic Theory and Introducing Advanced Empirics about Prosperity Macroeconomics and Growth 44 2. Since absolute convergence implies a clear negative 9. We include here. 1. 'Productivity Growth. The nice negative relationship has disappeared. consider the somewhat shorter period from 1960 to 2000. for data quality.Jacobsen: I Part 1 . except that we exclude countries for which the data quality is relatively poor according to tbe PWT itself.012x. One way to proceed is as follows: in order to have good standardized data for a large number of countries.0 15 Cl)Q.035 cuN ~ I ~ LO 3m e0>. and it is the uniquely determined line that minimizes the wm of the squared vertical distances between the points in the figure and the line. 113811 54. 0 0 0. This is what we will do. Bradford DeLong in an article in 1988. The selection of countries thus has a bias in favour of the hypothesis we are testing. Additional information for the statistically proficient: R 2 = 0. Source: Penn World Table 6. 78.37 (standard error of the coefficient is 0.: (ij~ cJ 0 cco 3~ ~ Q) Q) 0. Czech Republic and Poland. all countries above grade D. . which have been excluded due to lack of data.o 0.1 2000.1 is given a grade.045 Q. The est imated line is y= 0.5 9 9. 1951 Figure 2. CorD. 10 Figure 2. Bradford Delong.5 Log of GOP per worker. A.005 0 8 8.03 0.57. 10. e Q) 0.04 co o .002). American Economic Review.. but the line fits so poorly that we cannot attach any statistical signillcance to its slope. and include all countries lor which the relevant data for this period are available from the PWT. and I= 5. 2005 PART 1: BASIC THEORY AND EMPIRICS ABOU T P RO S PERITY AND GROWTH 0.4: Average annual growth rate of GOP per worker against initial level of GOP per worker. pp. This problem is known as the 'sample selection bias' problem. Each country appearing in the PWT 6. and at most places later on. except Germany. and was noted in the present context by the economist J. and Welfare: Comment'. Members in 2000. Some facts about © The McGrawHill prosperity and growth Companies. B. Hungary.02 0. J.0.
CHN .1 BEL I :o: • LKA~ + ~ + • ·~ DKN l=!::==:::•c====. Mauritania. .02+..?•C.. 1960 Figure 2. 90 countries 1 Note: The estimated line is y=0. Some details are: R2 = 0.0013 and I= 0. absolute convergence is too much to hope for..04 . A third structural characteristic likely to be important is population growth. .======±. Higher population growth means that a larger number of people will come to share the physical and human capital accumulated in the past.5 9 9.5 8 8.•. ..1999..___ _.5 10 Log of GOP per worker.:.06 BWA 0.1 relationship.05 0...0136 + 0...34.·~~~._•_. . TZA ~ + • + + SWE • '""•.=. 10.0006x. we have to draw the sad conclusion: the hypothesis ofabsolute convergence does not hold...... . Conditional convergence Thinking a bit more about it.0 1 +..~•'U:usA .5 6.. some countries have higher rates of saving and investment than others... some countries spend a larger fraction of GDP on education (investing in human capital) than others. 1. and suppose that the first country has more favourable structural characteristics than the second..eFtE+ o+~•~B~D~I~•L.~~ =~~•?. + ITA + • ~:~ '+. and capital is productive. Fiji and Botswana: 1960. Some facts about prosperity and growth © The McGrawHill Companies.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . this should pull GDP per capita down.__.5 7 7..0. 0. and education makes labour more productive. Other characteristics being equal. again preventing absolute convergence.__~~"""•""' ~ •.. Since the first country will then approach a higher level of GDP per worker than .. For instance.z""'M ""B"'V""6E'N• NIC NGA+ . Savings and investment accumulate as capital..Basic Theory and Empirics about Prosperity and Growth 2.0. Singapore: 19601996. We know that the countries in Fig.IRL • LUX .03 0 02 · o. Similarly. 2..•. Consider t\ovo countries with the same level of GDP per worker in an initial year zero. 5 diller considerably with respect to basic structural characteristics..GRBRR+•~~~. Source: Penn World Table 6.o1 KOR+ 45 HI\G + + SGP .5: Average annual growth rate of GOP per worker against initial level of GOP per worker.. We should therefore expect countries with higher savings rates to have higher GDP per worker.... 2005 2 SOME FACTS ABO UT PROSPERITY AND GROWTH • 0. but then GDP per worker ca1mot converge to one and the same level for all countries. • JP ._u N ~ZL~ • . Countries with higher investment rates in human capital should therefore be expected to approach higher levels of GDP per worker.
Nevertheless. countries with relatively low levels of GDP per worker ill an initial year will grow relatively fast after that initial year. The correct equation would not be (1) above. This could. it is only after controlling appropriately for structural diilerences that one should find a negative relationship between initial GDP per worker and subsequent growth. which is for now postulated. as presented in the coming chapters. Tile jitrther below its own longru11 growth path a country su1rts. So. tile jnster it will grow. be a policy of assisting a country in building up sound Hnancial and educational systems. According to the hypothesis of conditional convergence. Assume further that (for reasons that will become clear in Chapter 5) the appropriate . it will become as rich in due time. for instance. and the population growth rate. A country's i11come per worker (or per capita) converges to a countryspecific longrun growth path which is given f1y the basic structural characteristics of tile cou11try. What are the implications of conditional convergence for the relevance of foreign aid? Certainly. for example testing it against data. Which economic variables should be included in zi. 11i. and what should the function y look like? We h ave argued intuitively above that the rates of investment in physical and human capital and the population growth rate should probably be included. It does imply that if a poor country can manage somehow to achieve the same structural characteristics as rich countries. assume that (among) the relevant structural ch aracteristics of country i are the rate of investment in physical capital (the GDP share of gross investment in physical capital).Basic Theory and Empirics about Prosperity and Growth 2. conditional convergence is also a fascinating possibility. but this does not tell us how they should enter. but perhaps an equation like (3) where zi is a vector of variables capturing countryspecific structural characteristics. but will be rooted in theory later on. l11come per worker therefore converges to the same level across cou11tries conditional on the countries being structurally alike. since condition al convergence means that a country may be caught in poverty due to bad structural characteristics. To work seriously with an equation like (3). si. one must have an idea of how to handle the structural variables. The crucial addition compared to the absolute convergence hypothesis is the phrase 'other things being equal'. However. and y is a function expressing their influence. Such reasoning h as led to a weaker notion of convergence: The Hypothesis of Conditional Convergence. the conditional convergence hypothesis points to the importance of supporting poor countries in improving their internal structures. Some facts about prosperity and growth © The McGrawHill Companies. Rather. conditional convergence gives more room for foreign aid than does absolute convergence. The hypothesis of conditional convergence does not imply that poverty would disappear by itself in the long run.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 46 I Part 1 . it will also have higher average growth in GDP per worker over a period after year zero. the kind of policy indicated is not so much a traditional one of transfers to the poor countries ('<\lith a reservation like the one we made in connection with absolute convergence). A good answer will require some growth theory. 2005 PART 1 : BAS IC TH EO RY AND EMPIRI CS A BOUT PROSPERITY AND GROWTH the second. At this stage it will be illustrative to consider a speciHc version of (3 ). Again the convergence hypothesis implies a relationship between the initial level of and the subsequent growth in GDP per worker: other things (basic structural characteristics) being equal.
0.02 + HKG MUS KOR + IRL 0.::.. and n. 2.Jacobsen: Introducing Advanced Macroeconomics I Part 1 .In u:) . growth in GDP per worker is negatively correlated with the initial level of GDP per worker lor the countries of the world.fJ 1 In Y:l + (J 2 [ln s' . but now measure the growth rate adjusted for cozmtryspecijic structural characteristics. Finally. over the period 1960 to 2000 (these can also be computed from data in the PWT. Measures. create a diagram like Fig.6 do seem to be clustered around a negatively sloped straight line.02 6. 00 CJ~ 0 0. Mauritania. and does it in the right way (assuming the right y). up along the vertical axis..01 0 g>8 Q) Q.0. still with In y~0 along the horizontal axis. and we have not even controlled for all the structural characteristics that should be important on a priori grounds.1997. · · T "' flo. but will be later on. + 0. . 2 ..SorensenWhitta.020 (Chapter 5 will explain this value). c 0. (4) Consider the same countries and years as in Fig.03 • CHN 0.06 0 ~o <VO Q.01 >" <((5 ZMB 3 . (In Y~l. Education has not been taken into account.5. .('ol o.075)]. as the country's average gross investment rate and its average population growth rate.075)].05 +•  (ii"' ::J~ cc 0 .In y~ 0)/40 .5 10 10. Based on theoretical and empirical work. This accords with the hypothesis of conditional convergence. . Some facts about prosperity and growth Companies. 2. 1 .5 Log of GDP per worker. Figure 2.0. controlling for some structural dillerences makes the negative relationship between initial GDP per worker and subsequent growth visible again.(J 2 [ln s.ln(n' + 0 . Singapore : 19601996. 2005 2 SOME FACTS ABO UT PROSPERITY AND GROWTH 47 way they enter is: In y~.. then indeed one will end up with a significant and positive estimate of the fJ 1 in (3 ). . That is. Hence. 1960 Figure 2. This will make the picture even more compatible with conditional convergence.l . respectively.. if one controls appropriately for the influence of structural characteristics. 90 countries 1. and are included in Table A of Book One) .ln(n.. The points in Fig. most economists believe that if one puts the right structural characteristics into z.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill 2.5 8 • 8.6: Average annual growth rate of GOP per worker adjusted for structural characteristics 1 against the initial level of GOP per worker. Source: Penn World Table 6.04  ~  ::J ~ 2 0 ~ ~0 3'lii 0 m~ BWA 0. Fiji and Botswana: 1960.5 9 9. 5.6 then results.5 7 7. Assume that the appropriate value of the parameter fJ 2 is 0 .
What these economists have in mind is that there is a certain threshold value of GDP per worker (or perhaps there are several thresholds. it will converge to another path.club' is that countries that start oil' on the same side of the threshold value are in the same category. Some facts about prosperity and growth © The McGrawHill Companies. In this way history can have a permanent impact . but also on the economy's starting point. A country's income per worker (or per capitn) converges to a longrem growth path that depends on the country's basic structural characteristics and on whether its initial GDP per mpita is nbove or below a spec(fic threshold value. They argue that indeed the initial position of a country may have an influence on the level of the growth path that the country is approaching in the long run. The two growth paths may have the same constant growth rate.Basic Theory and Empirics about Prosperity and Growth 2.7: Club convergence . Stating the hypothesis comprehensively . it will converge to one growth path. Income per worker therefore converges to the same level across countries conditional 011 the countries being structurally nlike mtd on the countries startillB 011 the same side of their respective threshold values. and if it happens to start above. Some economists doubt that this i~ really true. Threshold value / =0 Figure 2. The Hypothesis of Conditional Convergence. Figure 2. such that if a country happens to start below that value. The further below the relevnnt growth path a country starts out. 2005 PART 1: BASIC THEORY AND EMPIRICS ABOU T PROSPERITY AND GROWTH Club convergence According to the hypothesis of conditional convergence.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 48 I Part 1 . 7 gives an illustration of club convergence. but they diller with respect to their levels: the first path lies below the second. the faster it will grow. The meaning of the term . but here we present the idea in its simplest version). It is independent of the country's initial level of GDP per worker: the starting point has no influence on the longrun growth path. what must be changed from the definition of conditional convergence above is that the longrun growth path depends not only on structural characteristics. which may be countryspecific. the longrun growth path of each country is given entirely by the country's structural characteristics.its influence is not washed out in the long run.
. as well as models supporting the idea of club convergence.vth in output per capita.... At present the issue of conditional or club convergence is unsettled.....~.P.... Some facts about prosperity and growth © The McGrawHill Companies. that is... An interesting implication of the idea of club convergence is that it may provide a rationale for traditional foreign aid policy.... Giving high transfers to a developing country over some period may bring the country's income per capita above the threshold level.~... and it also shows the linear trends... many countries in Westem Europe and North America have lwd relatively constant ann11al rates of growth in GDP per capita in ~he range 1....~?... However. A foreign aid policy of giving temporary transfers to the poor countries can thus have permanent beneficial ellects.....~. Fig. To separate convergence to the growth path from the longrun growth path itself. ~?.. Because of a lack of very long nm data lor GDP per worker. Steady longrun growth Figure 2... according to the idea of club convergence.. Likewise.. 2.~~~..8 shows the evolution ofthe log of GDP per capita in a number of Western countries for periods exceeding 100 years. even after the transfers are no longer given... We may sum up our discussion of convergence below in Stylized fact 4 ..... 2. we should look at very long series for GDP per person... The reader should be warned. Stylized fact 5 Over periods of more titan 130 years........ The previous section was concerned with the relationship between the growth processes in dillerent countries. there is good reason to believe that GDP per worker has also grown at .. It turns out that indeed many of the countries that industrialized early have had relatively constant growth in GDP per person over quite long periods......Basic Theory and Empirics about Prosperity and Growth 2.~.... on growth along the country's own longrun growth path.. T~.. some empirical analyses are in favour of conditional convergence...~ ~......~~~. Now we will focus on the longrun growth process within a single and steadily growing economy.. This accords with the idea that in the long nm income and GDP per worker converge to a countryspec~flc growth path which is given by the country's basic structural c/wracteristics.. 2005 2 SOM E FA CTS AB OUT PRO SPERIT Y AN D GROW TH 49 One can write down sound economic growth models that support the idea of conditional convergence..... Fig. and some are in favour of club convergence..~. ..... however. and thereby initiate a growth process eventually leading the country to higher levels of income than would have been reached without the transfers.~~?...~.8 shows GDP per capita.. that the issue of convergence is perhaps a bit more controversial than our use of the word 'fact' suggests: Stylized fact 4 Convergence: lf one controls appropriately for structural dij)'erences between the countries of tile world.... and possibly also by its initial position.... probably up to 200 years.. a lower initial val11e of GDP per worker tends to be associated with a higher subsetJuent growth rate in GDP per worker...SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1.. Disregarding shorter run fluctuations and the infl uences from great wars and depressions...52 per cent.8 shows for each country a remarkable tendency towards a constant and positive rate of grm.
.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies.3% ·5_ "' (. <» Average annual growth rate: 1.s 00 10 10 a.. 8...j 0 7."' (') . Western countries Source: Angus Maddison. 0 "' 00 00 . Monitoring the World Economy 1820.5 a..5 .5 / 7.5 E :§ 10 Year rl 9.5 ·~ Ol ~ 0 8 E c 8 Ol .. 9 Ol .. 00 .j 8.5 ·~ Ol  0 <» 9 = 7 8 "' a.5 7.s 1u a....8: GOP per capita.) 1u a.8o/o Denmark .. 9.1992.) 9 a.5 0 00 a. 1./' "' 00 7 0  0 ·~ Ol a.s ·a. 0 0 0 'iii 9 'iii 0 . 9 § 8. 0 'iii ~ 0 ~ .5 Ol 0 7 <0 (') Finland 1u .s ·a.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 50 I Part 1 .5 ~ 0 8. 'iii ~ <» E Sweden 0 0 00 8. a. <» <0 9..5 0 "' 00 Year 00 00 "'<» 00 ...5 ~ 9 0 ~ . a.s= . OECO. ~ 8.5% 0 0 .j '<t Q) E :E "'<» a.5 a."' 00 00 <» 00 (') '<t <» <» 0 Year 10 .j "'<» 10 ·g.j 10 "' <» CD C'J Ol 6.5 'iii ~ 8 ~ <» "'<» '<t 00 10 <» Ol <0 Ol 00 10 <0 0 00 <» <» <» (. E 7.5 rl 10 1.. 2005 2..'!: 10 0 .5 ·5_ "' Average annual growth rate : 1.5 <0 (') Year United Kingdom 10 ...5 0 0 7. Some facts about prosperity and growth PAR T 1: BAS IC THEORY AND EMPIRICS ABOU T PROS PERITY AND GROWTH I USA 10.5 0 8 = E 7..j 0 7 "' "' 00 00  00 0 "' 00 00 00 "'<» 00 '<t 10 <» <» <» 0 Year "' Average annual growth rate : 1."' (') '<t 10 <» <» <» 0 <0 (') <» "'<» '<t Year Figure 2. ~ 9. rl 9..5 0 <» <» 8 ·a. <» <» <» 0 00 <» <» <» . 0 <» Netherlands "'~ 9. 1995.
Consider a Western economy that has experienced steady annual growth in GDP per worker. This would direct portfolios away from bonds and towards real capital. so must the share of all other production factors since this latter share is one minus labour's share. Y JKu must have been changing by the same rates. If labour's share has been relatively constant. which would tend to equalize the rates of returns between bonds and capital. but also. The 'law' of stable income shares is illustrated in Fig. the average real wage rate must have been increasing by more or less the same rate as GDP per worker. 2. There are considerable shortrun cyclical movements in labour's share (not so visible from Fig. respectively. Figure 2. For purposes of longrun analysis. land and other natural resources. If the rate of return on capital (as defined here) over a long period increased by one hall' of a per cent per year. five of which also appear in Fig. real interest rates. constancy of capital's share implies that the real rate of return on capital.9). y 1 = YJL 1. Steady growth in GDP per worker. Over long periods there should be no systematic differences between the trends in the real rates of return on dillerent types of assets. but there is no tendency for real interest rates to be systematically increasing or decreasing over long periods: they have no longrun trend. 2 . say. The figure shows that real interest rates (as computed in the figure) fluctuate a lot. We can rewrite labour's share as wJ(YJL 1) = wJy(' Hence. labour's share of GDP has stayed relatively constant. may therefore be treated as if they were constant. 2. since participation rates h ave typically increased gradually.8 . Labour's share in year tis w 1 L)Y1. which shows the evolution of the income share of labour for six OECD countries. Let us call the other factors 'capital'. .9. if y 1 has grown at a relatively constant rate. Some facts about prosperity and growth © The McGrawHill Companies. including into this category not only reproducible physical capital. while over the same period the real rate of interest on bonds stayed constant. In the long run the trend of the rate of return on capital must.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1. investment in real capital would soon become much more advantageous than buying bonds.10 reports on the longrun behaviour of real interest rates over long periods in some Western countries. upwards or downwards. and (hence) the average real wage of a worker has grown by approximately the same rate as GDP per worker. and hence the real rate of return on capital. Not only does labour's share not show any longrun trend. The countries that have experienced relatively constant growth have typically had remarkably constant factor income shares over long periods as well. therefore. ru and the outputcapital ratio. where Y1 and L1 are the GDP and the number of workers in year t. for example. Stylized fact 6 During the long periods of relatively constant growth rates in GDP per worker in the typical Westem economy. and a constant labour income share implies steady grovvth in the real wage. Let the total capital input in year t be denoted by K1• If we denote the rate of return on capital by r1. then capital's share is r 1K1/Y1 = r J(YJK1) . 2005 2 SOM E FA CTS AB OUT PRO SPERIT Y AN D GROW TH 51 constant (slightly lower) rates in the countries considered. the relatively constant labour share turns out to be relatively close to 2/3 in all the countries. and labour's share has stayed relatively constant.Basic Theory and Empirics about Prosperity and Growth 2. but the longrun evolution in the labour income share is fairly precisely described as being constant. be anchored by the trend of real interest rates on bonds. Hence. where w 1 is the average real wage per worker.
5 0 "' E 0  Year  5 0..6 . 2005 2.7 ti 0..7 ::::::::::::::: z::::::::=. ..... 0.8 () ....i'!! () 0.5 iii Q) E 0. Sweden 0....."'"' "' "'"' "' "'.. "'. <X> . ..4 Q) (.. .....£  0.2 ~ 0.7 ti 0..2 0. .i'!! ·~ (X) (X) <» <» "' "' "' "'..£ (X) ..1 (/)  0.."'"' "'"' "'"'"' "'"' (X) 1 Q) "' E =:s.2 0..6 .<:" . Domestic factor income is compensation to employees plus gross operating surplus and gross mixed income.....1 .. .Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies.3 ~ (/) D enmark 0. . Remaining countries: OECD National Accounts. 0 <:" '<t .. Denmark: ADAM database....1 0 0 0 . Statistics Denmark...4 1 ()  United Kingdom Year .  0....6 .i'!! .£ '<t (X) () Q) E 0 '0 0 Q) (.3 0.9: Labour's share of domestic factor incomes Note: Labour income is compensation to employees. For the USA domestic factor income is GOP less indirect taxes plus subsidies... Some facts about prosperity and growth 1 "' E Q) 0 () .8 0.. .8 0 0..3 ti 0.. ~ (/) (X) (X) 0 0 0 '0 0. 0 <:" '<t . "' Q) '<t 0.9 Q) (.'<t "'. Source: USA: Bureau of Economic Analysis. .. <» 0 <X> <» <:" <X> <» '<t <X> <» "'<» <X> Year <X> <X> <» 0 0 0 "' "' "'"'"' "' "' "' "' "' "'"' "'0 '<t (X) Flgure2.2 0.4 0  0 0 0 <:" 0..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Part 1 ..9 0 0...<»<X> <» <» "' 0 <X> <» <:" <X> <» '<t <X> <» "'<» <X> 0 "'"' "' 000 "' "' "' "' "' "' "' "' "' <X> <X> <:" <» (X) '<t Netherlands  B elg ium  0...5 ~ 0 '0 USA 0 . <» <» <» "'<». (\1 "'"' <X> "'"' "'"'<» "'<» '<t <D Cl 0 <X> 0..... ~ = = 0..9 Q) E 0 ~ ~ c::=::: 0.. <» <» .
International Historical Sta6stics. Consumer price indices from B. The nominal interest rate used is the average annual yield on fong·term government bonds. and p is the relative change in consumer prices from the current period to the next period.10 . 17501993. i is the nominal interest rate. I SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Part 1 .20  0 0 0>  10 0 10 0> 0> 0> 0 0 10 0> 0> ('I 0 ('I C'l 0> 10 C'l 0> 0 10 0> 0> ~ ~ United Kin dom 0 10 0> 10 10 :»  Netherlands 0 10 0 10 0> 0> 0> 0> "' "' " " 0 <Xl 0> 10 CD 0> Year 15 .10 . Some facts about prosperity and growth © The McGrawHill Companies. Mitchell. A History of Interest Rates. Sylla. 0> ('I 0> 0> 10 <Xl 0> 0> "' " " "  CD 0> ~ CD 0> "<Xl 0> Year 30 25 20 15 10 % 5 0 5 .R. Western countries Note: The formula used for the real interest rate i s (1 + !)(1 + p) = 1 + i.20 0 C'l 0> C'l C'l 0> "' C'l 0> 0> C'l 0> ('I ~ 10 CD 0> 0> ~ 0> 10 ~ 0> ~ 10 0> " "'0 "' C'l 10 0> 0> 0> «> «> a. where r is the real interest rate. Homer and R. Source: Interest rates from S.10: Real interest rates. .10 . Macmillan.15 . 1998. Rutgers University Press. 2005 20 .15 . 1991.USA 15  Denmark 10 5 % 0 5 .Basic Theory and Empirics about Prosperity and Growth 2.Norway  Sweden 10 5 % 0 5 . 0> ('I 0> 0> 10 <Xl 0> 0> "' " " "  CD 0> ~ <Xl 0> " CD 0> Year Figure 2.15 0 C'l 0> C'l C'l 0> "' C'l 0> 0> C'l 0> ('I ~ 0> 10 CD 0> 0> ~ ~ 10 0> ~ 10 0> " "'0 "' C'l 10 0> 0> 0> «> «> a.
! = kjy . and the rate of return on capital. and the capital.Jacobsen: I Part 1 . Some facts about © The McGrawHill prosperity and growth Companies. consumption. . Finally. .output ratio.output ratio is constant. a constant annual growth rate. All the other features in our list of facts follow from these constancies (convince yourselfofthis). (1 s)Y.. 11 The growth process follows a balanced growth path if (Wl' per worker. You can fine this lecture by Nicholas Kaldor. n.C. F. We can rewrite the capital. 11 All three facts can be ~::xpre:. is growing at the same rate as y 1• total capital is growing at the rate. Lutz and D. g. the constancy of the capital. Hence total consumption. and k./(Y.. capital's share ar1d the rate of retum on capital have shown no trend. so GDP grows at a constant rate equal to the sum of the growth rate of GDP per worker and the population growth rate. _1 = (In y.. must also be grovving at the rate. . w hich is implied by. 12.In L1_ 1) i: g + 11. Constancy of K.ivdy by lhree fuudameulal cou:. We find the definition given here useful for how we are going to use the concept of balanced growth: as a comprehensive expression of the empirical regulanties that we want our growth theories to explain. Hague (eds). implies that the capital intensity grows at the same rate as GDP per worker.eu compreh eu:. the labour force (population) grows at constant rate.output ratio is constant. then the output. g + 11 ./Lt' by k. . g.ln y 1_ 1) +(In L. or capital intensity. in GDP per worker. r.vth path the capital.. and the rate of return on 'capital' is relatively constant./K. the capital. Our definition of balanced growth is a relatively stric t one.laudes: lhe growlh rate ofGDP per worker is relatively constant. r. Along a balanced gro\. capital. g + n. and the capital intensity has grown by approximately the same rate as GDP per worker. Balanced Growth. our definition. in The Theory of Capital.). Y JK. and for which total annual consumption is a constant fraction. and population all grow at constant rates.capital ratio.output ratio is a main motivating factor behind the definition of 11. where we have denoted the capital. The stylized facts have given rise to an idealized picture of the longrun growth process called 'balanced growth'. consumption. Often balanced growth is just meant to describe a situation w here GOP. KJY" must also have been constant. 1 . 1961. Consider an economy that fulfils the three constancies with an annual grovvth rate. have both been relatively constant. the functional income distribution between labour and 'capital' is relatively constant.output ratio as KjY1 = (KjL 1)/(YjL . 2005 PART 1 : B AS IC TH EO RY AND EMPIRI CS A BOUT PROSPERITY AND GROWTH If capital's share. sY1. Stylized fact 7 During the long periods of relatively constant growth in GDP per worker in the typical Western economy. St Martins. Balanced growth The empirical regularities in our Stylized facts 5. but does not itself imply. and capital grow at the common rate. GDP.7 are much inspired by a famous lecture given by the British economist Nicholas Kaldor. in the number of workers.s.In Y./Y. n. In fact. who pioneered the approach of setting up the stylized facts and constructing theories to explain them.. and tile rate ofretum 011 capital is co11stant.Basic Theory and Introducing Advanced Empirics about Prosperity Macroeconomics and Growth 54 2. Since Y1 = y . of total annual GDP (a realistic longrun feature for a typical western economy).L. and total investment. and the capital intensity all grow at one and the same constant rate. consumption per worker. K. New York..SorensenWhitta. the real wage rate.A.labour ratio. onehas ln Y. (therefore) tile capitaloutput ratio has been relatively constant. 'Capital Accumulation and Economic Growth'. since K1 = k 1 L1. g + n.:.
...... a relatively prosperous country will not auto matically remain so regardless of the policies it follows ..output ratio may be close to constant in the US in the long run. Direct estimates of the longrun evolution of capital.... In absolute terms poverty has therefore become less severe..... but at the same time the absolute international income differences between the richest and the poorest countries have increased. A country's GOP per worker may be used as a proxy for the average standard of living and the average productivity of labour in the country. The evidence seems to support the hypothesis that the world's countries converge in a conditional sense: if one controls a ppropriately for structural differences across countries. 1.... ...SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1  Basic Theory and Empirics about Prosperity and Growth 2. By the process of growing or declining quickly...':1:r.. The empirical evidence on levels of GOP per worker reveals enormous income d ifferences across countries....... If a growth model predicts that the economy \>\Till converge to or move a long a specific longrun growth path.. because of the empirical plausibility ofbalanced growth........ 2... most notably some by the British economist and economic historian Angus MaddL~on..... The evidence shows that growth rates vary substantially across countries....... We have argued indirectly for this constancy: it follows from constant income shares and the absence of a trend in the rate of return on capital.. growth can break in a country....... 2005 2 SOME FAC TS ABO UT PROSPERITY AND GROWTH 55 balanced growth. Angus Maddison. 4.....output ratios. or vice versa....r.l. Press.... 13 When we tum to growth models in the subsequent chapters.. but in several other countries the ratio seems to have been gradually increasing over the last 100 years........ Combining data on GOP per worker with data on population s ize.................. or from being re latively rich to being relatively poor... a lower 13. but there was no reduction in international inequality at the bottom of the world income ladder.. turning from a high rate to a low one.... However..... 3...... Focusing on GOP per member of the offic ial labour force rather than on GOP per capita is a rough way of adjusting for this...... The fact that the world income distribution has improved (slightly) in relative terms means that on average poor countries have enjoyed percentage increases in income per worker at or above the percentage increase in income per worker in the world in general.. balanced growth will be of theoretical importance. $. a country can move from being re latively poor to being relatively rich.. Moreover. Longrun accordance with balanced growth will thus be used as an 'empirical consistency check' of growth models.... In some countries the population participates much less in the formal market economy than in other countries......... These observations point to the importance of structural policies to promote or maintain economic growth.. it should be mentioned that there is some controversy about whether longrun capitaloutput ratios are real ly constant...... These facts indicate that the fight against poverty is not hopeless.... On the other hand. Oxford University ..... 1991.J:I_l_<:J:~Y.. where we argued for the latter by pointing to the longrun behaviour of real interest rates... Some facts about prosperity and growth © The McGrawHill Companies. we found that the relative crosscountry income differences have declined a bit over the last 40 years........ then this path should be a balanced growth path. Dynamic Forces in Capitalist Development: A Long Run Comparative View. suggest that the capital.
Furthermore. rather than GOP per capita. Exercise 3. The facts just mentioned have given rise to an idealized picture of the process of constant growth called 'balanced growth'.g. consumption per worker. Show that GOP per capita. Hence the average real wage of a worker has grown by approximately the same rate as GOP per worker. the country's average annual growth rate in GOP per worker from 1960 to 1998 (you can take these from Table 2. . Compute the Gini coefficient for each year. grows at the rate g each year over many years.6 Exercises Exercise 1. 2 per cent. many countries in Western Europe and North America have had relatively constant annual rates of growth in GOP per capita in the range 1.4 shows GOP per worker {in 1996 US dollars) and population size for 1960 and 2000 for the rich part of the world (defined here as the OECD countries for which data were available). capital's share and the rate of return on capital have shown no trend. The income distribution for the rich part of the world Table 2.output ratio has been relatively constant. e. During these long periods. Construct Lorenz curves for the rich world for 1960 and 2000. Comment on how inequality. Compare this to how the income distribution of the world has developed.ln this book balanced growth will describe a situation where GOP per worker. where g is written as a percentage.SorensenWhittaJacobsen: •~traducing Advanced lacroeconomics 56 I Part 1 . will then approximately double every 70/g years. or per worker. that is. if growth continues at the same rate as it averaged from 1960 to 1998. and the capital. Relate this to what you know about convergence from Section 3. Over periods of more than 130 years. GOP per capita versus GOP per worker G ive some arguments why GOP per worker.2 per cent. Exercise 2. In the long run income and GOP per worker thus converge to a countryspecific growth path which is given by the country's basic structural characteristics {and possibly also by its initial position).2). Some facts about prosperity and growth © The McGrawHill Companies.labour ratio all grow at one and the same constant rate. 6. among the rich has developed. 2005 PART 1: BAS IC THEORY AND EMPIRICS ABOU T PROSPERITY AND GROWTH initial value of GOP per worker tends to be associated with a higher subsequent growth rate in GOP per worker. 2. Use this to set up a table showing for each of the 15 richest countries in the world in 1998. should be used for a comparison of standards of living between countries. and probably up to 200 years.5. the (average) real wage rate. labour's share of GOP has stayed relatively constant. and the rate of return on capital is constant. and the number of years it takes for the country's GOP per worker to double. or equality. so the capital. the stock of capital per worker has grown at roughly the same rate as GOP per worker. 5. Time to double Assume that in a specific country GOP per capita.Basic Theory and Empirics about Prosperity and Growth 2. assuming that everybody in the population earned the GOP per worker of the relevant year. We will use longrun accordance with balanced growth as an empirical consistency check for the growth models to be presented in coming chapters. or worker. and draw the two curves in one figure. respectively.
9 49 5 4. 1960 and 2000.053 103.283 16.927 59.338 19.704 2.623 8.200 46.7 4 2 49. where along the horizontal axis you have the log of GDP per worker.242 24.457 7.373 3. Members in 2000 except Hungary.581 1 1.991 23.1 10 5.750 57. but in a final year.430 50.136 50.275 126.135 1.3 in Section 3 we said that it showed a tendency towards convergence to a common growth path w ith constant growth.480 27.730 93.752 6 4.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1. 4 23 3.568 52.357 15.Basic Theory and Empirics about Prosperity and Growth I © The McGrawHill 2.412 48.537 65.700 18.068 26. Germany and the Czech Republic.4.856 32. not in an initial year.732 27.005 10.304 30.4 91 10.206 24.391 16. Source : Penn World Table 6.043 26.4: Population size and GOP per worker. OECD countries 1 1960 Korea Turkey Japan Portugal G reece Spain Ireland Mexico Austria Finland Italy France Belgium Iceland United Kingdom Norway Sweden Denmark Netherlands Australia Luxembourg Canada USA New Zealand Switzerland 2000 Population (1 . and average annual g rowth in GDP per worker in a subsequent relatively long period along the vertical axis.920 30. 2000. Comment on your findings.008 35.252 27.826 9.542 2.1 13 4 4.4).756 281 8.737 39. Exercise 5.075 Turkey Mexico Portugal Greece Korea Japan New Zealand Spain United Kingdom Iceland Sweden Switzerland Finland France Austria Denmark Australia Netherlands Canada Italy Norway Belgium USA Ireland Luxembourg Population (1 .OOOi GDP per worker (1996 dollars) 25.360 4 4.157 15.47 1 5. 431 8. Perhaps we were too hasty. 1 Exercise 4.919 3.480 4. 2.254 12.904 10.919 20. Level convergence among the rich? Commenting on Fig.909 180.295 53.154 13.832 35. Construct an 'opposite' figure for the same countries as in Fig .221 10.588 35.000) GDP per worker (1996 dollars) 66. Data for this exercise can be taken from Table A at the end of Book One.874 30. 2.689 8.728 4. Growth against ultimo level In Section 3 there were figures showing across countries the log of GDP per worker in an initial year along the horizontal axis.185 5.254 275. 2. Poland.176 60.256 20.871 7.453 47.016 13.372 5.243 36. 2005 2 SOME FAC TS ABO UT PROSPERITY AND GROWTH 57 Table 2.530 8.649 45.850 38.055 45.293 7.787 4 41 15.582 7.591 50.032 56.069 4.1 1 1 20. The following .779 51. and along the vert cal axis you have the average annual growth rate in GDP per worker from 1960 to 2000.530 314 17.107 176 52. Some facts about prosperity and growth Companies.831 39.362 4.835 97.481 10. For the OECD countries there was a clear negative relationship (see Fig .230 52.840 20.327 30.558 4 7.
a tendency to 'convergence' towards a general intermediate height of everybody. one value for each country (you can do this on the computer or by drawing values written on pieces of paper from an envelope).edu/Downloads/index.econ.3. Wh::~t h::~s grown most. Some facts about prosperity and growth © The McGrawHill Companies.upenn. 14. where the latter is defined as the total value of exports and imports. There is at leas! one more important fact....1 yourself (http://pwt. 1993.000 to Country 1. Why is it that there must be a tendency towards a negative relationship? G iven that there must be such a negative relationship.htm). To find the data for this exercise. mace a similar fallacy in connection with heights of sons and fathers. or growth in GOP causes growth in trade? Exercise 7. Is it also true the other way around.000}.edu/Downloads/index. Firs! make sure that you understand what the variable 'openness' measures. { 10.000. For each country you will now have an initial 1960 value for GOP per worker and an eventual year 2000 value. work! GOP or worlri tr::~riP. Do ::~n OLS P. Galton's Fallacy 14 It is an implication of convergence that a lower initial GOP per worker should. 2. 2005 PAR T 1: BAS IC THEORY AND EMPIRICS ABOU T PROS PERITY AND GROWTH figure will make level differences more visible if they are there. that a clear negative relationship between initial GOP per worker and subsequent growth is necessarily an indication of convergence? The answer is no. 12.st fit. Exercise 6. To draw the following figure you should go to the PWT 6.SorensenWhittaJacobsen: •~traducing Advanced lacroeconomics 58 I Part 1  Basic Theory and Empirics about Prosperity and Growth 2. He concluded from this. and vice versa.. Scandinavian Journal of Economics.000. etc. against the average annual growth rate of 'trade volume'. 12. Galton noted a tendency that relatively tall fathers got sons shorter than themselves.? Can it be inferred from the figure whether growth in trade causes growth in GOP. linP. imply a higher subsequent growth in GOP pe· worker.htm). late in the nineteenth century.upenn. can we infer anything about convergence from it? 14. you will have to go to the PWT 6.1 yourself (http://pwt.000 to Country 2.000 to Country 3.econ. The US will appear as a horizontal line. . The term 'Galton's Fallacy' comes from the English statistician Francis Galton who once. Plot across countries the average annual growth rate in GOP per worker against the initial level of GOP per worker.000 to Country 26. 15. Comment on the issue if all countries other than the US seem to be approaching the US horizontal line. say) as follows: 10.15 To each of 26 imaginary countries attribute a GOP per worker in 1960 (measured in US 1996 dollars. Growth and trade This chapter has presented the stylized facts of growth that are most important for the theoretical developments in subsequent chapters. 60. plot the average annual growth rate of GOP (not per worker) from 1960 to 2000. up to 60. The exercise is based on Danny Ouah. pp. as this exercise should teach you. This was used in the chapter to test convergence: the lack of a negative relationship between initial GOP per worker and subsequent growth would imply a rejection of the hypothesis. 427443.sti m::~ti on of thP. of hP. 'Galton's Fallacy and Tests of the Convergence Hypothesis'. Across as many countries as possible. 95. compute GOP per worker relative to GOP per worker in the USA over the period and illustrate in a figure with relative GOP per capita against years. For this you w. For the same countries and years as in Fig. other things being equal. which relates growth to growth in trade. .ll need to go to the explanatory documents of the PWT. Now determine each country's GOP per worker in the year 2000 by picking randomly (without replacement) from the same set of values for GOP per worker.
70. 1 In the Solow model. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. 2 The model describes how capital evolves as a result of capital accumulation. 2005 Capital accumulation and growth The basic Solow model T he previous chapter raised a basic economic question: how can a nation escape from poverty and ultimately become rich? Or more precisely: how can a country initiate a growth process that \<Viii lead it to a higher level of GDP and consumption per person? This ch apter presents a fundamental economic model that delivers some first answers. Furthermore. The assumption that these inputs can substitute considerably for each other in the long run is empirically plausible. 1956. 2 . and the labour force is assumed to grow at a given rate. the stock of capital will increase by an amount equal to gross investment minus depreciation on the initial capital stock. but he was the first t o do so in a model w ith realistic long·run substitution possibilities between capital and labour. 6594. The essential additional feature of the Solow model is that it incorporates the dynamic link between the !lows of savings and investment and the stock of capital. 59 . competitive clearing of factor markets implies that output in each period is determined by the available supplies of capital and labour.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Chapter I Part 1  • • • • • • • Basic Theory and Empirics about Prosperity and Growth 3. In 19 56 the economist and Nobel Prize Laureate Robert M. Solow published a seminal article called 'A Contribution to the Theory of Economic Growth'. Robert M. The model is known as the Solow model of economic growth . Solow was not the first to suggest a coherent formal framework w ith an explicit description of the capital accumula· tion process. The model \<Viii show how the longrun evolutions of income and consumption per worker in a country are allected by structural parameters such as the country's rate of saving and investment and the growth rate of its population . The article presented a coherent dynamic model \<\lith an explicit description of the process of capital accwnulatwll by which saving and investment become new capital. 'A Contribution t o the Theory of Economic Growth' Quarterly Journal of Economics. how the labour force evolves as a result of population growth. pp. Solow's model accounts for the fact that between any two successive periods. total saving and investment is assumed to be an exogenous fraction of total income. and how total production and income evolves as a consequence of the evolutions of the total inputs of capital and labour. 1. Solow.
1.<~. in this market is the amount of output that a firm must pay to a consumer for leasing one unit of capital during period t. . I. that is. and for each commodity there is a market. The reason is that capital depreciates. Thus r 1 is a real rental rate. p 1. We assume that the use of one unit of physical capital for one period implies that an amount<~. we assume). This is the rate of return on capital comparable to an interest rate earned from a flnancial asset like a bond. is the return to capital net ofdepreciation. This chapter presents the Solow model in its most basic version. comes from the firms. The demand for capital services.. that is. The model's real interest rate. so other prices are measured in units of output. Time runs in a discrete sequence of periods indexed by t. The model is thus a onesector model and does not distinguish between production of consumption goods and production of capital goods. 2005 PAR T 1 : BAS IC TH EORY AND EMPIRI CS A BOU T PROS PERITY AND G ROWTH The model therefore involves a certain evolution of income per worker as well. the model's real interest rate. units of capital services during period t. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. The real price. capital services and labour services. firms. where 0 <c)< 1.. Hence output can be used either for consumption. or it can be transformed into capital via investment (at no cost.Basic Theory and Empirics about Prosperity and Growth 3. The commodities are output. There will be much more to say after this chapter. commodities and markets The economic agents in the model are households. We may think of the accumulated stock of physical capital as being directly owned by the households w h o lease it to th e firms. . A period should be thought of as one year. There are three commodities in each period.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 60 I Part 1 . also referred to as consumers. C. and total investment. In the market for output the supply consists of the total output of firms . Y 1• The demand is the sum of total consumption. but we will show that it can easily be interpreted as including government expenditure and taxation . K. p 1 = r. must be set aside to compensate for depreciation. units of capital can give rise to a supply of K. r 1.1 The basic Solow m odel The model presented in this chapter describes a closed economy (the next chapter will present a Solow model for an open economy). Initially we will not explicitly include the public sector in the model. It thereby contributes to answering the limdamental question of what determines 'the wealth of nations'. This rental rate has a close association with. but is not equal to. Considering the simple framework. 3. The price in the output market is normalized to one. but the basic Solow model makes a great contribution and is indeed an important workhorse model in economics. in th e market for capital services the supply comes from the consumers. Agents. and possibly a government. also called producers. The name for c) is the rate of depreciation . for the capital th at is worn out by one period's use. Hence. By a convenient definition of the unit of measurement for capital services (machine years) a capital stock of K. the model will take us remarkably far in understanding the process of economic growth and the sources oflongrun prosperity.
output 'viii a lso go up by 100 per cent. For our purposes it does not matter whether the user cost of capital.t). Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.Basic Theory and Empirics about Prosperity and Growth 3. L~1 • to produce output (value added). Finally.mction with respect to L. Y 1 = F(K. so F(J.L'1) = A. L'1) = FLK(K d. one can think of physical capital as being owned by firms who finance their acquisition of capital by issuing debt to consumers. in an alternative interpretation. where FKK is the second derivative '<\lith respect to K11 • and Fu. From microeconomic theory we know that a competitive profitmaximizing firm will want to employ a factor of production up to the point where its marginal product is just equal to its real price. We measure labour flows in man years. The variables Kf and L. In this interpretation one can trunk of the frrm's 'rental rate' as r 1 = p. We assume that FK and FLare positive for all input combinations (Kd. + o. the marginal product of labour is the partial derivative of the production fi. since one could simply apply the same production process t'<Vice. The firm laces the real factor prices r 1 and w 1.F(K". The marginal product of capital is the increase in output generated by an extra unit of capital. so a labour force of L 1 can give rise to a labour supply of L 1• The real wage rate in period tis denoted by w ~' All three markets are assumed to be perfectly competitive. the necessary first order conditions . In our model.' are the amounts of capital and labour demanded by the firm in period t. Formally. say. is the real interest rate on debt plus the depreciation on one unit of capital. 100 per cent. Ld) is homogeneous of degree one.rKf . Ld) < 0. The production sector The production side of the economy is modelled as if aU production takes place in one representative profitmaximizing tlrm.> 0 . while the demand. Ld) for all ?. L"l of F(K". There L~ a replication argument for this assumption: it should be possible to produce t'<Vice as much from a doubling of inputs. the marginal products are also assumed to be decreasing in the amount of the factor used. L1• comes from the households. according to the production function: (1) The production function is assumed to display constant returns to scale. when the input of one factor increases. .1. Ld) > 0. This means that Fn(K". L~1• comes from the firms . In the labour market the supply of labour services. ?. The firm uses capital input. L'1) < 0 and FLL(K". L'{). so economic agents take the prices as given. Kf. r 1. is a direct leasing rate or the sum of an interest rate and a depreciation rate. L. that they are utilized up to the 'natural rate' defined in Chapter 1. the marginal product of the other factor is assumed to go up. is the second derivative '<\lith respect to LJ.wLf. and labour input.K". In accordance with standard microeconomic theory. also called the 'user cost'. so FKK(Kd. subject to the technical feasibility constraint. ¥ 1.1 and Lf to maximize pure pronts Y. In the latter case the real price of the use of a unit of capital for one period. This implies that available resources are fully utilized in all periods or. if we increase both inputs by. L"). it is the partial derivative FK(K". and in each market the appropriate price adjusts so that price taking supply becomes equal to price taking demand.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . 2005 3 CAPITAL AC CUMU LA TIO N AN D G RO W TH 61 Alternatively.t. In mat hematical terms this means that the function F(K.t. L<IJ with respect to K11• Similarly. and it will want to choose ¥ 1• K. In words. denoted by F1JKd.
an optimal combination offactor inputs.K.Basic Theory and Introducing Advanced PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH (resulting from differentiation.'. Y.I.L. L. Because the factor markets are perfectly competitive. it will pay for consumers to supply all of their capital to the market. and the wage rate. L 1) w 1L 1 F 1JK. the stocks of capital and labour are predetermined and given..)L. Therefore (4) and (5) determine the period's rental rate. r1 . the rental rate. so labour supply must equal the labour force. If one combination (K.. so capital supply is equal to the capital stock. 3 Nevertheless.L1) for any A > 0.L = ()Y y .and the wage rate. as will be explained below. K1• Furthermore. L.') = w.()L y is the elasticity of output with respect to labour input. Capital accumulation © The McGrawHill Empirics about Prosperity and growth: The basic Companies.' and L.. w . etc.. If a function is homogeneous of degree k.())..) take the form: F iK1. = FiK. FK and F. L'{) = r1• (2) FL(K. r 1 (and hence the interest rate. F(K 1.'.)K1 Y. Hence. gives: F iK1. Since F is homogeneous of degree one.. F(K1• L1) = (6) ' (7) Note that FKK ()Y K =y ()K y is the elasticity of output with respect to the input of capital. L~) fits in the two equations. (3) Note that (2) and (3) do not determine the levels of K1 and L1 given r1 and w I' even though they are two equations in two unknowns. and '£ FI. (5) In each period t. This is a consequence of the assumed homogeneity of the production function . AL •) = F~K•. respectively. as these depend on the given levels of capital and labour. then its first partial derivatives are homogeneous of degree k. (4) F L(K. L•). we lind that the shares of capital and labour in total income in any period t are: r1K.1• :A. must fu lfil (2) and (3) to be optimal. Inserting K:1= K1 and L1 = L1 into (2) and (3). The distribution of income and the CobbDouglas production function Using (1). w 1• respectively.. We will argue below that as long as the rental rate is just slightly positive. 2005 Macroeconomics and Growth Solow model SorensenWhitta. adjusts such that labour demand becomes equal to labour supply. K.'. r. (4) and ( 5). adjusts to equate capital demand with capital supply. FK(J. L.3. so does (?.Jacobsen: 62 I Part 1 . This leads to a theory of the functional distribution of income. L. we are going to assume that consumers supply labour inelastically. are homogeneous of degree zero. L 1) = r. .) = WI . Thus equations (6) and (7) say that 3..K•.
capturing that these amounts of input may vary over time. . (10) Using (8). The total factor productivity. the labour income share is given as the output elasticity with respect to labour input. and L.(K/L. since then called the CobbDouglas production function: B > 0.' = L. (9) and (10).L. Note that the inputs of capital and labour are indexed by t.)K.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . Charles Cobb found that there exists one such function... 1) = F. and B are given parameters. L. L)/L]. and if pure profits were negative. we would all expect new llrms to enter the market until profits were competed away. O<o<l. The numerator only depends on K/ L. L)/ L = LF(K/ L..(K1. Since the production function is homogeneous of degree one. Notice also that the firm's pure profit measured in real terms is Y 1 . and B is often referred to as the total factor productivity. 1)/ L = F(K/ L.(K/ L. L. so F(K. because if pure profits were positive.(K. so F. A similar observation was already puzzling the economist Paul Douglas in 19 2 7. More precisely.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies. 4 Over time the capital. you should check carefully that the Cobb. this is the same as requiring the output elasticities '>\lith respect to capital and labour to be constant (and hence independent ofKJL. (8) Here o. we would expect some failing llrms to be driven out of the market until the remaining flrms were able to breal< even. 2005 3.) . L) = FK(K. L) = (1 . labour's share depends on the capital.. In Chapter 2 we saw that in the long run the factor income shares are remarkably constant. 1). This is an appealing property of a perfectly competitive economy in longrun equilibrium.labour ratio has been systematically increasing in developed countries. Capital accumulation and growth: The basic Solow model 3 CAPITAL ACCUMULATION AND G RO W TH 63 the share of each factor in total income is equal to the elasticity of output with respect to that factor.). implying that pure profits are zero. The subject of Chapter 6 will be the Solow model '>\lith technological progress where B is increasing over time.K. According to (7). B.)B K L: )" = ( w.)K. L. + w. + F1. Apparently this should have implied a changing labour income share. He asked the mathematician Charles Cobb whether a production function exists which has all the properties assumed above. As we have seen in {6) and (7). which is changing over time.)LJ.o. because F1 is homogeneous of degree zero. = Taking partial derivatives in (8). This is the sense in which the model in this chapter is the basic Solow model. we have F(K. according to our theory. How does this constancy fit with our theory of income distribution? At ilrst sight it does not seem to llt. 1).(r. we find that Eqs (4) and (5) take the particular form: K1 K. (9) FL(K. so it is assumed here that there is no change in technology over time.) = F(K. L. L) = L" F.. and which will produce constant income shares when factors are paid their marginal products. L. Write labour's share as F1(K. because F is homo· geneous of degree one..[FK(K1. + FL(K. . L)j[F(K. KJL.Douglas production 4. has no time subscript. (K1. and the denominator only depends on K/ L.. L.labour ratio.).)L1. and this will generally depend on the input combination. and inserting the equilibrium conditions.
Basic Theory and Empirics about Prosperity and Growth 3. The fact that the labour income share stays rather close to 1(see again Chapter 2) suggests that we have a reasonable value for the a appearing in (R).. the household sector saves the exogenous traction. which is predetermined from capital accumulation in the past. Therefore the supply of capital services in year tis inelastic and equal to the size of the capital stock K1• Third. C1. respecting his intertemporal budget constraint. The intertemporal budget . Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. Aggregate production functions will be assumed to have the Cobb. because each consumer considers r 1 as given.Both these features will be used extensively in the growth theory presented in this book. Hence the intertemporal budget constraint is: = (11) If we were deriving consumption and savings behaviour from explicit optimization. an inelastic supply oflabour would follow if the income and substitution eflects from a change in the real wage were exactly otrsetting each other. oK. so the assumption of an inelastic labour supply is not a bad ftrst approximation. as mentioned. we make a simplifying short cut. .on the capital leased to the 11rms during period t. According to this constraint. a consumer seeking to maximize his income will want to lease all his capital to the 11rms.Douglas form.. If households were trading off consumption against leisure so as to maximize utility.e. which is predetermined from the past evolution of the population..K. and that it does imply constant income shares. r 1 > 0. S 1 Y. however. we will consider ~ as a reasonable value. Second. and whenever we need an estimate of a . constant rate of saving and investment is empirically plausible in the long run. The behaviour of consumers is described by four features.+ 1 . longrun labour supply is not too sensitive to changes in real wages. and a labour share of 1 . The household sector The number of consumers in period tis L..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 64 I Part 1 . a utility function defined over streams of present and future consumption would be maximized under the intertemporal budget constraint (11).C. Here. s. each consumer must decide how much to consume and how much to save.. The representative consumer must decide on consumption. .a. namely a number around~. Empirically. We assume that the outcome of the underlying optimization is a constant savings rate. (12) We are going to show below that the assumption of a given. of total income in each period: where 0 < s < 1. 2005 PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH function does indeed have all the properties attributed to F.Douglas form is a reasonable approximation for an aggregate production function for longrun analysis. so the total supply oflabour is L. a capital share of a. Y. We assume that the household sector behaves as one representative consumer who earns all the economy's income. i. For any positive rental rate. The proximate constancy of income shares in the long run suggests that a production function of the Cobb.. each consumer is assumed to supply one unit of labour inelastically in each period. to the consumer's real stock of capital from period t to t + 1 must equal his gross saving in period t minus the depreciation. the consumers own the capital stock K. First. and hence on gross saving.. the addition. K.
bK.. nand b. (17) K. Given an initial input combination K0 . as illustrated in Fig. = BK~'Ll a . Eq. (17) gives the gross saving. or rather of the labour Ioree: where n > .Basic Theory and © The McGrawHill 3. (16) S. (13) The complete model The basic Solow model consists of the following six equations.rl . (19) determines the next period's labour supply. w 1. endogenous variables that can adjust during the period in circles. L. ~~ CD (15) (16) r. in period t. K. Y. (18) determines the next period's capital supply. these six equations determine the dynamic evolution of the economy.. + 1 . WithY. (14) (K..~ Figure 3..SorensenWhitta. all repeated from above: Y. s. (15) (Kr w.. Hence. the 'biological' behaviour of the households is described by an exogenous growth rate.+ 1 = (1 + n)L1• (19) The parameters of the model are a. B. and L. 3. of the population.. . Capital accumulation and growth: The basic Solow model Empirics about Prosperity and Growth Companies. and one can start all over again with period t + 1 etc.+ K. = sY. . r . = (1 . L. so the first of the above equations gives the supplydetermined output.Jacobsen: Introducing Advanced Macroeconomics I Part 1 . (11). Eq. Eq. from the labour supply L. directly while the next two equations give the rental rate. of period t.1.1: The dynamics of the basic Solow model Note : Predetermined endogenous variables in squares. and then from 8 1 and K. +1 . and the wage rate. is itself important in our model: it is the aggregate capital accumulatimr equatiorr stating that new capital originates from net savings. S. are predetermined. K. determined. respectively. Finally. Fourth.l G) (17) ~ ~ ® / ~ / ~ (19) ~ ~. n.a)B L: . 1 and L. +1 have been determined. rt = aB L.+ ~ L2J _________________________(_ 18_)+ .1 . 2005 65 3 CAPITAL AC CUMU LA TION AN D G RO W TH constraint. 1  = (18) S. In period t the inputs K. L0 in some initial year zero.
cR. It. 5 Combining the above equations one llnds that: q = cf(1 . respectively. Let expenditure behaviour be described by the ratio of public consumption to GDP.1.cf)(1. Yt . (20) Again. where: s = 1 . to private disposable income. C. where we also allow cf to be timedependent. Cf. In Chapter 16 we shall see that cf is actually quite stable in the long run.Ct.. minus depreciation. (18). Then: q = cf(Y. where 0 < if < 1. is equivalent to It = St.ss:cJrily :1 q + Cf. We could instead have explained the capital accumulation equation in the following more direct way: by dellnition. Inserting S.cf . and by the ratio of public investment to GDP. since S.. Total (national) consumption.cj0 + cfif. 2005 PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH We have chosen to present the fundamental capital accumulation equation.(5K.. Y.if)Y1• Thm priv:cJtP. The assumption we make is that the share of total consumption in GDP. We allow cf and if to be time dependent and assume cf +if < 1. The national accounting identity for the output market. all we need to do to interpret a government into our model (1 4)(19) is to think of the constant saving and investment rate.onsumption :1s :1 fr:1r.r.). r. Capital accumulation and growth: The basic Solow model Companies. respectively: Cf =cfY.. is: constant. = [cf + cf(l ..if)]Y. = I. and that government expenditure for public consumption and public investment is Cf and If.cf. = C. w hi r..K.if). S1 = sY. = Government The more direct way of viewing the capital accumulation equation is useful for extending the model interpretation to include a government.cf.if).. as the representative consumer's intertemporal budget constraint. is the addition to capital: K1+ 1 .Basic Theory and Empirics about Prosperity and Growth © The McGrawHill 3. If = iJY. national gross investment must equal national gross saving. for It in the defmition of gross investment gives (18). Assume further that the government balances its budget: Let cf denote the ratio of private consumption.cf(1 . where 0 < cf < 1.SorensenWhitta.if) = (1 . cf + cP(1 . c)K. if. = Ct + I.Jacobsen: Introducing Advanced Macroeconomics 66 I Part 1 . 11 = 8 1• Hence. is a given constant. gross investment. Then national savings must be a constant fraction of GDP as well.. appearing in (17) as being given by underlying parameters 5. . s.cf ..T. Y. Assume that the government collects a tax revenue of T1in period t.h is n o t nP.  T1• where 0 < cf < 1.tion of (~J)p is rf(l . cf.cf.
8 0.4 .CrlY.. Note that s.2 0 <0 0.6 ef1"f..4 Cf!Y.2 Netherlands 0. co 0 <» <» 10 <» <» 0 0 0 "' . Table A at the end of Book One collects some data relevant for growth analysis.Jacobsen: Introducing Advanced Macroeconomics I Part 1 .2 0 co Belgium 0. why if affects s through the term cfi1). For a typical Western developed country...Basic Theory and © The McGrawHill 3.6 United Kingdom 1 C/~ 0 <» <» 10 <» <» 0 0 0 "' 0 0 <0 <» 10 <0 <» 0 10 ._ <» <» 0 10 <» <» co co 10 <» <» 0 0 0 "' co 6 10 .. USA 1 0. the national investment rate. or somewhat above.4 0..2: The c onsumption share of GOP in several W estern c ountries Source: OECO Economic Outlook database.2 10 <0 10 <» 0... Figure 3._ <» <» 0 10 <» <» co co 0 <» <» 10 <» <» 0 0 0 0 1 10 <» <» co 0 10 .8 Cf!Yr 0._ . CPJY Cf!Yr 0._ <» 0 10 <» <» co co 0 <» <» 10 <» <» 0 0 0 0 "' 0. among them average investment rates (sKin the table) over the period 19 60 to 2000 for a large number of countries..2 10 10 <» Ctf Yt 0..2 indicates that the ratio of total consumption to GDP has indeed stayed relatively constant over long periods in developed economies.2 0 0 10 <0 «>  <» ~ 0.._ <» <» 0 10 <» <» co co 0 <» <» 10 <» <» 0 0 0 "' CrfY. for instance. 0 <» <» Denmark 0..f_f!Yr 0 <0  <» 10 «> ~ R <» . 22 per cent. 0.._ <» <» CrfYr 0. a reasonable value for the model parameters could thus be.. CjY.6 Cf!Yr  1•..8 .6 r. and Cf/Y1 to move in opposite directions. An essential assumption underlying the Solow model is that the sum of the GDP shares of private and public consumption.8 . say.6 0. relatively stable._ . 0 0 <0 "'  GriYr 0. with a slight tendency for Cf/Y._ ._CyfV 0. while less developed countries can have investment rates far below 20 per cent.6 0.4 0 10 .4 0. is decreasing in cf and cf.OIV <» <» (J) «> 0 10 .4 0 Cf/'ft Finland 1 <» co CPJY.SorensenWhitta._ .8 0. which is the share of total consumption in GDP. keeping the sum./ 0. and increasing in if (explain. 0. 20 per cent are typical for developed countries. is a given constant._ .8 0 Cf/'ft 0. 2005 3 CAPITAL AC CU MULA TION AND G RO W TH 67 as in (20). Capital accumulation and growth: The basic Solow model Empirics about Prosperity and Growth Companies. You will see that values around.._ <» <» 0 10 <» <» co Figure 3.
There is.2 gives an indication th at the Netherlands and Denmark experienced downward structural breaks in the consumption share of GDP in the early 19 80s. however. Monetary Theory and Policy. The figure does not prove that money growth has no real impact in the long run. that is. One should treat this statement with caution. the systematic way that monetary authorities respond to economic events. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. while monetary policy will be at the heart of the analysis throug hout Book Two. Fig. as we will explain later in this chapter. Carl E. in an uninterrupted stream lil<e the real numbers. many economists think that the quality of monetary stabilization policy. Both in the US and in Denmark a change in money growth of a given size goes hand in hand with a change in inflation of the same size in the longer run (note th at the figure focuses on the long run by considering averages over periods oflO years). We have formulated time as running discretely like the integers 0. a rather sophisticated econometric literature on this issue. The peak eflects occur after a lag of several quarters (as much as two or three years in some of the estimates) and then die our.. one important economic event that we want to be able to analyse in our growth model is a permanent structural change in the savings rate from one constant value to another. 3. We are not arguing that savings rates stay constant forever. 1. 2000. 2005 PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH We should emphasize that sometimes breaks in national saving and investment rates occur. Likewise.. On the other hand. for instance by allecting our parameters. do not affect real variables such as GDP in the long run.Basic Theory and Empirics about Prosperity and Growth 3. MIT Press. . the conclusion ' . or continuously. This should explain why you will hear no more of money in Book One of this text. on the shortnm eflects of money is th at exogenous monetary policy shocks produce humpshaped movements in real economic activity. This is because most economists think that in tile long nm money does I'!Ot matter for real economic variables. Walsh gives an account of this literature and summarizes as follows as regards the long run 6 : 'Money growth and intlation display a correlation of 1: the correlation between money growth or inflation and real output growth is probably close to zero . 2... Cambridge and London. given th e quality of the financial system. 3 . For instance. Figure 3. . or of the shortterm nominal interest rate set by the central bank.. only that they tend to stay sulllciently constant over su!Ilciently long periods to justifY a modelling assumption of an exogenously given savings and investment rate. can have economic consequences also in the long run. However. W alsh. Indeed. most economists think that the exact level or growth rate of the nominal money stock. 6.. Certainly the quality of the financial system can have a longrun impact.'. the economist Carl E. Time In one respect our Solow model does not quite look as suggested by Solow himself. Perhaps it is more in accordance with how most people think of time to assume that time runs smoothly. Money? There is no direct trace of 'money' in the Solow model. In his advanced textbook on monetary economics.3 gives a sharp illustration.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 68 I Part 1 .
I Part 1 . 2005 Macroeconomics and Growth Solow model SorensenWhitta. 0konomisk Vaekst: En 0konometrisk Analyse af Danmark 18701981. Friedman and A Schwartz. Denmark and the US. Capital accumulation © The McGrawHill Introducing Advanced Empirics about Prosperity and growth: The basic Companies. Often the Solow model is formulated in continuous time. For instance. n . if the investment intensity at timet is I(t). consumption. Sources: For USA: M . so if the investment intensity stays constant at I over the lUll year. saving and investment are smooth li.Jacobsen: 3 CAPITAL ACCUMULATION AND G RO W TH USA 8 • 6 ~ ~ 1940s 4 c 2 ·. then the total investment during the year is I . 1930s • 4 990s+ • 1900s 2 4 6 4 2 0 2 4 6 8 10 12 14 Growth rate in money supply(%) Figure 3. The University of Chicago Press. 1982 and IMF International Financial Statistics.Basic Theory and 3. Jurist·og 0konomforbundets Forlag. A Monetary History of the United States 18671960. 1991 and OECD Economic Outlook (through EcoWin). In that case flow variables such as production. For Denmark: Niels Kaergard.3: Money growth and inflation.mctions of time to be interpreted as intensities. £ 0 0 69 • 1 880s 2 4 0 4 8 6 12 10 G rowth rate in money supply (o/o) Denmark 10 8 ~ Q) ~ c 6 0 2 £ 0 ·. then total investment over the year from time t = 0 to t = 1 is I(t)dt. Money supply is measured by M2. 18702000 Note: Money growth and inflation rates are average annual rates over each decade.
The main argument in favour of continuous time is that it is sometimes mathematically more convenient.bK(t). K(t + ~t) . 3. for instance.Basic Theory and Empirics about Prosperity and Growth 3. that is important..§. since the dot indicates change over time anyway.K. y.Y. The first one. and it is easier to do simulation exercises with a discrete time model on a computer. +1 . The last one says that the growth rate of population is n at all times: Y = BK" L 1" .. is K(t) = l(t) . Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.~. not GDP itself. dK(t) •.2 !\~./L 1• = lCD ..~g.~~.(~K. In the Solow model we are therefore interested in output per worker. k(t) is the intensity of change in capital at time t. is the total capital stock at time t. Y.~~~g. (23) S = sY. the counterparts of (14)( 19) above..a)B (LK)".~~'~'~''''''''''''''''''''''"'''''''''''''''''''''''''''''''''''''''''''''''''''''''.~~?.~~. ) .~~.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 70 I Part 1 . With these conventions. This will help you to master both ways of modelling time. 2005 PART 1: BASIC THEORY AND EMPIRICS ABOUT PROSPERITY AND GROWTH Stock variables are also smooth functions of time \. In continuous time the change in capital must be expressed as a time derivative. The counterpart of the discrete time expression. The case for discrete time modelling is that it is directly applicable to empirical data which are typically reported lor discrete periods like a year or a quarter. (21) r = aB(~r • (22) 1 w= (1 .. · at>o ~t dt which measures the change in capital per unit of time (since there is a ~ t in the denominator) at a specific point in time (at a nanosecond.~. (26) L The exercises at the end of the chapter will ask you to analyse the Solow model in continuous time. says that the production intensity at any time is given as a CobbDouglas function of the intensities by which capital and labour are used at that time.. so to speak) . = l . L . for instance. stating that the intensity of change in capital at time tis the gross investment intensity minus the depreciation intensity at timet.=n. Output and capital per worker For the prosperity of a nation it is GDP per worker or per capita. one often drops the explicit reference to time and one simply writesK = l .. the Solow model in continuous time consists of the following six equations. It is a matter of choice whether to formulate the Solow model in discrete time or in continuous time. K. which will be impnrt:mt :mclusl\lill in your liJrthl\r shJclil\s. K=sY  (24) (25) f)K. So.c5K.K(t) = K1 t = lun .~~~. When formulating the Solow model in continuous time.Vith the obvious interpretation that K(t).
1 . also called the capital. KjL. Yt Figure 3. taking logs on both sides. a Bk. or the capital intensity: k. The per capita productimzjimction is illustrated in Fig.1. the (approximate) growth rate in output per worker is propor tional to the grmvth rate in capital per worker.1 to period t be denoted by gf and g~.4: The per capita production function . One feature revealed directly by (27) is that in the basic Solow model (where B is assumed to be constant). This production function exhibits diminishing returns with the marginal product. an increase in production per worker can only come from an increase in capital per worker.. (14) above. and the proportionality factor is exactly the capital share.labour ratio. respectively. and subtracting gives: (28) According to (28).". In the basic Solow model economic grmvth is thus completely linked to capital accumulation. 2005 3 CAPITAL ACCUMULATION AND G RO W TH 71 Define similarly capital per worker in period t.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . Let the approximate growth rates of y 1 and k 1 from period t .4. it follows from dividing on both sides by L 1 that: = (27) This is like a production function where output is GDP per worker and input is capital per worker. decreasing from infinity to zero as k 1 increases from zero to infinity. From the first equation of the Solow model. a . Capital accumulation and growth: The basic Solow model Empirics about Prosperity and Growth © The McGrawHill Companies.Basic Theory and 3. 3. Writing an equation like (2 7) for both period t and period t .
(k.S.) v il'l r. depreciation per worker in period t is ok.) by y. + (1 . Tn this Wl'IY thP.. Using also the definitions of k. given k0 . .Basic Theory and Empirics about Prosperity and Growth 3.volntion of l'lll thP... = Y. Then divide this equation on both sides by L1+ 1 . and finally there is a 7. we can substitute Bk. fort= 0.". Subtracting k 1 from both sides in (29) gives the socalled Solow equatio11: (32) This has a nice intuitive interpretation. the per capita (or worker) magnitudes. or ll1e lrtmsiLiun etJUaLiun. Total consumption in period tis C.. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.". which gave the dynamic evolution of Y..+ (1 . On the righthand side are the elements that can create such an increase: saving per worker in period t is sy. . = (1 . 2005 PART 1: BASIC THEORY AND EMPIRICS ABOUT PROSPERITY AND GROWTH The law of motion We can analyse the Solow model directly in terms of the variables we are interested in.s)Rk. follows from (k.) lists k . to arrive at: (29) T!Ji~ i~ lhe basic law of moliuu. the transition equation (29) determines the full dynamic sequence (k . For a given initial value.s)y r Therefore (r. (29) determines the capitallabour ratio. k... Finally. have thus been boiled down to the single equation (29) giving the dynamic evolution of k. and L. andy.) fo llows from (k . The notation (k) will be used for the full dynamic sequence fo r k..+ 1 = sY. andy 1. that is." lory. on the righthand side using that L1 + 1 = (1 + n)L .. which can then be inserted on the righthand side to determine k 2 . S. = (1 .o)k. First. With (k.. = sBk. rl ynl'lm i r.. onedimensional.. (27). one gets: kt +J 1 =  1+ 11 [sy. given an initial capital intensity. (y 1). variables of interest can be derived. nonlinear difference equation.J. k 1 • of year one. = Bkf. 1. K. In this way. 7 The lour equations (14) and (17)(19). which adds to capital per worker. . from the per capita production function. There is another illustrative way to state the economy's law of motion. which subtracts from capital per worker. On the lefthand side is the increase in capital per worker..o)K. etc. P. k 0 • of the capitallabour ratio in year zero. so consumption per worker is c 1 = (1 .s)Y.) determined. the rental rates '0\Till be given from (15) and (16): (30) (31) The sequence for output per worker.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 72 I Part 1 . 2. fur lhe cctpilal inleusily k 1 following from the basic Solow model. where one expresses the change in the capital intensity as a function of the current capital intensity. insert the savings behaviour {17) into the capital accumulation equation (18) to get: K.) of capital intensities. Note that (29) has the form of a firstorder.
implying that the slope of the transition curve falls monotonically from infinity down to a value less than one. at which the transition curve intersects the 45° line in the upper part of the figure. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.+ 1/L 1) = 1/(1 + n). The lower part of Fig. Now. Using the definition of k 1• we have in that case: k1 + Jk 1 = (K 1+ 1/K1)/(L... Usually annual depreciation rates on aggregate capital are estimated to be between 5 and 15 per cent (there is some uncertainty here because there are many kinds of capital with dillerent depreciation rates). and the slope of the curve falls monotonically from inHnity to zero.k. as drawn. You may wonder why 1'1 enters ( 3 2) exactly as it does. we have k1+ 1 . k 1 will be taken to the horizontal axis as shown. while most often the lower end of the interval around 5 per cent is considered most plausible for aggregate capital. divided by (1 + n). Table A at the end of Book One).= This shows that the slope of the transition curve decreases monotonically from inHnity to (1 . k 1+ 1 = k 1. ask1 increases from zero to infinity. This means that capital per worker v.'.viii converge to the specific value given by o o a. or 11 > which is plausible empirically.Tj]l be reduced from t tot+ 1 by an amount caused by population growth. Here is an explanation. 3. = 0 . Di!I'erentiating (29) gives: dkt+l so. Population growth rates never slump to . Then k 1 will be the vertical distance from the horizontal axis up to the transition curve. sin ce the ray has constant slope n + > 0 . It follows that over time.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . The latter slope is positive and less than one if n + > 0. The dynamic evolution of the capital intensity is given by the staircaseformed broken line. nk.5. 0) and is everywhere increasing.05 (see.[n/(1 + n)]k 1• This shows how capital per worker evolves when there is no compensation for population growth. has also been drawn. We can therefore safely assume n > <~. so k. 1+n . The two curves must intersect each other once to the right of k 1 = 0. . or k1+ 1 = k1• Hence the intersection must be at the same k. is k1+ 1 . and then k 1+1 . Hence the transition curve must have a unique intersection with the 4 5° line (which has slope one).k. Where this intersection occurs. You should now recognize the way r1 enters (32). \.1 + (1 .'. shows the transition equation as given by (29). It contains the curve sBk. The upper part of the figure. 2005 3 CAPITAL AC CUMU LA TION AN D G RO W TH 73 subtraction.5 illustrates the dynamics of the Solow model.k1 = . Convergence to steady state Figure 3 . Assume an initial capital intensity k0 > 0.Bk. stemming from the fact that there are more workers in period t + 1 to share a given capital stock (ifr1 > 0). the socalled Solow diagram. to the right of k 1 = 0 . and then vertically down. The iterations of (29) described above appear in the transition diagram as follows.0. the trnnsition diagram. The 45° line. and by that alone. according to (32) . illustrates the Solow equation (32) . k 2 will be the vertical distance up to the transition cunre. and the ray (n + f~)k 1 • The vertical distance between the first and the second.+ 1 = k/(1 + 11).o) dk. k.<~)/(1 + n). This curve starts at (0. Assume th at in period t saving and investment have a size that keeps the capital stock exactly unchanged from t to t + 1 : K1+ 1 = K. and by going from the associated point on the transition curve horizontally to the 45° line.Basic Theory and Empirics about Prosperity and Growth 3. for instance.
and it will do so monotonically.o)k1) ko k* Figure 3. getting closer and closer all the time and never transcending to the other side of the intersection point. and its slope is everywhere posit ive. 2005 Macroeconomics and Growth Solow model SorensenWhitta.5: The transition d iagram (top). Furthermore. everywhere to the left of this intersection (except in k . Mathematically.Basic Theory and Introducing Advanced PART 1: BASIC THEORY AND EMPIRICS ABOUT PROSPERITY AND GROWTH 1 1 n (sBk.3. and the Solow d iagram (bottom) the intersection between the transition curve and the 45° line. Capital accumulation © The McGrawHill Empirics about Prosperity and growth: The basic Companies. the convergence is global in the sense that it holds for any strictly positive initial kw 8 8.Jacobsen: 74 I Part 1 . = 0) it lies above the 45 • line and everywhere to the right it lies below. to establish global.> 0. .' + (1 . monotonic convergence we have been using the following properties of the t ran· sition curve : it has a unique int ersect ion with the 45 ' 1ine fork.
k 1 is therefore increasing over time. k 0 in ko k* Figure 3. = l. = k. k*. andy. when the curve is above the ray.5. > 0 . k0 . Comparative analysis in the Solow diagram Figure 3. we want to illustrate the properties of our model a bit further using the Solow diagram. + 1 > kt' so k. 3. the gross savings rate shifts to s' and stays at its new higher level thereafter. In conclusion. Since the old parameter values have prevailed for a long time. We imagine that the economy has first been characterized by the parameter values a .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . This value is the unique positive solution. s. This will.". converges toy* over time. the dynamics of the Solow model are such that from any strictly positive initial value. while no other parameters change. This k* is called the steady state value for the capital intensity. but then from some year zero. the economy should initially be at (or close to) the steady state corresponding to the old parameter values. as indicated by the arrows along the k. y* = B(k*)a. Capital accumulation and growth: The basic Solow model 3 CAPITAL AC CUMU LA TION AN D G RO W TH 75 In the Solow diagram. and two dilferent curves. B./Yt' we may as well say th at it is the gross investment rate that has increased.6: An increase in the savings rate in the Solow diagram . while to the right it is decreasing. explain the convergence process in economic terms. one must have k. To the lett of the intersection point. and vice versa.+ 1 = kin (29) or (32) and solving fork.6 is a Solow diagram just like the lower part of Fig. among other things. It contains the ray (n + f~)k . or k. The steady state is important and Section 3 below \>\Jill locus on it. where s' > s. which is obtained by setting k. nand(~ for a long time.k. There is an associated steady state value for output per worker. axis. 2005 3. Before turning to that. We have described the model's convergence to steady state in mathematical terms. must be increasing over the periods. + 1 . Remember that since s = Sj Y.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies. 3.5. sBk~ and s' Bk. the capital intensity will converge monotonically to a speci1ic value given by the intersection points in Fig.
output and income per worker are thus unaffected. so the capital intensity is k0 . The key endogenous variables in steady state The steady state capital. say. This will show us which fundamental characteristics of an economy can create a high level of income and consumption per worker in the long run according to the basic Solow model. etc. s y 0 in period zero against sy 0 previously. assume that the economy is initially in steady state at parameters a . . B. k* andy*. so k1 stays unchanged from period to period. one will also have y 1 > y 0 . Furthermore. n and b. Current savings will therefore more than compensate for depreciation and population growth. where the savings rate was s. . k 2 > k 1. because most of the capital is destroyed in a war or an earthqual<e during period zero. Suppose that the economy's capital per worker then drops to k of its old value at the beginning of period one. a larger part will now be saved.labour ratio is given as the unique constant solution . Hence capital per worker will increase again. In year zero. because it is predetermined and given by capital accu mulation and population dynamics in the past.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 76 I Part 1 . As above. (This may illustrate something lil<e what happened to Japan in the Second World War) . In this steady state current savings are just large enough to compensate capital per worker lor depreciation and population growth. s'Bk~ > s'Bk~. In the Hgure. You should do this in a Solow diagram and convince yourself that the economy will immediately start a recovery back towards its initial steady state.~~. Perhaps this recovery can partly describe the strong growth performance of Japan after the Second World War. B.~. . k0 . y* and other key variables of the steady state as functions of the model's parameters a. However. while the parameters stay unchanged.?. One can use the Solow diagram for another type of comparative analysis which does not involve a parameter shift. and so on. 2005 PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH the figure.Basic Theory and Empirics about Prosperity and Growth 3. y 0 . and out of the increased income per worker will come even more savings per worker.. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. We can analyse this situation in the Solow diagram by keeping the curves in it unchanged. k1 + 1 = k 1 = k. s. In the short run. It is of interest to express k*. andy 1 will converge to the new and higher steady state value y* = B(k*)".4Y. n and c).~~············································································································································. s'Bk~~ > (n + c5)k 0 • so capital per worker will increase from period zero to period one.~~. respectively. but reducing the capital intensity to (l )k 0 from period one.~. out of the unchanged income per worker. In the long run successive increases in the capital intensity and in output per worker will make k 1 converge to the new and higher steady state value k*. the initial capital intensity will be unchanged.. Our model predicts that in the long run the capital intensity and GDP per worker converge to particular steady state levels. it will enable us to confront the model's steady state predictions with the data. With k 1 > k0 . s. In an exercise you will be asked to analyse further the remarkable recovery of the Japanese economy in the postwar period and investigate whether this seems to accord with the Solow model. Therefore output per worker stays unchanged at y 0 = Bk~~ as well. when the increase in the savings rate occurs. These increases reflect the economy's increased capacity for saving.
(38) from w hich. b. 9.. What can make a country rich in the long run? The answer suggested by (34) is that a relatively h igh technological level.ln(11 + <~)]. a low population growth rate.a). a high rate of gross savings and investment. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.a 1.In B + . this is a meaningful expression.g. such a k must fulfil: sBk" k 1 . from 20 to 24 per cent." = sB/(11 + f~) . 2005 3 CAPITAL AC CUMU LA TION AN D G RO W TH to (29) or (32) above. 1 (l 1. Y1 = Bk7: y * = B1/(la) S ) a/( l a) _ ·_ . tend to imply relatively high GDP per worker in the long run. ( 11 + (~ (34) Consumption per worker is c1 = (1 . for instance. or .. B. . such as the rate of investment and the population growth rate.a In y"' = .Basic Theory and Empirics about Prosperity and Growth 3. 11. according to our model a 20 per cent increase in the investment rate. For instance. th e elasticity of y * with respect to sis <4(1 . in the long run. Taking logs on both sides of (34) gives: .[Ins . a/(1 . The fact t hat we have found exactly one positive solution confirms our earlier stat ement that there is a unique posi· tive intersection between the transition curve and the 45 ' line in the trarsition diagram. 'o\Jill imply an increase in people's average incomes of around 10 per cent in the long run.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . e. so: c* = B l/(l a)(1  s) _s( a/(1a) 11 + b ) (3 5) Likewise.In other words. Since a is around ·k. The importance of such a statement for matters of structural economic policy is obvious. or t he c urve and the ray in the Solow dia~ram. inserting k* into (30) and (31) gives the real factor prices in steady state: )1 r* =a s( 11 + (~ 111"' = (1 . (34) gives a handy formula for how an economy's GDP per worker should depend on a few basic structural parameters. and a low depreciation rate. From the latter. 9 The steady state output per worker is found by inserting the particular value k* fork 1 in the per capita production function. Denoting the solution by k*. (33) o Given the assumption 11 + > 0.a)B l/(1a) (36) ( S )"/(I .·ll + (~ a) (37) These simple and innocentlooking formulae contain some very sharp and important prediction~ about an economy's longrun performance.<~) (sh ow this by dillerentiation with respect to s). should be around f.s)y 1 in any period t. the elasticity. this gives: k* = B 1/( 1a) _s_ ( 1'1 + (~ ) 1/(1 a) (11 77 + f~)k = 0. s.
His theory of interest rat es is summarized in: K1ut W ic ksell. we do have good data for GDP per worker and for the investment and population growth rates for many of the countries in the world. and the greater will be the demand for capital.c5. 'The Influence of the Rate of Interest on Prices'.it in thP. Figure 3.Hnmc:P. 1907.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 78 I Part 1 .fl p.es a prediction for the longrun GDP per worker. 3 per cent. The parameter a measures the response of output to an increase in capital input. You will easily compute that these parameters result in a value for the natural interest rate per annum of p* = 3.22. Today he is considered as one of t he great figures in the history of economic thought. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. How well does the prediction of (34) fit \>\lith the data on these variables? Figure 3 . This is the influence of productivity on p*. while the steady state real interest rate is p* = r* . be: a = s = 0.7 plots the GDP per worker in 2000 against the average gross investment rate between 19 60 and 2000 across all countries for which data were available from the PWT and which got a grade for data quality h igher than D in the PWT. 2005 PA RT 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROS PERIT Y AND G ROWTH The 'natural' interest rate: productivity and thrift Just as (34) gi\. and hence for replacement investment to compensate for depreciation and population growth in steady state. Testing the model's steady state prediction of GOP per worker We can test our prediction of longrun GDP per worker empirically .nsity to Si!VP. reasonable values by western standards for the parameters entering into (39) could. 21320. and the lower the equilibrium real interest rate. However. Although we are going to revise our formula for the natural real interest rate somewhat in more sophisticated models later on. the price of capital. Here we run into the difllculty that it is not so easy to get reliable data for difl'erences between countries in technological variables such as B and c).b. 11 = 0. and (5 = 0. A higher value of n + b increases the demand for replacement investment as well. Recall that the r* of (3 6) is the steady state value for the rental rate for capital. (39) s The p* of (39) is sometimes referred to as the ·natural' rate of interest. .Basic Theory and Empirics about Prosperity and Growth 3. The higher the value of a. on an annual basis. (36) gives a prediction for the real interest rate. and hence increases the equilibrium real interest rate. The influence of thrift is rp.05 . th P greater the supply of capital. for instance by considering crosscountry data such as those in Table A. 17. Knut W icksell Nrote at the tum of the twentieth century.~: th P. t.hi ghPr th P. As argued above and from our Table A. of .. w We see that the natural rate of interest in the basic Solow model is determined by the forces of productivity and thr(ft.8 plots GDP per worker against average population growth rates between 1960 and 2000 10. it is promising that our extremely simple model predicts a longrun real interest rate that is so close to empirical estimates.005 . the higher is the productivity of capital. Economic Journal. pp. so: p* = a n+b .c:tP. i!ppP. This is very close to the longrun average values of annual real interest rates implied by the observed historical diflerence between nomin al interest rates on longterm bonds and the observed rates of inflation. following the Swedish economist Knut Wicksell who was the first to use this term.nHti onHI propP. ceteris pari/Jus. as evidenced by Table A.
. Capital accumulation and growth: The basic Solow model 3 CAPITAL ACCUMULA TION AND GROW TH ~ '5 "0 (!) m m 6 70000 IRL••USA 60000 NLO 50000 GBR + SWE+ ( • ~ 40000 a. •• + CHN + 0 ONK . In particular.2 0..25 0..IIIIP=<~"'1'~_ _ _ _ _ _ _ _ _ _ __ ~ + SWE +KOR 30000 +~~.7: Real GOP per worker against the average investment share. _ ••..+ • 'JOR ION. g 30000 ~ 20000 EGY 0.# • .02 0.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 ."'~~ •• SYR.005 0.•orvo. ...Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies.2000 0.15 0..__. The increasing relationship in Fig. we will investigate whether the model prediction of a longrun elasticity of output \Vith respect to In s .03 0. 19602000 Figure 3.045 0.~ • •• • 20000 +~·~~A~R~G~nn~·~. 1960. 2005 3. 0 0. 85 countries Source: Penn World Table 6.:+++ + .04 0.•.015 0. For now.~~~~~ ~~!~ BRA•~· . A high savings rate and a low population growth rate do seem to go hand in band with a high level of GDP per worker.ln(n + b) of around stands up against the empirical evidence (and we ask you to do part of the job in the important Exercise 9 at the end of this chapter). (ij & . +SYR ~ UK 10000 +•• ~ ~ 11rrrM~IO~N.8: Real GOP per worker against the average annual population growth rate.. 3. the t .. 85 countries Source: Penn World Table 6.8 are nicely in accordance with the basic Solow model.·.~n~ •..1 0. +USA 60 000 B~l: NOR + + + +NOL ++ 50000 FIN~<~•• + + +ISR 40000 GBR~.~ •• # o+.35 Figure 3.05 0.. Later we will discuss the goodness of these fits further.7 and the decreasing one in Fig.~··~~~~~~~. ++ 10000 0 • • • ?: " +tjE[ +~_yFIN 0 0 79 •• • NOR ~ •..01 0..025 0. .035 0.C_HN. _ . + ..3 Average investment share of GOP.1 for the same countries. • +KOR jpN +THA +ROM 0. 3. • •• ·. 1 70000 ~ IRL.J.05 Average annual population growth rate.
.... Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. a result arising from the explicit consideration of the capital accumulation process. Assume that government increases its propensity to consume... and hence Y 1.. Hence....... an increase in government demand has no influence on GDP... where K1 is predetermined by capital accumulation in the past according to the old investment rate.. where the national savings and investment rate...... reflecting the economy's decreased potential for saving and investing..... s.... Structural policy and the golden rule ....... and lor a number of reasons...........SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 80 I Part 1  Basic Theory and Empirics about Prosperity and Growth 3.... ensuring that GDP per worker stays unch anged in the first period...... etc. c~.. The decrease in capital supply caused by the government's increased propensity to consume also implies that the longrun equilibrium interest rate increases......... is given by (20)... as we sh all see in Book Two on shortrun economic fluctuations and stabilization policies.. leading to a further fall in saving and capital accumulation... it may be able to do so iH the short nm if the economy is not currently at the capacity constraint....... Over time Y.. and therefore there will be less capital.... In the long run....... In that case a rise in public consumption will lead to less than full crowding out of private spending....... This should remind you of lessons learned earlier........ Out of the (unchanged) output and income in the llrst period of the new policy... such th at s decreases permanently to a new and lower constant level as given by (20).. will be unaflected by the policy change: Y1 will continue to grow at the rate n... In the long run there is more than full crowding out............. because of its negative influence on capital accumulation? Certainly not.... as we noted in connection with (3 9) above.. will thus develop more slowly than without the policy change..... However............ both K.. 2005 PA RT 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROS PERITY AND G ROWTH accordance between the data and the steady state prediction of the basic Solow model is promising and a main reason why we claimed in the introduction to this chapter that the model takes us remarkably far in understanding the sources of longrun prosperity...... but private investment will be fully crowded out by the increase in government consumption... What 11Vil1 be the ellects of such a policy according to our model starting from a steady state? This is easy to answer since we have already studied the effects of an increase ins..... the policy ch ange considered will have an effect on GDP and on GDP per capita.. Crowding out Consider the model interpretation with a government. and the ellects are now just the opposite.. It is an important and general insight that a government cannot create a positive influence on GOP or GDP per capita in the long run by boosting government consumption. Should one conclude from the above analysis that in the long nm government expenditure can haYe only a negative in fluence on economic activity and consumer welfare......... and hence output.... in the short run (in the first period of the new and lowers).... in connection with the socalled classical macroeconomic model: in a period where output is determined from the supply side by the available amounts of resources.. ..... thus bringing output per worker to a new and lower steady state value. for example... and L....... a smaller fraction than before will be saved and invested... in the succeeding period. Here we are concerned with the long run and with structural policies...
Thus.g. the benefits to other citizens that a person is literate) may motivate public provision or linancing. say.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . For example. In addition.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies. Third if public consumption spent on. for example. as sh own by (20). For educational services. Thus part of public consumption may help to increase the productivity variable Bas well as the private savings rate 1 . a main conclusion is that a government wishing to promote its citizens' longrun incomes should follow structural policies that can somehow improve technology.in the Solow model. a ta.x financed increase in government's propensity to invest raises the national investment rate to a new and higher level. a taxllnanced increase in public spending on schools and hospitals will most likely reduce private consumption on education and health by a roughly similar amount. our model does not give any explicit description of the underlying incentive mechanisms. Whether government is assumed to influence the model's basic parameters one way or the other. consumers may be willing to accept a fall in aggregate national income arising from an increase in government consumption if this increase enables the public sector to provide essential public goods which cannot be provided in adequate quantities by private. infrastmcture. say. much of what is normally categorized as public consumption may be seen as delivery of productive inputs to the private sector production process. instead of increasing its spending on consumption. However. and such activities may be meaningful according to the Solow model. a positive . health and education services may improve the productivity of the labour force. through lhe government's direct control over public consumption and investment. and an increase in if has a positive influence on s. education and health is a nearperfect substitute lor private spending on these items. and reduce population growth. be a positive external efrect associated with the technology. if some technology is only adopted by private llrms because it is subsidized. However. unatl'ected. from an economist's point of view. both distributional concerns about equality of opportunity and the presence of positive externalities (e. the government could choose to spend an increase in tax revenues on investment in. Capital accumulation and growth: The basic Solow model 3 CAPITAL AC CUMU LA TION AN D G RO W TH 81 Government expenditure motivated by longrun concerns First. Fourth. proiltmotivated producers via the marketplace.. and public institutions guaranteeing law and order (examples of essential public goods) may also be productivityenhancing and encourage a h igher private savings rate by safeguarding property rights. that is. Certain ly many countries use various programmes to enhance technology. through incentive policies. we can imagine that the government can inlluence the private savings rate. s. 2005 3. Second. increase the national propensity to save and invest. cr Incentive policies We have cousiuereu policies Lha l iullueuce Lhe ualiou al savings rale. health services to be publicly provided or linanced. there may be distributional reasons for wanting part of. for instance policies that restructure the tax system. which has a positive longmn impact on GDP per worker. leaving total saving and investment and thereby total output. there should. or the population growth rate. In our model this possibility is represented by an increase in if rather than in cf. Finally.
is called the golden rule savings rate. point to some main issues. since the marginal increase in c* obtained from an increase in sis small if s is already close to a (explain this). As we notice from Table A at the end of Book One. y*. 2005 PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH spillover eflect on agents other than the 11rm investing in the technology. Policies to promote private savings can be of many types. We can. a. The golden rule of saving One may get the impression from our discussion that it is good to have as high a savings and investment rate as possible. This consumptionmaximizing savings rate. a system of welldellned. there is civil war in the country. preferring to consume now rather than later.Basic Theory and Empirics about Prosperity and Growth 3. but according to (3 5). what should the savings rate be? Using (35) to maximizec* with respect to s gives the savings rate: s** =a (Exercise 7 will ask you to show and explain this).SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 82 I Part 1 . down to around 10 per cent. Quite often these policies take the form of tax incentives lor savings and investment. and in some countries an improvement of institutions is the number one need lor initiating positive economic development. however. If s is close to one. is capital income overtaxed compared to labour income? Are the dillerent types of capital income taxed in a uniform manner so as to prevent distortions in the pattern of savings and investment? At the more basic level. we cannot conduct an explicit analysis of such incentive policies. However. As mentioned. the basic Solow model strongly suggests that a good system for handling savings and channelling them into productive investment is important for prosperity. Because of its emphasis on capital accumulation. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. However. poor countries typically have much lower savings rates. and since higher saving implies postponed consumption which represents a welfare cost if people have 'time preference'. Equation (34) certainly says that a higher s gives higher longrun income per worker. There must be sale banks and assets in which savings can be placed. is around 1· If policy makers are interested in consumption per person. many rich countries in the world have savings and investment rates somewhat above 20 per cent. To achieve such a goal may be very dillicult if. Perhaps they should not even go close to that level. If the ultimate purpose of production is to enjoy the highest possible level of conswnption per worker. secure property rights and a sound llnancial system are important prerequisites for largescale private savin gs. GDP per worker will be maximal. lor instance. Some of the East Asian 'growth miracle' countries had investment rates up to around 1at the end oft he hventieth century. consumption per worker will then be close to zero. The capital income share. More advanced technology is not preferable at any cost. Thls is an example of a general and fundamental economic insight: institutions are very important for economic performance. it does not require much rellection to see that a maximal savings rate is not preferable. For some poor countries this is a serious issue. they should not try to drive savin gs rates above t. and sometimes even considerably less than that. For the latter . and the most essential policy to promote savings in such countries is the establishment of safe property rights and a good flnancial system. since we have not derived private savings behaviour from optimization. We will return to the discussion of policies to improve technology later in cmmection with models with endogenous technological innovation. s**. For example.
k 1 < k* . since k.. The value of B is just a normalization.. in the basic Solow model there is transitory growth in GDP per worker on the way to steady state... Hence capital per worker will increase from one period to the next... and in steady state there is no growth in GDP per worker. There can be growth in the GDP itself. correspond to a high annual population growth rate and a o o . As noted earlier. sBkf..... growth in k1 goes hand in hand with growth in y. a = t. is increasing............. An example It is relevant to investigate how last or slow transitory growth is according to the basic Solow model.... increases. but only at the speed of population growth..... and the resulting gross savings per worker... and 11...... each unit of additional capital per worker \>\Till generate ever smaller increases in income and gross savings per worker (due to diminishing returns to capital). ultimately (in steady state) gross savings per worker.... will be growing..... for c~ assuming a period length of one year........... n = 0....... Consider an economy that is initially described by the following parameter values: B = 1. which would mean constant returns to capital alone... During the transition phase.......... \>\lith the assumed aggregate production function.... growth in GDP per worker has to be rooted in growth in capital per worker...... corresponding too... will exceed the amount (n + c~)k.08 ............... has a 'marginal product'. both k. If growth fades within a few years..' were instead a ray lying above (n + o)k... Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.' in period t. Production and income per worker will be Bk...... At the same time the additional savings needed to compensate for population growth and depreciation increase..Basic Theory and Empirics about Prosperity and Growth 3. 3.. Why is it that growth has to cease in the long run according to the basic Solow model? Here.. Assume that the economy is initially below steady state. the model could not deliver much of an explanation of growth over decades...) As the steady state is approached.. As k.... = 0.. \>Viii just cover what is needed to keep capital per worker unchanged.. a..' being curved \>\lith a decreasing slope.. 2005 3 CAPITAL AC CU MULA TION AND G RO W TH 83 countries policies that could work to increase savings rates would seem to be important for raising the longrun level of economic welfare... (Try to envisage the growth process if sBk....' l . respectively. but no growth in steady state...5 Economic growth in the basic Solow model ... Transitory growth The longrun prediction of the Solow model is the steady state... diminishing returns to capital plays a crucial role.. The production function lor output per worker....Bk.. Hence.. Therefore. needed to maintain capital per worker in the face of population growth and depreciation (look again at the Solow diagram in the lower part ofFig. while a and have been set at reasonable values.....SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 ......5)......03 and s = 0....05 . The macroeconomic dynamics outside steady state in the Solow model are of interest since this is where there can be economic growth...... The values for nand s......... = 1. (1'1 + o)k 1• In the Solow diagram diminishing returns is reflected by the function sBk.. y 1 = Bk.. which decreases with k 1 because a< 1........ sBk~... 3..'....
(k.92 2. It takes around 12 years to get half the way Table 3. Furthermore. .19 3. or approximate by g[ = In y. y.y .9 . there is considerable transitory growth for a long time.27 2.24._ 1 .918 0.93 4. **computed as (lny. say.048 2.SorensenWhitta.38 2. (Try to explain why the increase in y* is not 100 per cent.310 1. c 1 and g[ over some 50 years after the change in the savings rate.049 1. For instance. the savings rate increases permanently up to a level typically seen in rich countries.) But how fast is the transition to the new steady state? We can simulate the transition equation (2 9) with the savings rate equal to s and the other parameters as given above over a number of periods t = 2.762 1.732 1. It illustrates the growth perspectives. The simulation gives rise to the time series shown in Table 3.006 0.907 2. can be constmcted.) by using c0 = (1 .98 0.000 1.942 0.)j y .270 1.).' in all years. exact (o/oY gf.33 2. 3 ...298 1.ln y 1_ 1) x 100 . which depicts the evolution in y .1. For each year one can then compute the growth rate in GDP per worker. six to be precise (explain this).316 0 0 (new st._ .) by using y.797 0.) one can derive a corresponding sequence (y.196 1.463 1. Using (3 3) and (34) one tlnds steady state capital per worker and GDP per worker.28 3. The new steady state is easily computed: k*' = 5.832 0. 73 .00 4.184 2.97 5 . = k.1.Jacobsen: Introducing Advanced Macroeconomics 84 I Part 1 . approx (%)** 0 1 2 3 4 5 6 7 8 9 10 1. as an elasticity of y* with respect to s equal to ! would suggest. according to the Solow model.208 1.000 1.s)y 0 and c. Capital accumulation and growth: The basic Solow model Companies. .) Note: *computed as ((y1 .20. while the other parameters stay unch anged. 3. . which could be descriptive of a typical poor country (see again Table A).s')y 1 for t".In y 1_ 1 • These operations are most easily done on a computer using a spreadsheet.66 2. 3.760 0.97 2.323 0. there is a substantial initial sacrifice of consumption lor some years.69 3.76 3. respectively: kx = y* = 1.965 0.st.) In this way a sequence.135 1. and illustrated graphically in Fig.) x 100.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill 3.173 1.155 1.81 4.986 1. (In year one capital per worker will be at the old steady state level. Now suppose that !Yom an initial year one.16 1. 2005 PAR T 1 : BAS IC TH EORY AND EMPIRI CS A BOU T PROS PERITY AND G ROWTH low propensity to save.240 1.000 1. ./y.40 2. There are several interesting features to note from Fig. andy*' = 1.000 1.614 1.892 0.317 1. s' = 0. 'liVe assume that the economy is initially in steady state at these parameter values.>l.1: Simulation output Year kt tt ct gf.. = (l .63 2.18 1. an increase in the savings rate by 200 per cent implies a 420 per cent increase in capital per worker and a 73 per cent increase in GDP per worker.9 . lor a poor country bringing its savings rate up to the standards of rich countries. since the increased savings rate during period one only h as an ellect on capital per worker from period two. From (k.863 0.094 1. and a sequence (c. starting with k0 = k 1 = 1. In the long run.00 4.920 0.
4 New steady state value for c 1.4 0 5 10 15 20 25 30 35 40 45 50 Year 6 5 4 %3 2 Growth rate of y. Transitory growth in the Solow model takes place over decades for realistic parameter values. Capital accumulation and growth: The basic Solow model 3 CAPITAL AC CUMU LA TION AN D G RO W TH 85 2 1. it is evident that the relative sizes of the parameters a. and this is still purely transitory growth arising from the permanent increase in the savings rate 20 years earlier. 3.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 .9: The evolution of y" c. the less curved and the closer to a straight line the savings curve. (11 + o)k. such as the lower part of Fig.2 0. nand c} are decisive for how fast convergence to steady state will be. This relative slowness in the adjustment outside steady state is why it is justified to say that: the basic Solow model is a growth model. will be.5. Intuitively from the llgure.8 0. The smaller 11 + ois.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies.6 0. Exercise 3 below will give . and af after the inc rease in s from the old to the new steady state level of GDP per capita. The higher the value of a. Year Figure 3.8 1. ray will imply a long transition period and hence slow convergence.6 r=========::::====== New steady state value for y 1. sBk~. the lower will be the position of the ray. From a Solow diagram. After 20 years the annual growth rate of GDP per worker is still close to 1 per cent. a high and relatively flat sBk~ curve and a low (n + <~)k. 2005 3..
on both sides of (32) gives the modijied Solow equatio11: k t+ . the growth rate of y . 1 + 1'1 (40) In the modilled Solow diagram. k. 3. Capital accumulation and growth: The basic Solow model PART 1 : BAS IC TH EORY AND EMPIRI CS A BOUT PROSPERITY AND G ROWTH more formal substance to the fact that the speed of convergence can be measured by the parameter A= (1 . Dividing by k. ~ ko Figure 3.. In the numerical example above this feature appeared.a)(n +b) . In a modified Solow diagram such as Fig. since g[ = ag~. (Fig. the growth rate of the capital intensity is thus proportional to the vertical distance between the decreasing curve.. where the curve and the line intersect. A 'cousin' to the Solow diagram is useful for illustrating the growth process outside steady state. jumped up and then decreased monotonically back towards zero. the 11rst country \<Viii grow faster and converge to a higher 11nallevel of GDP per worker. conditional on countries being structurally alike.10). the countries being otherwise structurally alike. 3.1 . The same \<Viii be true for the growth rate in y .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 86 I Part 1 . The model feature that growth is higher the further below steady state the economy is accords with the idea of conditional convergence dL~cussed in Chapter 2 . k. If one country has a higher savings rate than another. It follows that the growth rate in k 1 will be higher (in absolute value) the further away from the steady state k1 is.L.kox (1 + n) kc.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies. However. the positions of the curve and the line depend on the structural characteristics of the economy. After the initial increase in the savings rate.10: The modified Solow diagram k* sBk[' . 2005 3.k 1 I t = .10. the decreasing curve of the 11rst country will be situated above that of the second country.[sBk.'.(II+ f~)) . sBkf. The steady state is. .1 • and the horizontal line at n + (~. of course. If the two countries start out \<\lith the same GDP per worker.
...SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1  Basic Theory and Empirics about Prosperity and Growth 3.... The basic Solow model studied in this chapter is characterized by the following features: • Aggregate output........Douglas production function with constant returns to capital and labour and constant total factor p roductivity. GOP....... 3... such .. the Solow model supports conditional convergence.. These models will contain some form of technological progress.. The model implies that capital per worker and output per worker converge to particular constant values in the long run......... According to the model..... the real interest rate. 5. The expressions for the steady state values of the key variables contain some sharp predictions of how income per wo rker and consumption per worker depend on underlying parameters in the long run . • Competitive market c learing ensures that the amounts of capital and labour used in production in each period are equal to the predetemnined available amounts and that factors are rewarded at their marginal products.. respectively....... • In each period g ross saving is an exogenous fraction of total income or GOP.......... respectively. or the real wage rate...... consu mption per worker. This is at odds w ith the stylized facts of g rowth (in d eveloped economies).......... consumption per worker.............. 2005 3 CAPI TAL ACC UMULATION AND G ROWTH 87 the Solow model predicts that countries will converge to the same level of GDP per worker through a growth process where the countries that are initially most behind will grow the fastest..... The key assumptions of a Cobb..6 §. or policies to improve technology. policies to make a nation richer should mainly be policies that can increase the investment share of GOP and bring population growth under control......... Capital accumulation and growth: The basic Solow model © The McGrawHill Companies.. The convergence point defines the economy's steady state.. 4.. Depreciation on capital is given by an exogenous depreciation rate. the change in the capital stock equals gross saving (or investment) minus depreciation on the initial capital stock.Douglas aggregate production function and a g iven rate of saving and investment are motivated empirically by the relative constancy of income shares in the longrun and the relative longrun constancy of the GOP share of total consumption (private plus public). To c reate a high level of consumption per person.... 2........ The steady state has some empirical plausibility with respect to the predicted value for the longrun real interest rate and w ith respect to the longrun dependence of GOP per worker on the investment rate and on the population growth rate............ The labour force changes over time at an exogenous population growth rate.... wh ich is equal to the elastic ity of output with respect to capital. 1... In other words. These predictions are of importance for the design of structural policies to raise the standard of living. In the model's steady state there is no positive growth in GOP per worker... • Capital evolves over time such that from one year to the next.......l. and the wage rate are constant as well.. Remedying this shortcoming is one of the main purposes of the growth models to be presented later... 3. An implication is that the income shares of capital and labour are equal to the elasticities of output w ith respect to capital and labour.l::r!. the savings and investment rate should not exceed the golden rule value........~~Y... is produced from aggregate capital and aggregate labour accordi'lg to a Cobb . In steady state....~I....
where a and a are parameters. but also for the reason appeari n~=J from the answers to the next two questions. Find the marginal rate of substitution between capital and labour for the CES production function. partly for simplicity. . This is in accordance with the observed conditional convergence between the countries of the world. Economics has much more to say about the sources of prosperity and growth than what can be inferred from the basic Solow model. Show that profit maximization given rand w requires that this marginal rate of substitution is equal to r/w.Douglas function. This is defined as the elasticity in the composition. The elasticity of substitution between capital and labour and the income distribution The Cobb. These factors will be dealt with in later chapters. when the firm maximizes profits (minimizes costs) given rand w. The CES function is more general than the Cobb. human capital (the education and training embodied in people). This transitory growth is relatively longlasting for plausible parameter values." (omitting time subscripts and superscripts d for 'demand') implies a certain degree of substitution between the inputs.Douglas production function. of input demands with respect to the ratio. and show that the elasticity of substitution is a. Capital accumulation and growth: The basic Solow model PAR T 1: BAS IC THEORY AND EMPIRICS ABOU T PROS PERITY AND GROWTH that total factor productivity increases over time. This openness implies that capital can flow between countries at least to some extent.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 88 I Part 1 . oil and metals are factors of production which could be as important as physical capital and 'raw' labour. Exercises Exercise 1. and show that the elasticity of substitution between capital and labour is . the faster it will grow. Nevertheless we have chosen to work with the more special Cobb. The socalled CES (constant elasticity of substitution) production function is: 0 <a< 0. rj w. * 1. K/ L. 1. independent of rand w. independent of rand w. Furthermore. 6. For instance.1.Basic Theory and Empirics about Prosperity and Growth © The McGrawHill Companies. since the particu lar value a = 1 (or the limit for a > 1) corresponds to the CobbDouglas function. Y= BK• L 1 . As a measure of this degree economists use the socalled elasticity of substitution. a> 0. the process of transitory growth is such that the further below steady state an economy currently is. The basic Solow model does give rise to (positive or negative) growth in output per worker during the transition to steady state. 2.Douglas production function. It is therefore worthwhile first modifying the model to take account of a fact we have neglected in this chapter. but the basic Solow model is important as it stands. A Solow model taking capital mobility appropriately into account and describing the accumulation of capital and national wealth in an open economy will be presented in the next chapter. between the rental rates.Douglas function. 2005 3. the evolution of technology has to be an important determinant for longrun prosperity and growth as well. Furthermore. and natural resources such as land. namely the openness of real world economies. Find the marginal rate of substitution between capital and labour (the marginal product of capital divided by the marginal product of labour) for the Cobb.
k 0 > 0. the income share of labour under competitive market clearing is: wL 1 a = . Illustrate the above Solow equation in a Solow diagram with k along the horizontal axis and the curve sBka as well as the ray (n + o) k along the vertical axis. How are r* and w* affected by an increase in s? Explain. Demonstrate (from the Solow diagram) that from any initial value. y * and c*. income per worker. r""" and w*. s.(26).1 )/o ·  a L + (1 a) 4. the transition to steady state. k*. We use (for any point t in time) the definitions k = K/ L andy = Y/L etc. Show that the growth rate. for the real rental rates for capital and labour and compare again to the model in discrete time. 2005 3 CAPI TAL ACCUMULA TION AND GROWTH 89 3. according to the above formula? This gives an argument for considering a= 1 most plausible in models for the long run. Show that developed countries must then also have experienced increasing capital . and when is the growth rate ink. It is assumed that n + o > 0. nand o are the same as in the model in discrete time. how should labour's share evolve over time in the case of a< 1. If K/ L increases and eventually goes to infinity. a = 1 and a> 1. respectively (use k= 0). Show that with the CES production function. 1. Also compute the steady state values.Basic Theory and Empirics about Prosperity and Growth 3. Then show that the Solow model in continuous time implies the Solow equation: k = sBk" . What are the growth rates of y and Y in steady state? 4.1 k (n + o).yfy. Show that (it follows from the model that) the growth rate in kat any time is: k . In Chapter 2 it was shown that developed economies have experienced relatively constant positive growth rates in GOP per worker (annual rates around 2 per cent) and relatively constant capital . at any time t is a times the growth rate. 8 . The basic Solow model in continuous time This exercise asks you to analyse the Solow model in continuous time as given by the six equations (2 1).. the capital intensity. and consumption per worker. and hence in y.output ratios in the long run. k. for tloe capitallabour ratio. k*. Show that from k = K/ L it followsk/ k =K/ K L/L.1. Assume that initially k 0 < k~ . Compute the steady state values. in the long run (as t + oo).SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 .labour ratios in the long run. that is. The restrictions on the parameters a. Explain and compare to the parallel expressions found in this chapter for the model in discrete time. 2. For instance. Exercise 2. (41) Compare with (32) for the model in discrete time and give an intuitive explanation like the one given for (32) in this chapter. (42) Illustrate this in a modified Solow diagram. (Hint: You can do the 'logdiftrick'. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. how do r and w develop along the transition.y (K)(<~. largest? . 3. Explain why. except we do not have to assume n > . of k. respectively. will converge to a specific steady state level.= sBk". in words. k/ k. Describe and explain qualitatively.(n + o)k. first take logs on both sides of k = K/ L and then differentiate with respect to time).
a= i. etc. Capital accumulation and growth: The basic Solow model lCD © The McGrawHill Companies. and show that (given an initial capitaloutput ratio z 0 ) the solution is: ( 1 . . and an initial value k 0 . Show that the steady state value of z is z* = (1 .output ratio is the weighted average of its steady In other words. In the following assume for simplicity that 8 = 1.08. the capital. An analytical solution to the Solow model in continuous time It was stated ir this chapter that continuous time is sometimes mathematically more convenient than discrete time. Note that (43) is linear. Then show by doing appropriate substitutions in (42). that follows from (41).01 ." . and s = 0.03. kjy. For instance. by z "' k 1 . that: t =(1 a)SAZ.output ratio. Assume that in period zero the economy is in steady state at these parameter values. The trick is to define an auxiliary variable z. although the differential equation (41) is not linear. and what keeps it alive for some time? Now consder the same parametrization of the basic Solow model as in the example of Section 5. n = 0. Use this to show that convergence will be o slower the larger a is. Show that the solutions fork andy are: From the expressions it is clear that the larger A is. the faster w ll convergence to steady state be. we can find the (closed form) function (for k as a function of t). and then (working first time between period zero and period one) the population growth rate drops to 0. Exercise 4. that is. o= 0. since eu will go faster to zero the larger ). The effects of a decrease in the population growth rate 1. 8 = 1.·. and the smaller n + is. that is. Compare this to what was said on more intuitive grounds about the speed of convergence in Section 5 of this chapter. other parameters remaining unchanged. (This is illustrative for a developing country getting population growth down to western rates).( 1 elf) + z 0 e lf 1  s (1 n+o elf) 1 z 0 c lf.. Explain the economic effects according to the basic Solow model of a decrease (at some time) in the population growth rate from one constant level to a new and lower constant level: how does this affect the transition diagram and the Solow diagram? How does it affect the steady state values of capital. income and consumption per worker? Explain qualitatively the transition from the old steady state to the new one: what initiates the growth process. t.a)(n + c5)..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 90 I Part 1  Basic Theory and Empirics about Prosperity and Growth 3. Solve (43) by finding the characteristic roots. 2005 PA RT 1: BAS IC THEORY AND EMPIRICS ABO UT PROS PERITY AND GROWTH Exercise 3. First show that z is actually the capital. the weight going more and more to the first of these. at any time state value and its initial value. or from (42).05. we can actually solve it. is.a)s/1 = s/(n + o).a·ls z = . (43) where A"' (1 .
_. (n + c5)k1. say. and show in a diagram the evolution of y. The event we are interested in here is that at the beginning of some period. However. nevertheless. income and consumption per worker. Illustrate the point in your Solow diagram: the slope of Bk. Carry out a simulation of the model over the periods 0 to 100. Explain the difference. How does th s affect the Solow diagram and the capital accumulation process? Exercise 7. 2005 3 CAPI TAL ACCUMULA TION AND GROWTH 91 2. and the ray. . Does there seem to be a systematic difference in the speed of convergence between an increase in sand an increase in B? Exercise 6. both before and after this event the population growth rate stays unchanged at n.). and an increase in the savings rate as considered in the example of section 5? For instance.. not completely satisfied with the basic Solow model's steady state as a description of longrun growth? . Why are we. Exercise 5. What is the requ irement on s to ensure that the steady state capital intensity. Does there seem to be any systematic difference with respect to the speed of convergence to the new steady state (or the speed by which the benefits arrive) between a decrease in the population growth rate as considered here.' and (n + (5)k 1 intersect. Balanced growth in the steady state Show that all the requirements for balanced growth. period one say.In y. k*.. The particular steady state capital intensity is where sBk'. L. starting in steady state at the prevailing parameter values. c 1.)jy. sBk~.. Golden rule In this chapter it was stated that the golden rule savings rate that maximizes steady state consumption per person as given by (35) is s** = a . increases once and for all by a certain amount.' and (n + c5)k1• Compute the particular value k** of k . as these are stated in Chapter 2. for instance due to emigration. Explain this. By how many per cent does the reduction by ~ in the population !=jrowth rate increase the steady state values of capital. as usual.' should equal the slope of (n + o)k.. for the two experiments (approximately from the figures) compare the number of periods it takes until half of the distance between the old and the new steady state values for y has been eliminated. respectively? 3. consumption per worker in period t is the vertical distance between Bk'. becomes equal to the golden ru le capital intensity. For any given k. and draw the curve Bk. First analyse the qualitative effects on the Solow diagram. Explain this. A one shot increase in L1 Analyse in the Solow diagram. in steady state consumption per worker is the vertical distance between Bk'. over the 1 00 periods. Comment with respect to what a developing country can obtain by b irth control according to the Solow model. and the growth rate {y. Exercise 8.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1. _. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. the qualitative effects of a one shot increase in L. _1 (or the approximate one In y.Basic Theory and Empirics about Prosperity and Growth 3. Then consider the numerical example and analyse an increase in 8 from one to two by doing the appropriate simulations.y..' as well. that maximizes this distance. k**? Illustrate again in your diagram. Show this. To illustrate.' and sBk~. draw a Solow diagram with the curve. Hence. are fulfilled in the steady state of the basic Solow model. The effects of an increase in technology Do all you were asked to do in the previous exercise (there for a decrease in n) only this time fo r an increase in B. explaining the transition mechanisms.
012. 7 and 3. + c5) across countries i (with data quality higher than grade D). The issue is how well the basic Solow model can explain these series and thereby how Japan was able to catch up the USA. this exercise will consider the two transformed series that appear for each year and country dividing GDP per worker by the GDP per worker in the USA and multiplying by 100. k* 11 and y"" 11 . Does the relationship (by reasonable standards) seem to be a linear one? Estimate (by OLS) and draw the line of best fit through the points. respectively. ./(1.2 shows GDP per worker for the USA and Japan for the period 1959 to 1996. In Figs 3.8 we tested the steady state prediction of the Solow model by plotting. Further empirically testing the steady state prediction of the basic Solow model This exercise anticipates some issues taken up in succeeding chapters.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 92 I Part 1 .05.015. where the first is just a list of 1OOs. 2005 PAR T 1: BAS IC THEORY AND EMPIRICS ABOU T PROS PERITY AND GROWTH Exercise 9. of capital and output per worker.075. Compute the series (yy)~!59 and (y{) ~!59 and plot each of them against years in a diagram. s.ln(n. Give some qualitative description of the improvement in Japan's GDP. The recovery of Japan after the Second World War Table 3. These parameter values are realistic annual values (for some as averages for the considered period). 7 and 3.34 and n 1 = 0. purchasing poweradjusted currency (1996 dollars). The table only goes to 1996 to avo d considering (in the present longrun context) the more recent years of economic crisis in Japan. we normalize for each year the GDP per worker in the USA to 100. We assume that for both countries a = kand c5 = 0. What is the estimated slope of this line? How does this slope accord with the theory? Exercise 10. For the savings and population growth rates of the USA we assume s u = 0. in Japan with this 8? What must k~1 (capital per worker in Japan in 1959) have been? Describe roughly how far Japan . in questions 2 and 3. We will take the view that the two transformed series are the relevant data on the two countries. using for yi. and setting c5 = 0. and the slope in this relationship should be o. there should be a linear relationship between lny* and Ins . In the following we take (or perhaps test) the view that each country can be described by the basic Solow model with countryspecific parameter values. We concluded that the directions of the empirical relationships were nicely in accordance with the basic Solow model. and n .a).21 and n u = 0. The resulting series are called (yy)~:59 and (y {)i:59. To eliminate from these time series any underlying technological progress. The data you will need can be taken from Table A. We will first. assume that the two countries had the same technological level 8 over the period considered. and compute for Japan the GDP per worker as a percentage of that in the USA. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. and that the USA has been in steady state in all the considered years. GDP per worker in 2000 against investment rates (averaged over 1960 to 2000) and population growth rates (also averaged over 1960 to 2000). 2. Test this by creating a figure that plots In yi against Ins. What must 8 be for the USA to be in steady state at y*u = 100? What are the steady state values. the same data that were used in Figs 3. measured in a common.ln(n + c5). For Japan we assume s 1 = 0. respectively. the steady state prediction of the Solow model is more precise than 'y* is increasing in s and decreasing inn': as shown by (38).Basic Theory and Empirics about Prosperity and Growth 3. and it is a very useful empirical exerc se. However.8 (appearing in Table A). across countries. 1. That is.
simulate the model (the transition equation) for Japan over the period 1959 to 1996 and thereby create a modelbased time series (y {1 )~:59 for GOP per worker in Japan.710 35. respectively.068 36. Capital accumulation and growth: The basic Solow model Companies.876 30.259 6.645 25.432 39.782 11 '167 12.520 32.Basic Theory and Empirics about Prosperity and Growth I © The McGrawHill 3.119 21.167 44.363 36.048 47. to the steady state levels?).962 Source: Penn W orld Table 6. Plot the actual series.017 21.217 44.362 36.260 31 .735 41 . Starting with the computed k ~~.037 38. 2005 3 CAPITAL ACCUMULA TION AND GROWTH 93 Table 3.976 44.583 45.252 25.023 40.477 42.161 36.920 33.164 9.529 53.510 24.960 30.743 26.469 42.681 22.184 41 .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 .254 26.756 42.800 10.184 13.436 35.2: GOP per worker. as well as .367 39.164 49. USA and Japan (1996 dollars) Year USA Japan 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 29.302 39. and assuming the computed 8.603 52.880 7.184 36.689 8.402 15.361 56.271 50.314 20.462 23. 3.887 52.867 55.268 53.577 9.368 28.077 31.304 30.741 44.159 48.065 57.900 37.020 38. (y{)~:59 .353 28.532 27.174 53.733 37.736 20.098 18.047 16.079 36.735 21.495 18. 1 was from steady state in 1959 according to these computations (what were the percentages in 1959 of capital per worker and output per worker.411 53.370 33.
') .(n(k) + o)]. 4. n in the figure. and that the modified Solow equation is: k k 1 '+' = k. (y{2 )~:59 . Simulate the model for Japan again. Data for capital are not the most reliable. in (19).(n(k. Plot (y(2)~:59 and (y{)~:59 against time ir a diagram also indicating the steady state values. as horizontal lines. 1 +n . (19) is no longer exogenous. less infant mortality. How fast has Japan actually caught up to the USA compared to how fast it should have caught up according to the Solow model under the assumption of a common technological level? The assumption of the same 8 in the USA and Japan may be heroic. = 1 (k) (sBkf. and thus create a new modelbased series. andy.labour ratio. For low values of k. this time starting with k~ and assuming the computed 8 1. = h(y1)= h(Bk. n. y*u and y* 1\ as horizontal lines. 1 1 + n(k. (y{ 1 )~:59 . In Questions 4 and 5 we therefore assume a level k~~ = y~9 /2 of capital per worker in Japan in 1959. increases from the limiting value. Exercise 11. What must 8 1 be? Compute the associated steady state values. y'"u and y* 12 . n(k) is decreasing. 2005 PA RT 1: BAS IC THEORY AND EMPIRICS ABO UT PROS PERITY AND GROWTH the modelbased one. and an additional equation. Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. n(k) never falls below a certain positive value. k* 12 and y* 12 . states how the population growth rate depends on income per capita and hence on the capital. From that value.=n(k.). How far from steady state was Japan in 1959 according to these computations? 5. However. Endogenous population growth and club convergence in the basic Solow model Consider a growth model that consists of the same equations as the basic Solow model except that then in Eq. for output per worker in Japan. which can be different from that in the USA (the 8 above).1 .Basic Theory and Empirics about Prosperity and Growth 3. n is replaced with n. This is true fork. because as people get richer birth rates fall for various reasons. but rather an endogenous variable that depends on prosperity. up to a certain value.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 94 I Part 1 .k. Comment on the goodness of the model's fit and on the speed with which Japan actually caught up to the USA compared with the speed with which Japan should have caught up according to the Solow model under the assumption of different technological levels. etc. a~=Jainst time in a dia~=Jram that also shows the steady state values. since higher k. Show that the Solow equation that can be derived from the model defined this way is: k/+ 1 . ' ' It seems reasonable to assume a shape of the function n(k1) as illustrated in the figure opposite. ask.). but some empirical material suggests that in 1959 capital per worker in Japan was about one half of GOP per worker in Japan in the same year. Hence.. mean better living conditions. or just n. for Japan.) [sBka. . Another way to proceed in simulations is to rely on direct data for capital per worker in Japan in 1959 and compute an associated technological parameter 8 1.) +o)k. over the years 1959 to 1996. the function n(k1) is perhaps negative all but increasing.
three intersections between the curve sBk~. there may be three steady state values fork.) starts below a certain thresho ld value it converges to a low steady state value. departs a little from the steady state in question. and the curve n(k. (and hence y . while if it starts above the threshold value it converges to a high steady state value. If this is the case. the dynamics of the model tend to bring it back towards the steady state. Show by an appropriate illustration in the modified Solow diagram that w ith such a shape of n(k. that is.1 .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 1 . 2005 3 CAPI TAL ACC UMULA TION AND G ROWTH 95 n(k). Capital accumulation and growth: The basic Solow model © The McGrawHill Companies. .).Basic Theory and Empirics about Prosperity and Growth 3. how does the (unmodified) Solow d iagram look? Which of the steady states are stable in the sense that if k. Explain intuitively why it is that if k. and wh ich one is unstable? Relate the curves you have drawn in the Solow d iagram (unmodified or modified) to the idea of c lub convergence explained in Chapter 2.) + o..
f.. This is a good approximation to reality as long as international capital flows are relatively small. The potential for capitalllows to generate shortrun macroeconomic instability will be discussed in Part 7.. and how capital mobility and the associated opportunity to incur foreign debt brings beneHts as well as potential costs.. many critics of globalization have argued that increased capital mobility may destabilize the world economy and expose poor countries to the shifting moods of international investors... in recent decades the international mobility of capital has increased dramatically. However.~~P~~~}.Jacobsen: Introducing Advanced Macroeconomics Chapter I Part 1  Basic Theory and Empirics about Prosperity and Growth 4.. The light columns associated with the left scale measure the volume of international trade (the sum of exports 96 ...... In these circumstances a country may Hnance part of domestic investment via capital imports.................. whereas the model developed in the present chapter will highlight the potential longrun beneHts of capital mobility...... The rapidly growing mobility of capital is one of the most striking features of the ongoing process of globalization.. In this chapter we will develop a Solow model of an open economy to study how the dynamics of wealth accumulation are affected by international capital mobility.............. 2005 • • • • ~• Wealth accumulation • ~ and capital mobility • • :The Solow model for a small open economy n the basic Solow model the only source of domestic investment is the volume of domestic saving... Figure 4.......... ~..o/..~y~~~..?.. Indeed..1 illustrates trends in the degree of international economic integration from the late nineteenth century until the end of the twentieth century.. We will then set up a Solow model of a small open economy whose domestic capital market is perfectly integrated with the world capital market... or it may invest part of domestic savings abroad through capital exports..~..~.. Most economists believe that capital mobility tends to improve the longrun elliciency of the world economy..SorensenWhitta.. but the liberalization of international capital llows also creates new risks. Studying this chapter along with those in Part 7 should thus give you a better understanding of some of the contentious issues in the current debate on globalization.. This model will be used to analyse how an open capital account 1 modifies the dynamics of wealth accumulation in the individual country. .. We will start the chapter by documenting the trend towards growing capital mobility.~~~..~?..~g.... I 4......1 T~.... Wealth accumulation © The McGrawHill and capital mobility Companies....
In Fig. especially considering that the technological means of communication were less developed than today. and imports) relative to world GDP.. we see that even as late as 1960 the world's economies were still considerably less open to trade and capital flows than they were before the First World War. The dark columns associated with the right scale show the ratio of the total stock of foreignowned assets to world GDP. S eptember 2002. Indeed. culminating in the Great Depression ofthe 19 30s and in the catastrophe of the Second World War. there was much less fmancial integration in 1..1 this is rellected in the sharp drop in trade integration and financial integration after 1914.Jacobsen: Introducing Advanced Macroeconomics I Part 1 . Wealth accumulation © The McGrawHill and capital mobility Companies. . Thus the light columns indicate the degree of trade integration.r.3 of World Economic Ooutlook. 4.1: Foreign trade and foreign assets relative to world GOP Source: Figure 3. Countries allowing free capital mobility are said to have an 'open capital account ' since their governments do not control the items on the capital account of the balance of payments.Basic Theory and Empirics about Prosperity and Growth 4. This is why critics of the economic dominance of the industrialized countries often refer to this period as the heyday ofWestern imperialism. Indeed...~ ' r f """I 10 1 1 5 0 '""= 1870 rI rI I ~ r  1 1 .  15 60 1 rr.. International Monetary Fund.. and the dark columns illustrate the degree of financial integration (or capital mobility).. The First World War marked the beginning of a long period of economic and political instability. the average ratio of foreign assets to GDP was above SO per cent in the early twentieth century. 2005 4 W EALTH A CC UM ULATI ON AN D CAPITAL MO B ILIT Y 20 97 80 r D Trade (left scale) D Financial (right scale) 1 . The tlgure documents an interesting fact which is often forgotten in the contemporary debate on globalization: before the First World War international capital mobility was very high. as a result of large capital exports from the most developed countries (parti cularly Great Britain) to the less developed regions in the world. 1930 1938 1 20  I 1g6o 1980 1995 0 Figure 4. During this dark period in history national governments retrenched into inwardlooking economic policies involving strong trade protection and controls on capital ilows.SorensenWhitta.r   I     1900 19 14 40 . In particular.
5 23. Table 4.0 3. But most of the increase in capital flows stemmed from a combination of llnancialliberalization. Wealth accumulation © The McGrawHill and capital mobility Companies.SorensenWhitta.6 1.vned company retains and reinvests its profits in the host country.3 32.7 2.3 32.2 22. but more often it involves the acquisition by a foreign company of a controlling m.7 60.0 16.2 43.5 6.4 6. International capital flows are often divided into foreign direct investment (FDI) and foreign portfolio investment (PI).1 8.7 62.5 74.6 112.6 18.vnership share in an existing domestic company.7 14.1: International investment.4 3.2 36.4 120.2 14. this is also counted as FDI.and the rapid improvements in information and communication technologies which made it easier and cheaper to move capital across borders.1 4.0 120.0 181.4 41 .9 62.3 220.7 28.7 13. Source: World Investment Report 2003.0 40.3 Definition of FDI: Investment in a lasting enterprise resident in another country. 1 1.5 22.5 89.4 73.7 13.3 14.8 20.1 4. stocks.6 30.3 18.3 5.Jacobsen: I Part 1 . because most countries still maintained extensive restrictions on private capital flows across borders.0 2.4 22.2 142. and from 1980 to 199 5 it almost doubled once again.6 78.5 7. UNCTAD. in an attempt to maintain national autonomy in the conduct of monetary policy. Definition of PI: Loans in foreign banks and crossborder investment in bonds.5 0.6 19. To some extent this remarkable increase in capital mobility was driven by the simultaneous increase in world trade.4 7.3 3.7 10.4 46.1 23.1 shows that FDI has become increasingly important relative to domestic economic activity in recent years.7 10.0 56. since a part of international capital flows takes the form of trade credits granted by exporters to the foreign importers.9 17.Basic Theory and Introducing Advanced Empirics about Prosperity Macroeconomics and Growth 98 4.7 20. between 19 60 and 1980 the degree of financial integration roughly doubled.9 19.8 147.8 2. Table 4.9 5.5 22.9 87.4 8.8 29. If an already established foreignm.4 9. The distinguishing feature of FDI is that the investor is an active controlling owner who often brings along new production technologies and management knowhow with the capital invested.8 5 1.3 84.9 31.5 66.9 7 1.6 81.3 20. Because FDI often implies an international transfer of technology and is typically undertaken by large multinational enterprises. reaching the highest level ever recorded.7 60. 19852002 Stock of inward FOI Stock of outward FOI Stock of inward PI Stock of outward PI o/o of GOP o/o of GOP o/o of GOP o/o of GOP 1985 2002 1985 2002 1986 2001 1986 2001 Australia Austria Canada Denmark Finland France Germany Italy Japan Netherlands Norway Sweden UK USA EU 14.3 33.7 10.8 17.4 14.5 37.4 52.7 3 1. financial derivatives and ot her money market instruments. .0 60.8 74.7 4.3 153.0 28.7 27.9 2. 2005 PART 1 : B AS IC TH EORY A ND EMPIRI CS A BOUT PROS PERITY AND G ROWTH 19 60 than in 1900.1 51 .6 43.1 95.8 45.4 3.2 30.8 12.0 44.5 5. IM F International Financial Statistics database.5 40. it tends to attract special interest.5 18.9 3. However.9 10.4 4 1. FDI may take the form of 'greenlleld' investment where a foreign multinational company sets up a new production plant in a country.2 58.\>\lith one government after another abandoning capital controls .2 20.8 11.8 11 .1 116.0 5.0 12.0 54.
g. 1 in a regression equation of the following form where Y. is domestic gross saving in period t. is domestic gross investment in that period (which includes the replacement investment needed to compensate for the depreciation f~K. SjY1 > IJY. implying a 0 = 0 and a 1 = 1 in Eq.. whereas a country with a savings deficit will run a current account . the greater will be the tendency for domestic savings to flow out of the country in response to any given drop in domestic interest rates.. and where the superscript i refers to country i: (3) In a closed economy aggregate investment must equal aggregate saving. K. and F.reen countries. implying that a 1 < 1. International portfolio investment involves international debt flows (e.1. + <5K. K. is the rise in tbe stock of net foreign assets between period t and period t + 1. Thus: (2) is the amount of capital exported during period t. as documented in Table 4.. may be negative. is real GDP. or it may be used to build up foreign assets. Wealth accumulation © The McGrawHill and capital mobility Companies. One way of measuring the degree of international capital market integration is to calculate the correlation between national savings rates and national investment rates across countries. =I.Basic Theory and Empirics about Prosperity and Growth 4. In an open economy without restrictions on international capital flows . international portfolio investment (PI) still accounts for the bulk of international capital flows. and this fall in domestic interest rates induces savers to invest abroad rather than at home.Jacobsen: Introducing Advanced Macroeconomics I Part 1 . S. In the open economy the intertemporal budget constraint of the domestic representative consumer (or of the domestic citizens together) is therefore given as: (1) where S. leading to capital export or import as shown by (2). We see that capital exports are given by the difference betv. This may be done by estimating the parameters o. of the preexisting capital stock). International portfolio investors do not acquire a controlling ownership share in foreign firms and therefore are not actively engaged in running a business firm abroad. The greater the degree of integration of the domestic and foreign capital markets. Part of an increase in domestic saving will typically be invested abroad. so the lower o. In that case part of domestic investment is financed by capital imports from abroad. = S. A country with a savings surplus. F.+ 1 . investment and saving may deviate from each other.F. Alternatively.K. Of course. S. the purchase of foreign bonds or the extension of bank credit across borders) or the purchase of foreign shares for purposes of portfolio diversification. In an open economy with free capital flows. . via FDI or PI.SorensenWhitta. 0 and o. + 1 .reen domestic saving and domestic investment: that part of saving which is not invested at home must be invested abroad. domestic saving. will have a surplus on the current account of its balance of payments. 1 will tend to be. 2005 4 W EALTH A CC UMULATI ON A ND CAPITAL MO B ILI TY 99 While FDI is a more visible form ofinvestment.I. (3). may either be invested in building up the domestic capital stock. I. The reason is that an increase in domestic saving tends to reduce tbe level of domestic real interest rates as the supply of capital goes up. one can measure the degree of international capital mobility by focusing directly on the imbalances on current accounts betv.
plus net asset income from abroad.. 'Domestic Saving and International Capital Rows'. 2005 PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH dellcit.2.. 8 1 Y 1 + FF1 . and X. 11/Y.Basic Theory and Empirics about Prosperity and Growth 4. Some years ago economists Martin Feldstein and Charles Horioka estimated a crosscountry regression equation of the form (3 ). consisting of the trade balance X. Martin Feldstein and Charles Horioka.SorensenWhitta. 90. suggesting a rather low degree of international capital mobility. 3 14329. M. which shows a new estimation of Eq. as explained in Chapter 2). GNP). then net asset income from abroad is FF1• We may rewrite (4) as: (5) The magnitude Y. rF. lor a sample of 16 OECD countries over the period 1960. The llnding that savers apparently had a surprisingly high inclination to invest at home rather than abroad soon became known as 'the Feldstein. Wealth accumulation © The McGrawHill and capital mobility Companies. the lower will be the estimated value of a 1 in (3). Sj Y. (3) for the group of 16 OECD countries considered by Feldstein and Horioka. consisting of gross domestic product.M1 + rF1 is the surplus on the current account of the balance of payments.89 . However. If a country started to build up a substantial current account dellcit.M. To see this. One possible explanation lor the high correlation between domestic saving and domestic investment was that govermnents were eager to avoid large current account deficits because the OECD countries tried to maintain llxed exchange rates up until19 73. National saving is dellned as national income minus total consumption.. using average values of the investment rate. The indicated line of best Ht (estimated by OLS. The larger the average deviation between S 1 and 11 across countries.. . its government would normally take steps to increase domestic saving and/or reduce domestic investment to redress its current account imbalance for lear that a large deficit would invite a speculative attack against its currency (by currency speculators believing that the country might have to devalue to rebalance its current account). and the savings rate. over the years the correlation between domestic saving and domestic investment h as weakened.2. 2. 1 Their results are portrayed in the upper diagram in Fig. has a slope as high as 0. is the stock of net foreign assets at the start of period t. we may therefore write (5) as: = (6) This shows the wellknown fact that the surplus on the current account and capital exports are just two sides of the same coin and both are equal to the savings surplus. is total national income (the gross national product. but using data for the period 1990.Horioka puzzle'. recall that goods market equilibrium in an open economy requires (4) where C1 is total (private plus public) consumption. According to (6) a low value of a 1 therefore implies a large average imbalance on the current accounts of the countries considered.99. is aggregate imports. 1980. pp.74 . This is illustrated by the Hatter regression line in the bottom part of Fig. Denoting the current account balance by CA 1. X1 . Y. Economic Journal. is aggregate exports.Jacobsen: Introducing Advanced Macroeconomics 100 I Part 1 ... 4. 4. plus net foreign asset income.C. If r is the international interest rate (assumed to be constant) and F. FF1• By definition. + rF.
thereby weakening the link between national saving and national investment. UK and USA. Ireland.. Table 4..15 0..Horioka estimate... The Economic Journal. Austria...101 + 0....40 .. Italy.1 . . Sweden. ... and the explanatory power of the regression is also considerably lower. ..508 0.....20 0.20 0. 314329.2: National saving and national investment in 16 OECD countries Not e: The countries included in the sample are: Australia... New Zealand. Netherlands. 'Domestic Saving and International Capital Flows'.. Japan.2 provides estimates of a 1 for three successive periods for a larger group of countries which are divided into the 'old' OECD (those OECD countries for which data are .25 0. t = 4..... Canada. Belgit.25 0.30 c s Q) (/) Q) > £ 0.... Greece.... ...30 Savings rate 0...35 0.40 Savings rate Figure 4.SorensenWhitta. ...Basic Theory and Empirics about Prosperity and Growth 4... 2005 4 WEA LTH ACC UMULATION AND CAPITAL MOBILITY 101 19601974 0. 90. Data source for bottom diagram: OECO and Penn World Table 6.601 (SlY). 1980. = 0....30 0. Finland..1........20 0...592...l • 0.... Both of these llndings suggest that international capital mobility has become significantly higher over time...35 19901999 0.15 0. Data source for upper diagram: Martin Feldstein and Charles Horioka.... .j 0. Wealth accumulation © The McGrawHill and capital mobility Companies. We see that the slope coelllcient a 1 for this recent period is only about twothirds the size of the original Feldstein. OLS·Iine: (I!Y)..Jacobsen: Introducing Advanced Macroeconomics I Part 1 ......m. Germany.. fil = 0.40 0...15 0.25 0.15 . . pp...35+. Denmark..35 Q) ~ 0. .
.... available for all the years back to 19 75 )... the economy is so small that the volume of domestic saving and investment does not have any noticeable impact on world interest rates.. In particular. The empirical evidence surveyed above suggests that capital mobility is neither absent nor perfect......36 0. indicating gro\o\ling capital mobility..2 A Solow model of a small open economy ..01 was 67 per cent ofGDP.... As examples of small open economies with free capital movements you may think of practically any of the countries in Western Europe....50 0.. 2....... of around 20 per cent... since even the larger economies like those of Britain.... Thus the 'truth' about capital mobility lies somewhere in between the two extreme cases studied in the previous and the current chapter........ 4. In Chapter 3 we considered the benchmark case of an economy with no crossborder capital mobility at all...... France or Germany are small compared to the size of the world economy..............41 0..35 0. measured by the (low) value of a 1.. In this section we develop a growth model of a socalled small open economy with perlect mobility of commodities and capital... Similarly..... Wealth accumulation © The McGrawHill and capital mobility Companies......... up from 23 per cent at the start of the 1990s...Basic Theory and Empirics about Prosperity and Growth 4............ this meant that about half of all gross investment in Portugal was being llnanced by foreign investors at the turn of the century. Capital is assumed to be perfectly mobile between the .........47 0.... in Greece the current account deficit in 2000...........2: The correlation between national saving and national investment............14 Source: Table 3 in Olivier Blanchard and Francesco Giavazzi: ' Current Account Deficits in the Euro Area: The End of the Feldstein·Horioka Puzzle?'... 2002... the 15 countries which were members of the EU in 2001...Jacobsen: Introducing Advanced Macroeconomics 102 I Part 1 ....... but \o\lith no labour mobility at all..5 1 0....... Again we see a clear tendency for a 1 to fall over time. The small open economy The open economy considered is so small that its own economic activity does not signillcantly afiect economic conditions in the rest of the world.. and those EU countries which had adopted the euro in 2001.....SorensenWhitta./Y.. With an investment rate I... This rellects the empirical fact that commodities and capital move much more across country borders than persons do...38 0..... 2005 PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH Tabel4. Brookings Papers of Economic Activity. pp. This was also reflected in the fact that a country like Portugal ran a current account dellcit as large as 10 per cent of its GDP in 2000... In the present chapter we consider the opposite benchmark case of a small open economy whose capital market is perfectly integrated with the world capital market...... 19752002 (estimates of a 1 ) Period Old OECD European Union Euro area 1975 2001 19751990 19912001 0. but looking at the two benchmark cases has the advantage of highlighting clearly what diflerence capital mobility can mal<e for the dynamics of economic growth.....55 0.. 14 7209.. capital mobility in the euro area seems to have become very high.... up from 12 per cent in the early 1990s.... In particular......01 .
With free trade in commodities this means that the prices of domestic goods must equal the prices determined in the world market (with the 1997 1998 1999 2000 2001 2002 2003 Year Figure 4. 4.3. Given that this arbitrage can occur instantaneously and costlessly. we assume that the goods produced in the small domestic economy are perfect substitutes for (some of the) goods produced in other countries. when exchange rat es are flexible the domestic nominal interest rat e may deviate from the world interest rate because exchange rate changes (currency gains or losses! are p art of t he total return on foreign bonds. For example. For concreteness. and subsequently against the euro. the real rate of interest on Danish longterm bonds has followed the Gennan interest rate quite closely in recent years. 3 Just as foreign and domestic assets are taken to be perfect substitutes.SorensenWhitta. bond returns (and hence ellective foreign and domestic interest yields) must be equalized at any point in time. Under perfect capital mobility domestic and foreign assets are perfect substitutes. However.Jacobsen: I Part 1. . 3. and investors can instantaneously and costlessly s\>Vitch between domestic and foreign assets. international Hnancial investors would immediately sell domestic bonds in order to buy foreign bonds. 2005 4 W EA LTH ACC UM ULA TION A ND C APITAL M O BILITY 103 domestic economy and the rest of the world. at least over the longer run (as will be explained in detail in Part 7). thereby driving dm>VIl domestic bond prices until the rates of return would be equalized. For the small open economy this means that the domestic interest rate is ellectively given by the world interest rate which the domestic economy cannot allect. Denmark is a good example of a small open economy in which the interest rate L~ more or less given from abroad. As we shall see in Part 7. Under perfect capital mobility all of these bonds must then pay the same rate of interest. w ith flexible exchange rat es there w ill still be a tendency for the domestic real interest rat e to follow the foreign real interest rat e. Wealth accumulation © The McGrawHill and capital mobility Companies. let us assume that foreign as well as domestic Hrms Hnance their investment in real capital by issuing bonds in the international capital market (alternatively we might assume that part of business investment is nuanced by issuing shares which are perfect substitutes for bonds in the eyes of Hnancial investors). A~ shown in Fig. Denmarl< has maintained fixed exchange rates vis·a·vis the German D·mark. EcoWin database. if domestic bonds h ad a lower return than foreign bonds.3: Real long·term interest rates in Denmark and G ermany Source: IMF international financial statistics database.Basic Theory and Introducing Advanced Empirics about Prosperity Macroeconomics and Growth 4.
0 < (/. will not play any important role in our analysis. < 1. (1).. denoted by ¥ 1. s. Since we are treating f as an exogenous constant. Setting up the model We are now ready to specify our longrun model of the small open economy.SorensenWhitta. since income from abroad can also be saved: O<s<l.Basic Theory and Empirics about Prosperity and Growth 4. As we have already seen. but now aggregate national saving is a constant fraction of aggregate national income rather than a fraction of domestic product. the intertemporal budget constraint may be written as: (8) stating the simple fact that current savings imply a corresponding addition to national wealth. For simplicity we will therefore set this exogenous relative price equal to unity. while net foreign assets are wealth invested abroad by domestic citizens minus the part of domestic capital owned by foreigners.e. we also assume a constant savings rate.Jacobsen: Introducing Advanced Macroeconomics 104 I Part 1 . with capital mobility it is also necessary to distinguish between GDP. (11) . Our intention here is to understand the implications of international capital mobility lor wealth accumulation and the potential gains from capital mobility. which countries end up producing which types of goods) and the welfare gains from such trade. 2005 PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH possible addition of some exogenous transport cost). National wealth. given our assumption of perfect capital mobility. at the beginning of period tis the sum of the capital stock and the country's stock of net foreign assets at that time: (7) The capital stock is wealth invested at home by domestic citizens or foreigners. denoted by v. and national income or GNP. (10) Following the basic Solow model. the ratio of the domestic to the foreign price level is exogenously fJXed. In accordance with the basic Solow model.•. aggregate output is given by the CobbDouglas production function: B > 0. This is equivalent to assuming that all economies in the world produce the same single good. Wealth accumulation © The McGrawHill and capital mobility Companies. V1. Since the rate of depreciation of the capital stock. then simplifies to: o. that is. From the viewpoint of the small domestic economy the prices of exported as well as imported goods and services are thus exogenously given from abroad. National income is defined as: (9) where f is the world real interest rate which is also the domestic real interest rate. With th is assumption our model cannot explain the international pattern of commodity trade (i. With this definition. The intertemporal budget constraint. In the open economy it is important to distinguish between the country's capital stock and its total national wealth. we \<Viii set it equal to zero. this variable does not carry any time subscript.
With Y1 and F. respec tively. is determined by (14).SorensenWhitta. V. determines national savings. K..' and S. the domestic capital stock adjusts to the level K.. The return to capital invested in the domestic economy would then be higher than the return oflered in the international capital market.. determined and with the predetermined L. with parameters a.. f .. The predetermined variables in period t are national wealth.... is not predetermined since it can adjust within period t by capital export or import. (7)(14). and K.a)B ( K )" (14) The complete model of the small open economy consists of the eight equations. and the marginal product oflabour.'./L. nand f.1. Suppose for a moment that the domestic capital stock were lower than this level. w. With K. L. in turn.+ 1 through U!J. = (l . B. Wealth accumulation © The McGrawHill and capital mobility Companies... The dynamics of the model are illustrated in Fig. National savings add to national wealth determining V. F1• Given L. w. through (10). Capital would therefore !low into the domestic country. endogenous variables that can adjust in the period in circles.. The capital stock. given by (13 ). s. Y.. S. K. Y. creating an increase in the capital stock until the marginal product of capital would just become equal to f . the net foreign asset position. and Y.'. national income. ensuring that the marginal product of capital equals the world real interest rate. L~ 105 gro\<ving at the constant rate n: n > .. and hence the real wage. and the labour force . given from past accumulation of savings and from past population grO\'ITth . K . the representative domestic firm employs capital and labour up to the point where marginal products are equal to the real factor prices. Since we abstract from depreciation. L. 2005 4 WEALTH ACC UMU LA TION AND CAPITAL MOBILITY and the labour force.4. while L. 4. determined. Hence we have: _ (K. implying a marginal product of capital above F.. is determined by (9). and with endogenous variables V. is associated with a lower value of net foreign assets.Basic Theory and Empirics about Prosperity and Growth 4.Jacobsen: Introducing Advanced Macroeconomics I Part 1 .. r = aBL (13) I L: .. F. K.labour ratio. follows from (7). In this way (13) determines the domestic capital. Given V. domestic output. ..+ 1 follo ws Figure 4. f. is determined by (11). Equation (7) shows that a larger capital stock. the real user cost of capital (the rental rate) is simply equal to the real interest rate. L.)"1. F" Y.4: The dynamics of the Solow model for the small open economy Note: Predetermined endogenous variables in squares. (12) Maximizing profits under perfect competition. . Y.
.... National income per capita Equation (13) can be written r = ccBk~'.. . National income per capita is y. We can then start over with period t + 1.. but note that the share of labour income in national income. From the production function..... is domestic product per capita.... 2005 PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH from L. andf.rk...... ..1 ... is not (necessarily) 1 .... The law of motion We described above how capital exports or imports ensure that the capital stock and capital per worker adjust within each period to ensure th at the marginal product of capital equals the international interest rate. From given initial values.....: 1• Thus GDP is hilly used to reward the labour and the capital employed in the country... = (1 7) Bk~.•= y.....a)y1• It follows that y.. and the share of labour income in GDP.k..•= y.'.. is national wealth per capita...... using the notation: y.. V0 and 1 0 • our model thus determines the dynamic evolution of all the endogenous variables........Jacobsen: Introducing Advanced Macroeconomics 106 I Part 1 ..... Wealth accumulation © The McGrawHill and capital mobility Companies..3 Analysing the model: dynamics and steady state ..1 • In other words.... 4........ w.... This reflects that the labour Ioree and the capital owned by the citizens In the country (not the capital employed Ir1 the country) are the country's sources of income. and hence rk 1 = ccBk~' = ccy1• Furthermore. k1 is domestic capital per capita... rkr/!J........1 . k..a ..... = y. = v.... These features are unchanged from the model lor the closed economy of Chapter 3. (11).... The constant (steady state) value for the capital intensity is thus: k* = Bl/(1"l Cl ) l/(1fl) ( f ....Basic Theory and Empirics about Prosperity and Growth 4..... L..... y...a..cc)Bk~ = (1 .. = Bk~'. = rk 1 + v.. = (1 ..a)Bk.. + ~ft . Using here that !Jr..•is national income per capita.. while the income share of domestic capital in GDP. r = ccBk~.. gives: (16) In the open economy......... Consequently..... As in Chapter 3 we will analyse our model in terms of per capita (or worker) variables. is 1 .= Y. + r(v...rk 1 + rv....kJ = y....... (15) Here. w 1 = (1 ...... national income (per capita) is the sum of wages (per capita) and the return to national wealth (per capita). domestic output per worker. is net foreign assets per capita... and the wage rate. wh ile y.... + rf....... is a ...SorensenWhitta.... wJy... by (12). y ..... wrfy 1... It follows from (7) thatf...... y. Likewise... and from (9) that y.. An exercise will ask you to analyse the income distribution further. while v... = 1v... adjusts immediately to the particular constant level k* given by r = aB(k*Y'....• ..... etc. from (14).
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 WEALTH ACC UMU LA TION AND CAPITAL MOBILITY
107
must jump immediately to the steady state levels:
a)'40"l,
y* = B1/(1"l .;_
(f
(18)
r(l
w* = (1  a)B 1/(1al ( ~
al,
(19)
and the capital output ratio must jump to:
(20)
The constancy of the domestic capital intensity, k,. and of GDP per capita. y 1, means that
the capital stock, K,, and the domestic product, Y,, both grow steadily over time at a
rate equal to the rate of population growth, and this occurs immediately without any
adjustment time required (for the closed economy similar conditions held in the steady
state, but not during the adjustment to steady state). In particular, K, + 1 = ( 1 + n)K,. so
I, = K, + 1  K 1 = nK1• It follows that the rate of accumulation of domestic capital. IjK,,
adJusts Immediately to the constant level 11, and, using (20), that the Investment rate,
I,/Y, = n(K,/Y,). adjusts immediately to:
(~J =
i* = 11;1 .
(21)
Thus, if crosscountry variations in the ratio S,/Y, are driven mainly by variations in s, our
Solow model of the small open economy suggests that the parameter a 1 in the
Feldstein Horioka equation (3) should be close to zero (since (I,/Y,) * is independent ofs) .
These properties contrast with the basic Solow model for the closed economy where
capital and output per worker as well as factor prices depend on the domestic savings rate,
s (in steady state and during the adjustment to steady state). and where capital and output
per worker, and hence the rate of capital accumulation, only gradually converge on their
steady state levels.
Instead of governing capital accumulation. the propensity to save now intluences
the accumulation of national wealth. To see this, start from Eq. (8), V<+ 1 = V, + S,, divide
through by L, + 1 = (1 + n)L,. and use (9) and (10) to obtain:
Now insert the expression in ( 16) for y;', and use w 1 = w* to get
 1)) ,
1 ( v,+ s(w * + rv
v,+ 1 = 1+n
from which we obtain:
v
<+ I
1 + sf
sw*
=  v 1 +  .
1 + II
1 + II
(22)
Equation (22) is the tran sition equation for the srnall open economy. We have chosen to
express the transition equation in terms of w*, which is constant as long as f remains
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
108
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPI RICS AB O UT PROSPERITY A ND GRO W TH
constant. but one can trace the transition equation back to parameters by substituting
(19) for w* in (22).
Equation (22) is quite intuitive. The sources of national wealth per capita in period
t + 1 are the national wealth carried over from period t. adjusted for population growth,
v /(1 + n). plus the savings per capita in period t, su;' = s( w* + ru,), adjusted for population
growth. s(w* + rv,)/(1 + 11).
Given an initial value. v 0 • of wealth per capita. the transition equation determines v"
which in tum determines v 2 , etc.. thus determining the full sequence (v,) . From (v,). the
sequences for all the endogenous variables may then be derived. For instance. the
sequence for national income. (u;'). follows from y;' = w 1 + ru,. (16) above, by using
w, = w* for all t. Then the sequence. (c,). for consumption per capita follows from
c1 = (1  s)y;'.
Convergence to steady state
The transition equation is illustrated in the transition diagram, Fig. 4 .5. The transition
curve is a straight line that intersects the vertical axis at sw*/(1 + n) > 0. and bas a slope
of (1 + sr)/(1 + 11). If this slope is smaller than one, as drawn. there is a unique positive
intersection with the 45° line at a value lor v, denoted by v*. From any predetermined
initial wealth, u0 > 0, v, v.rill converge towards v*. as indicated by the iterations illustrated
by the 'staircase' in Fig. 4 .5. The convergence point, v* , is the steady state value lor
national wealth per capita.
For the closed economy considered in Chapter 3 . the transition curve (describing the
accumulation of capital) was concave, reflecting diminishing returns to capital. In the
small open economy, independently of how much wealth has already been accumulated.
1 + sf
sw*
1 + n v, + 1 + n
sw*
1+n
45'
vo
v*
v,
Figure 4.5: Convergence to steady state in the Solow model of the small open economy
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 WEALTH ACC UMU LA TION AND CAPITAL MOBILITY
additional wealth always earns the international rate of return.
f.
109
Hence. although there
are dimini~hing returns to domestic capital. an increase Liv in current national wealth
always creates the same increase rtlv in current national income and the same increase
srtlv in current savings, thus creating the same addition (1 + sf).<iv/(1 + n) to national
wealth next period after adjusting for population growth. This is why the transition curve
is a straight line with slope (1 + sf)/(1 + n).
Figure 4.5 shows that the economy will converge on the steady state only if the
transition line is flatter than the 45° line. 4 that is, if (1 + sr)/(1 + n) < 1. This requires
fulfilment of the stability cm1dition:
sf<
11 .
(23)
In Western countries a longrun national savings rate around 0 .2 is fairly common, and
over the long run real interest rates in most western economies have tended to fluctuate
around a level of 3 4 per cent per annum, so a value of sf around 0.007 would not be
unrealistic. In a literal interpretation of our model, n is the rate of population growth
which is rarely as high as 0 .7 per cent per year in OECD countries (see Table A at the end
of Book One). This seems to suggest that the stability condition (23) may not be met in
praclice.
However, this is an artifact of disregarding technological progress. In a more realistic
setting with technological progress. we should measure the labour Ioree, Lt> in socalled
'el11ciency units', accounting lor the fact that the rising level of technology gradually
makes each unit oflabour more productive. Ellective labour input would then grow at a
rate larger than the growth rate of the population (labour Ioree) and the stability condition would be just like (23) \.vith n interpreted as the grmvth rate of ellective labour, which
is the sum of the rate of population growth and the rate of socalled 'labour augmenting'
technological progress. Chapter 5 will present this extension of the Solow model in detail
and explain, among other things, that the economy's longrun growth rate in GDP equals
the growth rate of effective labour supply. Since longnm GDP growth rates in Western
economies are considerably above 0 .7 per cent per year, the stability condition would
easily be met.
An appropriate way of evaluating (2 3) in the present context without technological
progress is this: assume lor a moment that the small economy considered were completely
closed to the outside world. It should then work according to the model of Chapter 3, converging to a steady state in which the key variables would be given by the formulae in
Section 3 of Chapter 3. In particular, the real interest rate established in the long run
would be (recall that (5 = 0):
(24)
where the subscript c refers to 'closed'. The world economy itself is closed, so in the long
run the international interest rate should be f = an/s. where nand s are the average international population growth rate and the international savings rate. respectively (and we
have assumed a common a) . If the small economy considered is structurally identical to
the world economy (n/s = fi/s), then r'}: = f, and (2 3) becomes equivalent to a< 1. which is
4. You may check this by doing a graphical analysis of the case where the transition line is steeper than the 45.1ine.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
110
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH
indeed one of our assumptions (realistically a is much smaller than one. around tJ, In
other words. a~ long as the small economy we consider is structurally not very dillerent
from the surrounding world economy, the stability condition, sf < 11, will be fulfilled.
Without apology, we will therefore assume that (2 3) is indeed satisfied, implying that
v 1 converges to the steady state value v*. Consequently, since y;' = w, + fv ,. national
income per capita must converge to y"* = w* + fv*, and consumption per capita. c ,.
converges to the steady state level. c* = (1  s)y"*.
The key endogenous variables in steady state
Setting v ,+ 1 = v , = v * in (22) and solving for v *, we llnd the steady state level of national
wealth per capita,
v * = s W*  's/_n_ IV *.
11 
sr
1  (s/H)f
(25)
which is positive under our stability condition. (s/ n)F < 1. The expression for w* found in
(19) can be inserted in place of thew* in (25) to trace v * back to parameters. Since w* is
independent of sand 11. we see that the larger s/11 L~ . the larger steady state national wealth
per capita will be, reflecting that more savings and less erosion of per capita wealth by
population growth both tend to create more wealth per person.
The net foreign asset position in steady state isf* = v* k*. From our solutions (17)
and (19) fork* and w*. it follows that k* = (a/r)w* /(1  a). Inserting this expression fork*
along with the expression in ( 2 5) for v* into .f* = v*  k*, we get:
1 r  (rn1js) *
.f* = 1 s IV .
1  a n f 1  (s/n)F
(26)
This equation reveals that under our stability condition. net foreign assets are positive in
steady state exactly if an/s < f, while.f* is negative if an/s > f . From (24), the condition
o.n/s < f translates into r'}: < f . Hence the small open economy will become a net creditor
if and only if the long run interest rate it would create if it were closed is lower than the
world interest rate. Th is is quite intuitive. If a country starts out from autarky. an opening
of the capital account should lead to a capital export. and hence to a positive net foreign
asset position, if the international capital market otTers a higher rate of return than the
domestic economy prior to the liberalization of capital flows . Conversely, if a country oilers
a rate of return above the international level before capital mobility is allowed, an opening
of the capital account should lead to a capital import. Consequently. a country with a
relatively strong propensity to save (s/11 > s/n and thus r~ < f), will end up as a net
creditor in the international economy, while a country with a relatively weak propensity
to save (s/n < s/IT) will become a net debtor.
The values for variables such as domestic capital per capita. GDP per capita, the wage
rate. etc.. that jump immediately to their steady state levels have been stated in (17)(21)
above. The steady state value for national income per capita is y "* = w* + rv*, which we
rewrite by inserting the expression in (2 5) lor v*:
.
.
s
.
.(
sf )
* n _ ,* ___(1...., w"' = w"' 1 + _ = IV   _  H
1 )~ ·
y "" = w"' + f n  sf
n  sr
n  sr
1 sn 1
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 W EA LTH ACC UM ULA TION A ND C APITAL M O BILITY
111
We can then trace all the way back to parameters by inserting our solution, (19), for w*:
y"* = (l  a)B L/ (1 a)
a)"/(1a)

(r
.
l
l  (s/n)r
•
(27)
Note that y"* is positive under our stability condition, (s/n)r < l . The terms in front of the
dot on the righ thand side of (2 7) reflect the influence from the wage rate, w*: the lower
the world interest rate, F, the more capital per worker will be built up in the domestic
economy, and the higher the wage rate 'vill be. The term after the dot in (2 7) includes the
influence of the income from wealth, rv*. Here a low value of P has a negative influence,
whereas a large value of the ratio s/n has a positive influence because it generates a high
level of wealth per capita, v*.
The sceady state value for consumption per capita may be found as c* = (l  s)y"*. In
this model there is no meaningful golden rule value lor the domestic savings rate. as you
will be asked to explain in an exercise.
Issues of structural policy
Among the model parameters that domestic policy authorities should be able to influence
ares and 11 . According to (2 7) a high value of s/n tends to create a high value of national
income per capita. Hence, with respect to our fundamental question, 'what can mal<e a
nation rich', our conclusions are qualitatively unchanged from Chapter 3. Attaining a
high national savings rate and getting population growth under control will have a positive impact on income per capita in open as well as closed economies. reflecting that a high
level of national wealth per capita 'viii always contribute to a high level of national
income.
Likewise, an improvement of technology (an increase in B) increases income per
capita in the open as well as in the closed economy.
However, there are some policy issues of particular relevance for the open economy.
The open economy exposes itself to one type of shock that a closed economy avoids,
namely interest rate shocks. Does this feature represent a serious risk? And will a country
benefit in the long run from an open capital account? The answers to such questions are
obviously important for decisions to liberalize capital flows .
4.4
Implications of an open capital account: advantages and
~r..Cl.~?~.~.~~ ..<?[J~l.<:>?..~.l~~Cl..t.i.C>.~................................................................................................ ..
Capital mobility, national income and income distribution
Will a country get richer if its government allows free capital mobility between the
domestic economy and the rest of the world? In the first decades after the Second World
War almost all governments in the capitalist world maintained strict capital controls. 5 As
5. By decoupling the domestic interest rate from foreign interest rates, capital controls enable a government t o pursue
an independent national monetary policy even if exchange rates are fixed.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
112
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH
we have seen, capital was later on allowed to move much more freely between countries.
Our question is: from the viewpoint of the individual small open economy, should we expect
the liberalization of capital movements to have contributed positively to national income
per capita in the long run?
We will analyse this important policy issue by comparing the longrun level of
national income under free capital mobility to the longrun income level attainable in a
policy regime of 'autarky' where the government does not allow international capital
llows. Our answer will be: yes, an open capital account does create a higher average
income in a small economy. but domestic wages will fall in consequence of an opening if
the domestic economy has a relatively high propensity to save. We will first give the
intuition for our results and then prove them formally.
If the small economy were closed, the real interest rate established in the long nm
would be r: = an/s, as stated in (24) above. Except for a knifeedge case. there are two possibilities. Either the interest rate that the small economy would experience if it were closed
is lower than the international interest rate, or it is higher. that is. either r~ < r orr: > r.
The first case materializes when the small economy has a relatively h igh savings rate (relative to population grovvth), while the latter case corresponds to a relatively low savings
rate.
Consider the case r~ < f, and assume that the small economy decides to open up and
allow free movements of capital and commodities. The domestic owners of capital will
then be able to earn more by investing their wealth in the international capital market
rather than placing it in domestic capital as they had to do before the opening of the
capital account. Hence, a country 'vith a relatively high propensity to save will benefit
from capital mobility by exporting capital, thus taking advantage of the higher international interest rate and earning more on its national wealth by becoming an international net creditor. After the international reallocation of portfolios has been completed
there 'vill no longer be a dillerence between the return on domestic capital and the international interest rate. As capital flows out of the domestic economy, domestic capital per
worker decreases. creating a higher domestic marginal product of capital. This does not
change the fact that domestic wealth owners benefit from the opening through a higher
rate of return on wealth. However. as capital per worker decreases, the marginal product
of labour goes down. Hence the capital account liberalization implies lower real wages in
the domestic economy. If the small economy has a high savings propensity an opening
should thus imply higher average income. but those who own no wealth and must live
from their labour alone '0\Till be hurt.
Next consider the case
> r. Now an opening of the capital account '0\Till benefit
the small economy via capital import, making it a net debtor. Since the rate of return on
domestic capital is higher than the international interest rate. domestic residents can make
money by borrmo\Ting in the international capital market and investing the borrowed funds
in domestic capital. This arbitrage is advantageous until the t:v.ro rates of return have been
equalized by the resulting increase in the domestic capital intensity which will imply higher
domestic wages. In this case capital account liberalization will beneiH domestic workers,
but domestic wealth owners '0\Till be hurt by the fall in the rate of return on their wealth.
In the knifeedge case,
= f . an opening of the capital account should have no
inlluence on (steady state) average income or wages in the home country. according to
our reasoning.
r;
r:
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 W EA LTH AC C UMU LAT ION AND C APITAL M O BILITY
113
To prove formally that an open capital account always implies a higher longrun
average income but lower (higher) wages if the economy has a high (low) savings propensity. we import expressions from Section 3 of Chapter 3 stating what national income per
capita and the wage rate would be if the small economy were closed:
y~ =
S )a/(1a)
B1/(1a) ;;
(28)
(
.
( S
)a/(1a)
w; = (1  a)BI/(Ia);
.
(29)
Define x as national income per capita under an open capital account relative to national
income per capita under a closed capital account, and use our expressions for y"* (in (2 7))
and lory~ to get:
x
( 1 (lrl)a/(la)
1
=y"*
.
= (1  a)  .
y;
f s
1  (s/n)r
From (24) we can replace an/s by r';. , and s/n by a/r;. , giving:
:\: =
1 )a/(1a) 1
(1  a) : ,
(r
1  ar
where
r =*r.. '
f
We establish three features of x as a function of f. First. iff = 1 (corresponding to f = r~ ),
we have :\' = 1, equivalent to y"* = y ~ . This confirms the intuitive insight that in case of
= r, an opening of the capital account should have no influence on income per capita in
the long run. Second, the derivative ofln x with respect to f,
r;
dlnx
1
r1
a
  =  + ,
dr
1
(1
r
1
ar
is zero fori' = 1. Third, the same derivative is strictly increasing in f, since we assume the
stability condition. (s/n)f < 1. now translating into rff < 1. 6 These three properties imply
th at x as a function off attains its min imum for f = 1, and the minimum value is one. In
other words, for any constellation of r~ and r other than
= r, y "* is larger than y~
(under our realistic stability condition). An open capital account thus implies a higher
average income in the long run, except in the knifeedge case, ~ = f.
Next consider the wage rate under an open capital account relative to the wage rate
under a closed capital account. From (19) and (29) you will easily lind that
r;.
W~ = (ai~S)'"/(1 a) =
w;
r
(r!
)a/(1a),
r
from which follows that w* < w~ exactly if r;. < f, which is the same as rm/s < f . The
latter f could be replaced by an/s. so the condition would read, n/s < n/s. confirming that
an opening of the capital account will lead to a fall (rise) in wages in a country with a
relatively high (low) savings rate.
6. Recall from (24) that r~ =an/s, so f= rfr: = srfan, implying ol= srfn. Hence the stability condition srjn < 1 is
equivalent to af < l.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
114
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH
We have shown that a liberalization of capital flows will imply an increase in
average income in the long run, but we have also seen that different types of income
will be affected in dillerent directions. If the world interest rate is lower than the initial
domestic interest rate. capital account liberalization will make capital flow into the
country. This will raise the domestic capitallabour ratio, inducing a rise in wages and a
fall in the rate of return on wealth. In contrast, if the initial domestic interest rate is lower
than the world interest rate. a liberalization of capital movements mal<es capital flow out
of the country. Domestic capitalists will then gain from a h igher return to wealth
whereas workers \>Viii lose from the drop in real wages resulting from the fall in domestic
capital intensity. If wealth is evenly distributed among the domestic residents, the drop in
wages will be more than ollset by the increase in income from wealth for each and every
person. However, people living from labour income alone will experience a drop in
income when a country with a high savings rate opens up its capital account. This may
be a perfectly 'rational' reason for workers in relatively rich (highsaving) countries to be
concerned about the implications of globalization. Note, however. that since average
income increases, it should be possible to design a policy of redistribution through taxes
and transfers such that everybody gains in all circumstances. An exercise \>Viii ask you
to analyse further the impact on income distribution of a liberalization of capital
movements.
Just as international trade theory shows that free trade in goods and services generally tends to increase economic efllciency by enabling countries to exploit their comparative advantages, the theory of economic growth thus suggests that free trade in capital
services (i.e.. free international mobility of capital) tends to increase average longrun
prosperity. This insight from our Solow model of the open economy provides one explanation why most economists take a favourable attitude towards capital mobility, while
realizing that redistributive policies may be needed to ensure that all groups in society
enjoy the fruits ofliberalization.
In a more realistic setting \>\lith risk and uncertainty, capital mobility may also yield
another type of welfare gain because it allows riskaverse savers to reduce the volatility
(riskiness) of the rate of return on their wealth by holding foreign as well as domestic
assets in their portfolios (assuming that the rates of return on foreign and domestic assets
are not perfectly correlated).
At the same time a country opening its capital account also exposes itself to a type of
impact from the outside world that it would avoid if it remained closed, n amely changes in
the international interest rate. socalled interest rate shocks.
A permanent interest rate shock
It is of interest to analyse the Jor1gnm economic consequences of a permarre11t increase (or
decrease) in the international interest rate, f . Such an increase could, for instance, be
caused by a permanent increase in government consumption in some of the larger countries in the world. The world economic system is closed and, as we explained in Chapter 3,
a h igher government propensity to consume implies a higher longrun interest rate in a
closed economy. Hence a permanent expansion of public consumption in the rest of the
world will allect the small domestic economy via a permanent increase in f . What are the
implications according to our model?
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 WEALTH ACC UM ULA TION AND CAPITAL MOBILITY
115
Our formulae (17). (18) and (19) reveal directly that steady state domestic capital
intensity. GDP per capita and the real wage all decrease in response to a pennanent
increase in f . Just after the interest rate shock (before any adjustments have taken place),
the marginal product of domestic capital, and hence the return earned on domestic
capital, will fall short of the new and higher international interest rate. Hence capital ilows
out of the country, implying a decrease in domestic capital per worker. The outflow ceases
when the decrease in k, has driven the marginal product of capital up to the level of the
new and higher f. This adjustment of k, takes place immediately and is ell'ected by capital
export. A capital export should imply that the country's net foreign asset position
improves. We will return to this issue below.
The fall in domestic capital intensity explains why domestic output per capita falls.
Furthermore, the lower value of k* implies a smaller marginal product of labour and
therefore a lower wage rate. The rate of return on domestic capital increases with f, but
note that capital income created in the domestic economy (rk*) actually decreases, as can
be seen by multiplying both sides of (17) by f .
How about national income? Consider the formula (2 7) for the longrun value of
income per capita. and remember that the term in front of the dot on the righthand side
is the steady state wage rate, w*. We see that this term decreases in response to the
increase in f. reflecting the fall in the marginal product of labour caused by the drop in
domestic capital intensity . The term to the right of the dot includes the return on national
wealth. rv*. and this term increases (under our stability requirement). There are thus two
opposing forces on y"*. To determine which one is the strongest, we can dillerentiate (the
log of) y"* with respect to f :
dIn y"*
(1
1
s/n
=+
.
df
1  cc r 1  (s/n)r
This expression includes a negative and a positive term, rellecting the two opposing forces.
Under the stability condition the condition for the derivative being negative is that
f < cm/s. This is exactly the condition lor the country being initially a net debtor. f < r';:.
On the other hand, the derivative is positive exactly if r > an/s, that is, if the country is
initially a net creditor. The conclusion is that in the long nm national income (and
consumption) per capita will fall in response to an increase in the world interest rate if the
country has a low propensity to save and therefore is a net debtor, while national income
per capita will increase if the country has a strong savings capacity and is therefore a net
creditor.
This is easy to understand intuitively. If a country has debt, its average income will be
negatively influenced if a higher rate of interest must be paid on the debt (and we would
expect the debt to decrease after all adjustments have taken place). On the other hand. if a
country has positive net foreign assets, the domestic wealth owners will benefit from a
higher rate of return on wealth (and we would expect the country to react by increasing
further its net foreign assets).
As already explained, domestic wage earners will experience a drop in their wage
income regardless of the country's net foreign asset position and regardless of their
individual asset positions: a person without any debt or wealth will lose income after an
increase in the international interest rate as a consequence of the domestic economy's
reactions to the interest rate shock.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
116
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH
How are the long run values for national wealth per capita, v*, and net foreign assets
per capita, f*, allected by the negative interest rate shock? In the derivation of (2 7) we
found th at
.
lV*
y""' =:
1  (s/n)f '
(last equation before (2 7)). so (2 5) may be written as
s
.
v * = y""
I!
This shows that v* is affected by f through the same terms as y"* , so in the long run
national wealth falls alter the increase in f if the economy has a relatively low propensity
to save, and vice versa. National wealth is allected in the same direction as national
income because wealth comes from income through savings.
Finally, we can use (24) and (25) to rewrite (26) as
J* = ( 1 ~a
)(f ~'~
)v*.
Thus .f* is influenced by f through the same terms as y"* and v* , nnd through the term
(r  r~ )/r. If the country is initially a net debtor (r < r~ ) . then this last term is negative,
and as r increa~es it becomes smaller in absolute value. This reveals that net foreign assets
per capita, .f*, altogether become smaller in absolute value. The domestic economy reacts
to the increased price of debt by reducing its debt. On the other hand, if the country is
initially a net creditor. an increase in f will make its positive net foreign assets even larger
(show this) .
In conclusion, a permanent interest rate shock where r increases has a negative
influence on the domestic economy's longrun national income and wealth if the country
is a net debtor and a positive infl uence if the country is a net creditor, while wages are
negatively allected in all circumstances. A permanent decrease in f has, of course, just the
opposite ellects. If a country wants to avoid the ups and downs associated '<\lith international interest rate shocks, this may be a reason for not wanting to liberalize capital
movements, alt hough such a liberalization should have a positive in fluence on national
income on the average.
Country risk
So far we have assumed that the 'hurdle' rate of return required and accepted by investors
for investing their wealth in the small economy is the same as the international interest
rate, f . However. in the real world investors will often require a risk premium on top of the
international interest rate lor investing in the assets of any particular small economy.
In August 1998 the American 'hedge fund', Long Term Capital Management, was in
trouble(thereason is not important here) and had to sell off assets. It turned out that it had
quite a lot ofDanish mortgage bonds in its portfolio. Although the hedge fund was a small
economic agent relative to the American or the European economy, it was relatively large
compared to the Danish bond market and therefore its selling off caused a drop in Danish
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
and capital mobility
© The McGrawHill
Companies. 2005
4 WEALTH ACC UM ULA TION AND CAPITAL MOBILITY
117
bond prices. This was bad news lor the other investors who had (planned) to sell Danish
bonds at that time. The wave could, of course, have gone the other way with investors
being fortunate to sell Danish bonds at a time where largescale buying by a big international investor had driven prices upwards. However, usually investors are risk averse,
preferring a sale return of one million euros to a 11ftyfifty chance of earning either tv.ro
million euros or zero euros. Since the return on Danish assets can fluctuate a little depending on the particular circumstances oflarge international investors. investors in Danish
bonds will require an expected return (slightly) in excess of the interest rate on the safest
bonds traded in the international capital market. The 'thinness' of Danish asset markets
may thus imply a risk premium in the Danish interest rate. Figure 4 .3 indicates that the
Danish real interest rate has indeed been a bit higher than the euro interest rate in recent
years.
There can be other (more severe) reasons lor risk premia in interest rates. For
instance, if in vestment in a country is not considered completely sale because of risks of,
for example, civil war. coup d'etats, nationalizations, or widespread bankruptcies,
investors will typically require large risk premia lor investing in the country.
It may also happen that international investors suddenly lose conlldence in a
country. A dramatic example of this was provided by the SouthEast Asian economic crisis
in 199798 when foreign capital flows into the 'tiger' economies ofindonesia, Malaysia,
South Korea and Thailand suddenly stopped, forcing these countries into drastic cuts in
domestic consumption and investment as they could no longer finance their large current
account dellcits. 7 After such a loss of conlldence investors sharply increase the risk
premium they require in order to be willing to lend to the country.
Denoting, as before, the average international real rate of interest by P, the hurdle real
interest rate that investors require and accept lor investing in a particular small economy
may be f + E:, where E: is a risk premium compensating for the risk associated with investing in the country's assets. In principle. E: could be negative if the country is considered
very 'safe', but we will think of E: as positive. 8
The risk premium. £, can be influenced by certain types of domestic policy. For
instance, if Denmark were to join the European Monetary Union, abandoning the Danish
krone, Danish bonds denominated in euros would be very close substitutes lor other euro
bonds, and the market lor Danish bonds would therefore no longer be thin. Hence, the risk
we described above would be reduced and the (already small) risk premium in the Danish
real interest rate should therefore also be reduced. Likewise, polices that promote the
general safety of a country's assets. lor instance an improvement of the financial system or
of the system of corporate governance that reduces the risk of bankruptcies or fraud
should reduce the risk premium as well. Of course, a certain (bad) policy could imply an
increase in the risk premium.
In our Solow model of the small open economy. (7)(14). the interest rate, f, appears
in two places. One L~ in the arbitrage condition, (13). where f is obviously a hurdle interest
7. In his highly readable book on The Return of Depression Economics [Allen Lane, The Penguin Press, 1999), Paul
Krugman gives a breatht aking account of the Asian crisis as well as other recent international financial crises.
8. Just adding >: to r in the definition of the hurdle interest rat e is an ad hoc way of dealing w ith country risks since it
does not model uncertainty explicitly. Under uncertainty investors expect different rat es of return from investment in
the country with different probabilities. Since they are risk averse, they will require an average rate of return above the
safe r . We try to capture this by our >:, but without considering probability distributions explicitly.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
118
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPI RICS AB O UT PRO SPERITY A ND GRO W TH
rate and should be replaced by f + t to take country risk into account explicitly. The other
place is in the definition, (9), of national income. Should the f appearing here also be
replaced by f + e?
This depends on the country's net foreign asset position. If it is a net debtor, F1 < 0, the
foreign investors who have financed part of its capital stock will require a risk premium on
their investment, so the interest rate that must be paid on F1 is indeed r +e. Hence, if the
country is a net debtor, our model. (7)(14), is amended to take country risk into account
by writing f + t instead of r wherever f appears.
If the country is a net creditor, it '.viii earn only the international interest rate. f
(with out any risk premium), on its positive net foreign assets. In this case the appropriate
model '>Viii therefore consist of (7)(14). with the r in (9) kept unchanged, and '>\lith the r
in (13) replaced by f +e.
We will consider a net debtor country that is initially in steady state. Our model is then
simply ( 7)( 14), '.vith f replaced by r +e. Mathematically the model is thus unchanged,
'>\lith r + e now taking the place of f . We are interested in the longrun consequences of a
permanent increase in the risk premium, e. This situation could. for instance, be representative for a country that has developed by borrowing. thereby becoming a net debtor,
and where some event suddenly causes a longlasting lowering of investor expectations
regarding the country's filture longterm performance.
To analyse this situation we do not have to perform any new formal analysis. For a
net debtor our model is unchanged with respect to mathematical structure, so in all our
formulae for steady state values, transition curve, etc., we should just replacer by f + t . In
the model with country risk, an increase in twill have the same ellects as an increase in f
has in the old model, and for the same reasons. We have already analysed this and found
that lor a net debtor, wages, national income and national wealth would fall in the long
run. Thus this will also be the longrun reaction of the small open economy to an increase
in the country risk premium, t .
However, this time we cannot conclude that if the country is initially a net creditor,
national income will increase in response to an increase in e. Recall that the model
including country risk is different for a net debtor and a net creditor. An exercise '>Viii ask
you to analyse the case of a net creditor. The analysis will reveal that national income
and wealth are negatively atl'ected by an increase in the country risk premium in all
circumstances.
The possibility oflonglasting interest rate shocks, caused by changes in international
interest rates or by changes in countryspecific risk premia, thus implies some risks for
small open economies. However, the main risks of free capital movements lie in the
economy's shortrun reactions. and we should emphasize that our open economy Solow
model cannot give a realistic description of the shortrun ellects of an interest rate shock.
For example, although our model implies that workers must tal<e a wage cut when the
country is hit by an increase in r or t, at least they keep their jobs. But in practice wages
are not very flexible in the short run, so a higher interest rate will throw some people out
of work for a while as firms reduce their investment.
As another example, in our model iirms react to the higher interest rate by reducing
their capital stock and output. but they all stay in business. However. in the real world a
serious interest rate shock will cause many llrms to go bankrupt. and investment '>Vill lall
more dramatically in the short nm than in the long run.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1 
Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 WEALTH ACC UMULATION AND CAPITAL MOBILITY
119
In summary, because our Solow model with fully flexible prices does not allow output
and employment to deviate from their 'natural' rates. it does not capture aU the drama of
the recession that will occur in the short run when a country runs into an international
llnancial crisis. In Part 7 of this text we shall develop business cycle models which can also
illuminate the shortrun fluctuations generated by such a shock.
The costs and benefits of capital mobility
Much of the ongoing international debate on globalization centres on the question
whether free capital mobility is desirable or not. The analysis in this chapter suggests that
capital mobility brings longrun benellts as well as potential costs. An open capital
account enables a lowsaving country with high investment needs to borrow more
cheaply in the world capital market. It also allows wealth owners in highsaving countries
to tal<e advantage of more prontable investment opportunities abroad. In these ways
capital mobility tends to raise a country"s longnm national income and wealth. At the
same time capital mobility also makes a country vulnerable to shills in the moods of international investors. If investors suddenly lose confidence in a debtor country. the resulting
increase in the country's borrowing rate of interest will cause a recession and sometimes
even a depression, as illustrated by the Asian crisis in 199 7 98. Recent economic history
suggests that developing countries and emerging market economies tend to experience
greater shifts in international borrovving conditions than the most developed Western
market economies. This may explain why many developing countries still maintain some
capital controls whereas the richest countries in the world have liberalized international
capital flows .
4.5
§.~~.~~¥....................................................................................................................................................
1. In an open economy which allows free international mobility of capital, part of domestic investment may be financed by capital imports from abroad, and part of domestic saving may be
invested in foreign assets. It is then necessary to distinguish between the domestic capital
stock {the stock of capital invested in the domestic economy) and national wealth {the total
stock of assets owned by domestic residents). A country's stock of net foreign assets is equal
to the difference between its national wealth and the domestic capital stock.
2. In parallel to the distinction between the domestic capital stock and national wealth, one must
distinguisr between gross domestic product {GDP) which is the total output produced by the
factors of production employed in the domestic economy, and national income which is the
total income earned by the factors of production owned by domestic residents. National
income is equal to GDP plus the return to the country's net foreign assets.
3. The growth rate of a country's capital stock is given by its rate of domestic investment,
whereas the growth rate of national wealth is determined by the national savings rate. In a
closed economy saving and investment are identical and therefore perfectly correlated. In the
open ecoromy capital mobility reduces the correlation between saving and investment. The
current account on the balance of payments represents the difference between national
SorensenWhitta Jacobsen:
Introducing Advanced
acroeconomics
120
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPIRICS ABOUT PROSPERITY AND GROWTH
saving and national investment. As an accounting identity, the increase in a country's net
foreign assets is given by the balance on its current account.
4. International capital mobility may take the form of international portfolio investment, where
households or institutional investors (including banks) acquire foreign assets to diversify their
portfolios, or it may take the form of foreign direct investment where the investor acquires a
controlling ownership share of foreign business assets. In recent decades both forms of
international investment have grown much more rapidly than world output and world trade, as
many governments have liberalized capital flows, and as improvements in information and
communication technologies have reduced the barriers to foreign investment. The growing
mobility of cap tal has caused a fall in the correlation between national rates of saving and
investment. In the euro area this correlation is now close to zero.
5. Capital is said to be perfectly mobile when domestic and fo reign assets are perfect substitutes and investors can instantaneously and costlessly switch between the two asset types.
In a small open economy which cannot affect the world interest rate or any other foreign
macroeconomic variable, perfect capital mobility implies that the correlation between national
saving and national investment tends to zero.
6. In our Solow model of a small open economy with perfect capital mobility, the domestic
capital intensity adjusts instantaneously to its steady state level determined by the international interest rate. By contrast, national wealth per capita evolves gradually over time, since
any increase in wealth generates an increase in capital income, thereby raising aggregate
saving which in turn generates another increase in wealth, and so on.
7. The process of wealth accumulation in the small open economy is dynamically stable if and
only if the (effective) labour force grows at a rate exceeding the product of the savings rate
and the international real interest rate. When the labour force is measured in technologyadjusted efficiency units, this stability condition will be met for all realistic parameter values.
8. By allowing free international capital flows, a country will be able to increase its national
income whenever the domestic steady state real interest rate in the absence of capital mobility,
r;, deviates from the world real interest rate, r. If r; > r, which will typically be the case in
a country with a relatively low savings rate, capital mobility benefits the domestic economy by
enabling firms to borrow at a world interest rate below the (initial) marginal return to domestic
investment. If
r:
<
r, which will tend to be the case in a country with a relatively high savings
rate, domestic wealth owners gain because capital mobility allows them to invest their savings
at a higher (initial) rate of return abroad than at home. If wealth is unevenly distributed, capital
mobility will affect the distribution of income and may in fact hurt some domestic agents even
though it raises aggregate income.
9. While capital mobility enables domestic agents to take advantage of rateofreturn differentials between the domestic and the foreign economy, it also makes countries more vulnerable to international interest rate shocks. In particular, a permanent increase in the real
interest rate at which a net debtor country can borrow abroad will reduce the country's steady
state level of national income and wealth. Moreover, if such an increase in the country risk
premium occurs abruptly, as is often the case, it may create a serious domestic recession in
the short run.
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 1 
Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
4 WEALTH ACC UMULATION AND CAPITAL MOBILITY
121
Exercises
Exercise 1. The elasticity of longrun national income per capita with respect
to the savings rate in the small open economy
This exercise investigates how strongly national income per capita is influenced by the savings
rate in the open economy compared to the closed. First show from (27) that the elasticity of
y"* with respect to s is:
(s/ n)r
e = ___:....:..._=:_
o
1 (s/ n)r '
and where the stability condition, (s/n)r < 1, should be assumed. In the closed economy
studied in Chapter 3, the elastic ity of national income (equal to GOP) per capita w ith respect
to the savings rate, s, was found to be
ec = a./(1 a.). It seems d ifficult to compare these two
elasticities since they involve different parameters. We can, however, apply a trick we used in
th is chapter as well. The real interest rate that would emerge in the long run in the small open
economy if it were closed (given
o= 0) as stated in (24) was: r~ = a.n/s.
Show that the elastic ity, e 0 , can be written o.s:
r
where
'"' 
r*c
,
and where the stability condition is now a.f< 1. Show that if the longrun real interest rate that
the open economy would produce if it were closed is eq•Jal to the international real interest
rate, then
e 0 = ec. Then show that if the considered small economy has a stronger propensity
r, then e 0 > e 0 and vice versa.
to save (as measured by s/ n) than corresponding tor~ =
Try to explain your result intuitively. (Hint: If the economy considered is 'savings strong',
so that r~ < r, then, when it has adjusted fully to the cond itions of free cap ital movements it
w ill have taken over a rate of return on wealth that is higher than the one it would produce itself
if it were closed.)
Exercise 2. Golden rule in the open economy?
Show that in the small open economy's steady state consumption per capita, is:
c* = (1 s)
1
1  (s/n)r
where the expression for
w*
'
w* of ( 19) cou ld be inserted
(but, since
w* does not depend on s,
th is is not of importance for the present exerc ise) . Show that as long as the stability cond ition,
(s/n)r < 1, is fu lfilled, an increase ins has two opposing effects on c* , and explain the origins
of these.
G iven the two opposing effects it is natural to ask if there is a specific (golden rule) value
of s that rraximizes c* ? Show that this is not the case by demonstrating that as s increases
up towards njr (which is realistically assumed to be smaller than one), c * goes to infinity.
Try to explain intuitively what goes on here. Explain first, for instance, what happens to
steady state national income when
s
increases towards
n/r?
To understand th is, in turn,
explain what happens to the transition curve as s increases towards njr, and explain why the
transition curve rotates as it does.
SorensenWhitta Jacobsen:
Introducing Advanced
acroeconomics
122
I Part 1  Basic Theory and
Empirics about Prosperity
and Growth
4. Wealth accumulation
© The McGrawHill
and capital mobility
Companies. 2005
PART 1: BASIC T HEORY AND EMPIRICS ABOUT PROSPERITY AND GROWTH
Exercise 3. The income distribution in the small open economy
w,/y, is equal to
rk,/y,, is a.. These
As stated in this chapter, in any period t, the share of labour income in GOP,
V'1 a., while the share of income earned on domestic capital in GOP,
features are unchanged from the model for the closed economy of Chapter 3 . However, from
the point of view of the small open economy the income shares of the resources owned by
domestic residents in national income (rather than GOP) are of more interest. That is, the
income distribution is more relevantly described by the share of labour income in national
income,
w,/y7, and the share of income earned on national wealth in national income, rv,/y7.
1. Show, using (16) and the fact that w, jumps immediately tow*, that labour's share in any
period tis:
w,

Y7
u,
v,
Y~
w* + rv,
= 1r = 1r  
and show that as national wealth,
'
v,, converges towards its steady state level, v*, from below
(above), labour's share will converge to its steady state level from above (below).
2. Show that in steady state the shares in national income of labour and wealth, respectively,
are:
w*
s
=1  r
y n*
n
and
rv* s

y n*
=  7.
n
(As usual we assume the stability condition, (sj n)r < 1). Show that in an open economy, a
relatively strong propensity to save (as measured by s/ n) will ten d to create a relatively low
labour share in the long run. W ill it also imp ly a lower longrun wage rate, w*? Explain your
results intuitively. Should wage earners in open economies try to resist national saving?
3. To compare the functional income distribution just found to the income distribution in the
closed economy, one can (once again) use the trick of rewriting some expressions in terms of
the interest rate that the small economy
r~
= a.n/s.
would have created ir the long run if it were closed,
Show that the steady state functional income distribution in the open economy is
given by:
and
r
where ; ,  ,
r*c
where the stability condition, af < 1, is assumed. Show that if an economy has a relatively high
propensity to save by international standards, an opening of the economy w ill imply a lower
labour share in the long run. Explain why. W ill an opening also imply a lower longrun wage,
w*? Do wage earners in savings strong economies have reasons to try to res ist a
liberalization of capital movements? Should they rather direct their political influence towards
other aims?
4. How is the longrun functional income distribution in the open economy affected by a permanent international interest rate shock, where
level? Explain your result.
r
increases permanently to a new and higher
Jacobsen: Introducing Advanced Macroeconomics I Part 1. that is. The importance of capital movements for convergence and a permanent productivity shock in the open vs. 2.Basic Theory and Empirics about Prosperity and Growth 4. by comparing to the relevant formula in Chapter 3. s'. We assume throughout that indeed r~ = anfs = r. 1. Focus first on the first period with s'.) and show that under our assumptions its slope is: 1 + (J. the wage rate. r~ = an/s. and that o= 0. Using your transition diagram. shifts permanently upwards to a new and higher level.n 1+n . then it does not matter if the economy is closed or open. s. Assume that the economy was in steady state before the shift in the savings rate. Explain the economic background for each of these two movements. in case it were closed. . 1. r. as this exercise illustrates. Show further that steady state national wealth per capita. 2. describe periodbyperiod the process of wealth accumulation. The shift is sufficiently small to ensure that the stability condition is fu lfilled after the change. the closed economy In this chapter it was shown that if the small open economy. How do GOP per capita and the wage rate evolve over those periods? Describe how national wealth per capita evolves. national income per capita. One may get the impression that if anfs = r. then its steady state national income per capita would be the same as if it were closed. The effects of a permanent increase in the savings rate Consider the Solow model of the small open economy and assume that as from some period the savings rate. What happens outside steady state depends very much on the economy being closed or open. for the open economy is equal to the steady state capital intensity. for the closed economy. Compare the old to the new steady state.. This is absolutely false. consumption per capita and savings per capita affected in that period? Then consider the subsequent sequence of periods. Wealth accumulation © The McGrawHill and capital mobility Companies. Show from (27) that under these assumptions steady state national income per capita is: Then show. Explain that this gives rise to a change in the transition curve that can be described as a combination of a parallel shift upwards and an anticlockwise rotation around the intersection point with the v t + 1 axis. How does national income and consumption per capita evolve? 3. for the closed economy. We will consider one and the same small economy as open and as closed. y~.SorensenWhitta. as a function of v. k~ . explain qualitatively the economy's reactions over time to the permanent increase in the savings rate. It is only the steady state that is unaffected. that this y"* is equal to the steady state GOP per capita. equal to the international interest rate. Consider the transition equation for the open economy (v. How (in what direction) are GOP per capita and national income per capita changed? Can we be sure in what direction consumption per capita has changed? How do the country's net foreign assets change from the old to the new steady state? Exercise 5. How are GOP per capita. 2005 4 WEALTH ACC UMULATION AND CAPITAL MOBILITY 123 Exercise 4. and the country's position as net debtor or creditor in steady state (the sign off*) does not change. were to create a longrun interest rate. u*.
v 0 = k 0 . y*. (y~) and (y. note that the variable indicated along the horizontal axis in Fig. start at a common initial value. Describe in princip le how the permanent productivity shock considered affects the transition diagrams and the accumulation of wealth and capital. as from period one. and in Fig. Compute the new steady state values for national income per capita and for wealth per capita for both economies. below steady state.SorensenWhitta.7 was investment shares of GOP.8 we found a negative correlation between average population growth rates and GOP per capita. Wealth accumulation © The McGrawHill and capital mobility Companies. 4. 3.02.+1 as a function of k.n)/(1 + n) as well. under these specifications. respectively.) its slope is (1 + a. n: 1 (J. Exercise 6.. fo r national income per capita. n = 0. We will now assume that (initially) 8 = 1. but be sure to have quite a few). that for a small open economy in steady state there is the following necessary link between GOP per capita. but in an open economy they do not. y*. 3. i*. so we assume that the annual international real interest rate is 2 per cent.. Illustrate the output of your simulations by plotting each of y~ and y 1 against periods in one and the same diagram that also has the old and the new steady states indicated by horizontal lines. 3. 7 and 3. which is just a normalization. the total factor productivity increases permanently up to 8' = 2. Does the empirical evidence indicate that capital mobility is not important? In this chapter we found that for a (small) economy with free capital movements.). y n* = r: and u* = k.Inn). This implies that r~ =an/ s = 0.Basic Theory and Empirics about Prosperity and Growth 4. From these sequences compute the associated sequences.) and show that at steady state (fork. First. respectively. and the growth rate of the labour force. see (18). Show.In 8 + . the closed or the open. etc.7 we found a relatively clear positive correlation between average investment shares and GOP per capita across countries. lny* = . thereby creating a sequence (v 1 ) for the open economy and a sequence (k1 ) for the closed. Assume that v 1 and k. is independent of sand n. illustrate (in princip le) the transition curves by drawing both in the same diagram (so you have both v 1 and k 1 along the horizontal axis. not savings shares.8 that real world economies behave more as closed than as open economies and that capital movements therefore are not so important for observed economic performance. Assume that both economies. In Fig.2.). longrun GOP (not national income) per capita. On the basis of your findings in questions 1 and 2. for the open and the closed economy.= k. s = 0. 1 a 1 a . respectively. do you think? Explain intuitively why the fastest one is the fastest. Explain in words differences and similarities between the series (y7) and (y. say.). the investment rate. using (18) and (21 ). and for the other parameters (except (on annual basis): othat has already been set equal to zero) assume some plausible values a = 0. start in their steady states and then.. The question we address in this exercise is whether we should conclude from the empirical evidence presented in Figs 3. Now simulate each of the transition equations from period zero and onwards (for a number of periods of your own choice. 3. Compute the steady state values.(ln i* . In a closed economy these coincide.01. For which economy will convergence towards steady state be fastest. in the open and the closed economy. the open and the closed.Jacobsen: Introducing Advanced acroeconomics 124 I Part 1. 2005 PART 1: BASIC T HEORY AND EMPIRICS ABOUT PROSPERITY AND GROWTH Then consider the transition equation for the closed econorTy that we found in Chapter 3 (k.4.
The appropriate model is therefore (7) . 9. The meaning of the economy considered being small is exactly that. 1. We will consider the most standard case of e > 0. and we denote the average savings rate in the world by s. Note that l n is larger than n.. Otherwise notation is as in the chapter. Exercise 8. As argued in this chapter. W rite down the complete model (for your convenience) and show that in any period t. and c5 = 0 in both). and at some point our assumption that the domesti c economy is too small to affect the w orld economy w ill no longer be valid. O n the other hand the 'hurdle' rate of return on d omestic capital that will make investors consider wealth placed in the small open economy and in the nternational capital market equally good should include a risk premium. with the r in (g) kept unchanged. and discuss whether the empirical evidence reported in Figs 3. and with the r in (13) replaced by r +e. . If we assume that technology is the same in the domestic country and in the world (a is the same. Exercise 7.9 Show. This means that the formulae in Section 3 of Chapter 3 should be valid for the steady state of the world economy. Section 3. national income per capita is the same in the open economy as it would be if the economy were closed. Wealth accumulation © The McGrawHill and capital mobility Companies. Longrun national income in the small open economy as depending on domestic and international savings propensities A certain restatement of (27) can be of interest. then the steady state value for the an s 7=. We assume that this happens too far into the future to be of practical relevance. GOP per capita is: y. then the domest ic economy w ill grow faster than the world economy. Country risk and a risk premium shock if the small open economy is a net creditor We argued in the chapter that to take country risk into account for a net creditor one should assume that the country earns the international interest rate without a risk premium on its net fo reiQn assets since these are p laced abroad.SorensenWhitta. using (27). that steady state national income per capita is: Show that in the particular case. What if s/n > sfn? Explain. so it should behave according to the model of Chapter 3 in the aggregate.7 and 3. and for simplicity assume 8 = 1. the world economy is a closed system.Jacobsen: I Part 1. although it contributes to world savings and population growth.Basic Theory and Introducing Advanced Empirics about Prosperity Macroeconomics and Growth 4. and the world population growth rate by world real interest rate is: n. its impacts on s and n are so small that in p ractice r is given.(14). sf n = s/n.8 indicates that capital movements are not too important for economic performance across countries. 2005 4 WEALTH ACC UMULATION AND CAPITAL MOBILITY 125 Compare this to the expression for In y* we found in Chapter 3. = w1 +(l'+ E)k1 .
Show that domestic capital per capita and the wage rate. [ r+r 1 a r+r 5.a) k =f'+E )'l/(1 a w* = ( 1 a) _( f+ E and ' .( 1 a) . You can then use state national income per capita.Basic Theory and Empirics about Prosperity and Growth 4.Sorensen. Show that the transition equation for national wealth per capita is: v 1 + sr /+ 1 s(w* + r k*) 1+n = .. Using the above expressions fo r k* and w* . v*. the wage rate.. Wealth accumulation © The McGrawHill and capital mobility Companies..) Explain your results intuitively. wealth per capita. + ek1 + rv. 3. d iscuss arguments for and against liberalizing capital flows. is affected. adjust immediately to: * ( a ')1/(1...Whitta. Is you r overall conclusion based on the theory (and empirics) of th is chapter n favou r of or against such liberalization ? . Show that the steady state values for GOP per capita.. Show that an increase in 1: imp lies a decrease in w* + r k* . How is the transition equation affected by an increase in r ? 6. (Hint: You don't have to solve the model for y"*. Exercise 9. etc. is affected. to infer how steady y"*. whenever 1: > 0 initially. 2005 PART 1: BASIC T HEO RY AND EMPI RI CS ABOUT PROSPERITY AND GROWTH while national income per capita is: 2.a) .v 1 + '' 1+n 4. wealth per capita and national income per capita all decrease in response to the increase in 1:. show that: 1: a ] ( a )a/(1 .Jacobsen: Introducing Advanced acroeconomics 126 I Part 1 ._ .a) w* +rk* = 1 + . respectively. From the shift of the transition curve you can infer how steady state y7= w. Free capital movements: pros and cons On the background of what you have learned in this chapter.
2005 xogenous rowth • • • • •• • • • • • • • • 5 Technological progress and growth: the general Solow model 6 Education and growth: the Solow model with human .Exogenous Growth I 5.t.. Technological progress I © The McGrawHill and growth: The general Companies.eI SorensenWhittaJacobsen: Introducing Advanced I Part 2.~P.~~·················································································· ······························································ 7 Limits to growth? The Solow model with scarce natural resources 127 .~.i.
Technological progress and growth: The general Solow model © The McGrawHill Companies.Sorensen.Exogenous Growth 5.Whitta. 2005 .Jacobsen: Introducing Advanced Macroeconomics Part 2.
The tests will I 1. B. Instead it will be given as an exogenous sequence. we are going to take the model through some more specific empirical tests. However. For your convenience. This chapter presents a growth model with the properties we are looking for: the steady state of the model exhibits balanced growth with positive growth in output per worker.eI Sorensen. Solow. we give the reference to this pathbreaking article once again: Robert M. pp. Hence. 70. Solow in his famous 19 56 article referred to in Chapter 3. We want to develop a growth model whose longterm prediction is a balanced growth path with strictly positive growth in GDP per person.Exogenous 5.Whitta. and it is reassuring for the application of the model to issues of economic policy that its steady state mirrors a robust longrun growth fact. 1 The essential new feature of the model is that total factor productivity is no longer assumed to be a constant. 1956. it is not trivial that the consequence of steadily arriving tech nological progress should be a balanced growth path. This is at odds with the observed longrun growth in living standards in Western economies. explaini11g the technological progress th at creates growth in GDP per worker is a matter of great interest and it is the subject of Part 3 of this text. Technological progress © The McGrawHill Growth and growth: The general Companies. 'what creates prosperity in the long run?' and 'what creates transitory and longrun growth?' . In that case the model's steady state will display balanced growth with steady positive growth in GDP per worker. 129 . The resulting model is close to the one actually suggested by Robert M. Quarterly Journal of Economics. Of course. 'A Contribution to the Theory of Economic G rowth'. 6594. This explanation of growth may not seem deep. which may be steadily growing over time. Having obtained in this chapter a model that llts the basic stylized growth facts better. according to the general Solow model (as we will call it) tile root of steady positive longrun growth in GDP per person is a steady exogenous technological progress. 2005 Macroeconomics Solow model Chapter Technological progress and growth The general Solow model n one respect the basic Solow model presented in Chapter 3 did not perform well empirically: its longrun balanced growth path displayed zero growth in GDP per capita. (B 1). We obtain this by a slight generalization of the basic Solow model. if these answers are grounded in a (growth) model that accords with the most basic empirical facts.Jacobsen: Introducing Advanced I Part 2. Obviously we can only trust the answers to our big questions.
.. (At).... is called labouraugmenting or Harrodneutral.Jacobsen: Introducing Advanced I Part 2.... (1) This time we insert from the beginning the feature that the inputs demanded and actually used. and their behaviour is essentially the same.... (B. If it appears as an 2... Since one particular assumption on (B. where B..h:1ngP.... 5....Sorensen.. Technological progress © The McGrawHill and growth: The general Companies.. With more general production functions it may make some difl'erence... We could have done some of these more specific tests already in connection with the basic Solow model of Chapter 3... and the real wage rate.\ in each period t...... :1ncl morP.. \1\'c may alternatively write the production function as: = (2) where A 1 B:/(J..... and the markets are again assumed to be perfectly competitive... The full time sequence........ In particular.. the production function in period t is: 0<(1<1...... output can be obtained from the same amounts of inputs...... . of the total factor productivity. With respect to the qualitative features of the underlying 'micro world'..mel Lf.. a representative profitmaximizing firm has to decide on the inputs of capital and labour services..cl from K:1. for inst:Jnr. K:1 and L..... It has the same commodities and markets...... or just the Solow model... in fact... There are also the same kinds of economic agents. technological progress that appears as an increasing variable. > 0 in all time periods t.... The general Solow model is indeed a very important growth model.. of the labour productivity variable (not to be confused with the average labour productivity. is constant.. morP.. or by an appropriately defined sequence.1 The general Solow model .. and L.. w 1• The only difl'erence is th at the production function .... the general Solow model... At....... although there will be some aspects still to be improved upon.....1and Li. Generally..... is a generalization of the basic Solow model...Exogenous Growth Macroeconomics 130 5. Exercise 9 of Chapter 3 asked you to do one of them yourself. The notation for it will therefore be B.....r timP...P. (B. Maintaining a CobbDouglas form.L1)....Whitta....... proclnr. but we think it is better to undertake them now that the model is in accordance w ith the basic fact of steady and positive long·run g rowth.... given the real rental rate of capital........ which tells how much output can be r..... With the CobbDouglas form of the production function it mal<es no difference whether we describe technological ch ange by a certain time sequence.....al.....so th :1t. the general Solow model is identical to the basic Solow model.. 2005 Solow model PART 2 : EXOGENO US G RO W TH focus on the steady state prediction of the model as well as its outside steady state prediction concerning transitory growth or convergence...)...2 Our conclusion will be that the general Solow model does quite well empirically.......P. K...... both with respect to its steady state and its transitory growth predictions. in a production function. must equal the (inelastic) supplies..) could be that B..... m:1y now r. K.) .. The production function with technological progress The total factor productivity will now be allowed to depend on time... F(K" A.......... because the factor markets clear... of total factor productivity is exogenous...). and. YJL.ovP.....
Dividing by L. We will again use the dellnitions of output per worker (average labour productivity) y .Jacobsen: Introducing Advanced Macroeconomics I Part 2.). The idea of the 'endogenous growth' models presented in Part 3 is exactly to change the description of tech nological progress. we can write down the fu ll model consisting of seven equations without further delay: ( 7) (8) . We will simply assume that the labour productivity variable.In A.Exogenous Growth 5.. = Y. so that it will be the outcome of a use of economic resources. (A. g~ = ln(1 + g)~ g (where we use the approximation./L. in B.). it is called Hicksneutral. .In k.1. (6) These expressions reveal that an increase in output per worker can be obtained in a.ln At1• one has from (3). A.. (5) or in the usual notation lor approximate growth rates: g. = (1 + g)' A O' For the approximate growth rate. K. when it appears as an increasing variable. . other than the production function. is changing at a constant rate: g > ./L. In y .. (3) Here g is the exact (not the approximate) growth rate of A. _1) + (1 ._ 1 = a0n k. A positive value of g corresponds to a steadily arriving technological progress that comes exogenously to the economy without this requiring the use of economic resources.' = ag~ + (1 .. L. The production function just becomes ever more elllcient period by period. There are now a.' = In A. taking logs and time diflerences. . the (approximate) growth rate in output per worker is the weighted average of the rates of growth in capital per worker and in technology. it is called capitalaugmenting or Solowneutral. F(K. in F(D. In v ~ v . . It follows that if the technological level in some initial period zero is A0 • then in period t it L~ A . by more capital per worker or by better technology.). B. It is sometimes said that technological progress comes as manna from heaven.vo ways..1.vo potential sources of economic growth: capital accumulation and technological progress. Technological progress and growth: The general Solow model © The McGrawHill Companies..eI SorensenWhitta. L. A full description of the production possibilities includes a specification of the exogenous sequence.. g.Douglas specification one can simply choose the formulation of tech nological progress that is most convenient.a ) respectively. Given the Cobb._1). the weights being a and (1 . identical to the basic Solow model. on both sides of (2i gives the 'per capita production function': (4) and then. D. The complete model Since the general Solow model is in all respects. In fact.In y.a)g: . For our present purposes this turns out to be the form with labouraugmenting technological change as in (2). while. . and capital per worker (the capital intensity) k 1 = K. 2005 5 TEC HNOLOG ICAL PROGRESS AN D GRO W TH 131 increasing variable.a)(ln A. as in Chapter 2).
A first guess regarding the economic evolution implied by the general Solow model could be that.~~~~~g~p~r.S.2 :'\J?:.1: The dynamics of the general Solow model Note: Predetermined endogenous variables in squares.. but in an exercise you will be confronted with the counterpart model in continuous time..~. We have chosen to give a discrete time fonnulation of the general Solow model. K0 . output .. 2005 Solow model PART 2: EXOGENOUS GROWTH (YJ S. Technological progress © The McGrawHill and growth: The general Companies. k1.~. The next two equations give the rental rates from the marginal products of the inputs. the fundamental capital accumulation equation. but now the expressions for the marginal products are slightly ditferent. One can see from equation ( 6) that if k1 reaches a constant level.. and there are no pure profits.?.Exogenous Growth Macroeconomics 132 5. since they are derived from the new production function . 5. so there is no change in the assumed behaviour concerning savings or fertility.~1. so the functional income distribution is still given by a.. As in the basic Solow model this follows from competitive clearing of the input markets.1.?.~.  (1 0} ® ~ (1~1}~ ~ ~ (~12~}+ ~ (~1~ 3)~ ~ ~ ~ ~ Figure 5. = sY. 5. Note that we still haver. has to be respected as before.. the capital intensity.'?.~.<l}Y. just as in the basic Solow model. The last equation is just the addition of the assumption on technological change. (12) AI+ ] = (1 .. = (land w.g)A.KJY. Of course. (11) and (12) are exactly as before. the model (7)(13) determines the full dynamic evolution of the economic variables as illustrated in Fig.SorensenWhittaJacobsen: Introducing Advanced I Part 2.1. (13) The first equation is a repetition of the production function with the input supplies inserted. = 1 . L0 . For given initial values. A 0 .a . (10) (11) Ll+l = (1 + H)L. LJY. will converge to a speciflc steady state value.~~·····································································. Equations (10). of the state variables. endogenous variables that can adjust in the period in c ircles. (11).
<5)K. This guess is wrong. cannot stay constant in the long run when g > 0. In the general Solow model something more subtle is going on..1. +1• On the lefthand side you will now havek<+ 1• On the righthand side use A<+ 1 = (1 + g)A . then it must also imply that the growth rates of k1 and y.+ 1 L. if k. so th at the capital. Let us lind out what really happens. will be increasing (since g > 0). would not be a balanced growth path. If the (approximate) growth rates of k1 andy. to get: k1+ 1 = ( 1 ( Y1 K.. The guess was wrong. but it is instructive to see why it is wrong. respectively.L. and y JA . with a constant k. we would have a steady state with economic growth . gf = g~.. ) )( ) s+ (1 . Dividing on both sides of the production limction (7) by A 1 L 1 gives: ( 15) Now. The law of motion We analysed the basic Solow model in terms of the variables k 1 and y 1 which turned out to converge to specific and constant steady state levels.b)k1) . then (6) implies that their common value has to be the approximate growth rate of A. and eventually this will have to imply more and more capital per person.+ 1 = sY. will be increasing approximately at the rate (1 . L. g¥ = g~ = g. = 1 _ )( )(sy. Second. we cannot use k 1 andy 1 as such variables. since both have to change over time when g * 0 (repeat the argument). income per worker. and an increasing y. andy 1 ch ange by the same rate. both converge to the exogenous growth rate of A. This suggests analysing the model in terms of the variables: and (14) often called the technologyadjusted capital intensity (or capital per ellective worker) and the technologyadjusted average labour productivity (or output per ellective worker). Hence. should be changing at the same rate in steady state.output ratio stays constant. y . So k. that is.a )g because of technological progress. assumingg> 0 . and L. Look again at (6) above. The general Solow model would not perform well by empirical standards if the guess had been correct. Technological progress © The McGrawHill and growth: The general Companies. One of the requirements of balanced growth. as de tined in Chapter 2.. A. are the same. ({the model should imply convergence on a balanced growth path. is th at k. the growth path according to the guess. as wanted. k 1 andy. Divide on both sides by A. What variables could be used? For the longrun equilibrium of the model to accord with balanced growth.Jacobsen: I Part 2. absolutely wrong. so that eventually k JA.. I Sorensen. and the ever larger income per worker will create ever higher savings per worker. As we have just seen.Whitta. First. will be constant. (1 + n 1 + g . stays constant at some level. We also want to analyse the general Solow model in terms of variables that will be constant in steady state. y . + 1 = (1 + n)L. 2005 Solow model 5 TEC HNOLOG ICAL PROGRESS AN D GRO W TH 133 per worker.+ (1 . + (1 .b) 1+n 1+g A. and why it is not what we want to lind. then indeed from (4) or ( 6) ... So.Exogenous Introducing Advanced Growth Macroeconomics 5. start again from the capital accumulation equation with the savings behaviour inserted: K..
= (1 . the slope of the transition curve shown in the upper part of the tlgure decreases monotonically from infinity to (1 . Subtracting k.o)/[(1 + n)(1 +g)]. Further. approaches inllnity is less than one if 11 + g + f5 + 11g > 0.rt th <lt fj . The limit slope as k. = k . = 1 (1 + rz)(1 +g)   (sk. on both sides of ( 16) gives the Solow equation:   k. This is the law of motion. +1 . The first order diflerence equation (16) will then determine k1 and then the full sequence. the sequences for the economic variables in which we are interested follow. from (8) and (9): (17) (18) w 1 = (1 . must be increasing. ku will converge to a specific value. =(1 + g) ' A0 • we have: !J 1 = k~(1 + g) A 0 • 1 and then consumption per worker follows as c. and vice versa. In the long run k.. Thus they involve k1 rather thank" and the expressions involve the parameter g. i<*. For instance. k0 . 5. Assume given initial values of capital.Sorensen. = k~' to finn:  1 kt + l = (  )( 1 + 1'1 1 + g )  (Hi) (sk. qualitatively and with respect to interpretation the laws of motion are similar to those in the basic Solow model. which is highly plausible. (19) The transition equation ( 16) and the Solow equation (19) look much like their counterparts for the basic Solow model of Chapter 3.Hence an initial k0 = Kj(A 0 L0 ) will also be given.2.Jacobsen: Introducing Advanced Macroeconomics 134 I Part 2.(11 + g + (5 + ng)k.). 5.). = i< . or the transitio11 equation. labour and technology: K0 .s)y c Given in itial values K0 • L0 . since income per worker is y 1 = y. follows from fj.Whitta. so k. Technological progress and growth: The general Solow model Growth Companies. Hence.o)k. A 0 • the transition equation (16) and other model elements determine the full dynamic evolution of the economy. only now they incorporate technological change appropriately.k. given by the unique positive intersection between the transition curve and the 45° line (this follows simply from 'staircase iteration' in the llgure). so from any given positive initial value. as illustrated in Fig. 2005 PART 2 : EXOGENO US G RO W TH Fi nr~ ll y insP. However.. in the state variable k. and g1 = y JA 1 converges toy* = (k*)'' . L0 and Ao. Convergence to steady state The transition diagram associated with (16) and the Solow diagram associated with (19) also look much like their counterparts for the basic Solow model./A 1 converges to a speciHc steady state value i<*. increases from zero to inllnity (show this). (fj . This implies that eventually both k. under this realistic stability condition the transition curve has the properties illustrated in the upper part of Fig. As long as the curve sk ~ lies above the ray (11 + g + f5 + 11g)k 1.A" and A. The corresponding adjustments in the Solow diagram have been indicated by arrows in the lower part of the figure. the technology adjusted capital intensity. we must have k1+ 1 > k.'.2.). from these.'.). An associated sequence.' + (1 .a)A/< ~' . ask. and !lt will increase at the .Exogenous © The McGrawHill 5. In particular. (k .
since othenvise k JA . 2005 5 TEC HNOLOGICAL PROGRESS AND GROWTH 135 k* Figure 5./y.. implying a new and higher steady state level. = KJY. and y JA.output ratio k. k*'.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2. will also be constant as required for balanced growth.3 illustrates the effects of an increase in the savings rate from s to s in the Solow diagram. could not be constant. namely the growth rate g of A. .' shifts upwards to s'k .Exogenous Growth © The McGrawHill 5. The curve sk .2: The transition diagram (top). Comparative analysis in Solow diagrams Figure 5.'. and the Solow diagram (bottom) same rate. Technological progress and growth: The general Solow model Companies. of capital per ellective worker. Hence the capital.
of k1 initially jumps up to a positive level and then decreases monotonically back towards zero. 5. 2005 PART 2: EXOGENOUS GROWTH (n +g +i5 +ng)k1 k* k ** k:. isg~ =Ink. and output per worker. 5..4.3: An increase in the saving s rate in the Solow d iagram In Fig. The approximate growthrateofk.ln A.In k1_ 1 = (lnk. is equal to a relative increase ink. As long as k1 increases faster than A 1.. one can see that the growth rate.( kl  1 ( 1 + 11) 1 +g) .k. which has been derived from (19) by dividing by k. = . respectively. both change over tin1e at the rate g which we assume to be positive.. Figure 5.a)g~ . The growth path corresponding to the old steady state is illustrated by the lower upward sloping straight line in Fig. A . by a rate larger than g. In this situation capital per worker. Figure 5. From the modified Solow diagram. k 1• must be increasing by a rate larger than the rate by which the denominator. represented by the upper straight line in Fig. but gradually falls back towards g.5. on both sides: kl + l .. = k JA 1 increases. (20). Fig. The same must then be true for y 1• The transition from the old to the new steady state is illustrated by the curve going from the lower to the upper straight line in Fig. 5._ 1) g~ + g~. . 5. Technological progress and growth: The general Solow model Growth Companies.. k1 begins to increase as illustrated in the Solow diagram as well as in the modified Solow diagram. 5. k 1.(n + g + c~ + ng)). that is. . When k.'. y 1.ln k1_ 1) +(In /11 . 1 (sk. (k 1 + 1 . From the first period after the savings rate bas increased. (20) Assume that the economy is initially in the old steady state where k1 and [J 1 are constant and equal to k* andy*. in excess of the relative increase in 11. . the numerator. the capital intensity must be increasing all the time at a rate which is larger than g.soany relative increase ink.6 shows the adjustment of the growth rate ofy 1 during the transition from the old to the new steady state.5. increases.4 we illustrate the shift ins in the nwdijied Solow diagram associated with the modijied Solow equatioll. k*' and fJ*'.Exogenous © The McGrawHill 5.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 136 I Part 2. 5. so does y 1• since g¥ = ag~ + (1 .k1)/k. Along the adjustment to the new steady state.
In the Solow model with technological progress the growth path of y . first jumps up above g. . 5.Exogenous Growth 5. 5. and then falls monotonically back towards g.5: Transition from the old to the new steady state In the basic Solow model an increase in the savings rate implied that in the long run output per worker. Technological progress and growth: The general Solow model © The McGrawHill Companies.5. y" changed from one constant level to a new and higher constant level. Then the growth paths of Fig.lth rate being the same before and after. Figure 6. as shown in Fig. We are then back in the basic Solow model. This happens through a transition where the growth rate of y. One particular case is when g = 0. as illustrated in Fig.6. 2005 5 TEC HNOLOGICAL PROGRESS AND GROWTH 137 a. 5. shifts from one level to a new and higher level. 5.5 will be horizontal. namely g. with the longrun grov.1 S kI k* k*' k.6 will occur around a longrun growth rate of zero.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2. and the growth jump in Fig.4: An increase in the savings rate in the modified Solow diagram Figure 5.
... Furthermore. or: k1 = A......... k1 and [JI' we are interested in.... k1 = kJtl.Exogenous © The McGrawHill 5... This gives: _ ( k* = and h·om . ifg > 0 .. the steady state of the Solow model accords with the concept of balanced growth in some respects: the growth rates of capital per worker and of output per worker are constant and equal to each other.6: Ad.....SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 138 I Part 2...... The key endogenous variables in steady state The technologyadjusted capital intensity in steady state... A......... We are going to show that the steady state is in accordance VITith balanced growth in all respects.... 2005 PART 2 : EXOGENO US G RO W TH g( Figure 5..... both growth rates are positive........ yJA............. Technological progress and growth: The general Solow model Growth Companies... respectively. As we have seen.... [/" = S n + g + (~ + ng k*.. k1 + 1 = k1 = kin (16) or (19) and solving fork. = A1 [J*... n + g + (~ + ng ) 1/(1a) ......' we obtain the corresponding steady state value of y 1 ( 1 S 11 + g + (~ + ng ) a/(1a} : (22) · It is not really the technology adjusted variables...3 Steady state in the general Solow model .......ustment of the growth rate during the transition 5..... From the dellnitions............................. ( S ......................... is found by setting )1/(1 a} (21) ' y = i< ...... it follows that when the economy has reached its = steady state. capital per worker and output per worker in period twill be ki = y'.......k* and (23) .................... and [1........
The appearance of the g.a)( l'l+g +· c5+Hg S (26) )a/(1a) . Output per worker is given by a technological variable (A 1 in the general Solow model and B l/(1 al in the basic Solow model) times a certain fraction raised to the power a. (27) In steady state the rate of return on capital is a constant. is constant as well. K1 = k1 L1). in accordance \>\lith Stylized fact 5 of Chapter 2. = sy. The labour force grows at the constant rate. The steady state evolutions of the rental rates follow from inserting k* fork.. Note that substituting A0 (1 +g)' for A. In fact all the requirements for balanced growth. are fullllled in the steady state: GDP per worker. a(~~+g:0+_n_g wi . With respect to structural policy. as dell ned in Section 4 of Chapter 2.r1). does. and assuming g > 0 there is positive growth in GDP per worker in the long run. Technological progress and growth: The general Solow model Growth © The McGrawHill Companies.' = A. g..(1. times a constant. 2005 5 TEC HNOLOG ICAL PROGRESS AN D GRO W TH Yi = At S (n . r* .L. 11. you will see that they are reminiscent of the expressions we found in the basic Solow model. will trace the three steady state growth paths above back to parameters and initial values. as we have shown. = y.<5. In conclusion. tile Solow model passes tile el'npirica/ check that its long run. the steady state consumption path is: c'. w'.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2. r*./(1 . increases with A. consumption.s)y 1. The similarity between the models implies that their policy implications and the relevant empirical tests of their predictive power are also quite similar. and therefore (from Y..All. C. are the only dillerences. investment and capital a ll grow at one and the same constant rate. I. at rate g. the general Solow model points to one type that could not be considered in the basic Solow model: a policy to increase the growth rate of . respectively: r* = r. The real wage rate. Structural policy for steady state Looking at the above expressions for output per worker and consumption per worker in steady state. in (17) and (18).Exogenous 5. capital per worker.L. consumption per worker. and hence they all grow at the same rate gas A. approximately g + n. GDP. + g+ <5 + ng )a/(1a) · 139 (24) Since consumption per worker in any period is c1 = (1 . Finally.' . where the fraction has sin the numerator and 11 + (5 in the denominator (plus a gin the general Solow model). (25) The steady state expressions for the three variables above are all of the form A. the rate of return on capital and the real interest rate are constant. so the real interest rate.L. and the associated tact that now the technological variable in front of the parenthesis possibly increases over time. and the real wage rate all grow by one and the same constant rate. = c. steady state prediction accords with balanced growth.s) s)a/(1a) ( n+g +<5+ng . = S.
on y 1 than just their directions (the same was true lor the counterpart equation lor the basic Solow model) .. this is the first of two important relationships resulting from the Solow model. . where s. 5 .Sorensen. A 1• the steady state prediction of the model is that In yf should depend on [ln s .[In s .. that the countries considered were all in steady state in 2000 and bad the same technological level. and the relationship should be a linear one with a positive slope equal to a/(1 .) The figure is in fairly nice accordance with a positive and linear relationship. plotting GDP per worker against gross investment rates across countries. We have thus set g +a+ ng at 7. 1. (29) . In Fig. are mainly policies that can increase the savings rate or decrease the population growth rate. 3. we could have excluded ng in the above formulae. 3 . 5. Technological progress and growth: The general Solow model Growth © The McGrawHill Companies. a. Eq. policies suggested by the general Solow model to increase income per person. Taking logs on both sides of (24) gives: • (l In y7 = ln A 1 + . since the capital share. an estimate that is often used.8. + 0 . In y:>o• against Ins. 7 and 3. is around }. A!x>· in that year. (Constructing Fig. The slope should therefore be around~. is the average gross investment rate in country i over the period 1960 to 2000. and 11.ln(11.5 per cent for all countries. .7. Since ng is the product of two rather small growth rates. plotting GDP per worker against population growth rates.Whitta. now intended to elevate the whole growth path of income per person. Empirics for steady state Taking A 1 and gas given. The sample of countries is basically the same as in Figs.ln(11 + g + c} + ng)].Jacobsen: Introducing Advanced Macroeconomics 140 I Part 2. The straight line that has been drawn is a line of best fit resulting from an OLS estimation of the regression equation: In Y:m = Yo+ y[ln s.c1) . Thus we tested the influences of sand 11 on GDP per worker separately and found the model's prediction of the directions of these influences confirmed by the data.ln(n. The golden rule savings rate is unchanged from the basic Solow model. perhaps heroically.i we have put this prediction to a direct test a~suming . 7 we are more or less doing what you were asked to do yourself in Exercise 9 of Chapter 3. (24) implies a more precise theory of the influences of s and .Exogenous 5. without sacriHcing much precision. 3 . Thus. across countries i. the other one (to be derived below) will concern convergence to steady state. Otherwise. The figure plots. However. 2005 PART 2 : EXOGENO US G RO W TH technology.ln(r·r + g + (5 + ng)]. The figure's legend gives the details of the estimation. This one concerns steady state.a (28) As far as empirics are concerned. the log of GDP per worker in 2000.075). + 0 .8. and in Fig. and we will (again) defer further discussion of policies to allect technology to later chapters that consider models with endogenous technical innovation. the implication of (24) is that higher sand lower 11 will generate higher steady state output per worker. It is not easy to see from the present model what kind of policy could achieve this.075)]. In Chapter 3 we tested this prediction in Fig. is the average population growth rate over the same period. it is very small. Given the technological level. Hence we are assuming that g + c}"" OJJ75. Note also from (2 5) that the value of the savings rate that maximizes consumption per person (by taking the consumption growth path to the highest possible level) is still s** = a..
if the fraction a/(1 .JP"' • 8 HL • ZAF ++# +P~RC 10 9..7: Logarithm of real GOP per worker in 2000 against structu ral variables.(1) should equal 1.Jacobsen: Introducing Advanced I Part 2. the steady state prediction of the general Solow model must be said to perform quite well empirically. .5 7 6. However.5 ~· !. 86 countries Not e: Luxembourg is included in this figure. An a of around 0. giving a conHdence interval of [1 . Some detail3 for the statistically proficient: R 2 =0. is 1. In fact. Show this.5 ./11 + BRA EGY + ~· / ~ /• • • / •BDI+ +NGA M2G 1 . the model considerably underestimates the strength of the influences of the key parameters on GDP per person compared to what is found in the data. 7 and 3.47. but was not in Figs 3. but could be even better.6 is way above the capital share (around ~) . 2005 Macroeconomics Solow model 141 5 TEC HNOLOGICAL PROGRESS AN D GRO W TH (/) ~ :g <0 ~ g "'~ 0 "' 12 11 ~ a.14. ~PER • .8. PAK + •r Cflrq+ZW •co G •KEN •zMB • •TzA 0 0. 5 it should be according to our theory. then a would have to be 0 .eI SorensenWhitta.f['~~A ~~A 10.5 7.55. Source: Penn World Table 6.j 8 ~ CAN •·~~~OR GBR •• ~ . (28). +++:: _. 5. 9 Cl 8.ln{n' + 0. The 95 per cent confidence interval for the implied a is (0.0.5. THA ~ + .55 and I= 10.7 that the standard error of the estimate ofy is 0. Moreover.1.Exogenous 5.81 + 1.1.5 0 3 '" 11. In summary. 75].2 (standard error of the estimated coefficient is 0. :r:. Technol ogical progress © The McGrawHill Growth and growth: The general Companies.5 (. This does not include the value 0 . implied by the model are in nice accordance with the data.14).20. The estimated line is y= 8.5 C) u. 3. that a should correspond to. so the estimated slope is signifimntly larger than 0. the model's prediction of the direction of this influence and the linear form of the relationship.•.4 7. 3 The overall conclusion seems to be that the steady state equilibrium of the Solow model matches crosscountry data quite well (even under an assumption of a common technological level). 0. The 95 per cent conHdence interval lor the estimate goes approximately from two standard deviations below to two standard deviations above the estimate of y.5.5 . Note from the legend to Fig.64]. the slope of the line.47x.075) Figure 5. The data conHrm that the key parameters tend to influence GDP per worker through the way they a!Ject s/(n + g + c~) .5 2 In s'.5 1.1 The estimate of y. This slope is considerably larger than the 0 .6.
4 Economic growth and convergence in the gen eral Solow model 4 It is reassuring that the Solow model can create balanced growth with a positive growth rate in steady state. 5. it is a matter of great concern whether it tal<es 10 or 50 years to go half way to the new growth path. Technological progress and growth: The general Solow model © The McGrawHill Companies. For those who d o not yet know the technique. the savings rate) to approach a new and higher growth path. this will imply that it has t he highest growth rate in y . n. since these sections are concerned w ith convergence to steady state. We will now examine the convergence process in detail. we had to postulate an equation stating h ow the structural characteristics should enter (Eq. Consider two countries that have the same structural characteristics. that is. the Solow 4. just as important as convergence itself is the rate ofcol'! vergence. When we wanted to control lor structural diflerences between countries. and still read the remainder of the book. g. g. s.41or such a diagram) . Convergence according to the Solow model o. This will state bow a country's average growth rate in GDP per worker over a period should depend on its structural ch aracteristics and on its initial level of GDP per worker. Furthermore.. Some readers may find this section more difficult than the material presented up to now. A 0 . as well. but now that we have a real growth model we can. Through the process where each country approaches its own steady state. but there were things we could not do. Now that we have a real growth model we can explain howEq. and a corresponding one in Chapter 6. From the Solow model we will derive a 'convergence equation'. (4) of Chapter 2) . It follows that the country which is initially furthest below steady state will have the high est growth rate in k1• As we showed above. there should be a negative relationship between GDP per worker in an initial year and the growth rate in subsequent years. It is possible to skip this section. In Chapter 2 we could not estimate rates of convergence. 2005 PART 2 : EXOGENO US G RO W TH 5. but the model bas transitory growth as well. Hence the Solow model supports conditional convergence. It would be a pity. while the economy converges towards a steady state from below. w hich is just as important an element of our theory as steady state itself. the speed by which a country converges towards its longrun growth path. a. we w ill give a cookbook recipe for the solution. For a poor country that attempts to improve its structural characteristics (e. the growth in GDP per worker will exceed the steady state growth rate. in such a way that the growth in excess of g decreases and goes to zero in the long run. the two countries will converge towards having the same level of GDP per worker. They will then have identical modified Solow diagrams (see Fig. An implication is that if one controls appropriately for the structural dillerences between countries. since this may be decisive lor bow fast poverty can be fought. (4) is rooted in theory. In Chapter 2 we began to test the idea of convergence empirically by setting up growth in GDP per worker against the initial level of GDP per worker across countries. and faster so the further away it is from steady state. though. The only additional mathematical element used in this section is how to solve a firstorder linear difference equation.Exogenous Growth 5. that is. .g. As we have seen above.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 142 I Part 2. The monotonically decreasing transitory growth means that the general Solow model accords with the idea of co11ditional col'!vergel!ce explained in Chapter 2.
take the place of j. G(k 1). will be preserved. Our next step is to convert (30) into changes in natural logvalues rather than changes in absolute values. as long as z is not too far from x. the convergence equation wi11 have a coelllcient on the initial GDP per worker that is linked onetoone to the rate of convergence. which can be used whenever wanted. applying that from iJ 1 = k ~'. where the function G(k .Jacobsen: Introducing Advanced Macroeconomics I Part 2. stated in (23) fultlls k* = G(k*). (32) We convert into logdillerences of [J 1 rather than k.x ~ f'(x)(z . Using this on both sides of (32) gives: (33) This is still a linear. and then In y. so k.f(x) ~ .ln k* = (1/k*)(k 1 .x). and for any given x.Ink*). The steady state va1ue.Whitta.o)] ~ 1 + n (1 + {] G'(k. (or in the absolute discrepancy between k1 and its steady state value). this time letting the function In k. measuring at the steady state value.f'(x)(z . If x is such that x = f(x) . then we get f(z) . This is easily done using (again) f(z).. The linearized dynamic equation includes G'(k*). ~ k* as t ~ oo according to it.Exogenous 5.k* = k*(ln k. first order dillerence equation.) is given as the expression on the righthand side of (16). and since 0 < G'(k*) < 1. where k* is given in (21):  1 1 ) [sa(k 1) '' .f(x) ~ f'(x)(z . Using these properties lor the dynamic equation. We can derive an expression for G'(k*) in terms of model parameters by dillerentiating the righthand side of (16) with respect to k 1. This implies that the linearized dynamic equation (30) is stable. Before doing so we will ta1<e it through some rewriting. we have f(z) . Write it in the generic form. Start from the transition equation (16) .Ink*). f'(x) . . This is really just using the definition of the derivative.In [J* = a(ln k. (31) In the following we will still use the term G'(k*) for short. Technological progress and growth: The general Solow model Growth I © The McGrawHill Companies.x).Ink* = G'(k*)(ln k1 . 2005 5 TEC HNOLOGICAL PROGRESS AND GROWTH 143 model's convergence equation will rationalize (4) of Chapter 2. For any dillerentiable functionf(x). As long as we are not too far from steady state.x). Using this on both sides of(30) gives (note that k* cancels on both sides): ln k1+ 1 . or k 1 . Furthermore. which means that we can solve it. From (31) it is easy to demonstrate that 0 < G'(k*) < 1 (do that). but the linear structure..+ (1 . the new dynamic equation approximately represents (16 ). and then evaluating at k 1 = k* . the new difference is linear. we obtain: (30) This is a new dynamic equation.k*). but G'(k*) has now been expressed in terms of model parameters. k*.c))]. it implies thatfj 1 ~ y* as t ~ oo.eI Sorensen.Ink*). ..) = ( G'(k*) = ( ) 1 ) 1 + n (1 + {] ) o [a(n + {] + + 119) + (1 . k*. a first order difference equation ink. k. Here is how to proceed. now in ln [J 1.+ 1 = G(k 1). one has In [J 1 = (l ink. and hence the solvability. Importantly. and measuring at k*: In k 1 . so an estimate of this coelllcient will provide us with an estimate of the rate of convergence. The advantage of having transformed it into logdillerences . We now lir1earize the trnnsitim·1 equatio11 arormd the steady state.
A)' fulfils the homogeneous equation for all I. and g are small. il. and we call il. then each year 2 per cent of the remain ing distance to steady state is covered. is a time independent measure of the rate at which [J .) . = In [t* (check th is). must depend on timet for (3 6) to be fulfilled in all periods. A reasonable lowerbound estimate for A would be around 4 per cent.a)( ll + g + o).1.075 our model predicts a convergence rate of 5 per cent.. . Hrst subtract In [J.a)(n + g +b). corresponding to a = l and 11 + g + c) = 0 . the remaining relative gap. In y*. one particular solution (a time path for In {I 1 th at makes (3 6) hold for all years) is the constant one: In {I.lny 1) . since 0 < G'(k*) < 1.il/. The characteristic polynomial associated with this is: Q(x) = x . Second. and its steady state value) is that the rate of convergence appears in a welldeHned way. y*. arising from setting the constant term (A In y*) in (3 6) equal to zero. where Cis an arbitrary constant.Exogenous Growth 5. Hence.)ln y. Technological progress and growth: The general Solow model © The McGrawHill Companies.. First. and add In fj* on both sides of(33) to obtain: lny. The complete solution lor the homogeneous dillerence equation is then: In li. From ( 31) we can express the rate of convergence in terms of parameters as: il. irrespective of C. C heck that lnji. with root x = 1 .a)(n + g + (5 + ng) ~ (1 .In [J. (34) Here we deHne A= 1 . To arrive at the Solow model's convergence equation we have to solve (36). (1 + n)(1 +g) (3 5) For the latter approximation we have used the facts that since both .G'(k*).G'(k*) in (34) and rearranging a bit gives: lny. converges to fj*. Note th at (34) has the 'convergence property' : the growth rate of y1 is larger (in absolute value) the further away from steady state fj 1 is.06. the socalled homogeneous equation. Thus we may conclude that.g. = C(l .A)lnfj1. = (1 .. is: In y1+ 1 . tlrst order diilerence equation in In {1. if A is 0 . Inserting . the rate of convergence is approximately A= (1 . and 119 is close to zero.il)1. the fraction 1/[(1 + n)(1 +g)] is close to one. irrespective of how large the remaining gap currently is. that is. Equation (34) then reads In Y r +J . the lefthand side) is the fraction A of the current relative deviation between y..(1 . 5 Third.lny. the 'rate of convergence'.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 144 I Part 2. = C(l . saying that in any period t. for 1 . this time that the stated solution fuHils (36) for all t irrespective of C. Check again. the complete solution to (3 6) is the sum of a specific solution and the complete solution lor the homogeneous equation: In y. where we must have 0 <A< 1.(1 .ln y1. 6.+1 = illny* + (1 . = 0 . In other words. (36) which is a standard way of writing our linear. To see th is.In fj 1 = il(lny* . 1 =1 .il). according to the Solow model. For instance. determine how ln y. For realistic parameter values such as a ~ t and n + g + f5"" 0. 6 5. the relative change in y1 (from period t to t + 1. is reduced by the fraction A in any period. = In fl* + C(1.G'(k*) = (1 .G'(k*))(lny* .02.il. This is done using a standard 'cookbook procedure' involving the following steps.+1 . and its longrun equilibrium level.A (the x that implies Q(x) = 0). 2005 PART 2 : EXOGENO US G RO W TH (or relative discrepancies between fl . e.
) T)(ln y* .< 1• We simply postulate a common rate of convergence A and see how good our estimations become. You may think that it looks complicated. according to (39) (or (38)). for y . We do that for the same sample of 90 countries considered in 7.In A.A. along the steady state growth path.Whitta. and the actual GDP per worker. The reason is that if we accounted for the influence of n on A.A.Jacobsen: Introducing Advanced Macroeconomics I Part 2.A.ln[/ 0 )../1 1 and P2 in (36) all depend on A.A. having the same g and A 0 . When we do that and also insert that (In Ar . and dividing by Ton both sides gives: lnyr . y 0 .lny 0 ln Ar .lu(n+y +<h ny)] . we hold ft 0 . The other depends increasingly on the initial relative gap (the year zero logdifference) between the steady state growth path. as T goes to inllnity. g.r • l . the resulting non·finear equation would require too sophisticated methods t o estimate.In jj*.8 0 .lnA 0 1 . This is intuitively as it should be for an economy monotonically approaching an m>\111\ongrun growth path '>\lith a constant growth rate.lny*)(1 . ignoring that cross·country differences in n 1 should imp ly c ross·country differences in . This growth rate is.A.:y + luA0 + . 2005 Solow model 5 TEC HNOLOGICAL PROGRESS AND GROWTH 145 Fourth. In specifying this regression equation we are making a short·cut: according to (39) the parameters .[Iu s . fJ 1 . suggest a regression equation across countries i: (40) This is exactly (4) of Chapter 2.ln [/ 0 = [1 .ln A0 )/T is approximately equal to g.lny 0 ) .1 1 . g. which in turn depends on n as stated in (35). for both t = Tand t = 0. taking the value A 0 !i* in year zero. Technological progress © The McGrawHill Growth and growth: The general Companies. the second term will contribute very little to average growth.Exogenous 5.(1 . Nevertheless.r1 This is the convergence equation of the Solow model. given the initial value y0 . but its content is simple: on the lefthand side is the average aruma! growth rate in GDP per worker over a period from year zero to year T. .)']ln y* + (1 .(1 .luy 0 • (39) T T 1 . for the solution to be compatible with a given initial value jj 0 for t = 0. is: lny 1 = ln[J* + (lny 0 . (3 8) Here we can (finally) insert from (24) how the steady state value y* depends on the structural characteristics.eI Sorensen. A 1 [ / . we must have C = In y 0 .(1 . /1 1 and . Testing the model outside steady state and estimating the rate of convergence The theoretical convergence equation. and fJ 2 by OLS. The dependence is such that as the number of years considered becomes large.8 2 constant across countries in our estimations.. (39). Therefore. given as the sum of two terms: the tlrstis simply the growth rate.)r _* T = T + T (lnA 0 + ln y . the unique solution to (36). and an assumption that countries are technologically similar.) 1 = [1 . we get: In I/ .A.In IJ 1 1 .) 1 In Yo (37) Evaluating for a speciHc year Tand subtracting In y 0 on both sides gives: In Br .)r ( r1 ) . Now using In [J 1 = In !. 7 With appropriate data we can estimate the parameters fJ 0 .(1 .
}. Inserting T = 40 and our estimate of fJ 1 of 0.0 2 • 0>2 (ii o Q) 0. Taking uncertainty into consideration.05 Q.020 in the construction of Fig.T(J 1) J/r. (se=0. we already did this estimation._ Q) Q) ~ :. 5. using the same data where. For convenience the figure is repeated here as Fig.Whitta.5 VEN+ CH 9 9.8.o.063 . Singapore: 19601996. resulting from our estimation .6 of Chapter 2. 1 oo (9~ 0 .020[ln s. . Fiji and Botswana: 19601997..0.OOJ 5) y. Importantly.. + 0..c: ~~ KO~~MUS IRL • 0. Mauritania.6.. are: g~. tor example.(1 .In( I!. to create Fig. According to (39). adj.. etc..006 gives }. The details of the estimation..060 = · 0 .5 10 10. 2.+ 0. is the average investment rate between 1960 and 2000 of country i.7 per cent.0 2 6.0025) (41) This is why we chose the parameter value fJ 2 = 0. Even under an assumption that the many countries considered are technologically similar.5 • LUX • 8.063 . In fact.5 8 •• • ZMB .0. 0 + 0 . 2 . 0 • and we indicated the straight line.: . We were then plotting (g1ku. = 1 . our estimation in (41) implies an estimate of the rate of convergence. Thus our estimate of the annual rate of convergence based on our model and data for the 'countries of the world' is 0 . the parameter {3 1 is equal to (1 . the model's convergence equation stands up quite well when confronted with crosscountry data. R2 = 0 .. Source: Penn World Table 6. 0.075)]) against In y~..)r)/T.(1 . . 2005 PART 2 : EXOGENO US G RO W TH Figs. Technological progress and growth: The general Solow model Growth Companies.04 H~G • Gf:JN 0.. with standard errors in parentheses below estimates.Exogenous © The McGrawHill 5.ln(11.1 .40.006 In y~.6. ('l 0. One may take this as evidence that the assumption that countries have equal access to technology is not as farfetched as it may 11rst seem. the 9 5 per 0. 90 countries 1 1.0.0 1 Cllo 0 ::J c= c 0 "'~ Q) c "' 0 (» (» ~ ~ 0 :.020 [ln s.075)] . = 0. s.0 . an issue we will return to in the next chapter where the convergence equation derived will perform even better.5 Log of GOP per worker.006 In (se=O.007. 2.Jacobsen: Introducing Advanced Macroeconomics 146 I Part 2.8: Average annual growth rate of GOP per worker adjusted for structural characteristics against the initial level of GOP per worker.06 0 ~o Ol O BWA 0. 1960 Figure 5. 5 and 2.03 ~ n ::J .0 1 NGA+ TZA • 7 7.0. implying that A. 0.Sorensen. The 11gure and the details of the estimation reported in (41) show that the explanatory power of our regression equation is not bad.
.1 per cent.014(1n s1.. This suggests that the Solow model exaggerates considerably the speed of convergence compared to real world convergence...... We will still assume the production function to be of the Cobb... of 0....~...002..... The follow·up paper is: Robert M.... pp.... 0 .3 per cent to 1.. and a standard error on the estimate of p . 'Technical Change and the Aggregate Produc tion Function'... the general Solow model stands up quite well confronted with real world.. 5.... = Bi/(l a). 2.. computed (again) the simple way: go two standard errors down and two standard deviations up from the estimatedj3 1 to lind (approximately) the 9 5 per cent confidence interval [0. 8 Our estimates of annual rates of convergence based on our model and crosscountry data are thus.. 9 Growth accounting uses just one of the equations of the Solow model. 39. In summary.~~~?. . annual rates of convergence should. remembering that A...... 9. the estimate of the rate of convergence is 2.~... The estimate of /1 1 of 0.....~. the aggregate production function .2 per cent.... but one could equally well express it as labour augmenting changes in an appropriately defined A . both with respect to its steady state level prediction and with respect to its outofsteady state convergence prediction. Above we found that for plausible parameter values. Solow... 2005 5 TEC HNOLOG ICAL PROGRESS AN D GRO W TH 147 cent confidence interval for A goes from approximately 0.009] for {3 1• Then insert the two end points of the interval into the formula for A to lind the 9 5 per cent confidence interval for il. be no less than 4 per cent. 131 .Exogenous Growth 5.0 13y'60 + 0.Tfl 1) 1/T...020.~~~~~.. in the range 1..(1 .. using the formula A = 1 .... we have: (43) 8....... Growth nccounti11g was suggested by Robert Solow in a paper that followed up on his famous 19 56 article...~..195 1 = 49.~. . Technological progress and growth: The general Solow model © The McGrawHill Companies..005...0 13 gives an estimate ofA of 0. Compute the 95 per cent confidence interval yourself. according to the Solow model...eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2. So the aggregate production of a country in year tis assumed to depend on the inputs of capital and labour and on the level of technology as: (42) For another and later year T > t..5 9.. crosscountry data..... In growth accounting exercises one often expresses technical progress as (Hicks neutral) changes in the total factor productivity.... In Chapter 6 we will present an extension of the Solow model that implies substantial improvements with respect to both of these empirical shortcomings.4.. For both of these predictions there is. with an adj. however.. 1957......075)]..003....Douglas form .... 2. An estimation of Eq........ 3 12320..... now with T = 2000.. Review of Economics and Statistics. a problem of magnitudes.0 per cent.0. (40) for the 24 OECD countries considered in Fig..66. It is a problem for the Solow model that these estimated annual rates of convergence are substantially smaller than what they should be according to the model for realistic parameter values.....?. R' of 0... for which we found the very nice convergence picture displayed in Fig... B..4 over the period 1951 to 2000 gives: g'00•5 1 = 0. and on /3 2 of 0. .. (Do th is yourself..) If we lim it attention to the OECD countries... more or less.ln(n 1 + 0..
over some period? In a way. or between two different countries for the same period. and divide on both sides by T .. may be the first warning that productivity is not developing as favourably as it used to.t T.Sorensen. of growth in the total factor productivity can be computed residually. (44) can be used to compute the contributions to total growth from each of the growth components. Solow himself has expressed this by saying that the Solow residual may be viewed as 'a measure of our ignorance·.:___ ____: = + (l lnKr .lnL1 T. i. and setting a = } . = B.t.1 reports on a growth accounting exercise for sLx dillerent countries.t).vth rate in GDP per worker has been split up in components according to (45) assuming a = that is.t For B. T. Taking logs and time differences. 2005 PART 2 : EXOGENO US G RO W TH If we take logs on both sides in each of the two above equations. from data on GDP per worker and capital per worker. represents. at }. this equation may be rewritten in terms of A1 to give In Yr . although we do not know exactly w hat an increase in B.ln !J 1 (45) T. Lower growth in B. Again.t . and if one sets the capital share o. or as favourably as in some other country. (In Br . We have simply put all factors ofimportance lor aggregate production other than physical capital and labour into B. etc. one can determine the growth contribution from technology (or other factors) residually. the contribution.Exogenous © The McGrawHill 5.___ __.)/(T . . Table 5.t _ = (1 .. How informative is a calculation of the average annual growth rate in total factor productivity of. one could compare the Solow residuals between two dillerent periods for one country. not very informative. according to gY = g 6 + l/3 in our usual notation.ln y1 __.t _ ___.Whitta. T. and k. T. For instance.t T. growth accow1ting exercises may be useful.' in the usual notation. The righthand side is the approximate average annual growth rate of total factor productivity plus a weighted average of the approximate growth rates of capital and labour.ln k1 +a .:. If one residual is considerably larger than the other. However. we get: lnYr . lnBr . . One can also do growth accounting per worker.a) ln Ar . This component is called the Solow residual. The production function (42) implies that y. since we do not really know what B. the contribution to growth coming from unknm>\Tl1 factors has changed from one period to the other. say 3 per cent.. (46) expressing the growth rate of y 1 as the weighted average of the growth rates of A.lnB.lnK. = A: _. gives: In Yr . is.t + (1  (l) lnLr . the average annual grm.lnY. in accordance with (5) above. Technological progress and growth: The general Solow model Growth Companies.Jacobsen: Introducing Advanced Macroeconomics 148 I Part 2.t • (44) The lefthand side is the approximate average annual gro\<vth rate of GDP between t and T. If one has data lor GDP and for the inputs of capital and labour for two years t and T. subtract the first from the second. Equation (44) thus splits up the growth in GDP into growth components. T.ln A 1 In kr .1965 and 19952000.k. For each country and lor liveyear periods between 1960. Equation (45) splits up the growth in GDP per worker into contributions from growth in capital per worker and growth in totallactor productivity. or has been larger in one cow1try th an in the other. In particular.In B.
90 1990.6 0.8 1.65 1965.9 0.9 0.6 0.1 4.2 1.0 2.4 2.5 1.6 Belgium (%) Ireland (%) Period Period 1960.8 2.65 1965.0 0.6 3.6 2.6 4.1 1960.6 1.75 1975.2 3.85 1985. Economic Outlook Database. 2005 5 TECHNOLOG ICAL PROGRESS AND GROWTH 149 Table 5.4 3.9 4.1 1.4 1.2 0.90 1990.7 2.4 0.00 1.1 1.2 0.7 2.4 0.9 2.5 1.70 1970.90 1990.3 0.2 0.70 1970.0 1.7 2.8 3. .2 1.9 1.1 2.4 1.0 1965.95 1995.9 1965.8 1.3 2.6 0.7 1960.2 2.3 1.8 2.00 1.0 0.3 0.2 3.80 1980.85 1985.8 0.2 4.3 2.3 2.1 3.3 1.8 0.4 1.4 0.00 2.2 3.4 0.7 0.7 5.8 1.2 0.65 1965.2 3.6 1. Penn World Table 5.6 1.8 0.2 2.0 1.4 1.2 1.6 0.0 0.6 0.9 0.1 0.8 1.8 0.8 1.7 2.7 0.9 0. For Denmark 1960.9 0.7 2.8 1.4 2.0 0.0 0.75 1975.3 2.70 1970.3 1960.0 3 .3 1.7 0.1: Growth accounting: the great productivity slowdown United Kingdom (o/o) USA (%) Period gY gk/ 3 g B= gY_ gkj3 Period gk 1960.80 1980.1 1965.90 1990.4 0.9 0.1 2.3 2.6 3.70 1970.9 2.2 4.85 1985.7 3.70 1970.0 1.4 Sources: OECO.1 2.80 1980.4 2.4 3.5 1.6 3.6 1.6 0.8 1.5 1.2 0.2 4.4 0.65 1965.3 0.95 1995.9 2.00 5.5 0.5 0.0 0.1 1.2 1960.80 1980.5 0.95 1995.3 1.6 3.00 3.9 2.4 1.7 0.75 1975.0 0.9 1.80 1980.7 0.2 1.9 1.8 0.85 1985.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2  Exogenous Growth I © The McGrawHill 5.00 2.1 0.7 3.6 0.4 0.7 0.4 1.6 1.5 2. Technological progress and growth: The general Solow model Companies.8 0.8 g B= gY_ gk/3 Period 2.4 1.3 0.9 2.3 0.3 1.0 2.7 1.85 1985.2 3.7 0.8 2.2 2.0 3.8 0.5 1.5 0.65 1965.95 1995.3 0.4 1960.6 0.5 0.2 1.95 1995.85 1985.6 0.75 1975.4 2.7 1.75 1975.7 2.6 0.19 70.8 5.5 0.3 1960.3 2.3 0.70 1970.00 6.7 3.90 1990.9 0.8 0.6 4.3 2.8 gY gk/ 3 gB = gY_ gk/3 Sweden (%) gY Period Denmark (%) gkj3 1960.90 1990.3 0.9 0.5 3.65 1965.1 1.8 2.7 2.0 1.80 1980.75 1975.00 1.4 1.0 2.00 1.00 2.95 1995.6 1.2 0.5 1.9 1.0 1.6 3.8 1.3 1.00 0.8 0.5 1.2 1.0 0.00 2.6.
note that a gB of 2. = Btf( Lfl) it follows that gA= g8 /(1 . often referred to as a human capital function . What we want to emphasize is that the evidence from a growth accounting exercise has induced people to think about wh at was going on. In the ftrst liveyear period TFP (total factor productivity) growth in the US was 2.4 per cent for a = i (from A.output ratio either decreased or increased substantially). and doing so will reveal that high TFP growth was typical for a longer period following the Second World War. As you will see.1 only go back to 1960. The period of high TFP growth in the US was followed by a considerable slowdown with quite low rates of TFP growth in the late 1960s and 1970s. while Ireland seems to have been developing quickly throughout the period. and under balanced growth. Rather it is the high TFP growth in the 1960s that is unusual and needs to be explained.). Some have suggested that the oil crises of 1973 and 19 79 were responsible (but note that the slowdown in the US started before that). This illustrates that growth accounting is a useful tool lor economic analysis. Still others have claimed that declines in the quality of the labour force or ofinfrastructure caused the slowdown. and you will see from Table 5. The internationally comparable data used lor Table 5. It is often argued that growth accounting should allow not only for the quar1tity oflabour input. The downward movement in TFP growth in the middle of the period has been called the 'great productivity slowdown'.9 per cent per year. Economists have searched lor reasons for the international productivity slowdown described. but also for the quality of labour input. This indicates that the US was close to balanced growth over the period (and so was Belgium. We will not go into detail about the explanations oiJered. Growth accounting only uses a production function and some knowledge of parameters (like our a). 11 1• that people in the year t labour Ioree have spent on education (in school.1 that the UK. The production function does not have to be as simple as (42) above.a). early industrialized countries. followed again by a partial return to higher TFP growth in the 1980s and 1990s. over the full period from 1960 to 2000. One of their findings (most . Here some results from labour economics can be of use. The quality of labour would then be given by an increasing function l!(u . measured by the average level of education. It is natural to measure the average level of education in year t by the average number of years. 2005 PART 2 : EXOGENO US G RO W TH Consider first the ligures for the US. capital and GDP (per worker) grew at almost the same rate. while for the other countries the capital. Annual growth rates in GDP per worker of around 4. Many labour economists have studied the relationship between education and wages. which you should show).Whitta. Others have stressed the relative shift in the production mix away from manufacturing towards the more labour intensive production of services where productivity gains may be harder to achieve.Jacobsen: Introducing Advanced Macroeconomics 150 I Part 2. There are pros and cons lor all these hypotheses.9 per cent (per year) corresponds tog A of 4. some have argued that it is not the low TFP growth of the 19 70s and l 980s that need explanation . Moreover. Technological progress and growth: The general Solow model © The McGrawHill Companies.). gAis directly comparable to gu.Sorensen. and the aggregate production function would be modified to: (47) We cannot do growth accounting based on (47) until we know more about the function h(u. measured in hours or manyears. which is high. Belgium.Exogenous Growth 5.4 per cent are very high and unusual as averages over longer periods in Western. say). To illustrate. but for the US one can go further back. Sweden and Denmark experienced similar slowdowns.
one more year in school say. The East Asian tiger economies have typically experienced huge increases in educational ellort..(1)1/J . and a is~.. one additional year of education yields a certain relative increase (1/J) in the level of human capital independently of the initial u 1• Let us now try to estimate 1/J. Technological progress and growth: The general Solow model © The McGrawHill Companies. An 10.a)!fJuJ.ln k 1 ) llr . and average years of schooling of the labour force in two dillerent years t and T. and the residual growth in TFP.t (19) Derive this formula yourself and note that it is not the average relative change in schooling from year t to T that enters. but for some other countries as well) is that the percentage increase in pay generated by a oneyear increase in schooling is around 7 per cent. gives (check this for yourself): dw 1 .a)ljJ is roughly equal to 0 ." = (1 ... Experience. Columbia University Press./w 1 = du 1 (1 . we can do growth accounting again. T. If(l . growth in education.Jacobsen: Introducing Advanced Macroeconomics I Part 2. one can again split up growth into the contributions from growth in capital per worker. . w 1• should be: w. if a country has experienced a large increase in education over the period considered.a 1/J .a)1jJ. In per worker terms the appropriate formu la following from (48) is: In In y T. seems to give rise to a specific relative change in the wage rate. but the average absol11te change. capital per worker. one may tend to exaggerate the contribution from TFP growth if one does not include education in the growth accounting.eI Sorensen.Whitta.__"". ( Taking logs on both sides and differentiating \Vith respect to u.ln B1 In kr . If labour is paid its marginal product.u + (l + (1 . then !fJ must be approximately 0 . A pioneering contribution is Jacob Mincer.= ln Br. we obtain an increasing human capital function with exactly this property Oust take logs on both sides and dillerentiate with respect to 11 1) : That is. In particular. A rather robust tlnding in empirical studies (mainly lor the US.a)B 1 Kr'h(u 1) 1 '" L. = (1 .:. Thus a oneyear increase in schooling gives a relative increase in pay equal to {1 .t T.t Yr1 _:.1 . K L: )'" exp((1 . This may be particularly relevant for an assessment of TFP growth in the growth miracle countries of East Asia. then (47) implies that the wage rate.. and Earnings.a)B. Schooling.07. 2005 5 TEC HNOLOG ICAL PROGRESS AN D GRO W TH 151 often based on microeconomic data) is that a specific absolute change in the amount of education.t T.Exogenous Growth 5. With the production function: (48) and knowledge of the parameters a and 1/J. 1974. 10 By assuming the functional form: 1/J > 0 . From data on GDP per worker.1 (10 per cent).
Both observations will be true......... An exercise will ask you to figure out exactly what a growth accounting exercise lor an economy in Solow steady state will say.. A growth accounting exercise can split up growth in GDP or in GDP per worker into components.... one can go on looking for other neglected factors of production to explain the difference. so technological growth is the ultimate source of longrun growth ........... The general Solow model developed in this chapter generalizes the basic Solow model by introducing steady exogenous technological progress.. where output per person. However.... so part of the growth observed comes from capital accumulation.. consumption.. One explanation which has been suggested is that over the periods most often considered..........................6 Summary ...... 1.... it may be capitalaugmenting........ 5.......... there will also be growth in capital per worker.. 2005 PART 2 : EX OGENO US G RO W TH exercise will ask you to try to get an impression of the TFP growth in the East compared to the West. The basic S olow mod el cannot g enerate the longrun growth in GOP per capita that we observe in the data.. Technological progress and growth: The general Solow model © The McGrawHill Companies.......... corsumption per person.... In the basic Solow model the only source of longrun growth in economic activity is population g rowth.. one will fmd that the growth in GDP per capita comes partly from technological growth and partly from growth in capital per worker... taking education into account....... Assume that the growth process of some economy can be described fully by the steady state of the Solow model with appropriate parameter values....SorensenWhittaJacobsen: Introducing Advanced acroeconomics 152 I Part 2  Exogenous Growth 5. Technical progress may be labouraugmenting........ increasing the efficiency of labou r...... when there is (steady state) growth in GDP per worker (and hence in technology) .. or it may take the form of an increase in total factor productivity. if one conducts a growth accounting exercise. In that case there will be positive growth in GDP per worker if and only if g > 0 ... 3... controlling appropriately for differences in education.. This enables the model to generate longrun growth in G OP per capita......... taking all relevant factors into account....... the West has experienced a larger shift of the production mix from manufacturing to services than East Asia.. and the . In the long run the general Solow model converges on a steady state with balanced economic growth...Douglas form....... Without technological growth there would be no growth in GDP per worker.... It is an open question whether. Here is why. The tiger economies do indeed tend to exhibit very h igh TFP growth even after one has controlled lor education.. raising the productivity of capital. where total output..... all of these types of technological progress are equivalent and have the same implications for the evolution of the economy. 2. thus attributing total growth to various sources. East Asia has bad aboveaverage TFP growth. investment and the capital stock all grow at a rate equal to the sum of the exogenous growth rates of population and labouraugmenting productivity. or whether its high growth is simply due to aboveaverage factor accumulation. this does not identify the causes of the growth.. As long as the aggregate production function has the Cobb. However.. all growth in GDP per worker will then be rooted in the exogenous growth of the technological variable... On the other hand....... capital per person. In a causal sense. If one ends up finding that a certain country has had considerably higherTFP growth than another..............
7. one can estimate the contributions to aggregate output growth stemming from increases in factor inputs and from increases in total factor productivity. Hence the longrun economic predictions of the general Solow model accord with the basic stylized facts about longrun growth.eI SorensenWhitta. . and where the real interest rate is constant. 2005 5 TECHNOLOGICAL PROGRESS AND GROWTH 153 real wage rate all grow at the same rate as labouraugmenting productivity. Using an aggregate production function.(!5_)"1' r.apital per worker <. The general Solow model implies that structural economic policies which succeed in raising the economy' s savings rate o r in reducing its rate of population g rowth w ill gradually take the economy to a higher steady state growth path characterized by a higher level of steady state income and consumption. sin<. in a causal sense the ultimate soun. the theoretical Solow model underestimates the observed quantitative effects of these structural characteristics. Such a decomposition of the overall growth rate is called growth accounting.an only occur if there is continued growth in total factor productivity. but for reasonable parameter values the theoretical model si gnifi cant~ overestimates the rate (speed) at which economies converge. The data for countries around the world support this prediction.. Structural policies which increase the savings rate or reduce the population growth rate cannot permanently raise the economy's growth rate. Outside tr e steady state the general Solow model predicts conditional convergence: controlling for crosscountry differences in structural characteristics. Technological progress and growth: The general Solow model Growth © The McGrawHill Companies.. The general Solow model implies the same golden savings rule as the basic Solow model: longrun consumption per capita will be maximized when the savings rate equals the capital income share of GOP. Empirical data for a large sample of countries around the world confirm the steady state prediction of the Solow model. Outside steady state the general Solow model accords w ith the observation of conditional convergence. 8.Exogenous 5.hnologi<. In basic growth accounting. the lower its initial level of real GOP per worker. a country's growth rate will be higher. that real GOP per worker will tend to be higher. Still. growth in output per worker can originate from growth in the capital stock per worker or from growth in total factor productivity. for reasonable values of the capital income share of GOP.Jacobsen: Introducing Advanced Macroeconomics I Part 2. The general Solow model in continuous time In continuous time the general Solow model consists of the seven equations: . This can only be achieved via a policy that permanently raises the growth rate of factor productivity.. 4.a AL w ( K)" = (1 . Exercises Exercise 1. However. 6.. the higher the investment rate and the lower the population growth rate.e of growth is te<. but the general Solow model is silent about the factors determining technological progress.al progress. 5.a) AL A.e longrun g rowth in <..
2. Compare to the parallel expressions found in this chapter for the model in d iscrete time.a . Show that the growth rate.output ratio. Demonstrate {from the Solow d iagram) that from any initial value. g and oare t~e same as in the model in dis crete time. k 0 > 0.. For instance. k Illustrate this in a modified Solow diagram. r* and w*. the transition to steady state. Show that {it follows from the model that) the g rowth rate ink at any time is: k  . Technological progress and growth: The general Solow model Companies. Define z "' k 1 . and show that {given an initial capitaloutput ratio z 0 ) the solution is: z= (1 . It is assumed that n + g + o> 0. k*.Exogenous Growth © The McGrawHill 5.e•~ + z0 e" 1 s .. 4.a)(n + g + o).. n+g+r) ."' {1 .e. To solve the general Solow model analytically one can apply a trick as in Exercise 3 of Chapter 3 . {50) where A. K= sY(jK. Solve (50) by finding the characteristic roots. 2005 PART 2: EXOGENO US G ROW TH S = sY.(1 . in the long run (as t > =). h K/(AL) andy a Y/ (AL) . k. k"' K/ L andy "' Y/ L. w ill converge towards a specific steady state level. 3. k/y. show that the law of motion fork following from the general Solow model in continuous time can be expressed in the Solow equation: k = gj(a  (n + g + o)k Compare to {19) for the model in d iscrete time.1 (n + g + o). for the real rental rates and compare again to the d iscrete time model.Sorensen.. capital per effective worker. Note that {50) is linear. n. Compute the steady state values.. etc. .1 and g > 1 . L L = n. We use {for any point tin time) the usual definitions.. Then show that: z = (1 a)s)z. = sk". Applying methods similar to those used in Exercise 2 of Chapter 3. Assume that initially k0 < k*. of output per worker at any time t is a times the growth rate of k plus g . Also compute the steady state value and path {respectively). as well as the definitions. 1. Illustrate the above Solow equation in a Solow d iagram w ith k along the horizontal axis.Whitta. Show that z is the capital..a)s (1 . Show that the steady state value of z is z* = {1 . for capital per effective worker and income per effective worker. s. where the restrictions on the parameters a. in words. and when is the growth rate in y at its highest? 5.l~ + z0 eM. Describe and explain qualitatively. Yfy.a)s/J = sf (n + g + r5). except that we do not have to assume n > . k* and Ji*.Jacobsen: Introducing Advanced acroeconomics 154 I Part 2. how do rand w develop along the transition.
is constant in steady state and equal to: z* = ___ s__ n+ g + o+ ng This g ives the idea that the model cou ld alternatively have been analysed in terms of z r In fact.) ' .Then show. Exercise 2. we were looking for some variables that would be constant in a steady state with balanced growth. . by writing this equation for period t + 1. 2." . that: From this equation show by appropriate manipulations that the law of motion for z. 1 "' K.= 1/ z. that z.output ratio. and we analysed the model in terms of these variables. had this property. Show.. _. 1. =k./ A . Technological progress and growth: The general Solow model Companies. and then using the model's equations. and show from the results of Section 3 that indeed z.. firstorder d ifference equation. as just defined fulfils: z./k. and outside you get a the (1 / A 1) ..to make z. In steady state the capital. . Analysing the Solow model in discrete time in terms of the capitaloutput ratio {This exercise may be a useful preparation for Chapter 7.Exogenous Growth © The McGrawHill 5. 2005 5 TECHNOLOGICAL PROGRESS AND GROWTH 155 This says that at any time t. out of the middle parentheses and then utilize the fact that inside the parentheses you get y. "' K. is: (5 1) (Hint: Start by taking k. is a measure of the speed of convergence and relate to what was found for the rate of convergence (by considering a linear approximation around steady state) in Section 4 of this chapter./y.. argue that A. technically). appear).Jacobsen: Introducing Advanced Macroeconomics I Part 2. using the per capita production function. Show that z. Cl ear~. When analysing the general Solow model in this chapter. the capital . Show that the solutions for k andy are: From these expressions./Y. so k:" that you can put together w ith we d id not get solvability. .IL . z.. and therefore solvable. using the definition k.= (k. = i<:" with definitions as in this chapter..eI SorensenWhitta. We found that k1 andy. w ith the former variable carrying an increasing weight over time. the previous exercise showed that analysing the Solow model in continuous time in terms of the capitaloutput ratio had the advantage that the law of motion was a linear differential equation that could be solved. This exercise asks you to analyse the model in discrete time in terms of zP among other reasons to see whether the law of motion becomes a linear. = k. must also be constant for the steady state to exhibit balanced growth./y. (5 1) is a nonlinear d ifference equation.output ratio is the weighted average of its steady state value and its initial value.
. Compute the s1eady state values k* and ji* for k1 and ji 1 in the base scenario (all definitions are standard and as in the chapter). You have not yet been given suffic ient information to do Diagrams 2 and 3.In ji 1= A. monotone convergence to z*. what size is h? How large is h for our d irect estimates of A. Let A 0 = 1. and do Diagrams 2 and 3 . Technological progress and growth: The general Solow model Companies. Time to halve In Section 4 we arrived at the dynamic equation. (Hint: Use the solution (37) to our d ifference equation). thus c reating computed time series for relevant variables according to the alternative scenario.).In ji 0 . g = 0. Show that it follows from this dynamic equation that the number of years. wh ile all other parameters remain unchanged. is defined by the economy being in steady state at these parameter values from period zero through period 100. and establish properties of the transition equation that imply global. Exercise 3. Diagram 2 should plot In y . Do the parallel extensions of Diagrams 2 and 3. andy. and that the common approximate growth rate must be g. all parameters. as if the economy had been in the new steady state in all periods. In the following. In Diagrams 2 and 3 you should include the lines corresponding to the new steady state. from any given initial z 0 > 0. has been covered is h = . the base scenario §.Jacobsen: Introducing Advanced I Part 2. and A 0 .In 2/ ln(1 A. h. Now derive the steady state value z'" d irectly from (51). (5 = 0. against t.G' (k*). A. against t.. both according to the base scenario and accord ing to the alternative scenario.Sorensen. For realistic parameter values and the expression (35) for A.Whitta.12.12 to s' = 0. Illustrate the base scenario in three diagrams. inserted for 1 . working initially in period ten. Diagram 1 should plotji1 against timet (the points are just situated on a horizontal line). Simulate the model using appropriate parameters and initial values..? Exercise 4. n = 0 .In ji1). Sketch the transition diaQram for z . Extend Diagram 1 so that it shows the evolution in ji. This exercise asks you to do simulations and graphical rep resentations to illustrate numerically the qualitative findings. 1. and the economy is in steady state at these parameter values). are as in the base scenario fro m period zero to period nine. must be increasing at the same rate.0 1. Now construct a Diagram 4 . In ji* .(In y* .Exogenous Growth 1croeconomics 156 © The McGrawHill 5. The effects of an increase in the savings rate and convergence time In th is chapter an increase in the savings rate was analysed in the Solow diagram and in the modified Solow d iagram. and th is is independent of how large the initial gap is. that is. In .05.) Finally show that in steady state k. (For this you must establish properties as listed in Footnote 8 in Chapter 3 (p. Let the parameters (before the shift) in the general Solow model be g iven by: a = s = 0."' (1 . It follows from the chapter that the speed by which the economy adjusts from the old to the new steady state should (approximately) be measured by the rate of convergence. lines for In y.a) (n + g + J). 2005 PART 2: EXOGENO US G ROW TH 3.Y. 2. which is (34) w ith the rate of convergence A. 74). before half of an initial relative gap in year zero. ln period ten a permanent shift in the savings rate from s = 0. Explain why. and Inc. and Diagram 3 should plot Inc. The alternative scenario is defined by the economy evolving according to the base scenario from period zero through period nine (that is. and equivalent to (36). 1 . and the growth jump arising from the increase in s was described qualitatively.01 .24 occurs.
2005 5 TECHNOLOGICAL PROGRESS AND GROWTH 157 that shows the evolution in the growth rate. (same periods. . given a rate of convergence. and from there increases monotonically towards g' (here g>. In doing this."' (1 a) (n + g + J). show that just after the change. 100.. is useful). Illustrate the adjustments of y.ln y. We could . g>. = ag: + (1 . g' > 0.c1. Do all of the above once more.. 5.05 and A 0 = 1. 3. and construct a Diagram 5 that shows the evolution in the capitaloutput ratio./ Y. g. In y. s = 0. First. wh ich was low compared to the theoretical expression. both scenarios). in a (t. while the alternative scenario is given by an increase in the savings rate to 0. w ith the economy being in steady state at these parameter values in all periods.eI SorensenWhitta. Relate the halflife you have just found to the one computed from the formu la of that exercise. following from k. but now w ith g = 0.= k. The y in that equation was estimated to be 1. So.Jacobsen: Introducing Advanced Macroeconomics I Part 2. illustrate in a d iagram how the approximate growth rate g: of k 1 evolves./A 1.. Deriving an estimate of equation a from the estimation of the convergence In this chapter we derived an estimate of a from our estimation of the steady state regression equation (29)._. J = 0.k of k. Why can there be a (slight) d ifference between these halflives? 4. both according to the base and according to the alternative scenarios. g. Compare the two experiments in particular with respect to the halflives of the adjustment from the o ld to the new steady state and relate to the 'theoretical' rate of convergence. illustrate how k..7). A= (1 . etc. n. show that the approximate growth rate. and ng' is approximately zero. g = 0. From our estimation of the convergence regression equation (40) reported in (41) we derived an estimate of the rate of convergence. using A. so we had to conclude that in this respect the model did not accord perfectly well with the data. Finally.. Show that the value it first falls to is approximately (g' . s.6 .a)g~ may be useful). Exercise 5. Exercise 6. just after the change jumps up from g to ag + (1 . The effects of an increase in the rate of technological progress Assume that the economy is initially in steady state at parameter values a. 7 per cent. and then increases monotonically towards g' (here the formu la g: = g/ + g~..04.k falls to a certain negative value and then it increases monotonically back towards zero. A. 12. how are the Solow d iagram and the modified Solow d iagram affected by the change? In d iagrams with t along the horizontal axis.. and since y = a/(1 a) according to the model. b.)g'. K. A.Exogenous Growth © The McGrawHill 5. Since c1. but both n and g and g' are relatively small. Then from some period the rate of technological progress increases permanently to a new and higher constant level. since this is realistically around 5 per cent. is capital's share of GOP. it should be around ~. h. and found a value of 0.g).04 throughout.a )(n + g + m.47 (see the legend for Fig. so 1/ [(1 + n)( 1 + g')] is approximately one. In Exercise 3 you were asked to derive a formula for the halflife. From your simulations find the (approximate and rounded up) number of periods it takes forji 1 to move half the way from its old to its new steady state value. adjust over time (all definitions as usual). You should find that in the first period after the change. n = 0..) diagram. the base scenario is now g iven by a = ~. This exercise asks you to conduct a qualitative analysis of the growth effects of the increase in the rate of technological progress. g{ "' In y. Next. Technological progress and growth: The general Solow model Companies.0 1. A. the value for c1. g: stays close to the orig inal g. implied by the estimation had to be 0 . and the growth rate g. over the periods t= 1.24 as from period ten.
513 32. Exercise 7.267 26. k.6 Next use the same formu la to do growth accounting for the subperiods 1966.646 Norway Sweden Finland Year Y1 kI Y1 kl Y1 kl 1966 1978 1990 17.604 21.982 41. and average years of education in the population aged 15 or more.1978 and 1978.125 16. Use formula (45). to decompose for each country the average annual growth rate of GOP per worker over the full period 1966 to 1990 into components from growth in capital per worker and from TFP g rowth.137 32.056 19.362 15.734 26.705 18.824 33.459 24. . Comment again w ith respect to how your country has been doing.Jacobsen: Introducing Advanced acroeconomics 158 I Part 2  Exogenous Growth © The McGrawHill 5.730 16.124 21 .784 28. Some figures are substantially changed in version 6.142 31.409 14. assuming a = ~. Growth accounting: have the growth miracle economies experienced abovenormal TFP growth? Table 5. y 1.2: GOP per worker and nonresidential capital per worker.2 shows GOP per worker.849 24. 2005 PART 2: EXOGENOUS GROWTH also have derived an estimated value of a.811 21 ' 179 Ireland Belgium Netherlands Year Y1 kI Y1 kl Y1 kl 1966 1978 1990 10.6 of Penn World Table.350 18.538 15.244 27.496 45.935 28. capital per worker.Sorensen. Table 5. for some East Asian and some Western countries for 1961 and 1990.875 48.380 18.474 31. 1985 .058 7.3 shows GOP per worker.242 15.374 17.674 27.284 26.984 29.935 23.389 16.755 9. and to the value implied by the estimation of the steady state equation.341 26.173 39. Comment on how your own country has been doing (if your country is not there.152 32.793 23. Technological progress and growth: The general Solow model Companies. from our estimation of the convergence equation by using the estimated values of fJ 1 and fJ 2 and the theoretical equation (39).549 34.771 18.330 29. Find this implied value of a and compare it to the 'correct' value of around ~.660 20.895 21. for some Table 5.958 29. Exercise 8. 1 Source: Penn World Tables 5.135 20.767 Not e: This table is based on the older version 5.971 15. just pick one as your home country for th is exercise). Growth accounting: does your country keep up? y.1990 for each country. k. countries and years.248 39.792 36.773 36.Whitta.international prices USA United Kingdom Denmark Year Y1 kI Y1 kl Y1 kl 1966 1978 1990 29. and nonresidential capital per worker.
022 2. k.034 28.111 30. The World Bank Economic Review. 2001. but have they also had relatively high technological growth? We argued in Section 5 that to g ive an answer via g rowth accounting.7 15. u. u. Robert J. 2. u. u.467 80. International Data on Educational Attainment.445 56.242 35.2 9.8 8.026 4. u. east and west. This exercise asks you to do some growth accounting with education included. 2000. Sources: W illiam Easterly and Ross Levine. k. 1961 1990 2. 1961 and 1990 Japan East Hong Kong Year y.971 28.5 11 .670 4.3 8.9 5.3 6. y.560 78.0 9.225 24. 15.599 64. k.294 30.501 31.480 9. 1961 1990 17. We all know that the East Asian countries included in the table have had very high growth in GOP per person compared to the West over the period considered. y. y. 1961 1990 5. u.214 49. y. y.234 5.393 4.771 52.987 30.357 26. and k. No. S arro and JongWha Lee.601 65. k. k.0 4. A 1 • . u. 1961 1990 14.630 22.369 8.0 11.4 7.554 24.3: GOP per worker.194 5.7 6.624 7. is average years of ed ucation of people older than 15 years. y.077 26.0 West United Kingdom USA Year y. u. and years of education. u. k. capital per worker. Technological progress and growth: The general Solow model 5 TECHNOLOGICAL PROGRESS AND GROWTH 159 Table 5.431 36.755 21. are measured in 1985 . Let the aggregate production function for each country be: where we have chosen this time to express technology in terms of the labour augmenting productivity variable.827 7. Center for International Development at Harvard University (CIDl.2 Singapore South Korea Year y. as also evidenced by the numbers in the table. 'It's Not Factor Accumulation : Stylized Facts and G rowth Models'.8 Italy France Year y. one wou ld need to include education because some of the East Asian countries have had large increases in education standards. k.419 8.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2  Growth Exogenous © The McGrawHill Companies.227 7.724 16.5 Netherlands Denmark Year y.6 Note: u. u.113 80.741 7.553 5. k.international prices.2 9.7 15. 2005 5.797 27. 1961 1990 24.747 72. k. k.372 22.8 9.
5.) .Exogenous Growth 5. 5. For each of these groups. respectively? Again. Derive an appropriate growth accounting formu la that splits up the average annual growth rate of GOP per worker into components com ing from growth in A. but it has a c lose relationship to the analysis in Chapter 6 as well.Jacobsen: Introducing Advanced acroeconomics 160 I Part 2.1. still in connection w ith the emp irical success of the steady state prediction of the Solow model. 2005 PART 2: EXOGENO US G ROW TH 1. Exclude countries w ith grade 0 for data quality. Can th is growth accounting exercise. that is. where the assumed growth rate of A 1 is g. All required data can be taken from Table A at the back of Book One. assuming c1. If you wish you can. be said to give a full answer to the question raised? Exercise 9. 5 . by reasonable standards. One thing to learn about empirical analysis is that there is always reason to be critical and to ask hard questions. Comment on your findings.Sorensen. Discuss whether there seems to have been excessive technological progress in East Asia. 7.Whitta. 2. = ~. what is the true source of growth in this economy? Exercise 10. construct a figure like Fig. what w ill the average annual growth rates of y. (Exercise 1 of Chapter 6 will illustrate another standard type of critical question one can ask. that the steady state prediction of the Solow model did quite well empirically (but w ith some shortcomings concerning magnitudes of estimates). 7 and the estimation of (29) going w ith it (details reported in the figure legend). and countries that had GOP per worker below 50 per cent of the US level. Use the formula to split up the average annual growth rates of GOP per worker from 1961 to 1990 into the components for the countries shown in the table. for your own interest. 3. that is. what wou ld be the fractions of growth in GOP per worker attributed to growth in capital per worker and to g rowth in A. already do most of that exercise now. consider the same countries as in Fig. One (more) critical look at the empirical success of the Solow model's steady state prediction In this chapter we concluded. Proceed as follows.7. and growth in education. Where growth comes from according to the Solow model and according to growth accounting Assume that some economy behaves completely in accordance w ith the steady state of the Solow model presented in this chapter. based on Fig. Split these countries into two subgroups: countries that in 2000 had a GOP per worker above 50 per cent of the US level. From what you have learned in this chapter. 1/J = 0. This exercise illustrates one standard type of critical question: is the conclusion arrived at robust w ith respect to an appropriate division of the available data. k 1 and A 1 be over some period from year t to year T? What creates growth in GOP per worker? If one did a growth accounting exerc ise for this economy. Technological progress and growth: The general Solow model © The McGrawHill Companies. and estimate by OLS the steady state equation (29) for each subgroup. growth in capital per worker. find the lines of best fit in your figures.
and that this capital were accumulated in the same way as physical capital {and therefore delayed convergence similarly).o. who presented it in an article entitled 'A Contribution to the Empirics of Economic Growth'. Gregory Mankiw. pp. but a is one minus labour's share. 1 We will now explain the idea intuitively. and this stock (per worker) serves to increase the productivity of workers. If o.Exogenous Growth 6. Weil. which can be thought of as the sum of all the education and training the workers have had. In the Solow model convergence is slower the larger the elasticity of output with respect to capital. cannot be set much higher than t. Education and growth: The Solow model with human capital © The McGrawHill Companies. David Romer and David N. cp say. could be chosen freely. and the model seemed to substantially overestimate the speed by which economies converge to their steady states.. Then. However. Furthermore. David Romer and David N. It is due to the economists. suppose the income share of this form of capital were accruing to workers. around i· This suggests the idea of adding human capital to the model: each year there is a certain stock of human capital. suppose there were also another kind of capital '>\lith an output elasticity and 'income share' of its own. we could simply set o. and the idea behind it is brilliant.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2.. Wei!. The income share of human capital goes to workers. N. sutllciently large to bring A in accordance with the empirics. 107. but we identified two systematic discrepancies between the international evidence and the predictions of the model: the steady state of the model seemed to substantially underestimate how strongly savings and population growth rates influence GDP per person. 'A Contributior to the Empirics of Economic Growth'. Is there a single modification of the Solow model that can redress both of its two empirical weaknesses? Indeed there is. Consider the problem concerning the speed of convergence. N. so realL~tically o. o. 407437. total human capital is assumed to accumulate by a certain exogenous fraction of GDP being added to it each year. one would have a slower convergence since o. + cp > ~.a )(n + g + f~) . lor a maintained around ~. 2005 Chapter Education and growth The Solow model with human capital T he Solow model of Chapter 5 performed quite well by empirical standards. because human capital (embedded 1.and still a labour's share. We explained this fact in Chapter 3 and again in Chapter 5 where we found the rate of convergence to be A= ( 1 . 161 . The Quarterly Journal of Economics. Just like physical capital. Gregory Mankiw. 1992. 1 .
.g.. Thus.~... increases.. Because the output elasticity of physical capital is unchanged... What about the first problem concerning the model's steady state prediction? Assume that the rate of investment in physical capital. So the services traded on the labour market will no longer be units of 'raw' labour. so labour's share will also be unchanged... one for the services of physical capital with real rental rate.. or as investment in h uman capital.. We assume that each unit of output can be used either as consumption.~. The fraction earned by human capital goes to the workers. Since human capital accumulates like physical capital... being equal).. There are the same kinds of economic agents. ~~.. a possibility that will be considered in an exercise) . This level is assumed to be the same for all workers. Employing one more worker means hiring an additional unit oflabour equipped with h 1• The Solow model with human capital thus has two new features compared to the original Solow model: the inclusion of human capital in the production function and the accumulation of h uman capital by the consumers... = .. a . The production function of the firm is now different as it includes human capital.. The introduction of human capital sh ould thus work to mitigate both of the empirical failures of the Solow model.. e.. that goes inseparably with the worker..~~~?... adding human capital should work to overcome the second of the empirical problems.. and one for labour with real wage rate.. a representative profit maximizing Hrm and consumers (government can again be interpreted into the model.. L. There are also the same competitive markets as before. the increase in GDP per person caused by a higher investment rate in physical capital is larger in a model including human capital than in one with only physical capital (other things..~~.. ~g~~...~~.. one lor output with real price normalized to 1.... manifests itself as each worker in the labour Ioree......~}. w . The 'micro world' of the model to be presented is similar to that of the Solow models considered previously. How well a job it does in this respect is the subject of th is chapter.. Hence we do not assume th at there is a separate market for the services of human capital with a rental rate of its own.. investment in physical capital... If a certain fraction of GDP is invested in human capital. Education and growth: The Solow model with human capital © The McGrawHill Companies...Exogenous Growth 6... and we will see that it does a very fine job.~..... !... manyears. This will create higher GDP per person in steady state because more physical capital per person is accumulated.P.~. Hence the total stock of human capital. and the consumers will also have to decide on how large a fraction of income to accumulate as human capital each period..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 162 I Part 2... r 1. Hj L.... then the higher GDP first caused by the higher rate of investment in physical capital will also imply that more humnn capital per worker is accumulated.~. This is because human capital cannot be separated from workers..~~. Thus. Its services have to be sold together with labour. holding a human capital of h...~..... H. . 2005 PART 2: EXOGEN OU S GRO W T H education) is linked to workers and its services therefore sold inseparably together with work hours. the owners of physical capital will earn the same fraction of GDP.. This reinforces the increase in output. the inclusion of it will lower the model's speed of convergence. ~~. but manyears endowed with a certain level of human capital or education.?.....
growth in GDP per worker can come from three sources: more physical capital per worker. meaning that physical capital doubles to 2Kl' and each also being endowed with the same amount.)".K'. a+ <p < 1. each being supplied with the same amount. YJL~' we get the 'per capita production function ': = (5) from which. respectively. KJLt. Education and growth: The Solow model with human capital Growth 6 ED UCAT ION AND G ROW TH 163 The production function with human capital In period t the amounts of physical capital. are predetermined by past accumulation. it is easy to verify th at the income shares of the O\>Vl1ers of physical capital and of the workers are rtKJY. The (approximate) growth rate of y tis the weighted average of the growth . and using the usual dellnition of output per worker. as given in the computation of the latter) : H. taking logs and time dill'erences. K1.1h"'A1 _'"_"'L1 " = a .L. k. The production function (1) assumes constant returns to the three inputs K. 2005 6. Ht = ht L. more human capital per worker. t t t t L ( A.In y.L.L 1 into (1) gives: = (2) Profit maximization and perfect competition in the markets lor the services of physical capital and labour imply that the rental rates. so human capital doubles to 2H.. is again assumed to follow an exogenous path given by a constant rate.(1 . Lt.._ 1. h 1. Given a technological level A. one gets: = ( 6) In the Solow model \. The aggregate production li. (1) where A. the technological variable. g.8 I SorensenWhittaJacobsen: I Part 2.) as being given.)" ( ' H . Hiring one more (marginal) unit oflabour now means hiring one more unit endowed wttb the average amount.. ' w = (1 .. and human capital. a + <p<l. of human capital. for example Bl' In y . of technological progress. Hence.vith human capital. where g > . Ht and L t ' This accords with the replication argument. it should be possible to 'double production by doubling factor inputs'.a)K"fl'PA taqLa = (1 a) . Inserting H1 = h.( ' 1 ' t L ( AtLt t L (3) A.)"' A .1.)'~' r = aK".. h../Y. Hence A. H. AtLt t (4) From (3) and (4) as well as (2). of human capital per worker. = A 1p + g) ' for all t.Exogenous Introducing Advanced Macroeconomics © The McGrawHill Companies. .. r1 and wt' are given by the marginal products of Kt and L1 (taking h.K'. It follows that there are diminishing returns to the two reproducible capital inputs. and better technology. and using the usual notation for the approximate growth rates. = 1 . y. a flrm cannot Increase the input of 'raw· labour. of physical capital. = a and w 1 L.mction is assumed to be: 0<a<1. O<<p<l. When we compute the marginal product of labour we should therefore consider 11. by h iring 2Lt workers instead of Lt. (not H. Dividing on both sides of (2) by L1. that is. without increasing proportionally the input of human capital.
Omitting human capital may be a major source of error.). of income is invested in physical capital: () < SK < 1.Exogenous 6. then the implied income share of human capital would have been cp. households supplies one unit oflabour inelastically in period t.. = S. h. and supplies all accumulated physical capital as long as the real rental rate. but simply assume that the results of the consumers· decisions are that a given and constant fraction. the weights being the appropriate exponents in the CobbDouglas production function . Education and growth: The Solow model with human capital Growth © The McGrawHill Companies.C. If we can interpret the minimum wage as the compensation for labour without education. and A.. If. and hence how much to save. Again we do not consider any explicit dynamic optimization. = Y. ''\lith both of the two latter shares going to workers. where 11 > .. The intertemporal budget constraint that the choices of C.C. n. but it does give some impression oft he magnitude ofcp. Is there an empirical observation that could be used to give an impression of the magnitude of cp? Mankiw. and the other half is a return to human capital. 2005 PART 2: EXOGEN OU S GRO W T H rates of k. C. Romer and Wei! suggest one in their paper. given Y. r. perhaps somewhat larger. .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 164 I Part 2. etc. and the average wage as the compensation lor raw labour endowed with average education. Consider again the production function (1). If the share of human capital of ~ computed above can be equalized with this implied share. Note that this calibration makes h uman capital as productive as physical capital in aggregate production. 8. The income shares of physical capital.a . Given these gross investments. is: (9) If+ I:' = Y. is positive. In US manufacturing the average wage has typically been around twice the minimum wage. (or S.. The representative consumer must decide how much to consume. (10) . .. It may not give as Hrm a determination of cp as the relatively constant labour share does lor a.. this suggests that around one half of what a typical worker earns is really a compensation of raw labour. but this time also how to distribute total savings between gross investment in physical capital. human capital and raw labour would then all be around~. and gross investment in human capital. The accumulation of human capital Each of the L. then cp should be around ~ · Therefore. the evolution in the capital stocks are: (7) (R) where we have assumed that physical and human capital depreciate at the same rate. It is evident that had we assumed that there was a separate market for the services of human capital with marginal product compensation.1.cp. The empirical observation that the income share of workers is around ~ gives an estimate of a of around t. are around}.. Again the number of households is assumed to change by an exogenous rate. I~' . sK..If and I~' must obey. a realistic calibration of parameters could be that both a and cp. and hence 1 .
'H'f(A 1 L1) ~a 'l'. 2005 6 ED UCAT ION AND G ROW TH 165 and a given fraction. The model is a gel'leralization of the Solow model. g. [J. K )"('H A..c)K. = YJ(A 1 L 1) . H0 . The first is the production function repeated from (1) above. = Ktf(A 1 L 1). 6. A0 . s K' s1.. y JA. h J A. The way we define consumption here does not correspond exactly to the usual national accounts definition where consumpt ion is the excess of G OP over investment in physical capital only (and net exports). A 1 L1 (14) Kt+J. the model of Chapter 5 appears (now denoting by s K what was called s in Chapter 5). = sKY. also derived above in (3) and (4).L. (11) The complete model Since we have now described all the new model elements.. Dividing both sides of (12) by A 1 L 1 gives: = = = (19) 2. physical capital per effective worker. < 1. human capital per effective worker. the model will determine the dynamic evolution of the economy as illustrated in Fig. inserted. (18) The parameters of the model are a.1• c~. (15) H 1 + 1 .Exogenous 6. = Htf(A 1 L1) . the basic Solow model of Chapter 3 appears. (16). 1'1. K0 . If we set cp = 0 . of the state variables. we can write down the complete model consisting of the seven equations below.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2.+ 1 = (1 + n)L. f r . (16) L. respectively.H 1 = s11 Y. L0 . Education and growth: The Solow model with human capital Growth © The McGrawHill Companies. /1.1. (10) and (11). and consequently drop Eq.. The next two are the expressions for the rental rates. Y 1 = K. s H• is invested in human capital: 0 < s.c~H1 . The last two equations contain the assumptions of given exogenous rates of change in population and technology. And if we also set g = 0. cp. (17) A 1+ 1 = (1 + g)A. The following two are the capital accumulation equations th at follow from (7) and (8) above with the assumed investment behaviour.K.. and output per ellective worker. • (12) (13) r w1 = (1 a) ( 'A 1.a (Kt~ \AILt 1 ( r HtAIL.. The law of motion To analyse the Solow model with human capital it is again convenient to define 'technology adjusted' varia bles. k1 kJA . . Given initial values.
the sequence ([J . Given K0 . lzt + l .1: The dynamics of the Solow model with human capital Not e: Predetennined endogenous variables in squares. In this way the full sequences. 1 + 11 1 + g 11 Inserting into these the expression for [1 1 in (19) gives a law of motion for k1 and li 1 consisting of the following two equations: . These two values can again be inserted into (20) and (21) to computek 2 and ii 2 • etc. Then k1 can be computed from (20) and/1 1 from (21)..s 11)y 1. .c5)k1). + (1 .). endogenous variables that can adjust in the period in circles. = ( )( .] (1_a_)___ .+ 1 = s K Y1 + (1 .: Yt = (1 + g) 1 A)<~•Jit. and Ji1• and likewise lor Ji 1 • 1• Given the sequences (k 1) and (li .b)h. etc. (1_7_)_____ [. and Ht +I = sN yt + (1 . and the sequence (c1) for consumption per worker follows from C1 = (1 . [n fact. = [J 1A . L0 .s K . .). +. +. the next k1+1 depends on both k.+1 = ( 1 )( ) (s !/. (15) and (16). (20) .it + (1 .A. are determined.b)K. Education and growth: The Solow model with human capital Growth PART 2: EXOGEN OU S GRO W T H ~~I K/ +1 1 (1_6_)~ IH. H0 . providing a global stability analysis of a twodimensional system of 1!011lillear difference equations is not easy. The associated sequence (y 1) for GDP per worker follows from y.) follows from (19).o)H(' and dividing by At+l L l +I on both sides in each of them gives: kt+l = ( 1 + ll)( +g) (SI<l.O)kt). The new feature is that in each period t. (s 11k~'h t + (1 . Writing the capital accumulation equations. 1 1 h. (k 1) and (11 1).SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 166 I Part 2.(5)1!1) . This time there is no simple diagrammatic way to demonstrate global convergence to a steady state... (21) ) (sKk~'Jzt + (1 (1 + 11)(1 + g ) 1 + 1'1 1 + g This law of motion is a bit more complicated than the transition equations you have seen before. In the following we will limit ourselves to . Figure 6. kt+l = 1 1 . 2005 6. since it consists of two coupled firstorder dillerence equations (or a twodimensional system of firstorder dillerence equations). asK.Exogenous © The McGrawHill Companies. A 0 one can compute k0 = Ko/(A 0 L0 ) andfi 0 = Ht/(A 0 L0 ) .
.k.... and by considering the linearization of the system arow1d steady state to show convergence to steady state for the linearized system..9) o n + g + + ng )a/(1aq... = 1 )( ) (1+11 1+g 1!....' hf. in this case........... two particular positive values fork.....~.... which are such that if these values are ever reached.'h.. The key endogenous variables in steady state For the computation of steady state and for comparative static analysis it is convenient llrst to rewrite (20) and (2 1) as 'Solow equations' by subtracting k..... also using (19)..Exogenous © The McGrawHill Companies...2 Steady state and comparative statics in the Solow model y~~. 1 +rl 1 +g) (23) A steady state is then a positive solution ink 1 and /i 1 to the two equations that appear by setting the lefthand sides equal to zero: si7Frf.. By definition a steady state is an economically meaningful 'resting point' for the dynamic srstem considered.(n + g + <~ + ng)h........ One can compute the steady state without first having established convergence to it over time....).......~.+ 1 .....(n +g+c~+ ng)/! 1) .....(n + g + <~ + ng)k... = ( (sJ< fhf.. and li ..(11 + g + <~ + ng)k....~~. (22) 1 )( (sHk.. from both sides of the second one: k..~...........~. 2005 6....... = 0..) ( + ng s"s1a ' K' II Sl·l 1'1 + g + <~ + ng )q..j(J a......~~ .. Before doing so we will assume stability and study the steady state in detail and develop a diagram that is useful for comparative static exercises./(J a9) · (28) ...~......... = 0..'.il.}?:~... the steady state value for output per ellective worker is: y* = 51"'s' "'II )a/(1a<p)( (n + g + c) + ng ' K SK = ( 1'1 + g + <~ )q.. (24) sHi{.~P~..eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2.. 6.... Education and growth: The Solow model with human capital Growth 6 EDUCATION AND GROWTH 167 demonstrating convergence towards a steady state of the system (20) and (21) by doing a numerical simulation for realistic parameter values (and an exercise wUI ask you to do several more such simulations to check convergence)..~. the economy stays at them forever... lrom both sides of the first equation and /1 ..... (25) Indicating the solution by an asterisk one has: k* = ) 1/(1a<p) 51"'s"' H 'K ( n +g+b + Hg J1* = (2 6) ' s"s1a )1/(1a<p) 'K' II ' (n + g + c~ + 11g (2 7) from which.+ 1 .
c"' = A 1(1 + g)'(1 . while below the same curve it must be decreasing as indicated by arrows. function (demonstrate this) . Education and growth: The Solow model with human capital Growth PART 2: EXOGEN OU S GRO W T H (As an exercise. should look like a {k. so in this area k. k*. one has k. is where the two curves intersect. With the constant steady state values of the technology adjusted variables in hand. (2 5) gives all the combinations of k. Ji 1 th at follows from requiring k1 + 1  k1 = 0.s 11 )y 1.sK. so all expressions above and below are meaningful. Likewise. = 0] in Fig. li*./(1ri<p) + g + b + ng . You should check by isolating fi.2: Phase diagram for the S olow model with human capital . [t. We can draw these combinations in a k1. 6. It is also easy to see from (22) that for every combination of k1.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 168 I Part 2. 6. A. = 0]. = (1.s . 11. or in other words. 11 1 such thatk 1 stays unchanged. fi:'om !J 1 = fJ.Exogenous © The McGrawHill Companies. ever)'\\rhere to the right of r k* Figure 6. 2005 6. The first of th ese gives th e restriction on k1. > 0. L )a/(1a<p)( S )(<p/1a<p) H • (n+g +Kb+ng n +g+ b + ng (29) where A.k.li 1 such that li 1 stays unchanged..k.. realistically. (30) The phase diagram and comparative statics Consider again the two equations (24) and (2 5) from which we computed the steady state.s11) · ' < ' K ( SK 11 )a/(La<p)( + g + b + ng Sn 11 )q. We assume. it gives all the combinations ofk 1. and the curve for it. we can ilnd the steady state growth paths of the more interesting variables such as k1. I! . For instance. Likewise. that 11 + g + b > 0 . The steady state combination. + 1 . it follows that in steady state: Yi =A. the growth path for consumption per worker in steady state is: . diagram (see the curve labelled [t. do the derivations leading to the second equality).2). fi:'om (24) that indeed this curve should look like a k~ function in thek t> li 1 diagram for realistic parameter values (e. (l = <p = t)..h.2. since c. is growing according to A 0 (1 +g)'.g. Likewise. and y ..k1 = 0] in Fig. h1 above the curve [t. must be increasing.
Education and growth: The Solow model with human capital Growth © The McGrawHill Companies. shifts upwards as illustrated by (Mi. more physical capital \o\Till also be accumulated because of the constant rate of investment in physical capital. /1. s 11 • increases permanently./1. and on this boundary k. the curve [~ ii t = 01 one has from (2 3) thadi t + 1 .2. 2005 6 EDUCATION AND G ROWTH 169 r. Assume. but we emphasize that this does not follow just from the figure. decreases as indicated by arrows.2 is called a phase diagram and it is useful for comparative static analysis. so in this area increases over time. 6.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2. The (t heoreticaO possibility of such 'big jumps' is one reason why stability analysis is difficult here.3. Figure 6. the economy \o\Till be in the region where both k1 and 11 ./1 0 • like the alternative points shown in Fig.2 is drawn it suggests convergence to the steady state. 6. lor instance. This explains why. As the phase diagram has now changed.Exogenous 6. = O] in Fig. is increasing and ii. along 1. These trajectories must respect the directions of change indicated by the arrows in the different regions. the indicated trajectory. Once fi . and 1 r h*' h* k* k*' Figure 6. = 0] and [~ 11. the old steady state pointE will be situated exactly on a boundary between two regions. that the economy is initially in steady state as illustrated by the intersection pointE between the two curves [~k. 6. > 0. is decreasing.3. increase and there could be convergence to the new steady state E' along the trajectory indicated. k1. while the other parameters remain the same. t here will not be such big jumps. for realistic parameter values. has increased. Suppose then that the investment rate in human capital. Note that we are assuming that the initial jump upwards in ii.. 1 What happens is that the increased s11 in the very beginning implies more accumulation of human capital (and only that).3: An increase in the investment rate in human capital 3. increases. both k.6. increase. You will easily show that this implies that the curve corresponding to unchanged Ji. does not change while ii. Exercises will ask you to do numerical simulat ions t hat suggest that. while to the left of the curve. As Fig. but as soon as the increased stock of human capital begins to generate increases in output. so the reaction to a paramet er change is indeed as indicated by t he grey curve in Fig. 6. = 0]' in Fig.11. From initial points k0 . will move along the trajectories as illustrated.3 . . is not so big that one ends in the upper region where k.
2005 6. For s~ we will use the average gross investment rate in physical capital between 1960 and 2000. output per worker. We have good data lor y:Jo· s~ and 11. but which falls back towards g as the new steady state is approached. is now stronger than the influence of s K (and ofs 11 ). so fj. we will take data from the PWT 6. since n appears in both of the explanatory variables. y . s~" and 11 . The jump in the growth rate of y.. that the rate of population growth affects GDP per worker more strongly th an investment rates is realistic (as you will be asked to verify in an exercise). g.~) (show this). This is as we wanted and it is due to the effect described in the introduction to this chapter.ln(n + g +c))]..~) on the frrst of the explanatory variables.ln(n + g + c))]. Notice also that the inlluence ofr·1on y. = k ~'lif must be increasing.a ~ [In sK . one will be able to estimate the regression equation: (3 2) suggested by the theoretical equation (31). (3 1) The relationship between In Hl and the two explanatory variables.(). caused by the increase ins 11 is thus much like the growth jump caused by an increased sK in the general Solow model. the elasticity of y*. Education and growth: The Solow model with human capital Growth PART 2: EXOGEN OU S GRO W T H During the transition to the new steady state both k.a . If both a and ~ are around ~ · then both the coelllcients in (31) should be around one. and/1. for many countries. You may think this is a heroic assumption and we will discuss it further below. Note that the coefficient a/(1 . the latter appearing by setting ~ = 0..Exogenous © The McGrawHill Companies.1 and include countries with at least a C in data quality according to this source. The formulation of (32) assumes the same technological level A00 in year 2000 in all countries considered. An exercise will ask you to do simulations suggesting that responses to parameter changes do have the 'growth jump character·. Empirics for steady state To test the model's steady state prediction empirically first take logs on both sides of (29) and use ng "" 0: In Yi "" In A. Since [J. = y tf A. This feature. must. The reason is that increased population growth now gives a thinning out of both physical nnd human capital.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 170 I Part 2. be increasing at a rate which L~ larger than g. with respect to sK (show this). is larger than the a/( 1 a) of the general Solow model. is a linear one.In(/!+ g +c))] and [Ins 11 . which is larger in absolute value than both of the elasticities appearing directly in (31). during the transition. With data lor GDP per worker across countries i in year 2000 as well as data lor s~. increase. As before. A higher s K gives higher GDP per worker in steady state because of greater physical capital accumulation. The elasticity of y 1 with respect to (n + g + o) is (a + ~)/( 1 . the average population growth rate between 1960 and 2000. . and the relationship generalizes the one found in the general Solow model. + a 1 . but the increased GDP per worker now also implies more human capital per worker.a~ [In s 11 . reinforcing the increase in GDP per worker.a . and depreciation rate. nn sK . and for n. as well as the same technological growth rate.ln(n + g + c))] + ~ 1 .
Our proxy for s His thus 'the fraction of the school age population (12. the result of an OLS estimation of (32) is: s. including those in education. It is not easy to get data lor the fraction just described. and for such countries one might consider measuring s11 as public expenditure on education divided by GDP. an estimation of (32) based on the investment rates in human capital dellned should give correct estimates of y 1 and y 2 . 5. Our figures for sH are taken from the data set from B en S. then with the rates. The adjusted R 2 . in important respects. There are 1:\vo problems with this. as defined. and in all countries the main part of the investment in education that consists of forgone income during education is paid privately. but if a correct measure could be assumed to be a constant times the constant being the same in all countries.. 2005 6. has the GDP that is not being produced because some people are in education.59 [ln s~. + 0 .7. better than the corresponding one lor the Solow model reported in the legend for Fig. If we use this fraction as a proxy for the 'degree of education· among those who are of school age in a country. One can. This fraction. 'Is G rowth Exogenous? Taking Mankiw.79 .. we can compensate for dillerences in the compositions of populations by multiplying it by the fraction of all people of working age who are of school age. s. y 0 • by the same amount for all countries.70 + 0. we would arrive at the correct relative investment rates between countries.ln(11.9 7 [In s~1 . 199 5). more or less the fraction of the variation in GDP per worker between the countries explained by the equation. R2 = 0 .4 These rates are not correct measures with respect to their levels since they ignore direct expenditure on education. s. NB ER Working Paper 8365. s11 .55 to 4. This is how the investment rates. participated in production.0 7 5)]. For the latter we use the fraction of people between 15 and 19 to all people of working age (again because of available data). get data across countries for the fraction of all people of school age (12.7 5] (se=O.. GUrkaynak. in human capital?This is required for an estimation of (32) and hence lor testing the steady state prediction of our model. reported in Table A across countries. This estimation is. one would have a good proxy. i.0 I SorensenWhittaJacobsen: I Part 2. In many other countries much educational expenditure is paid privately. the GDP lost due to education relative to potential GDP. In some countries most education costs are paid for by local and central public authorities.. July 200 1. . l 4) + 0 .. but the fraction of people of working age in education to all people of working age should have a close relationship to it.17) attending secondary school times the fraction of the working age population that is of secondary school age (15. Romer.Exogenous Introducing Advanced Macroeconomics Growth © The McGrawHill Companies. Education and growth: The Solow model with human capital 6 ED UCAT ION AND G ROW TH 171 But how can we get data lor the investment rate.ln(n..i. are computed (averaged over the years 1960. Hence. In the regression equation (32) th is constant times y 2 would enter the intercept variable. has increased from 0. and as its numerator. The data available are such that even this latter fraction cannot be obtained directly.17) who attend secondary school. Bemanke and Relet S. l l ) (33) where standard errors are indicated in parentheses below the estimates. Considering a sample of countries for which all the described data are available (77 countries). This leads to the idea of measuring s11 by a fraction that has as its denominator the GDP that could have been produced if all people of working ages.. In y:lO = 9 . however. and Weil Seriously'. would not include direct expenditure on education. but if this can be assumed to vary proportionally to lost GDP.19)'.. 1965 . + 0 . (se= O. adj.
Our estimates of y 1 andy 2 are not exactly equal and only the estimate of y 2 is truly close to one. + 0. Since y 1 sh ould be a/(1 . 79. are of equal size and both around one. are now in much better accordance with the values we would expect based on the empirical observations that suggest an a of around }. Education and growth: © The McGrawHill The Solow model with Companies..5 as it should be. It is indeed impressive that the simple equation we are considering should explain 79 per cent of the per capita income variation in the world.Exogenous Growth Macroeconomics 172 6.5. the 0. but not along one with a slope of one. The grey line is an OLS regression line through the observations. the accordance can be made even better. 38. but taking uncertainty into account. 1.47 with the 9 5 per cent confidence interval going down to 1. Fig. Another way to illustrate the better accordance with the theory is to compute the estimates of a and <p implied by the estimates ofy 1 and y 2 . We could also h ave chosen to plot In y :>o against 0.Jacobsen: Introducing Advanced I Part 2. 5 [In s~ .ln(J·I.. y 1 and y 2. = Yt 1 + y l + y2 an d <p = Y2 1 +yl +y2 Inserting the estimated values for y 1 and y 2 gives a = 0.a We tested this in Chapter 5 by plotting In !J:>o against In s ~ . If t hey were bot h 0. Consider first the steady state equation of the Solow model without human capital arising by setting <p = 0 (still using ng ~ 0): (1 lny. Assuming a value of t lor capital's share.ln(n. the difference being the addition of the factor 0. which is in both 95 pet cent confidence intervals.) We can also give a graphical illustration of the improvement in our model obtained by including human capital. (Compute the 95 per cent confidence intervals lor a and <p. 5. 5 Note that a similar accordance between theoretical and estimated values cannot be obtained for they of the Solow model considered in Chapter 5.<p).3 say (rather than 0. The estimate was 1. a.ln(n + g + (5)].97.).[In sK. respectively. 2 3 and <p = 0. Yet.Sorensen. we have two equations in two unknowns that can easily be solved to lind: o. the plotted points should be close to a straight line with n slope of one for the model without human capital to be correct. Based on empirical observations we can only say that a and <p should be around ~ . + 0 .Whitta.075).4 shows the plot for the 77 countries for which we also have the data relevant to the model with hmnan capital. 6. 2005 human capital PART 2: EXOGEN OU S GRO W T H 0. + . The main reason is that the sizes of the estimated parameters. which are not too far from the 'correct' values of around } and t.a . one cannot reject the hypothesis considered.<p)..20 which L~ still far from the theoretical value for y of0. The black one is a straight line with the same intercept as the grey one and \>\lith a slope of one. 5. The 95 percent conl1dence intervals fory 1 and y 2 (each of which goes approximately from two standard deviations below to two standard deviations above the estimate) overlap and the interval for y 1 reaches up close to one. As mentioned. =InA. the increase in R 2 is not the main reason lor considering the present equation and estimation an improved one. but not with a slope of 0 . then both of y 1 and y2 should be 0. and (with less precision) a <p also around !. By including a bit of uncertainty w ith respect to the t rue values of a and <p.07 5)]. this implies a hypothesis that the coelllcients.59 and 0. finding a nice linear relationship.333 . 75. The flgure shows clearly that the data points may be situated along a straight line. and y 2 should be <p/ ( 1 a. .
0
I SorensenWhittaJacobs en: I Part 2 Exogenous
Introducing Advanced
Macroeconomics
Growth
6. Education and growth:
The Solow model with
human capita I
© The McGrawHill
Companies, 2005
6 EDU C A TIO N AN D G RO W TH
0
0
G
('I
173
10.5
~
Q)
""
0
~
Q;
a.
9.5
a.
0
"~
8.5
0
E
.r.
:e
(ij
Ol
0
7.5
'
6.5
4
3
2
1
0
2
~[In sK  ln(n + 0.075))
Figure 6.4: Logarithm of real GOP p er worker in 2000 against structural variables, sK and n, 77
cou ntries
Note : The figure includes all countries with available data (also for human capital) that fulfil the data quality criterion.
The estimated line is y= 9.19 + 2.42x, and has R 2 =0.66. The standard error of the estimat e of the slope is 0.22.
Source: Penn. World Table 6.1.
Doing the same kind of graphical test for the Solow model with human capital consists
of plotting In Y:m against [ln s~  ln(n; + 0.07 5)) + [In s~1  ln(n; + 0 .075)], where we
have set both coell1cients at one. which follows from a = cp = ~ . The result is shown in
Fig. 6.5 . Compared to Fig. 6.4, the points are now clustered around a grey OLS line that
has a slope that is relative[y close to one (like the black line).
We conclude that the steady state prediction of the Solow model with human capital
accords remarkably well with empirical evidence. It is a substantial matter we are trying to
exp[ain, namely why some countries are poor and some are rich. We have provided a high
degree of explanation using a lew explanatory variables, s K· s H and n. which have been
constrained to appear in our regression equation as required by our theoretical model.
The good fit of Fig. 6 .5 does not prove that the Solow model with hmnan capital is
correct. For instance. the problem of 'reversed causality' could be relevant here: is it really
s K• s 11 and n that explain y. as the model says, or is it rather a high y that tends to give rise to
high values of s K and s w and a low n? This cannot be answered from the evidence presented.
but it is reassuring that the prediction of the model is not contradicted by this evidence.
The entire discussion above assumes (among other things) a common technological
level, A00, in 2000 among the coun tries considered. It is an important detail that this
assumption does not seem to be strongly contradicted by the evidence: the points in
Fig. 6.5 do seem to be situated along one and the same straight line with a slope close to
one and this is evidence in favour of a common A 00 . It may be surprising th at in explaining
the income diflerences in the world we do not have to rely largely on technological
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
174
I Part 2 Exogenous
6. Education and growth:
The Solow model with
human capital
Growth
© The McGrawHill
Companies. 2005
PART 2: EXOGEN OU S GRO W T H
0
0
0
"'ai'
10.5
""0
3
Q;
a.
••••
•
•
9 .5
a..
Cl
C)
(ii
~
8.5
0
=
E
·~
0>
0
7.5
...J
6.5
4
3
2
1
0
2
[In .~k  ln(n + 0 075)] + (I n .~h ln(n + 0.075))
Figure 6.5: l ogarithm of real GOP per worker in 2000 again st structural variables, sx, sH and n, 77
count ries
Note: The figure includes all countries w ith available data that fulfil the data quality c riterion. The estimated line is
y= 9.53 + 0.81x, and has R 2 = 0.78. The standard deviation of the estimate of the slope is 0.049.
Source: Penn World Table 6.1. Estimates of investment rates in human capital from Bernanke and GUrkaynak (2001).
differences, but can. to a large degree, see income ditTerences as reflections of differences
in investment rates and population growth rates. We should emphasize that this result
was obtained by leaving out many of the world's poorest countries as these typically have
data quality gradeD in the PWT. Nevertheless, a rough way to state our conclusion (or
hypothesis) is that computers and education in poorer countries are as advanced as in
richer countries, there are just fewer computers and less education per capita in the poorer
countries because of lower investment rates and higher population growth rates. 6
Structural policy for steady state
Since the model's steady state prediction Hts as well as it does with the evidence. one can
feel somewhat safe in using the model for recommendations concerning longrun structural policies. With respect to policies intended to affect investment rates in physical
capital and population growth rates, the conclusions go in the same direction as in
previous chapters. The present model may be more correct with respect to the sizes of the
underlying elasticities, which is important for policy, but qualitatively the policy implications are not aHected. It is worth repeating, h owever, that according to the present model,
6. Systematic investigations of the extent to which the income gap between the poorest and the richest nations must
be explained by ' technological differences' typically find extents higher than suggested here. See, for instance,
Robert E. Hall and Charles I. Jones, 'Why do some Countries Produce so Much More Output per Worker than
O thers?', Quarterly Journal of Economics, 114, 1999, pp. 83 116.
0
I SorensenWhittaJacobsen: I Part 2 Exogenous
Introducing Advanced
Macroeconomics
Growth
6. Education and growth:
The Solow model with
human capital
© The McGrawHill
Companies. 2005
6 ED UCAT ION AND G ROW TH
175
the elasticity of GDP per person with respect to population growth (or rather ton+ g +b)
is larger in absolute value than each of the elasticities with respect to the investment rates.
SK and SJ1.
It is with respect to investment in human capital that the model may have new policy
implications. First note that according to (29), steady state GDP per worker is unambiguously increasing in the investment rates 11• This means that more investment in education
may have desirable ellects. but not under all circumstances. First, given the model's other
parameters, the 'golden rule' values for s K and s H that will maximize the altitude of the
growth path for consumption per worker ares K =a and s11 = cp (an exercise will ask you to
show this). Raising the investment rates above these values is not meaningful.
Second, a public policy to provide for or subsidize education should be justified by
specilic reasons why leaving educational decisions to private individuals cannot be
expected to yield optimal outcomes. In the case of education there are many such reasons.
One is imperfect credit markets. If education were a completely private matter, people who
would like to educate themselves would have to borrow money against their expected
higher future labour incomes to provide lor the costs of education and to compensate for
the forgone income during education. However, there are strict limits to how much private
banks will lend to a (poor) person based on this person's expected higher future income.
Another reason is imperfect insurance markets. More education does not yield a
specillc increase in future income with certainty, but rather with some probability. A decision on education may fail. so connected to education is an insurance problem. Few
private insurance companies are prepared to sell insurance policies that pay out money in
cases of failed education.
There are probably also strong positive externalities associated '>\lith education. Part
of the return on an investment in education goes to other people. For instance. an
engineer in a private firm is more productive if the other employees in the llrm can read
and write and do calculations. This means that part of the social return to the education
invested in these employees appears as higher productivity of the engineer and is not
appropriated by those who acquired the education. If left to private decisions, people will
therefore probably educate too little as compared to a social optimum.
You may come up with additional reasons lor public policies to support education
(one could be imperfect private knowledge about the private and social returns to education) . In any case. the reasons stated together with the model of this chapter should be sufllcient for the conclusion that the issues of how and to what extent to provide lor and
subsidize education are among the most important concerns of public policy.
6.3
.g?.~Y..~.~.~~~~~..~.~ ·~·~·~··§.~.~?.~..~.~~.~.~..~.~~.~..~~~~.~.. ~.~P..~~.~~·~·······"""""'
The steady state of the Solow model with h uman capital accords in all respects with positive balanced growth if g > 0 . It is left to you as an exercise to demonstrate this in detail.
but note that since k,. 11, and g, are all locked at constant values in steady state. k,, h, and
y , must all be increasing at the same rate as A,. that is. at rate g.
7. Some readers may find this section d ifficult and it is possible to understand the remainder of the book without
reading th1s section. Until the subsection 'The Convergence Process ... ' there should be no difficulties, however.
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
176
I Part 2 Exogenous
Growth
6. Education and growth:
The Solow model with
human capital
© The McGrawHill
Companies. 2005
PART 2: EXOGEN O US GRO W T H
The Solow model with human capital thus passes the 'longrun empirical growth
check', but this time we have not shown that the steady state equilibrium is stable. Taking
stability for granted (for the time being). as the economy converges to steady state there
will be transitory growth in addition to the underlying technological growth , and the
transitory growth seems to be more complicated than in the general Solow model since
there are now two state variables. k, and i1 ,. in the dynamic system. For instance, outside
steady state one variable could be below and the other one above its steady state value.
By conducting a numerical simulation exercise with the system (20) and (21). we will
start by indicating that this system does imply convergence to steady state. We will then
study the linear approximation of the system around steady state. Although this is a
twodimensional linear dillerence equation ink, and i1 1• it turns out that it implies a onedimensional linear difierence equation in [J 1, which has the same form as the corresponding
equation for the general Solow model. This has three implications. First, the convergence
process in terms of [J 1 is like the one we have studied for the general Solow model: the
model with h uman capital also supports conditional convergence. Second, we can show
analytically that the linearized system is stable towards steady state. Third. we can test the
convergence equation empirically and we will lind that the rate of convergence predicted
by the model is now in much better accordance with empirical evidence than was the case
for the general Solow model.
Illustrating convergence to steady state by a simulation
Let us specif)r some plausible parameter values: a = <p = f,s K = 0.2, s 11 = 0.15, f5 = 0.06. n = 0
and g = 0.015 . The steady state values k* and li* are easily computed from (26) and (2 7):
k* = 14.22 and/i* = 10.67. We consider a starting point where one of the state variables is
above its steady state value and the other below: k0 = 16 and li 0 = 2. Starting from these
values we simulate the system (20) and (21) Vlrith the chosen parameter values. The result
is shown in Fig. 6.6, a kt> fi , diagram where each dot corresponds to one period. The figure
also has the two black curves [~k 1 = 0] and [~li , = 0]. which makes it a phase diagram.
The sequence of dots in Fig. 6.6 converges to the steady state. An exercise will ask you
to do simulations from other starting points and using other parameter combinations.
You should lind convergence to steady state. We may now reveal that to construct
Fig. 6.2 we actually simulated the model with a lternative starting points and the same
parameter values as considered here, so Fig. 6.2 contributes to the impression of general
convergence and it suggests the type of picture you are supposed to lind when you attempt
the exercise.
The convergence process according to the Solow model with human capital
To obtain a linear approximation of the system (20) and (21 ), first write it in the generic
form:
k,+ 1 = G(kJi,).
fi,+1 = L(k,, i11),
where the functions G and L are defined as the righ thand sides of (20) and (21),
0
I SorensenWhittaJacobsen: I Part 2 Exogenous
Introducing Advanced
Macroeconomics
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
Growth
6 EDUCATION AND GROWTH
177
ii,
18
16
r
14
12
[~k, = 0)
ii*
[~ii, = O)
10
8
6
4
.,
2
I
__t
I•
0
I
I
I
k_• •
5
0
10
20
15
k,
Figure 6.6: Simulation of convergence in the Solow model with human capital
respectively. Linearizing around steady state turns the first of these equations into:
k, + 1  k* = Gk(k*. ii*)(k I  k*) + Gll(k* ,Ji*)(/1 I  /i*),
where Gk and G11 are the partial derivatives ofG with respect tok, and 11,. respectively, and
similarly for the other equation. If we use k 1+ 1  k* ""i<*Hn k 1+1  Ink*) etc. Oust as in the
previous chapter). we get that. up to a linear approximntwn around steady state. the system
(20) and (21) is equivalent to:
•
•
• •
Ji*
 •
ln k,+ 1  lnk* = Gk(k*,l!*)(ln k, ln k*) +=: G11(k*,fl*)(ln h, ln h*),
k"'
.

k*
 


 


ln ill+ 1  ln I!* = ::: Lk(k*, h*)(ln k,  ln k*) + L11(k*, h*)(ln h, In h*).
h"'
(34)
(35)
This is a twodimensional linear system of diflerence equations in the relative discrepancies (from steady state): ln k 1  ln k* and ln 11 1  In li*. The coell1cients in the system
can be computed from the partial derivatives of G and L. Doing the dillerentiations and
using [J* = (k*)"(li *)"' gives:
where we have used R to denote 1/[(1 + n)(1 +g)]. From the steady state expressions (2 6)
to (28) we get:
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
178
I Part 2 Exogenous
I © The McGrawHill
6. Education and growth:
The Solow model with
human capital
Growth
Companies. 2005
PART 2: EXOGEN OU S GRO W T H
n+ n+ b+
.ti*
k* =
nn
SK
n + g + c)+ ng
[J*
Ji* =
I~* = :: .
k*
(
)
Inserting into (34) and (35) then gives the linearized system:
In k1+ 1  Ink* = R[a(n + g + c)+ ng) + (1  b)](ln k1  Ink*)
+ RqJ(n + g + b + ng)(lnli 1  ln ii*),
o
ln/1 1+ 1  lnii* = Ra(n + g+ + ng)(lnk1  lnk*)
+R[q;(n + g + O+ ng) + (1  b)](lnii, lnli*).
It turns out that this system implies a specillc onedimensional and linear dillerence equa1f, it follows that in
tion in In [J 1  In y*, to which we will limit our attention. From y1 =kf1
any period, In [J 1 = a Ink1 + <P In II,. Hence, multiplying the tlrst of the above two dynamic
equations by (1, and the second by qJ, and then adding the two gives:
In [!1+ 1  In y* = R[(a + <P)(n + g + <5 + ng) + (1  o)]a(ln k1  Ink*)
+R[(a + <P)(n + g +<5+ ng) +(l  c5)]q;(ln ii 1  lnii*),
and then, moving the common factor on the righthand side outside the bracket we get:
ln[J 1+ 1  In[!* = R[(a + q>)(n+ g+ c5 + ng) + (1  <5)](1nj),  ln[J*).
(36)
This is a llrstorder linear difference equation in In [J ,  In [J*. The coefflcient on the righthand side is positive and you will easily (and should) verify that it is smaller than one if
a+ <P < 1, which is among our assumptions. Hence (36) implies stability, so according to
it [J , ~ !i* as t ~ oo. You will recognize that (3 6) has exactly the same structure as (3 3) in
Chapter 5.
Subtracting In [! 1 and adding In [J* on both sides of (36). and after that inserting for R
gives:
ln[/1+ 1  ln[/ 1 = { 1  R[(a + q>)(n + g +c) + ng) + (1  b)])(ln[J*  In[!)
(1 + n)(1 + g)  (a + q>)(n + g + b + ng)  (1  b)
_*
_
=
(
)(
)
(lny  ln y,),
1+n 1+g
which, USing (l + n)(1 +g)  (l  0) =II + g + <5 + 119 (check) gives:
_
_
lny1+ 1  lny1 =
(1  a q>)(n + g+o + ng)
(
)(
)
l+n l+g
_*
_
(lny  lnuJ
(37)
Just as in Chapter 5, we have arrived at a dynamic equation of the form In [J 1 + 1  In fj 1 =
A.(ln fj *  In [J 1) . now with the rate of convergence:
A=
(1  a  q>)(n + g +c)+ ng)
.
'
l1 + n)(1 + g)
'
"" (1  a  <P)(n + g + b),
(38)
where 0 <A< l (since A. is one minus the coefficient we found above to be between zero
0
I
SorensenWhitta Jacobsen:
Introducing Advanced
I
Part 2 Exogenous
6. Education and growth:
© The McGrawHill
Growth
The Solow model with
Companies. 2005
Macroeconomics
human capital
6 ED UCAT ION AND G ROW TH
179
and one). Consequently our dynamic equation, (3 7), has the convergence property, that
t he further below steady state the economy is. the larger is the growth rate of fJ ,. and
hence of y ,. The model th us supports conditional convergence.
In Chapter 5 we found the solution to the dynamic equation, In fJ 1 + 1  In fJ, =
il(ln [J*  In [J ,) after having written it in the standard form. In [J , + 1 = il ln [J* + (1  il)ln fj,
(equation (3 6) of Chapter 5) and applied a cookbook solution procedure. The solution we
found was (3R) of Chapter 5, repeated here:
lnAr  lnA 0 1  (1  il)r
*
T
+
T
(In A 0 + In [I  In y 0) .
Inserting the steady state value y* from (29) and (In Ar  ln A0 )/T ~ g we get:
In Yr ln Yo _
1  (1 = g+
T
T
+
+
AY
lnA 0 
1  (1  il) r
T
lny 0
1  (1  il)r
a
T
1  a  q.>
1  (1  AY
9'
'1'
1  o.  q.>
[In SK  In(rr + o + <5+ ng)]
On
S 1. 
'
ln(n + g+
o+ ng)].
(39)
This is the convergence equation for the Solow model with human capital. It tells how
growth should depend on the structural characteristics and on the initial position. Below
we will test it empirically, but note already now from (38) that we have obtained a lower
rate of convergence than before. (1  a  q.>)(n + g + c~) instead of (1  a )(n + g + c5).
Realistic parameter values, e.g. n + g +c) between 0 .06 to 0.075 and a and 9' both around
now imply a rate of convergence between 2 per cent and 2.5 per cent, which should be
compared to the 4 to 5 per cent we obtained from the general Solow model. The modelbased rate of convergence has thus come much closer to our earlier empirical estimates.
but these were not rooted in the present model. We should now do an estimation based on
the new convergence equation (39) and see how well the equation describes the data and
how well the estimate of il, implied by the estimation, fits with the modelbased values
between 0.02 and 0.025.
t.
Testing the convergence equation and estimating the rate of convergence
The theoretical equation (3 9) suggests a regression equation across countries i :
O~· o = {3 0

fJ 1 In Yi1 + {3 1 [ln s~ 
ln(ni + 0.075)] + {3 3 [ln s~1  ln(ni + 0 .075)],
(40)
where the coelllcients fJ 1• fJ 2 and fJ 3 should be linked to the parameters of the model as
given by (39). and. in particular. fJ 0 should depend on the underlying rate of tech nological
progress, g, and the initia l level, A 0 , as in (3 9). Hence. (40) assumes this latter technological influence on growth to be the same lor a ll countries.
An OLS estimation of (40) based on SO coun tries with available data and at least the
grade C in data quality according to the PWT gives:H
8. We have a few more countries than for t he steady state equation since we have included some countries for which
we do not have the G OP per worker in :2000 (so t hey were not in t he steady state estimation), but for which we have
data for, e.g ., 1999, and w ithout much loss of precision we used 9~. 60 for 9~0• 60 •
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
180
I Part 2 Exogenous
6. Education and growth:
The Solow model with
human capital
Growth
© The McGrawHill
Companies. 2005
PART 2: EXOGENOUS GROWT H
f/:,
0 . 60
= 0.118  0 .011 lny:,+ 0.014 [ln s~  ln(lli+0. 075)]
(se =0.001)
(se =0 .003)
+ 0 .012 [ln s\1  ln(d + 0 .075)], adj . R2 = 0.49.
(;e=O.OO.!)
(41)
There are several nice features about this estimation. First note that the adjusted R 2
is 0.49, so apparently the simple equation explains around half of the crosscountry
variation in growth rates. This is pretty good, and better than the 40 per cent we arrived
at in the general Solow model. However, the increase in R2 is not the main reason for
considering this estimation a much improved one. The main reason is that the estimated
coefllcients are now in better accordance with the underlying model.
First note that according to the theory, still assuming that (J. and qy are of equal size, the
two coelllcients on the square bracket terms should also be of equal size. They are indeed
estimated to be roughly equal. particularly if one takes uncertainty into account. Furthermore. we can derive the magnitudes of 1, a, and qy from the estimation. The parameter {3 1
should be (1  (1  J..)7)/T. implying (as in Chapter 5)).. = 1  (1  T{3 1) L/r. Inserting T = 40,
andf3 1 = 0.011, gives).. = 0.014. Hence, our estimate of the (annual) rate of convergence is
now 1.4 per cent. The 9 5 per cent confidence interval for).. ranges from 0.8 per cent up to
2.3 per cent (compute this yourself). Hence. including uncertainty, our estimate of the rate
of convergence is now in relatively good accordance with the 'theoretical' 2 to 2.5 per cent
arising from the present model's rate of convergence. (1  (J.  qy)(n + g +b). Remember
from Chapter S that the estimated ).. of 0. 7 per cent from the general Solow model (with
confidence interval from 0 .3 per cent to 1.1 per cent) did not accord 'vith the 'theoretical' 4
to 5 per cent arL~ing from that model's rate, (1  (J.)(n + g +b).
It is left to you as an exercise to lind a way to derive (1 and qy from our estimation. and
to ilnd the implied magnitudes. but the solution is a = 0.38 and cp = 0.32. Again. these
values are in nice accordance with the assumed values based on empirical observations.
Figure 6. 7 illustrates the gradual improvement of the explanation of growth and
convergence we have achieved. All three panels include the 80 countries considered in
this section. The top panel tests (again) unconditional convergence and simply plots
average ammal growth in GDP per worker from 19 60 to 2000 against the log of GDP per
worker in 19 60. It is thus very similar to Fig. 2. 5, but considers the smaller sample of
countries. The conclusion is also the same as in Chapter 2: there is no evidence that the
iuilially poorest countries grow Lhe fastest uncurulilicmally.
The middle panel illustrates conditional convergence according to the general Solow
model. It is close to, but not the same as, Fig. 2.6 and Fig. 5.8. Based on an estimation of
the convergence equation of the general Solow model. Fig. 5.8 plotted the 'growth rate
controlled for structural characteristics' a:,0 _60  0 .020[ln si  ln(ni + 0.075)] against
In y~0 for 90 countries. The middle panel in Fig. 6. 7 does almost the same for the 80
countries we are now considering. but plots the 'growth rate controlled for structural
characteristics' (defined the same way) plus the avernge of structural influel!ce over all the
countries against ln 0 . The dillerence lies in the addition of the average. which makes the
scales along the vertical axes in the panels of Fig. 6. 7 comparable.
The bottom panel does for the Solow model with human capital what the middle
panel does for the general Solow model. that is, it plots the growth rate a:x>.60 minus the
structural influence. 0 .014[ln s~  ln(ni + 0.075)] + 0.012[ln s~1  ln(ni + 0 .075)]. plus
the average of these structural influences over all the countries against In y~O'
y:,
0
I SorensenWhittaJacobsen: I Part 2 Exogenous
Introducing Advanced
Macroeconomics
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
Growth
6 ED UCATION AND G ROWTH
:;;
0
""'
0.06
3
Q;
•
00
C)';'
_o 0 .02
O<D
.,m
• • ••
~~
~
0 .00
•
0
0
• ••
•
a.o 0.04
(LO
 0.02
7
6
~·
·~
181
......
•••
•••
1. ••• .t_ ,l~·~~
\
•
•••••
:
•
9
8
:
•
10
11
10
11
In Yso
ti
""'
0
3
:;;
0.06
•
a.o 0.04
0..0
oo
eJ<;'
of6
.,m
0 .02
~
0 .00
~
~
0
•
 0 .02
7
6
•
..
9
8
In Yso
.:
Q)
""'
0
3
:;;
0.06
• ••
a.o 0.04
o..O
•
00
eJ<;'
_o
0(!)
.,m
0 .02
..
•
~~
~
0 .00
0
0
 0 .02
6
7
9
8
•
.~
10
11
ln rso
Figure 6 .7: Convergence, unconditional, conditional on sK and n, and cond itional on sK, sH and n,
80 countries
Source: Penn World Table 6.1. Estimates of the savings rates in human capital from Bernanke and GUrkaynak
(2001).
In the two lowest panels the data points do seem to be clustered around the negatively
sloped straight lines. and mostly so in the bottom panel. This panel is evidence that countries that start off poor tend to grow faster (structural characteristics being equal).
Particularly. in so far as the points in the bottom panel can be said to be 'represented' by
one and the same straight line, it gives supporting evidence that what above was called
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
182
I Part 2 Exogenous
Growth
6. Education and growth:
The Solow model with
human capital
© The McGrawHill
Companies. 2005
PART 2: EXOGEN OU S GRO W T H
the technological inlluence on growth, is roughly the same in all countries. This is a
striking conclusion, but remember again that we have excluded many of the world's
poorest countries from our sample.
Discussing the Solow model with human capital and
looking ahead
................................................................................................................................................................................
We may conclude that the Solow model with h uman capital does remarkably well
empirically. If we can assume that the countries of the world were close to their steady
states in the year 2000 then. according to the Solow model with human capital, their
levels ofGDP per person in 2000 should be relatively well explained by their technological
levels in 2000. their investment rates in physical and human capital and their population
growth rates in a way that obeys Eq. (31). We have found that indeed they are, even
under a strong assumption that all countries have the same 'year 2000 technological
level'.
Furthermore, still according to the Solow model with human capital. the growth
rates (of GDP per worker) of the countries over some period, 1960 to 2000 say, should be
explained by the countries· technological levels in 1960, their growth rates oftechnology,
the countries' investment and population growth rates, and their levels ofGDP per worker
in 1960 in a way that obeys Eq. (39). embodying conditional convergence. We have
found that grovvth rates across countries are explained relatively well this way, even
assuming the same influence of the technological variables in all countries. In particular,
this means that we have found support for the idea of conditional convergence: countries
with similar characteristics with respect to investment and population growth rates do
seem to converge to relatively close growth paths.
The good lit of the Solow model with human capital is reassuring for the use of the
model as a tool for understanding the world and for making economic policy recommendations. But it does not mean that there is not more to do.
Although the fit of the model is good. it is not perfect. This is evidenced by Fig. 6.5 and
the bottom panel of Fig. 6. 7. In both figures the data points do seem to be clustered around
the appropriate straight lines, but in a rather cloudlike way. For instance, if one were to
roughly draw two other lines in the bottom panel of Fig. 6. 7, one above and one below the
indicated line and both parallel to it, in such a way th at 'most of the cloud' was to be
situated between the lines. then the 'th ickness' of the area between the two lines would
have to correspond to no less than around two percentage points of economic growth per
year. A dillerence of two percen tage points in the annual growth rate of GDP per worker is
large. The convergence equation of the Solow model with human capital may do a good
part of the job of explain ing growth differences between countries, but it certain ly does not
explain everything. A good explanation of the remaining growth differences would be
more than welcome.
Furthermore. although we may conclude from the model studied in this chapter that
the rates of investment in physical and human capital and of population growth are
important for prosperity and growth. the model is completely silent about how these rates
are determined. We will not go into explanations of these basic rates at the moment,
0
I SorensenWhittaJacobsen: I Part 2 Introducing Advanced
Macroeconomics
Growth
Exogenous
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
6 EDUCATION AND GROWTH
183
but there is economic theory that seeks to provide such explanations. Book Two of this
volume contains chapters that present deeper explanations of investment and saving
behaviour.
Finally. according to our model the technological variables A 0 and g are very
important for longrun prosperity and growth. In the long run the altitude of the (steady
state) growth path that the economy evolves along depends on A0 , and the growth rate
along the path is equal to g. We may conclude from the analysis of this chapter that A 0 and
g seem to be more equal across countries than most people would probably expect a priori,
but still A0 and g are exogenous and thus unexplained. Theories to explain g, that is,
growth models with an endogenous rate of technological progress, would be highly
interesting. Part 3 presents such theories.
Thus, what we have been doing up to now naturally leads further to the theories of
endogenous growth dealt with in Part 3. Before turning to endogenous growth. however,
there is one important aspect of real world growth that can fruitfully be studied in a model
of exogenous growth. namely the growth consequences of scarce natural resources such
as land or oil. As we will see, the dependence of the economy on natural resources may
have a growth hampering ellect in the long run. It is a natural idea to confront this effect
with the growth promoting ellect of technological progress. that is, to study a model with
both natural resources and exogenous technological progress and investigate which effect
is strongest. This is a central theme in our next ch apter.
6.5
§.~~.~.~!:¥....................................................................................................................................................
1. The quality and skills of the labour force are of obvious importance in human production. In
this chapter we amended the Solow model to inc lude human capital, defined as the stock of
skills accumulated by workers through education and training. The stock of human capital
was included among the essential inputs besides physical capital and 'raw' labour.
2. In some respects human capital was treated just like physical capital: we assumed that in
each period there is a certain stock,
H,. of human capital, that human capital enters into an
aggregate production function w ith constant returns to scale, Y,
= K;'Hy(A,L,) 1  "  "',
parallel
to the way physical capital enters, and that human capital accumulates by a certain fraction,
sH, of GOP being added to it each year. While empirical estimates of investment rates in
physical capital are d irectly available from national accounts, nothing similar is true for Sw We
argued that since the most important part of investment in human capital is the GOP and
income fo rgone while people educate themselves, the fraction of people of working age in
education should g ive a good approximation of Sw Using this idea, empirical estimates of sH
can be obtained across countries.
3. In one important respect human capital must be treated d ifferently from physical capital: since
human capital is embodied in persons, a firm cannot change its input of human capital and of
(raw) labour independently. We made the simplifying assumption that each worker is
equipped with the same amount of human capital, so each worker embodies a human capital
of h, = H,/L,, and hiring one worker means increasing raw labour input by one unit and
increasing the input of human capital by h,. Hence there is no separate market for the services
SorensenWhittaJacobsen:
Introducing Advanced
lacroeconomics
184
I Part 2 Exogenous
Growth
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
PART 2: EXOGENO US GROWTH
of human capital: these are traded together w ith raw labour, and the return to human capital
appears as a part of the real wage which is thus really a mix of a reward to raw labour and a
return to human capital. Furthermore, in the computation of the marginal product of labour that
determines the real wage, one should take human capital per worker, not total human capital,
as given.
4. Our considerations on how to deal with human cap ital led to a model characterized by the following features:
e
The aggregate production function was as stated above. From empirical evidence on
labour's share and on average wages compared to the wages of unskilled workers, we
inferred that plausible parameter values could be a.
e
e
= cp = ~·
The prices of capital and labour were given by the marginal products.
The stocks of physical and human capital accumulated as implied by given investment
rates, sK and sH, respectively, and a common depreciation rate.
e
The labour force as well as the technological variable A to grew at exogenously g iven rates .
5. The model implied convergence to a welldefined steady state in wh ich GOP per worker
follows a specific growth path w ith constant growth rate (equal to the exogenous growth rate
of
A,). The 'steady state equation' told how the growth path depends on model parameters.
Accord ing to the steady state equation:
•
An increase in sK has a larger positive impact on GOP per worker than in the general
Solow model because the rise in GOP occurring as more physical capital is accumulated
will cause more human capital to be accumulated, reinfo rcing the increase in GOP.
e
An increase in the population growth rate has a stronger negative impact on GOP per
worker than was the case in the general Solow model, because population growth now
means thinning out of both physical and human capital.
•
An increase in sH has a positive impact on GOP per worker similar to that of an increase
in
SK .
These features were promising for an empirical test since they appeared to remedy the
empirical shortcomings of the general Solow model's steady state prediction that we had
identified.
6. Confronting our model's steady state prediction w ith crosscountry empirical evidence under
the heroic assumptions that the countries considered were in steady state in year 2000 and
all had the same technolog ical level in that year, we found that the model d id remarkably well
and actually went a long way towards solving the empirical problems with the general Solow
model's steady state prediction. This was evidence that countries are perhaps not as technologically d ifferent as usually thought.
7. The steady state analysis supported policy recommendations of ensuring adequate savings
and investment rates and keeping population growth rates down, the latter now being even
more strongly supported than earlier. The really new issue was government polic ies to provide
for or subsidize education. Such policies are recommendable according to our model in so far
as one cannot fully trust private decision making and markets to devote a suffic ient amount of
resources to educational purposes. We gave several reasons why a marketdetermined
allocation would not be optimal, implying that educational policies should be at the forefront
of structural policies to promote prosperity and growth (as indeed they are in many countries).
0
I SorensenWhittaJacobsen: I Part 2 Exogenous
Introducing Advanced
Macroeconomics
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
Growth
6 EDUCATION AND GROWTH
185
8. The general prediction of economic performance derived from the Solow model w ith human
capital is the convergence process implied by the model. From the model one can derive a
'convergence equation' telling how a country's average annual growth rate of GOP per worker
over some period should depend on parameters and on the country's initial GOP per worker.
The structure of the equation is such that the growth rate equals the underlying technological
growth rate adjusted upwards for how far below steady state the country was initially, and for
the speed by wh ich the country converges to steady state. The rate of convergence depends
in a simple way on model parameters and (ceteris paribus) it is smaller for the Solow model
with human capital than for the general Solow model. The reason is that capital accumulation
takes time and delays convergence, and there is more capital to accumulate when there is
also human capital. This was again promising from an empirical point of view since we have
found the general Solow model to overestimate the rate of convergence.
9. Estimating the model's convergence equation on crosscountry data gave a remarkably good
result and in particular a much improved accordance between the estimated rate of convergence and the one predicted by the model. The good result was strongly in support of the
idea of conditional convergence among the countries of the world and gave an indication that
countries are perhaps also more similar w ith respect to their technological growth rates than
usually thought.
10. We closed this chapter w ith a discussion of where to go fro m here. We concluded that a main
further job is to explain the technological growth rate that is the source of longrun economic
growth. This points further to theories of endogenous growth.
6.6
Exercises
Exercise 1. One more empirical improvement obtained by adding human
capital
We opened this chapter by summarizing two main empirical shortcomings of the general
Solow model. One was that its steady state prediction tends to understate the influences of
saving and population growth rates on GOP per person. There may be another one that also
concerns the general Solow model's steady state prediction which is (setting ng ~ 0):
In
a
Y7 =In A,+  [In SKln(n + g + o)].
1 a
The elasticity of
y7
with respect to
s K is the same as the elasticity with respect to n + g + (5
in absolute value. Show that both are a/(1 a), and hence should be around 0 .5 . To check
empirically whether the two elasticities are approximately equal, construct two figures, taking
data from Table A (the figures are reminiscent of, but not the same as, Figs 3 .7 and 3.8). One
figure should plot In y~0 against In s~ across countries i leaving out countries with a 0 in data
quality. Estimate (by OLS) and draw the line of best fit through the points and identify its
slope. The other figure should plot In y~0 against ln(ni + 0.075), and include the line of best fit
through the points. The slope of the line should be identified. Do the two slopes seem to be
of equal size (in absolute value), or does one seem to be greater than the other? In case of the
latter, which of the estimated elasticities is largest? How do your findings accord with the
SorensenWhittaJacobsen:
Introducing Advanced
I Part 2 Exogenous
acroeconomics
186
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
Growth
PART 2: EXOGENO US GROWTH
general Solow model's steady state prediction. How does it accord w ith that of the Solow
model w ith human capital?
Another way to proceed is to do an O LS estimation of the regression equation:
In Y~o =
K +y In s~ + v ln(n; + 0 .0 75),
where we have not restricted the coeffic ients of Ins~ and ln(n; + 0.075) to be equal. If you
know how to do an OLS estimation w ith more than one explanatory variable on the righthand
side, do an estimation of the above equation and compare your estimates of y and v. Are they
of about equal size in absolute value (taking uncertainty on the estimates into account)? If one
is larger than the other, how does that accord w ith each of the two models?
Exercise 2. Government in the Solow model with human capital
When we were setting up the basic Solow model in Chapter 3 we were careful to explain that
the savings rate, s, in S, = sY, could be interpreted as s = (1 c P){1  c9), where we now
assume
if= 0,
and constant values for the private propensity to consume out of d isposable
income, cP, and for the spending share, c9, of government that balanced its budget each
period. The model cou ld thus be interpreted to include a government and obviously a gov
ernment could be interpreted into the general Solow model considered in Chapter 5 in exactly
the same way. This exercise asks you to show how government can be included in the model
of this chapter.
Assume that the government always taxes all income by the constant rater, and in each
period spends !he constant fraction sf, of the tax revenue on education and the remaining tax
revenue on government consumption (thus balancing its budget each period and not investing in physical capital). What the government spends on education accumulates as human
capital with the households.
Assume that out of their disposable incomes households invest the fractions~ in physical
capital and the fraction sf.J in human capital. In all remaining respects the model is as in this
chapter. Show that the model presented in th is chapter can be interpreted to cover the case
with a government if the parameters of the chapter are: sK
= (1 
r) s~ and sH = rs~ + (1  r)sf.J.
Exercise 3. Balanced growth in steady state
Show that growth in the steady state of the Solow model w ith human capital in a// respects
accords w ith the concept of balanced growth. In particular, derive an expression for the
r*, and show that th is is constant. Also compare the expression
in the general Solow model. How is it possible that r* does not
steady state real rental rate,
to the corresponding one
depend on sH ?This is d ifficult to explain intuitively, but we only ask you to try to identify some
counteracting factors that together can imply that
r * is not sensitive to Sw
Exercise 4. Golden rule
Show that, g iven the values of the model's other parameters, the sizes of the investment rates,
sK and sH, that w ill maximize the altitude of the growth path of consumption per worker in the
steady state of the Solow model with human capital are sK =a. and sH = cp.
0
I SorensenWhittaJacobsen: I Part 2 Exogenous
Introducing Advanced
Macroeconomics
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
Growth
6 EDUCATION AND GROWTH
187
Exercise 5. Further simulations to investigate stability
In Section 3 we ran a simulation of the Solow model w ith human capital to check stability of
k,. h,
towards the steady state
k*, h*. We only considered one starting
point,k0
= 16, h0 = 2,
which resulted in convergence to steady state, but in an interesting way that included some
nonmonotonic ity
(k,
first decreases, then increases). Run some simulations yourself where
(first) you consider the same parameter values as in Section 3, but you choose d ifferent initial
points. Make d iagrams like Fig. 6 .6 including the two curves corresponding to constant state
variables. Be sure that you have initial points in each of the figure's four reg ions (preferably
several in each reg ion wh ich are spread out in a 'representative' way) and that you simulate
over a sufficient number of periods to be able to judge whether there seems to be convergence to steady state. Choose some changed, but still realistic parameter values (obeying all
stated parameter restrictions), and do the same again. Do your simulations seem to support
convergence to steady state generally?
Exercise 6. The effects of various parameter changes
Let a
base scenario be defined by the economy being, in all periods from 0 to 200 say, in
steady state of the Solow model with human capital at parameter values c1. = (/) = ~. s K = 0.1 2,
sH = 0.2, c5 = 0 .055, n = 0, g = 0.02, and w ith an initial value of the technological variable A0 = 1.
We will consider various alternative scenarios, each one defined by the economy
evolving according to the base scenario from period zero through period nine, while (working
first time) in period ten a permanent shift in one (or several) parameters occurs, while other
unmentioned parameters stay unchanged.
For each alternative scenario the qualitative effects of the parameter shift should be
described and illustrated by showing how it affects the phase diagram and by explaining
intuitively the transition towards the new steady state. The quantitative effects should be computed and illustrated by conducting the appropriate simulations and creating the following
figures:
e
A Diagram 1 showing the evolution of In y, against periods t = 0, 1, ... 200 according to (1)
the base scenario (the old steady state), (2) the new steady state (as if the economy had,
in all periods, been in steady state at the new parameter values and w ith A 0 = 1) and (3) the
alternative scenario.
e
A Diagram 2 showing In c 1 the way Diagram 1 illustrates In y,.
• A Diagram 3 showing the growth rate
gf"'
(y, y,_,)/yt1 over periods t= 1, ..., 200,
otherwise as Diagram 1.
• A Diagram 4 showing
K,!Y, and H,/Y,, otherwise as D.agram 1.
Alternative 1 is defined by the investment rate in physical capital increasing to s~ = 0 .24.
In particular, what is the half life for ji 1 (the number of periods it takes for ji1 to go half the way
from the old to the new steady state)? Compare to the half life you found in Exercise 4 of
Chapter 5, and explain the difference.
Alternative 2 is defined by the investment rate in human capital increasing to s'r; = 0.25.
From your two analyses comment on the development perspectives from increasing the
investment rates, in particular w ith respect to the duration of the consumption sacrifice.
Alternative 3 is defined by the investment rate in physical capital increasing to s~ = 0 .24,
and the investment rate in human capital decreasing to s~ = 0. 15.
DoP.~
thP. P.r.onomy :=~lw:=~y~
~P.P.m
to
rP.:=~r.t
to
r::~r::~mP.tP.r r.h::~ngP.~
with
::~
'growth jumr'?
SorensenWhittaJacobsen:
Introducing Advanced
I Part 2 Exogenous
acroeconomics
188
© The McGrawHill
Companies. 2005
6. Education and growth:
The Solow model with
human capital
Growth
PART 2 : EXOGENO US GROWTH
Exercise 7. Inferring a and
qJ
from the estimation of the convergence equation
First find a way to estimate values for a. and cp from the information gathered in (39), (40) and
(41 ). Then show that the implied values are a= 0.38 and cp= 0.32. Find the 95 per cent
confidence intervals for the estimates of a and cp, and comment.
Exercise 8. Illustrating further the improvement of the convergence equation
obtained by adding human capital
In this chapter we created Figs 6.4 and 6.5 to illustrate the improvement in the steady state
equation of the model with human capital over that of the model without human capital. A line
of best fit should have a slope around one for the model to be correct. The line in the model
with human capital had such a slope, but the model without human capital did not. This
exercise asks you to make a similar illustration with respect to the convergence equations.
You should consider the same 77 countries considered in Figs 6.4 and 6.5. Data can be
taken from Table A at the back of Book One.
1. First consider the convergence equation of the general Solow model:
9{o ~ V
1 ( 1 
1 (1 ))T
))T
T
lny0 +
T
0.
·   [lnsK In(n+9+ o) ],
1 a
where we have denoted the 'technological influence' by V and set n9 = 0. Assuming as usual
a=~ and n + 9 +
o= 0.075 (and remembering A,= (1  a )(n + 9 + o)), what are the 'theoretical'
/J 1 = (1  (1  J)T)jT and /J 2 ={31a./(1  a), in this equation if
T = 40? In a figure, plot 9~0, 60 against P1 In y~0 + /J 2 [1n s~  ln(n; + 0.075)] with fJ 1 and {3 2 so
computed and check if the line of best fit seems to have a slope of one.
sizes of the two coefficients,
2. Then consider the convergence equation of the Solow model with human capital:
9~ 0 ~ V
1 (1 A) T
·
+
T
lny0 +
1 ( 1 ))T
cp
T
1 a.cp
1  (1A) T
a
T
1  a  cp
[lnsK In(n + 9+ o)]
[In SH In(n + 9 + o)].
o
Assuming a= q. =~ and n + 9 + = 0.075 (and remembering )= (1  a  cp)(n + 9 + o)), compute
the 'theoretical' sizes of the three coefficients {3 1, /J 2 and {33 in the equation given T = 40. In a
figure, plot 9~, 60 against P1 In / 60 + {J 2 [1n s~  ln(n; + 0.075)] +{J3 [1n s~  ln(n; + 0.075)] and
check if the line of best fit seems to have a slope of one.
Exercise 9. An alternative Solow model with human capital
You may have been wondering why we have built human capital into the model of this chapter
in a way that is different from how it was brought into the picture in Chapter 5, Section 5 on
growth accounting. In fact, there is an alternative growth model with human capital that
incorporates human capital in the way it was done in Chapter 5.
This alternative model with human capital consists of the same equations as the general
Solow model, that is, Eq. (7)  (13) of Chapter 5, with the following exceptions: the production
function (7) is changed to:
Y, = K;'(A,hL,) 1  " ,
.ln(n+g + o + ng)] + 1/JU. and k. Education and growth: The Solow model with human capital Growth 6 EDUCATION AND GROWTH 189 where: h = exp(!fJu)..::~st C for wh ir.1 ui. The 'rental rate equal to marginal product' conditions (8) and (9) should be modified appropriately.ln(n. etc.5[1n s i .h data on s i. 6.[In s.::~lly ::~s follows. Estimate again the line of best fit and see if it has come c loser to having a slope of one. s )"/(1<t) exp(!fJ u). are output per worker./L. 2005 6. = i< .(In Yo .( 1. By linearizing the above law of motion around steady state. Show that k. This is almost what we did in Fig. ConsirlP. 1a . The 'steady state equation' derived from the above is: a lny7 = In A 1 + .. k. Show that Y. converge to constant steady state values and that in steady state output per worker evolves as: y~ =A. Nevertheless. Next p lot In y~ against 0 .. is simply a constant. . how this constant depends on education as measured by u may be important. show that the 'convergence equation' of this model is: lny . and ji.r r. and ui are available from Table A. =y.' + (1 . and 1/J is a given productivity parameter in the production of human capital and u is the average number of years that people in the labour force have been to school (the same in all periods).A)r Sf g + '::::': T x (In A 0 + ___!2_ [Ins ln(n + g + (5 + ng)].). Assuming a =~.ln(ni + 0. Estimate the line of best fit and judge if it seems to have slope one. 1 + n 1 +g 2. but not for exactly the same countries./(hA 1) and K./(hA . + 0 .Exogenous Introducing Advanced Macroeconomics © The McGrawHill Companies. but rather is simply a given constant throughout.mri rir.).o)k.' and that the law of motion for k1 is: k. h. but are not used in the following. Is the improvement of the steady state prediction comparable to the one obtained in the model of this chapter? 4.4.lny T T 0 1. where y. 1 a TP. Consider first the model without human capital (1/J = 0).1/Ju)). = ( 1  )(  ) ( sk ./L. 1. plot In y~0 against 0. n .. respectively. ( n+g + (~ + ng 3. respectively. y.ountriP.075)) across coun tries i.s i w ith rl::~t::~ qu::~lity ::~t IP. The difference between this alternative model and the one considered in !his chapter is that in the alternative model human capital per worker is not being built up gradually through an accumulation process where additions come from investing a given frac1ion of GDP each year..075)) + 0 .st this P. y. and capital per worker.0 I SorensenWhittaJacobsen: I Part 2. Note that in this model human capital per worker. 1 (see again Section 5 of Chapter 5). Both 1/J and u are positive and constant parameters and to be thought of as rates like a and s.5[1n si . Realistically 1/J should be around 0. The full model is conveniently analysed in terms of the auxiliary variables i(..
Find the elasticities of y7 with respect to sK.a)(n + g + o).. (1 . O): lny7 . r.. which data do you think are most reliable.. + 0. respectively. An empirical shortcoming of the Solow model with human capital? Exercise 1 abo'le pointed to an{other) empirical shortcoming of the Solow model without human capital. The steady state prediction of the Solow model w ith human capital is (setting ng s. . and v accord with what the elasticities should be? Take the uncertainty of your estimates into account when you answer. Taking data fro m Table A and leaving out countries with 0 in data quality. that can be derived from the estimated parameters? By introducing human capital. or ln(n.SorensenWhittaJacobsen: Introducing Advanced lacroeconomics 190 I Part 2. sH and n + g + o. Let's now take this latter model through a similar direct ct>allenge.075).075)]. In s~. The model's prediction that the investment rate in physical capital and the population growth rates should affect GOP per worker equally strongly (but in opposite directions) does not seem to hold true empirically. those for u. + 0.075). For instance. Rather. 60 =Po  fJ .(nA 1 + a 1 a.1 u) + {J 2 [1n s . Estimate this by OLS. A re!=jression equation SU!=J!=Jested by this conver!=Jence equation is: 9~ 0.+ 0. while the elasticity with respect to n + g + (5 should be twice as large in absolute value. for realistic parameter values) and the estimated rates as compared to what was found in the general Solow model? Is the improvement comparable to the one obtained in the chapter's model with human capital? Why not? 5.t.~ [lnsKIn(n+g+())] + ~ 1 a. 2005 6. used in the model of this exercise or those for s~ used in this chapter? Which of the models do you think has the nicest properties? Is there a trade off with respect to which model to prefer? Exercise 10. How do your estimates of y. Are the estimated parameters significant and do they have the right signs? What are the values for a and A. where we have not imposed any modelbased restrictions on the coefficients to Ins~.~ [lnsH. does this model bring a better accordance between the model's rate of convergence {an evaluation of A.In(n +g + (5)] . do an OLS estimation of the following regression equation (requires that you can do an OLS estimation with more than one explanatory variable on the righthand side): lny~ 0 = K+y In s~ + c In s~ +v ln(n. a feature indeed found in the Solow model with human capital.0. taking data from Table A.ln(n.(In Y~o.. Discuss the advantages and disadvantages of the two models with human capital. there seems to be a stronger influence from the population growth rate than from the investment rate.Exogenous © The McGrawHill Companies.. Education and growth: The Solow model with human capital Growth PART 2: EXOGENO US GROWTH where . Show that if a = ~ = ~ (as we more or less believe) then the elasticities with respect to sK and sH should be of equal size and both should be one..
labour and land. today land is still in fixed supply and remains an important input to aggregate production. I SorensenWhittaJacobsen: I Part 2. Suppose further that land is in llxed supply. Asswne now.it should be clear that 'land' is vital for economic activity. that there are three inputs in the aggregate production function: capital. Assume further that the economy manages to build up capital at the same speed as the population increases. However. so it is no wonder that the economists of the time considered land to be essential. Because of constnnt returns to cnpitnl and labour. leaving output per worker constant (possibly at a high level).Exogenous Introducing Advanced Macroeconomics Growth 7. Moreover. Adam Smith. if we interpret the services of 'land' in a broad sense to include all the life support services of the natural environment . From the replication argument this production function should have constant returns to capital. implying diminishing returns to the combinntion ofcnpital nnd labour. In a nutshell. They thought that the presence of land as a fixed factor in production implied a severe constraint on the economy's longterm growth potential. In modern terms their argument was the following. The theory of economic growth was initiated by the great classical economists. this is the classical argument why income per capita has to decline in the long run as a consequence of 191 . Not only is agriculture still of importance. Assume there is some population growth and disregard technological progress for a moment. output will grow less than proportionally to labour and capitaL and hence output per worker will decline in the long run. The analysis of the classical economists taught them that the presence of an irreplacable factor in fixed supply would imply a tendency towards longrun decline in income per capita and eventually stagnation at a low income level. Note that this L~ exactly what we found in the basic Solow model's steady state: capital and labour were growing at the same rate (the capital. while land stays fixed. but any kind of production requires at least some space and hence land. labour and land. In their time agriculture accounted for a much greater share in total production than today. as the inputs of capital and labour increase proportionally. Thomas Malthus and David Ricardo. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. Now. this implied that output was also growing at this rate.labour ratio was constant).such as its ability to generate clean air and water and to absorb the waste products of human activity . realistically. 2005 Chapter :Eimits to growth? The Solow model with scarce natural resources S o far our presentation of growth theory has been a bit like 'playing Hamlet without the Prince of Denmark'. as we assume they do.
technological growth is a source of longrun growth in income per capita. This would eliminate population growth. A slight acquaintance with numbers will show the immensity of the first power in comparison of the second . 1 59. 2005 PART 2: EXOGENOUS GROWTH population growth. Nordhaus. 1 tal<ing up a discussion about the limits to growth provoked by the pessimistic views expressed in some famous books from 19 72 and 199 2 known as the 'Limits to Growth' reports. The possibilities of prolonged technological progress and birth control was not emphasized by the classical economists. As income per capita falls. Brockings Papers on Economic Activity. In the words ofMalthus: Population. land. the modelling of thls chapter will show that the classical economists were in some sense righ t: land does imply a growth drag such that population growth leads to negative growth in income per capita in tile afJsence of tecill'lological progress. 2. Furthermore. population growth would also eventually be brought to a halt. its effects are . 2 Like Nordhaus. but because of the miserable living conditions of workers. pp.. In fact. . W illiam 0 . The economic history of the last 200 years suggests that the pessimism of the classical economists was unwarranted. but in the 'end state' income per capita would have stagnated at a subsistence level. This will enable us to take a stand on the essential issue whether fixed land or technical progress has the strongest influence on longrun growth. this book predicted that economic growth and population growth would be brought to an abrupt halt some time in the present century.vo countervailing influences on longrun growth . The ultimate root of this decline L~ the diminishing returns to capital and labour arising from the presence of a fixed factor. The model we consider comes from an article by the economL~t William D.rith exhaustible natural resources. 1798). misery and vice. we will go one step further and also present a model v. An Essay on Population.Exogenous Growth 7. or as a consequence of an ecological breakdown arising from man·made pollution. (from Chapter I of Thomas Malthus' famous book. Income per capita has been growing at rates around 2 per cent per year over very long periods in many Western countries. with decreasing incomes. 'Lethal Model 2: The Limits to Growth Revisited'. 'land·.. As we saw in Chapter 5. This would imply an even faster economic decline. In neo·Malthusian fashion. the supply of which stays unchanged as it is 1. mortality would be so high that the labour force would stay constant. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. either as a result of acute shortages of food or raw materials. 2. only that there were perhaps some other factors to which they did not pay enough attention. This implies a strong and constantly operating check on population from the difficulty of subsistence .. Which one will be the strongest in the long run is a crucial issue and a central theme in this chapter. Among mankind. Fertility might stay unchanged. This is not even the end of the story according to the classical economists.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 192 I Part 2. the original source of declining incomes. as an input in the aggregate production function . If a fixed factor such as land. The labelling of economics as the 'dismal science' arose from this idea. Fixed natural resources and technological progress seem to be a.. We will study a Solow growth model that also includes a fixed factor. and in Chapter 6 as well.. The most wellknown of these reports was The Limits to Growth published in 1972 by a group of scientists gathered in the Club of Rome. Thls does not mean that the classical economists were wrong in their reasoning. savings per capita would also fall. Nordhaus.. 1992. Subsistence increases only in an arithmetical ratio. when unchecked. so capital would not really be able to increase at the same speed as labour in the long run. increases in a geometrical ratio.
Equation (2) is the capital accumulation equation. (3) (4) .K) is less th an one. O<b<l. measures how much the returns to capital and labour diminish . its services could be interpreted broadly to include all the life support services of the natural environment. The competitive market lor the services of land implies. by adjustment of v 1. The degree to which a + fJ is below one. and g reflects the rate of labouraugmenting tech nological progress. and the amount of land is flxed and does not change with its use in production. Ll' and for the labouraugmenting elllciency factor. (1) The input X of land does not carry a time subscript since land is in fixed supply. 0. labour and land . and also to the combination of any two factors. one should expect that exhaustible resources such as oil and gas. assuming given rates of gross investment in physical capital (now disregarding human capital) and of depreciation: K1+ 1 = sY1 + (1 . A 1. x: > 0. (2) Finally. profltmaximizing ilrm will now include three inputs: capital. Hence there will be diminishing returns to any single factor. The model Equation (1) is a Cobb. wage income and rent from land.Exogenous Introducing Advanced Macroeconomics Growth © The McGrawHill Companies. {3 and K: a > 0 ..1 The Solow model with land The production function of the representative.Douglas aggregate production function with land and exogenous. respectively. We should note that although the llxed factor is referred to as 'land' for convenience. since a + {3 (= 1 . A1+ 1 = (1 + g)AI' n. O. The consumers own the land (as well as the capital and labour). We will again use the model to address the essential question: is the growth drag so strong that it will ultimately bring growth in income per capita to a halt despite continued technological progress? 7.. We assume that both of these rates are at least zero: Lt+ 1 = (1 + n)L1. Eqs (3) and (4) assume given growth rates for the labour force. The consumers as a group (or consumers and government together) save a fraction s of total income which is the sum of interest earned on capital. Consumers sell the services of capital. 0 <s < 1. labour and land to the firm in competitive markets where the real factor prices are denoted by r 1• w 1 and vi' respectively. a doubling of both K 1 and L1 \viii imply that output.o)K1.as it should according to the replication argument. Our model will show that this intuition is indeed correct. which disappear as they are used. can imply even more of a growth drag. constant technical coefficients a. 2005 7. Limits to growth? The Solow model with scarce natural resources 7 LIM ITS TO G RO W TH ? 193 used in production.{3 > 0. Thus 11 is the growth rate of the labour Ioree. Yl' is multiplied by 2 "+P. which is less than two. I SorensenWhittaJacobsen: I Part 2. or the size of K. The production function exhibits constant returns to the three inputs capital. can imply a drag on growth. that all land supplied is also demanded.. labour and land. In particu lar.+ {) + K = l.
SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 194 I Part 2.K. that the dynamics of Eqs (1)(4) are independent of the real factor prices.. let us llrst see what can be said about the sources of economic growth from the production function alone. = a . the broader interpretation of X remains valid.In Y t1' etc. Eq. The dynamics of the Solow growth models we have considered earlier were such that the economy converged towards a steady state in w hich the capital. If KrfY1 converges towards a constant steady state level./Y. Note. so lor given initial values the four equations above will determine the full time sequence for the state variables and lor Y. each factor will earn its marginal product which depends on the predetermined levels of K" L" A 1 and X: 3 . etc.)". 2005 PART 2: EXOGENO US G RO W TH In each period the values of the state variables. where land is of no importance. must fulnl: or= (9) Equation (8) shows that an increase in income per worker can now be obtained (only) by an increase in capital per worker.. and you can easily verify that in the special case of K = 0 .. one gets: = = = (8) and hence. Let us assume for a moment that the dynamics of the Solow model with land has the same implication.L1 1 = (5) 1 1 ~ K' )" ( ' . in each period. by better technology. ol'· is (approximately) the weighted average of g~. so must 3. f:J and K..A.output ratio was constant. the model boils down to the Solow model of Chapter 5. In the introduction and in this section we have opted for a broad interpretation of 'land' to include nature's ability to absorb waste. Defin ing output per worker asy 1 Y. In connection w ith a market for the services of land a more literal interpretation is appropriate since nature's ability to absorb waste is typic ally not owned or marketed by private people. Limits to growth? The Solow model with scarce natural resources Growth © The McGrawHill Companies. so the model is again one of constant income shares. etc.(X)" AL 1 1. capital per worker as k 1 K j L. g and . are predetermined. You can also easily show that the income shares of capitaL labour and land (r. (K. K.)" Au ( AILI AtLt (6) K(_!S_)"(_!i__)"J AL A L. however. or by an increase in land per worker..) are a. (9) shows that the approximate growth rate of output per worker. The competitive factor markets imply that. so as long as we include only Eqs (1) (4) in our model../L 1.n.Exogenous 7. .. 1 1 1 These three latter equations complete the description of the model. and A. with weights a . and land per worker as x 1 X/L" and dividing on both sides of (1) by L. respectively. by taking logs and time diflerences. L. the approximate growth rates In y 1 . With our assumptions the latter ell'ect can only give a decline in growth since the supply of land is constant and we have assumed that population growth is at least zero. fJ and K . (7) W 1 ={:J V . In fact. Some production function arithmetic As with the other Solow models we have analysed.
+ (1 . We have thus arrived at a transition equation for z.output ratio in steady state (and constant growth rates for technology and the labour Ioree).o) 1 ( (1 + n)(1 + g) z1 la A~Px~" )1 a k aAilx• 1 1 )1a z 1 ' '' where for the last equality we used the expression in (11) for z. It was derived from the production function and an assumption of a constant capital. We will do exactly that in the next subsection.6) k.. y.())z. = k~'A~x:. Convergence to steady state You may have the idea that we s hould (as usual) analyse the model in terms of technology adjusted Yariables. However. Inserting this equality into (9) above and solving lor gf shows that gf converges towards a constant gY fullllling: #g .. +1 1 =((1 + n)(1 )il[(s+(1 . and where we have exploited the fact that a + f3 + K = 1. K. we get z.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2.f5)k 1) )p(s!J 1 1 +(1 . ( (1 + n)L. and then the steady state growth rates of capital per worker and of output per worker must be identical. z. 2005 7. it cannot possibly converge towards a constant value.a (J+K (10) We have already arrived at an expression lor the steady state growth rate in output per worker.. = X/(A. k.).+ LA. In other words. ' (1 + n)L.•. or: = (11) Leading this expression for z. Limits to growth? The Solow model with scarce natural resources Growth 7 LIM ITS TO G RO W TH ? 195 k jy .k. If the assumption we built on in the previous subsection is correct. sincex. is one such variable./y...1a (J . )p(s + (1 ./A./A.b)K.. which does not interest us. +l .=k.= .< _ z. +g) (12) . For a proper analysis we should now demonstrate convergence of the capital.E . = ( (1 + = ____!_!_!_ K ) n~(l + g) 1 ( (1 + n)(1 + g) r (sy 1 + (1 . the capital. most conveniently written as: z. It is not entirely correct to say that this was derived only from 'production function arithmetic' . [J../(k~'A~xn.=yJA. except in the case g = n = 0. = k.output ratio to a constant steady state level..)] H z..Kll j3g . one period and inserting appropriate model equations gives: _. = k../Y.Kll gY.)l a( 1 + g) fJ A P( X )1( . We will therefore analyse the dynamics of our model directly in terms of z 1• From the per capita production function .+ lx r+ l   = ( Lt+ l 1 " (J _ _ A t+ I ( X )" Lt+ I sY. and x. g~ = gf. Having established a llrm foundation for our expression. we will discuss it at length.Exogenous © The McGrawHill Companies. = kjy 1.output ratio.L. =x. 1. We are looking for variables that could converge towards constant steady state values.
and convince oneself by 'staircase iteration' that z. the slope of the 45° line.output ratio. but here t he procedure suggested is the simplest. We show this by solving the transition equation (12) for z. where a . ' {J + K {J + K (14) An exercise at the end of this chapter will ask you to find the exact growth rate of y . Above we found the expression (10) for the common value of the approximate growth rates.Exogenous 7. from the appropriate equations in Section 3 of Chapter 5).c~)z)]la{=} +g) . It has exactly one strictly positive intersection with the 45° line. = 0. in steady state and show that this is indeed close to the expression in (14).J. to a particular and constant steady state level. (13) where we have used 1 . It is everywhere strictly increasing. we get: dzt+l ( 1 T. The steady state growth rate Summarizing. g ~ 0 . 0) which is strictly larger than one. One can now sketch the transition diagram for z1 as it must look given the four properties. 4 By differentiation of (12) with respect to z. . 2. and the inequality follows from n ~ 0. = 0.output ratio z. + 1 = z. We could alternatively show that the slope at z* is smaller than one (do thai and explain why t his is also sufficient). 0). 4. z = z"' = [(1 + n)(1 + g)]sPf(P+~<l. This means that the growth rates of k.<~) > 0. = z.o)[(s + (1 . 2005 PART 2: EXOGENO US G RO W TH To show that this transition equation implies convergence of the capital.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 196 I Part 2. z*. Limits to growth? The Solow model with scarce natural resources Growth © The McGrawHill Companies. z. It has a slope at (0.(1 .g . indeed converges to z* in the long run from any initial z0 > 0 (do this) .a)(1 . the steady state capital. It passes through (0. We have thus found a unique value. The same exercise will ask you to verif)r that the steady state of the Solow model with land meets the requirements for balanced growth formulated in Chapter 2. This goes to infinity as z1 goes to zero because of the presence of z~·. z*. This follows directly from (12) setting z. = k 1/y 1 converges towards the constant steady state level z*. the same as that in the Solow model Oust compute k'. In the particular case of K.n.. and c~ < 1.(~)z .1 ] . This gives: zl"=((1 + 11)(1 1 )p[(s+ (1 .o)z.W« z~ + [(s + (1 . This value was: fJ I( gil "" . we have established that the capital.a = {J + K../y'. 4. = (1 + n)(1 + g) X )P [(1 .1 < 0 . This also follows directly from inspection of (12 ). it suffices to demonstrate the following four properties of the transition curve: 1. z0 .. from any S[fictly positive initial level. 3.)] H a z~.. for the intersection with the 45° line.output ratio z* becomes s/(n + g + o + ng). and y 1 must converge towards the same rate.
then 0 < gY <g.Exogenous Introducing Advanced Macroeconomics Growth 7. and hence savings per worker. but will it. which was equal to the rate g of labouraugmenting technological progress. while capital. the condition . which should be no surprise. and the more important land is {the larger K is).vth will be negative according to (14) . would be approximately 0 .r~ils~cl ilmnnnts n f ~n~r. g > 0. an unfortunate state of all'airs much emphasized by the classical economists. There will be growth in income per worker. As land becomes more and more scarce relative to labour (and capital). Is this likely to be fulfilled? A lower bound for labour's share. or: (15) It is illuminating to compare the steady state growth rate gY in output per worker that we have arrived at here to that of the real Solow model. The classical economists thought that decreasing wages would ultimately bring population growth to a stop. realistically. 11 > 0. K.. Again this reflects the diminishing returns to capital and labour which imply that income per worker. 11 = 0. (J. economic grm. Working people would be poor while the landlords would live rich and idle lives. the approximate growth rates of Y. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. be enough to outweigh the negative influence from population growth? The condition for strictly positive growth in GDP per worker is (Jg > Kn. the gY in (14) is smaller than g (whenever g > 0 or n > 0). while the real rent on land will be increasing at the same rate as total output..tiv~ li!hnnr i!ncl r. = k. which would leave 0. Technological progress does have a positive influence on growth in income per capita according to (14). L. diminishing returns will imply that income per worker will decline. but the growth rate will be smaller than the rate of labouraugmenting technological progress. When that happened.i!piti! l w ill rr~ss nn t h ~ fix~cl ilmmmt nf li!ncl. in steady state are both close to gY + 11. . A. it also implies a drag on growth in the sense that technological progress is less ellective in creating economic growth than it would have been if land had no importance lor aggregate production. However. through diminishing returns to capital and effective labour. As you will be asked to show in an exercise. (the 'modern case'). the more g11 will fall below g. In the 'classical case' where there is no technological progress. but at a somewhat lower rate. will increase at a somewhat lower rate than g (see again (15)). Note from (15) that the input of capital will not be growing at the same rate as labour. The growing input oflabour will inevitably press on the limited amount orland. the growth rate gY in (14) also tal<es the value g. in an absolutely worst case. w hen land has some importance in production so that K> 0. g = 0. So.. that income per worker cannot keep pace with labouraugmenting technological progress. If there is no population growth. the real wage will fall by the same rate as output per worker. increases at rate g. and total capital is K. Y. In the particular case of K = 0. We can conclude that the classics were right on their own assumptions and perhaps more right than they thought. I SorensenWhittaJacobsen: I Part 2. implying.4 to the sum of capital's share a and land's share K . 2005 7 LIM ITS TO G RO W TH ? 197 Since total output is Y.6. At the very most. Not only does land imply a drag on growth in the presence of population growth. K could thus be close to 0. but some population growth.L. wages would be at a subsiscence level while the rent on land would be at a maximum. and K.4.. = y. but there is some technological growth. cannot keep pace with the labour force. Technological progress means that the ejfective labour input. The in r. L.
the calculations above show th at the classical pessimism may be sadly relevant for many developing countries that have not got population growth under control.6 and a = K = 0.output ratio. as stated in (8) . The fact that GDP per capita has grown at annual rates of around 2 per cent in many Western countries may suggest that technological growth rates (our Model's g) have been somewhat above 2 per cent per year. y 1 = k. and ~ per cent is well below empirical estimates of longrun annual growth rates for labouraugmenting technological progress in Western countries. We can see this as follows. then gYwould be 1. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. a larger amount of land. that if a country has little control over population growth and experiences annual values for 11 of around 3 per cent (which is the case in some developing countries.2 for land's share. which is a decent longrun growth rate and certain ly far from zero. This leaves another 0. The steady state balanced growth path One may get the impression that to have much land is bad. Note.Exogenous Growth 7. above 1 per cent would be needed just to overcome the negative influence of population growth. say. an annual growth rate of labouraugmenting technological progress of ~ per cent would sull'ice for positive longrun growth in income per worker according to our model. The amount ofland does not affect the &rowtl! rate in output per worker along the steady state growth path. and rates of technological growth . In steady state the capital. g = 2 per cent and n = 0. For given inputs of other factors. which is between plausible and high by Western standards. 5 per cent. in principle.'Afx~. but it does allect the level of this path. and a capital share of 0. right. If we consider more realistic parameter values such as {3 = 0. z. We have just derived a worst case condition for longrun economic growth. Again.(! )n.'A~(X/LY'. divide both sides by y~ . Isn't there anything good about an abundance ofland? A glance at the production function will reveal that there is. 5 we get g1' = (i)g . \<\Till imply higher output and average income. see Table A). then the drag on growth caused by population growth could be as large as ~ of a percentage point. .output ratio on the righthand side. In the article by Nordhaus cited earlier in this chapter (see footnote 1). X..375 percent per year. on the other h and. g.2 are considered realistic. a labour share of 0. Two hundred years of economic history in the West have shown th at they were not suH!ciently aware of the possibilities lor longrun technological progress and for birth control not caused by 'misery and vice'.2 . thereby getting y:a = z.6. our analysis suggests that the classics were. To get the capital. Start again from the per capita production function. which could be typical lor a Western economy. Land implies a drag on growth (compared to a llctitious situation where land is not important) and more so the more important land is. With an annual population growth rate of 1 per cent. 2005 PART 2: EXOGENO US G RO W TH lor positive economic growth would beg> (~)H. However. Hence.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 198 I Part 2. so the steady state growth path for y. is (remembering that 1 .a = f3+K): (16) 5. if the annual rates are. is constant and always equal to the z* of (13) .
38 ln(X'/L~4) .ln .075)] + 0. Strictly speaking. (18) suggests the following regression equation across countries i : In yi. 7. here also using ng ~ 0 . lny7.ln(n + g + (5). We can test these predictions empirically using data for the value of natural resources. we can insert A. 61. 2005 7. Furthermore (16) shows that in any given year steady state income per worker. ~ . L~4 is the labour Ioree in 1994 from the PWT. With our usual shortcuts. the growth rate along this path is alfected by both g and 11.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2.Inz". In fact. while s.71 + 1. Yi .ln(n + g + b)] + aInK In A1 + {3 . and 11.ln(n. Limits to growth? The Solow model with scarce natural resources 7 LIM ITS TO G RO W TH ? 199 If we want to trace the growth path back to parameters (including those entering z*) and initial values. and. in which case In z* ~Ins . + K.2 say. we may crudely approximate the exponentP/({3 + K) to be 1. = y0 + y 1[ln s. .6..ln(11 + g + b) is also estimated to have the 'correct' . that is. that is. Since f3 is realistically at least 0 . = {3. the more land there is per worker in some base year. = A0 (1 +g)' and L1 = L0 (1 + 11) 1 to get: X)"/(P+><l (1 + g)Pf<P+Klt(1 + rr) Kf<P+~<lt ..0 75)] + y 2 ln(~J Here X. (18) suggests using the same year for GDP per worker and lor land per worker.l l ) (19) There seems to be a positive and significant inlluence from land per worker on GDP per worker. An estimation based on 73 countries with at least grade C for data quality in PWT and for which data are available from the PWTand the World Bank gives: y.. ( Lo (17) Yi = (z*y 4<P+K) A~I<P+<l  Note that in line with our discussion in the previous subsection. so we will have to use these. Hence from the above formula for In Yi : lny7..ln(11. R 2 = 0 . depends positively on X/L.35 [lns.14) (se=O.lnA 1 +  {3 + K {3 + K {3 +K ~ Here we should insert the expression for z* from (13). are annual averages over the period 1960 to 2000. are the same in all countries. from the World Bank (see the legend of Fig. is the 1994 valueof 'agricultural land' for country i according to the data from the World Bank.. as usual. (se=0. among them 'agricultural land'. and K is relatively small. but 1994 is so close to 2000 that our inaccuracy should not be serious. Tal<ing g and 11 as given. 0 is the GDP per worker in year 2000. Taking logs on both sides of (16) gives: (X) a .Exogenous Growth © The McGrawHill Companies. + 0 . +K ~+K (lR) ~ We have arrived (more or less) at the steady state equation for the general Solow model corrected for the presence of land (setting K = 0 in (18) gives the steady state equation of Chapter 5 with ng = 0). The data on natural resources only appear for 1994. and the influence from Ins . adj .(3 13 + K (X) a [In s . it appears that the growth rate must be the gY of (14).1 for the exact source) in association with data from the Penn World Table (PWT). but taking the log of z* does not give a handy expression because of the exponent in the denominator of z* . + 0 . assuming that the basic parameters a. the growth path lies higher the larger X/L 0 is. 0. lny:x1= 2. f3 and K as well as A.
. with one exception: the natural resource owned by consumers is no longer a constant amount ofland but rather..Exogenous © The McGrawHill Companies. li. R. The model will abstract from the presence of land.ln(ni + 0. th e amount of which stays constant when used in production. 7. that is. +ZMB Q) 0::: 6 13 14 15 16 17 18 19 Log of land per worker. nonrenewable resources are depleted as they are used.l. The part of this stock that is used as energy input during period twill be denoted by E1• An important resource depletion equation will state that the stock of oil will be reduced from one period to the next by exactly the amount used in production in the 11rst of the periods.1: Residual log of GOP per worker against log of the value of land per worker. Estimates of land from World B ank: Wealth Es:imates.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 200 I Part 2. The 11gure gives an indication of a relatively tight positive relationship.. Intuitively this should imply even more of a drag on growth than does land. 9 0 0 0 B . . ln(Xi/L~ 4 ). sign. against land per worker. but in the next section we will consider a model with both land and oil. are a lso important. a. across countries. An exercise will ask you if the estimates of y 1 and y 1 seem reasonable in view of plausible values for the parameters a.: 10 Q) ~ 0 :. coal.. oil. We will investigate this intuition by analysing a Solow model with 'oil'. a. at the beginning of any period t.K.. 1. the total remai11ing stock. Nonrenewable resources such as oil. In y.075)]. gas. etc. metals. Unlike land. the part not explained by the structural parameters si and ni. It plots the 'residual value' of the log of GDP per worker. Figure 7. 73 countries Source: Penn World Table 6. 2005 7. using the term 'oil' generically for exhaustible resources. The underlying micro world one should now have in mind is like the one for the Solow model with land.35[ln si . {3 and K . 1994.2 The Solow model with oil Land is not the only natural resource of importance for aggregate production. Limits to growth? The Solow model with scarce natural resources Growth PART 2: EXOGENOUS GROWTH 0 0 0 11 GRB SWE C'< .1 gives an illustration. of an exhaustible resource.Q 'iii :::> '0 • THA JAM Q) MWI • 7 ·c. 1994 Figure 7.
The supply would. on the other hand. for simplicity. c >O. per unit of oil. of the exhaustible resource will be used in production as the input.. (23) The depletion of the natural resource is described by: (24) Tllis leaves us with the question: how much of th e rem ain ing stock. a competitive one with a real price. let us say. of course. and E. that is: (25) Alternatively.005 is considered reasonable.. The possibility of placing a revenue from the sales of oil in interestbearing capital speaks. population growth and tech nological progress are as before: K. a+f3 + c = l. E.+ 1 = (1 + n)L. we can view our analysis as simply exploring the consequences of a constant extraction rate.. Limits to growth? The Solow model with scarce natural resources 7 LIM ITS TO G RO W TH ? 201 The model The production function of the representative llrm is now: a>0. this production function has constant returns to scale v. which is.f3>0.c~)K. corresponding to one half of a per cent of the remaining stock of exhaustible resources being used each year.. (22) A1+ 1 = (1 + g)AI' g ~ 0. u 1. which speaks for not selling too much off today.Exogenous Introducing Advanced Macroeconomics Growth © The McGrawHill Companies. in each period t? For a proper microfounded answer (as given for the inputs oflabour and capital) we should assume a market for oil. Since the stock of oil is gradually used up. 2005 7. and each of whom would h ave negligible influence on the current and future prices of oil and on the total future stock of oil.. (21) L. n ~ 0. energy prices would and should be expected to increase over time. an assumption of sli < c~ is plausible as annual depreciation rates lor aggregate . In (2 5) we have assumed s E < 1. The demand in this market would come from the representative frrm and would in any period obey a usual marginal product (of energy E. s Ii • without bothering about the micro foundations for such a rule..8 I SorensenWhittaJacobsen: I Part 2. K.. This problem is complex and we will not present a solution here. that can be varied at the firm level within a period. O<c}<l. Instead we will postulate that the counteracting forces described can result in the 'rule' that in each period a certain and constant fraction s li of the remaining stock is used in production. but for saving it for later use. In Nordhaus' article an annual value for the extraction rates E of0. L. The interaction between demand and supply involving current and (correctly foreseen) future prices would solve the full problem of how a given remaining stock of the exhaustible resource would be allocated over time. R. come from many small suppliers (possibly 'represented' by one) who would in total own R. lor selling off more now. The equations describing capital accumulation. 0 <s < 1.rith respect to the three inputs. (20) In line with the replication argument. + (1 . At any rate.+1 = sY. under the not quite realistic assumption of a perfectly competitive market. A supplier would be faced with a basic tradeoff. required.) equal to real price condition.
Th is means that just like technology.e. It should be no surprise that the income shares of capital and labour will be constant. the growth rate of the stock of the resource is constant and equal to . into the production function gives: Y 1 = K~(A . Kt' Ll' A... The growth rate. respectively.1.. and u..n . it is easy to complete the model with equations stating the (rental) prices of the factors . Dividing both sides by L. 2005 PART 2: EXOGENO US G RO W TH capital are usually estimated to at least 0. respectively.. and population. Assuming again that the dynamics of our fu ll model are such that the capital. g. approximately given by: j3 f3 +c' £ £ f3 +c f3 +t gY .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 202 I Part 2. Note that the two equations (24) and (25) imply that R.. which will be of relevance for what you are asked to demonstrate in Exercise 5 at the end of the chapter. remains constant at e in all periods as the stock of oil is run down. The steady state growth rate Inserting E. gY.)il(sE R. u 1 EJY./y 1. It is done in much the same way as for the Solow model with land. in terms of the state variables.. Second. the six equations determine the full evolution of all the endogenous variables (convince yourself of this). Similarly. the growth rates of output and capital per worker must converge towards the same rate.Exogenous 7. of labouraugmenting technological progress is less eflective in creating economic growth than in the absence of natural resources.s E) ' R0 from an initial value R0 • Having assumed a specific rule for the use of the exhaustible resource.s liRJY1. converges towards a constant steady state level.s 1) R. energy's share. L. r 1. Setting gy = g~ in (2 7) and solving shows that both rates must converge towards a common constant level. k. It is easy to see that for given initial values of the state variables.output ratio. An exercise will ask you to do this yourself. Rt' lives a dynamic life of its own and evolves over time as R1 = (1 . j3j(j3 + c) is less than one. A. = s"R. a given rate.s". of (28) is in two respects just lil<e the one obtained in (14) for the Solow model with land.sli. Limits to growth? The Solow model with scarce natural resources Growth © The McGrawHill Companies.. An exercise will ask you to do so and to think about how the factor rewards evolve over time. I ' ' (28) We should demonstrate (rather than assume) that the dynamics of our model imply convergence towards a steady state with a constant capital. Our model consists of the six equations above.g . (26) From the latter equality.)'. one gets in the now familiar notation: (2 7) where it was used that g~ is approximately equal to . L.05. l t ( Ll t I l I . and R.: ..+ 1 = (1 . First... w. gives: s R )' = s'1i k"AilR'Lc y t = k"Ail ..s1.output ratio. We will therefore make this assumption. a and j3 . Given our resource extraction rule. = u . i. the stock of resources. population growth implies a drag on lfl .1L... taking logs and time diflerences.
so) 1 R0 • and of course the correct expression for z*. The last term. the capital. Clearly. and to lind the appropriate expression for z*. but no exh austible resource. we cannot give all of the income share that is left after capital and labour's shares to each of them. z*. (29) If one wants to trace this back to parameters and initial values. but a very important one.j<fJ+t) L: ( .a = j3 + E../(/3 + t. However.. In the real world there are both fixed natural resources and exhaustible ones and with constant returns to all factors. First we considered a model with land. We could proceed as in the model with land. as in the model with land. since it demonstrates the isolated effects of the presence of an exhaustible natural resource. the growth path will be higher the higher R0 /L 0 is./L..6) and then gave the remaining share 0 . doing so would result in an 'error' that tends to overestimate the negative influence on growth of land and exhaustible resources. given the rates 11. and the faster the negative influence from diminishing returns to the other factors '<\Till grow. raising to the power of 1/(1 . Now we consider a model '<\lith an exhaustib[e resource but without land. Dividing on both sides of the Hrst equality by y.[t. gives the steady state balanced growth path for y 1: R )r/(fJ+e) y ~ = (z*)'4(fJ+e) A~f(fJ+cls. You can view the model of this section as a building block on the way. The steady state balanced growth path Consider again (26). the natural resource is in even more limited supply as it disappears gradually through its use in production . is positively allected by R. Limits to growth? The Solow model with scarce natural resources 7 LIM ITS TO G RO W TH ? 203 economic growth of the size .2 to the remaining factor.8 I SorensenWhittaJacobsen: I Part 2. sIi • the faster the exhaustible resource '<Viii be depleted. for a really meaningful numerical exercise there is no way around the job of setting up a model with both land and exhaustible resources. . Increasing amounts of ellective labour in association with increasing amounts of capital will press on the limited amount of the natural resource and therefore. and we are again about to give all of the remaining share. This is exactly what we do in the next section.j(j3 + t. .' one gets. etc. 2005 7. on the righthand side: Yi a = z~A~s~(R . through diminishing returns./y.[E. one can insert A.output ratio z 1 = k. S r.2 and fJ = 0./L1) '. An exercise will ask you to show that in this model z. We assumed plausible shares of capital and labour (a = 0. Inserting z* for z 1. also converges towards a constant steady state level.al on both sides. to this last factor. These two features occur lor the same reasons as in the model with land. The reason is as follows.2. Hence. One can also see directly from (29) th at y". This time. the diminishing returns to capital and ellective labour ari~ing from the scarcity of the natural resource become more and more severe. implying even more of a drag on growth than in the case of a fixed factor. however.) ]s IJ• in (2 8) is explained by this increased scarcity of natural resources. R. = (1 . imply a slower (and possibly negative) growth in income per worker than with no natural resource.Exogenous Introducing Advanced Macroeconomics Growth © The McGrawHill Companies. trying to make a numerical evaluation of gll assuming realistic values lor a. = (1 +B)' A0 • L1 = (1 + n) ' L0 . the larger is the extraction rate. In other words. land. In any case. and using that 1 . the 0 . £ and s Ii• etc. This means that over time. You should do such numerical exercises around (28) yourself.)]n. j3.
. By this they mean that in many resource·abundant countries.SorensenWhittaJacobsen: Introducing Advanced I Part 2.in(n + g + b) should hold as a crude approximation. and an exhaustible resource such as oil. as in the model with land. a +{J + K+t: = 1. (30) suggests a regression equation across countries i : . Spectacular examples of this were given in Fig.+0.Exogenous Growth Macroeconomics 204 7. land. . opposing interest groups have often been in politic al and even armed conflict to secure the rents from the exploitation of the resources for themselves. and n..03) Figure 7. + t . adj .2 illustrates by plotting the 'residual log of GDP per worker' (that part of GDP which is not explained by the structural parameters s. l6) (se=0. the steady state balanced growth path can be expressed as: In y: {3 In ~ {3 + t (R ) A. For w hatever reason.3 The Solow model with both land and oil The model The aggregate production function with both a factor in Hxed supply. {3 > 0. and not about the impact on the growth rate. etc. The conclusion from this empirical investigation of the partial influence of 'oil' per worker on income per worker is that indeed there seems to be a positive and signillcant influence in accordance \0\Tith our model..1 which showed that oil· rich courtries like Nigeria and Venezuela have experienced negative growth in income per worker since 1960. There is also a danger that resc<Jrce abundance in a country w ill reduce the perceived need to invest in growt h·enhancing activities such as education. Hence.075)] + y 3 ln (R~4) . coal. the error from mixing year 2000 data for GDP per worker with 1994 data for 'oil' per worker should be small. . K > 0. Again.075)] + 0. (se=O. In y. Note t hat this is a statement about the influence of nat ural resources on the level of income.[In s . {3 + t {3 +t fJ + t L. fmding the right expression for z* is left to you as an exercise. R 2 = 0. {3 and taswell as both A . 2.26 [lnsi . 2005 natural resources PART 2: EXOGENOU S GROWTH according to our model. etc. it is good to have an abundance of exhaustible resources. gas. but it should be no surprise that. 61. lfl . (31) L 94 For R ~ 4 we use World Bank data for the value of 'subsoil assets' (including oil. and sli are the same in all countries). land and energy.!.m= y 0 + y 1[ln s' . An estimation of (31) based on 60 countries gives: lny:lo = 6. research and development. in a base year or left over in the current year. There are constant returns to the four factors. 7.ln(n+ g + f~)] + t . capital. taking logs on both sides in (29). t: > 0.) in 2000 against the log of the value of subsoil assets per worker in 1994.lnsli+ a. Referring to the latter impact..). it is a fact that many resource·rich countries have had a poor historical growt h performance. In z* ~ In s . is (in the obvious notation): a > 0.14 ln(R~ 4 /L~ 4 ) .ln(n' + 0 . labour. Hence the 6. Limits to growth? The © The McGrawHill Solow model with scarce Companies.83 + 1. some economists have spoken of 'the curse of nat ural resources'. metals... (30) With the usual shortcuts (assuming that the basic parameters a.In . 6 As stated.. other circumstances being equal.ln(n.
LKA+ a.. sum of the income shares of the two latter factors .eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2. Cl 8 0> 7 0. v 1 and ul' from which constant income shares would follow (show this). Estimates of subsoil assets from Wortd eank: Wealth Estimates. where we continue to assumes E < c~.. in terms of the . now gives: R )' = s' X " k" AilR' L _.. t t t (L. 2005 7. The explanations of the formula. rl' w.output ratio converges towards a constant steady state level..!. 'iii 0:: 5 +TAN 4 6 8 10 12 14 16 18 20 Log of subsoil assets per worker.. Therefore..otmlriAs Source: Penn World Table 6.r1 = fJ + K + t ...(K + t)g:· ~ ag~ + {Jg  t:sli  (K + t )n. Setting ol' = g~ in (3 3). One can complete the model with equations for the factor reward rates. 1994.2: Residual log of GDP per worker against log of the value of subsoil assets per worker..x)K(s:. must be what is left over after capital and labour's shares. as you are again asked to show in an exercise. and calling the constant solution gu gives: (34) This formula is. 60 r. 0 0 THA NPL ZW E+ +ZMB . L. The remaining equations of the model are again (21)(25) above... K + t . 11 = k" Afl  't ! t E (3 2) I From the latter equality we get. and dividing both sides by L.. gf and g~ converge towards the same rate.Exogenous © The McGrawHill Companies... remembering that 1 . in a sense. Limits to growth? The Solow model with scarce natural resources Growth 7 LIM ITS TO G RO W TH ? 0 0 0 "'. tal<ing log dillerences and in the usual notation: gj' = og ~ + {Jg~ + egf .: 205 11 FIN 10 "< 0 3 •••• • +•JAM 9 Q._. (33) In the long run the capital.. a perfect combination of the two gY formulae from the model with only land and that with only oil. 1.Q (ij ::J o 6 . The steady state growth rate Inserting into the production function that E1 = s ER.!£. 1994 Figure 7.
For instance. 2005 PART 2: EXOGENO US G RO W TH diminishing returns arising from the presence of land and the increasingly severe diminishing returns arising from the presence of oil. 75 percentage points. the drag on growth arising as the number o. tile observed dWerences in annual populatim1 growth rates between developed and developing countries o.0.6 and K = E: = 0.1 to be realistic.12 Sse If we set n = 0. it is an effect of potentially serious magnitude. other things being equal. (36) This equation can be said to involve two growth drags. a rate of technological progress of. This i~ a big dillerence taking longrun compound interest eflects into account. This is indeed a major conclusion from our analysis.2 in the model with only oil. one would have to set E: = 0.003. it is also the growth rate along this path.Exogenous Growth 7. so a g of around 1 per cent would be needed just to overcome the negative influence of population growth. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. This is an ellect completely dillerent from the level ellect we have discussed before. possibly growthpreventing magnitude.()()25 . are also the same. would imply that the second of the above growth drags would be around 0 . There is an important difference to the model with just oil with respect to the negative influence of sI!' Taking capital's and labour's shares as given.75g . sc' is potentially much smaller. 75g .6. 2.2 and f3 = 0 .2.8 percentage points.25n .f'people presses on the limited resources is ofa serious. arriving. arising from the combined eflect of increasing population and exhaustion of the limited resources. 4 per cent per year will still contribute 1. Clearly.1. f3 = 0 . Our model's steady state prediction of how economic growth should depend on population growth can be confronted with the usual type of crosscountry data. However.g. the negative influence on growth frmn the extraction rate. set s E = 0. One can lfl . e.0.01 (which is a plausible upper estimate for developed countries. none of the drags 011 growth arising f rom natural resources seem to be of a growthpreventing magnitude as long as there is a reasonable amoum of technological progress and population growth is under control as typically seen in the most developed part of the world.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 206 I Part 2.000625 ~ 0.0. The most serious one seems to be that (labouraugmenting) technological growth is only three quarters as ellective in creating economic growth as it would be in the (fictitious) absence of natural resources.0. e. With both land and oil in the model. The other one. as seen in some developing countries.g.f' up to three percentage points would.75g . One should again bear in mind that a population growth rate of 3 per cent. taking again a plausible value from Nordhaus's article.. according to our model.g. Furthermore. 75g) of at most 0.8 percentage points to growth in GDP per capita. e. K = t: = 0. imply d!{f'erences in annual growth rates o. It is not (just) the level of the longrun growth path for GDP per worker that is negatively allected by population growth.0. Indeed. from (28).005. Equation (3 5) above says that a one percentage point change in the population growth rate gives a quarter of a percentage point cfrar1ge in the growth rate of GDP per capita. With these values (34) becomes (35) gY= 0. lfpopulation growth is not under control. we get for the longrun growth rate of income per worker: gY= 0. a = 0.2. many of which are considerably below) and.3 percentage points.f'GDP per worker of 0. at a coeflicient on sE of ~· In the present model Kandt should share the remaining 0 . appears as a further drag (on the 'reduced' growth rate 0. in which case the coetllcient on sEbecomes k. Nordhaus considers the values a = 0.
The steady state balanced growth path If we divide both sides of the Hrst equality of (32) by y.16.0 .01 0.eI SorensenWhitta. Note that the estimated slope is around .5.05 Average annual growth rate of population.05 ~"' ~0 0. 19602000 Figure 7.01 OlO. we obtain: a = zr'A~(X/L.0.(RjL. 0. Some details for the statistically proficient: R 2 = 0.vth rates as in Fig.1 2 and I= 3.07 a. 90 countries Not e: The estimated line is y = 0. but there may well be other mechanisms which tend to generate a feedback from higher income growth to lower population growth . the existence of such mechanisms might help to explain why our estimated correlation between population growth and income growth is somewhat higher th an suggested by our theoretical model.0. ~0 0. 2005 Macroeconomics natural resources 7 LIM ITS TO G RO W TH ? 207 simply plot longrun grm. KOR+ QlC) < .02 0. with a standard error of 0 . . 7.02 «! Q) 0.0..5 (standard error on estimated coefficient is 0.~ 0.03 0. but significantly negative.16). where z1 KjY1• In = y: 0.04 3~ 0 0.00 0. again to get the capitaloutput ratio on the righthand side. Our model provides theoretical reasons why the causality should run from population growth to economic growth.5584x. Does Fig. 1. Details of the estimation of the straight line are given in the Hgure' s legend. Source: Penn World Table 6.02 0. Limits to growth? The © The McGrawHill Growth Solow model with scarce Companies.2 5 predicted by our model (see the coe!Ilcient on n in (3 5)).vth rates in GDP per worker against longrun population grm.06 _. Thus the estimated slope is within two standard deviations of the theoretical value of . 0. 7.03 _~0 o c o Q)C'< me 0.00 ~ ~ c ~ .3.Exogenous 7. or is it rather the other way around that a better growth performance helps to get population growth under control? One simply cannot tell from the empirical evidence presented here.)' .3 really illustrate a negative in fluence of population growth on economic growth across countries.a. Hence the direction as well as the magnitude of the intluence of population growth on economic growth in our model's steady state are not contradicted by the data. We should remind you once again of the warning concerning reversed causality that is very often important lor the interpretation of empirical relationships in economics. The 11gure and the estimation point to a relationship that is perhaps not too tight. Indeed.• .) " sf.Jacobsen: Introducing Advanced I Part 2.04 0.01 .3: Average annual growth rate of real GOP per worker against average annual growth rate of population.0.0294.
Hence the steady state balanced growth path for y 1 is: y'. 0 .ln(n + g + o)] (R') ~ In.48 + 1. according to the data from the World Bank.ll ) ~. has reached a constant value.+ t ln sli+ {J {J + K+E: a + K +E: [lns .25 ln( (se =O.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 208 I Part 2. Limits to growth? The Solow model with scarce natural resources Growth PART 2 : EXO GENO US G ROW TH steady state z.. R= .23 [In .sc0. Our main conclusions are repeated in the follm<Ving double statement: • ~f underlying teclmological progress and population growth rates can be l!eld at levels that have typically been seen over long (but recent) periods in Western countries. +:0}2 In R. while R ~ 4 is the value of 'subsoil assets' in 1994.<t . by taking logs on both sides of the equation above. (3 7) This suggests a regression equation: (3 8) across countries. but it should be no surprise that again In z* "" In s . Hence. = (z*)"'<fl+< +c) A~l<fJ +K+r) L.Exogenous © The McGrawHill Companies. the influence is in the expected direction and statistically signiHcant. lfl . • ~f population growth rates are at the highest levels seen in developing countries. z*. l6) . Once again we take X. + 0. tile limited natural resources imply a serious drag on growth that may eliminate most or all oftile positive influence of technological progress on income per capita. s'j<fi+K+r) Getting an expression lor z* requires a dynamic analysis like that conducted for the model with only land. permanent economic growth seems to be sustainable.6 3.adJ.04) (L94 (se=O. 2005 7.+ {J {J + K +t (X) ~ t + K +E: In . . X)</({I +K+t) ( (RL:)t/(fl+<+t).0 75)] + 0. the steady state balanced growth path can be expressed as: inyi "" f3 + fJ + K+t K In A.4 This chapter deals with the very big issue whether sustained growth in per capita income can essentially go on for ever despite the limitntions given by the natural envirorunent.94). An estimation based on 60 countries gives: In y:)() = 3. to be the 1994 value of 'agricultural land' lor country i. ) L94 (3 9) One should note that for both of the types of natural resources considered.ln (n + g + <~) can be used as a crude approximation (an exercise will ask you to demonstrate convergence and find z*). 7.ln(H. not in conjlict with nature's finiteness. that is.
According to this view the modelling assumption of unlimited substitution should not be taken literally. With this chapter we do not pretend to have given a definitive answer to the question: 'Are there limits to growth?' However. Passing a fully qualified judgement on these fundamental issues requires insigh t into the natural sciences as well as into economics. reached on the premises of the models we have considered. The mainly optimistic conclusion is. its relative price will tend to go up. On the other hand. providing strong incentives for the development of alternative production techniques and consumption patterns which rely less on the scarce factor. of course.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 2.Exogenous Growth 7. which is clearly not an innocent one. Much oft he public debate on the possible limits to growth is really about the validity of this assumption. Influenced by the economic h istory of Western countries over the last two centuries. With just one or two types of natural resources in our production function we cannot describe this process explicitly. but the assumption of unlimited substitution can be seen as representing it. w hen the world has only one ton of copper left it is not that production will literally still use small amounts of the remaining copper in association with extremely sophisticated labour in accordance with an old production function. we hope to have shed some light on the basic assumptions one needs to mal<e to warrant either grmvth optimism or growth pessimism. but should be seen as representing the described incentive elfect. copper will have been replaced by a substituting product invented. They point out that whenever a particular natural resource becomes scarce. many natural scientists and environmentalists argue that there are in fact limits to the possibilities of substituting other factors for certain essential raw materials and life support services offered by the environment. The more moderate critics doubt that the market mechanisms will always ensure the development of new substitute techniques in time to prevent serious disruptions stemming from environmental degradation. as copper. The same (optimistic) economists also argue that the deterioration of the natural environment caused by the polluting activities of firms and consumers can be held in check by intelligent use of 'green' taxes and other economic instruments in environmental policy. 2005 7 LIM ITS TO G RO W TH ? 209 The second of these statements means that lor the poorest parts of the world the classical ('dismal science') views are still sadly relevant. became extremely expensive and the economic gains from developing a substituting product became very large. . It is now time to face one critical modelling assumption explicitly: our CobbDouglas production functions assume that there are no limits to the technological possibilities of substituting capital and technologically augmented labour lor scarce natural resources. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. most economists tend to be technological optimists. we implicitly assumed that production can continue to grow even as the input of natural resources becomes infinitely small relative to other inputs. believing that the substitution possibilities in production are essentially unlimited in the long run. Nevertheless the overall conclusion with respect to the possibility of sustainable growth is optimistic because of the first of the statements above. In other words. Rather. Indeed. because of its scarcity.
......SorensenWhittaJacobsen: Introducing Advanced I Part 2  Growth Exogenous acroeconomics 210 7.. The international empirical evidence is consistent w ith this prediction. In such an economy positive economic growth is sustainable for parameter values typical for Western countries... the international evidence shows a positive relationship between the stock of subsoil assets and the level of real GOP per worker. but also in the opposite direction............. but the need for natural resources implies d iminishing returns to the combination of labour and capital. 5.. ceteris paribus...5 .... a fixed supply of natural resources would not realistically prevent sustained positive growth in income per capita... 1...... technological progress serves in part to offset the negative influence of increasing land scarcity on income per worker. This is particu larly the case if ' land' is interpreted in a broad sense to include all the life support services of the natural environment..... 2005 PART 2: EXOGENOUS GROWTH 7....... the gradual depletion of the natural resource creates even stronger d iminishing returns to capital and labour than in the model with a fixed stock of land. The international crosscountry evidence does indeed show a negative relationship between population growth and economic growth......................... Confirming this. lfl ........~Y.~... all other things equal... Hence the growth rate of per capita income is lower than the rate of labouraugmenting technological progress even if there is no population growth..... for parameter values characteristic of developed countries. other things equal.... However. and the quantitative relationship is roughly in line with the prediction of the Solow model....~... an increase in all factor inputs yields constant returns to scale.. By the replication argument.... 2. causality may not only run from population growth to economic growth.. In a threefactor Solow model where labour and capital are comb ined w ith a fixed supply of land. Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies... However..... Positive population growth implies a further drag on growth in per capita income when land is in fixed supply and could even bring growth to a halt.... The Solow model with a fixed supply of land predicts that a country w ith a higher initial stock of land per worker will have a higher level of GOP per worker.. The Solow model w ith o il implies that a larger initial stock of the exhaustible resource per worker should... We also developed a fourfactor Solow model where production uses a fixed stock of land as well as inputs of an exhaustible natural resource in combination w ith capital and labour............... yield a higher GOP per worker.. but may not be attainable in poor developing countries which have not managed to bring population growth under contro l. Land and natural resources are important inputs into production.. 6. The crucial issue is whether technological progress is suffic ient to generate perpetual economic growth despite the scarcity of natural resources.... 4. 3. This chapter has developed an extended Solow model where capital and labour are combined with natural resources in production... The Solow model w ith land and oil predicts that the average rate of growth in GOP per worker w ill be lower the higher the rate of population growth.............. Hence a higher rate of technical progress is requ ired to outweigh the negative effect on growth arising from population growth and increasing natural resource scarcity..~~~. 7. as predicted by the classical economists... In a threefactor Solow model where labour and capital are combined w ith an exhaustible natural resource ('oil')...
but g = 0. i n.. but should be seen as reflecting a belief that growing scarcity of a particular natural resource will generate an economic incentive to develop new substitute inputs. grows by the same rate as both output per worker and capital per worker. (Hint: Use repeated ly the a. grows faster than the real wage rate. Show that the steady state of the Solow model w ith land is in accordance w ith balanced growth.output ratio is constant. how w ill their life conditions evolve compared to the conditions of the workers? Exercise 2.0 I SorensenWhittaJacobsen: I Part 2. Consider the following continuous time version of the Solow model w ith land (in obvious notation): Y = K"(AL)PX". Describe how the real rates. Much of the debate about the possible limits to growth is really about the validity of the assumption of unlimited substitution possibilities.K> 0 ic =sr.1. CJ. . not only in the respect that the capital. Limits to growth? The Solow model with scarce natural resources 7 LIM ITS TO GROWTH? 211 8. 3. gr of (14) in Section 1. d ivide one by the other and proceed in terms of growth factors. Show also that the real rate of rent on land.bK.1.) 2. r. Resolving this issue requires insight into the natural sciences as well as into economics. A ll=g. this time do not take logs. If there is a fixed number of landlords owning the land. Exact growth rates.. common growth rate gr• of output per worker and of capital per worker in steady state is: gr• = (1 + g)~/(~H) 1 )K/(/1+•) .Exogenous Introducing Advanced Macroeconomics Growth © The McGrawHill Companies. w.6 Exercises Exercise 1. a. but analysis in continuous time often runs more elegantly. v 1. Assu me that n > 0.=ll. and that whenever n > 0..{J. and balanced growth.1 ~ I n a. c5 < l. evolve over time. A crucial assumption in the Solow model w ith scarce natural resou rces is that there are unlimited technological possibilities of substituting capital and (educated) labour for increasingly scarce natural resource inputs. r. This assumption should not be taken literally. but write (8) also for period t. w. is constant. The Solow model with land in continuous time Discrete time may be easier to g rasp. Show that in the Solow model with land the exact (in contrast to the approximate).) Show that gr• is approximately equal to the approximation. but also in the following respects: the real wage rate.o c. and the real interest rate.. u1 . 7. and v.. more or less as considered by the c lassical economists. grows at the same rate as total output (not per worker). (1+n (Hint: Start again from the per capita production function (8).. 2005 7.+f3 +K = 1. () < s. in the steady state of the Solow model with land 1. the real rent on land. The latter may be particu larly true for the models considered in this chapter.
Golden rule in the Solow model with land For the Solow model w ith land.) 4. Does the introduction of land imply a similar improvement? 10 . Show that the per capita production function and the capitaloutput ratio (still in obvious notation.z. 2005 7.. Limits to growth? The Solow model with scarce natural resources PART 2: EXOGENOUS GROWTH 1. of what magnitude is the golden rule savings rate you have just found? Compare w ith empirical investment rates in typical developed countries (from Table A).. x ""' X/L) are. e..SorensenWhittaJacobsen: Introducing Advanced I Part 2. etc. found in Section 1 of this chapter. Find these and judge their reasonableness. Are the estimated coefficients in the Solow model with land reasonable? Consider the estimation in (19). respectively: y = k" AIIx'. Show that the steady state value. use. e. Take logs and differentiate w ith respect to time. Why is the result nevertheless d ifferent here? From reasonable parameter values considered in this chapter. Exercise 4. (For the latter you can just draw t as a function of z or you can solve the differential equation as in Exercise 3 of Chapter 3. z* . Remember that 1 a = f3 + x} 3.output ratio is: s z* = (]3 ~ K) (n + g) + b ' and show that the above d ifferential equation implies convergence of z to z * . s.Exogenous Growth acroeconomics 212 © The McGrawHill Companies. that will imply the highest possible position for the growth path of consumption per worker in steady state. Show that the grow1h rate of y in steady state is exactly: {Jg Kn f3 + K o and compare this to the approximate grow1h rate./ k=K/K... Compare your result to the golden rule of the Solow model of Chapter 5. find the value of the savings/in'lestment rate. Exercise 3. Show that the law of motion for z following from the model is the following linear d ifferential equation in z: t = (/3 + K )s . In the equation you arrive at.a A11 x•. where J. for the capital . that k. 2. z = k 1 . (Hint: Start from the expression for z you found in 1. the estimated coefficients in the model's steady state equation became much more reasonable.g ./. f3 and K . A major point in Chapter 6 was that by introducing human capital into the Solow model. and explain the similarity. Are the estimated values of y 1 and y 2 reasonable given (18) and plausible parameter values? Why is y 1 estimated to be smaller than they in Chapter 5? Does the fact that it is smaller tend to make it more reasonable? Is it enough? From (18) and (19) one can also find implied values of a.g. gY.. (f3 + K)(5 + f3(n +g).n..
Show that the approximate growth rates of Y1 and K1 fu lfil: Y K fJ fJ +c C g = g = . Find expressions for the approximate growth rates. How does u 1 evolve compared to w1? Exercise 8. in this model? Comment..8 +c)~ 0 and/ o r 1 . r11 the real wage rate. = not only fJ/(fJ +c) ~ 1. = u1sER1/ Y1 . and show that this transition equation implies convergence of z 1 towards a constant level z* . 2005 7.. u 11 respectively. s.. z 1 "' K1/ Y1 = k 1/y1 in usual notation. Balanced growth and the behaviour of the factor compensations in the steady state of the Solow model with oil 1. g '. fJ +t: Show that there is balanced growth in steady state (also) in the respects mentioned in Exercise 1. Show that the approximation we used in Section 2. is equal to t: in all periods. . Consider now only the steady state of the same model. imply a higher or lower position of this path? What is the golden ru le value for the savings/ investment rate. in a base year or left over in the current year. in particular that the share of energy (or oil).SE.. w1 and u 11 as depending. in the sense that the steady state growth path for y 1 (involving the steady state value z* for the capitaloutput ratio) w ith appropriate approximations leads to (37). given all other parameters.(g+n). L11 A 1 and R1• Show that the income shares are constant. in steady state. u 1 E1/ Y. 3. r1 . of the real interest rate. Golden rule in the Solow model with oil State the steady state growth path for consumption per worker of the Solow model with oil. Does more oil per worker. Exercise 6. in order to establish convergence to z* and to find the value for z *). set up the transition equation for the capitaloutput ratio. in any period. g "' and gu. State expressions for the factor reward rates. and the steady state growth path. you are asked to justify equation (37) of Section 3.Exogenous Introducing Advanced Macroeconomics © The McGrawHill Companies. State the expression for z* . in the Solow model with both land and oil For the Solow model with both land and oil. (Hint: You can go through a number of steps which are counterparts of the steps we went through for the model with just land in Section 1. Consider the Solow model w ith o il. (Note that for the latter you need one more appro:<imation than we used in Section 1. w 11 and the real oil price. etc. should not be farfetched. Question 2.s E 1 should be used). do what you were asked to do for the Solow model w ith just oil in Exercise 5 above. but also c/(. on the state variables K1 . Limits to growth? The Solow model with scarce natural resources Growth 7 LIM ITS TO GROWTH? 213 Exercise 5. Exercise 7. 2. Convergence to steady state. In z* ~Ins ln(n + g + c5). In particular.0 I SorensenWhittaJacobsen: I Part 2. Convergence to steady state in the Solow model with oil For the Solow model w ith oil.
etc. which assumes further n = 0. 2. show that the per capita production function is: and that: g{ E. {3+K+E {J +K+ E {3+K+f (Hint: 1 a.g. .03. sE = 0..SorensenWhittaJacobsen: Introducing Advanced I Part 2. e.2. . discuss if the introduction of both land and oil can substitute for the introduction of human capital. {3. In the text we reminded you that a considerably higher popu lation growth rate.output ratio.0 15 were more realistic? Why is the trebling of the relevant rate less important for s Ethan for n? Exercise 10.. f3 = 0. e 1 e E1/ L 11 and g{ e In y. Although the expressions are algebraically identical there is a difference. As in Exercise 3 above. Why? 10 .g. y 2 and y 3 reasonable for plausible parameter values? Back out implied values for a.. A full analysis of such a model is beyond the scope of this book.f {3 K+E f.)Compare this expression for gr with the corresponding one. Define the physical capital.!Y./L 1 . 2005 PART 2: EXOGENO US G ROW TH Exercise 9.. (3 4).f ag~ + cpg~ + {3g . n = 0. q 1 e H. and (36). / L 1 . Limits to growth? The Solow model with scarce natural resources © The McGrawHill Companies. the approximate growth rate of income per worker is: gr E. Are the estimated values of y 1 . s E· 1.ln y.ESE. The Solow model with natural resources and human capital In some exercises above you may have found that our parameter estimates based on the Solow models with natural resources are not too plausible. e K. gnsE. Assume that A 1 grows at cor stant rate g. Show that along a balanced growth path where both z 1 and q 1 are constant. and that the exhaustion rate is a constant. k 1 e K.Exogenous Growth acroeconomics 214 7. e.!Y. z. as well as the human._.(1< +l:)n. Are the estimated coefficients in the Solow model with land and oil reasonable? Consider equation (37) and the estimation in (38) based on it.005. but finding the balanced growth rate should not be too complicated. W ith usual notation like y 1 e Y. Further numerical evaluation of g Y in the model with both land and oil Consider equation (35). It is an obvious idea to build in human capital as well to see if this will make parameter estimates more reasonable. In a Solow model with both natural resources and human capital the production function would be: in an obvious combination of the notations of this chapter and the previous one. which assumes a= 0. L 1 at constant rate n. h 1 e H 1/ L1 . would change the conclusion that the growth drag from population growth is of a modest size. K and E from the estimation and judge their reasonableness.6. x1 e X/ L 1 . = 0. K = r. 1. Is something similar true for the growth drag from depletion of the exhaustible resource? What would this drag be if. from the model without human capital.cp= f3 + K+ <.01 and sE = 0. Exercise 11.
has the model become more realistic by introducing human capital? .6 w ith raw labour and human capital havin~ equal shares fJ and rp. also s E= 0. (Insert.005 and consider values of n typical for developed and developing countries.) In particular. how is the coefficient on n affected by also having human capital in the model? In view of what we found in Fig.2. {J + rp. and that capital's share is 0.Exogenous Introducing Advanced Macroeconomics Growth © The McGrawHill Companies. for instance. respectively.0 I SorensenWhittaJacobsen: I Part 2. Assume that labour's share. while land's and oil's shares are both 0 . 7. 2005 7. respectively.1 (as we assumed in this chapter). is 0 . Limits to growth? The Solow model with scarce natural resources 7 LIM ITS TO GROWTH? 215 3.3. Rewrite the formula for gr using these values and comment with respect to growth drags.
.......~g~·~························ 9 R&Dbased endogenous growth: macroeconomic ~P.....!....~?.....:?....E?.t.~......~.~....:~...Endogenous Introducing Advanced Growth I 8.. Productive externalities I © The McGrawHill and endogenous growth Companies..i......~g~~......... 217 ....~~....~~....Jacobsen: I Part 3.....~~~~~..Y~.t......I Sorensen........?.. ~~...~E?.~................ 2005 lfl ndogenous rowth 8 r................i......~~~........~...~......~~.Whitta......
Productive externalities and endogenous growth © The McGrawHill Companies.Whitta. 2005 .Endogenous Growth 8.0 I Sorensen.Jacobsen: Introducing Advanced Macroeconomics Part 3.
Hence.technology'. Solow models therefore belong to the class of socalled exogenous growth theories. 2005 lfl •• • • =Productive externalities and endogenous growth !though the Solow models studied so far are quite successful in accounting lor many important aspects of economic growth. In both categories there will be aggregate production functions involving a variable :1 1 that describes .vth in output per capita in the long run. that is. A 219 .vth rate of technology depends on basic model parameters such as the investment rates in physical and human capital.+1 . but there will be no assumption of exogenous technological progress such as A. they have one major limitation: by treating the rate of technological change as exogenous.Endogenous Growth 8. Since we think of the production of technological progress as ari~ing from research and development. feature of endogenous growth models: structural economic policy has implications for grm. In this chapter and the two following ones we \>Viii study endogenous growth theory. makes this rate depend on basic model parameters.explains' the longrun rate of growth in GDP per worker is one that endogenizes the underlying rate of technical change. These are the subject of the next two chapters. the population growth rate. they leave the economy's longrun tsteady state) growth rate unexplained. by an endogenous growth model we mean a model in which the longrun grm. A model that . Productive externalities and endogenous growth © The McGrawHill Companies. This is an important. A. is produced through a specific production process th at requires inputs of its own.+ 1 = ( 1 + g)A . or other fundamental characteristics of the economy. and some will say the defining.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Chapter I Part 3. In this part of the book we will try to answer a very big question that remains unanswered: how can we explain the rate of technological change which is the source of longrun growth in income per capita? The search lor an answer to this fundamental question takes us to the modern theory of elldogellous growth where the longrun rate of growth in GDP per person is truly endogenous.A. in period t. One category contains models that include an explicit description of how technological progress. where g is exogenous.. we call such models R&Dbased models of e11dogenous growth. An endogenous growth model therefore allows an analysis of how economic policies that affect these basic parameters will ailect longrun growth in income per capita. The models to be presented can be divided into two categories.
0 I SorensenWhitta.. 2005 PART 3: ENDO GENOUS G ROWTH The other category does not have an explicit production process for technological improvement. will have increasing returns to scale.........vill be a constant and positive growth rate in output per worker..~~~. The production function was F 1 ...K f L: a. Wh at happens is that as long as capital increases faster than labour there will be more and more capital per worker and this implies.~~.~i.Endogenous Introducing Advanced Macroeconomics 220 Growth 8. As a consequence.~.. and in GDP... simply because of increasing returns. 8.i~.I?.....viii be constant and equal to 1 per cent. the (approximate) growth rate of capital per worker. then. However.. there '. as opposed to that of the individual firm . '.'. assume that there is some growth in the labour force.... And in this case growth would not have to cease in the long run ... Since a< 1.. This implies that the aggregate production function.. growth in capital. gf. This reasoning suggests that if there were increasing returns to capital and labour. say at 2 per cent...l.~. only if capital increases by more than 1 per cent per year.~.. due to diminishing returns.Jacobsen: I Part 3..~. output would be increasing by more than 1 per cent per year.. Let us tal<e the explanation once more in a way that is well suited to our present purposes..~. The explanation was related to constant returns to capital and labour and the associated diminishing returns to capital alone. and they are the subject of this chapter. will be gf = ag ~. As we will see. but assumes that the A 1 of the individual llrm depends positively on the aggregate use of capital. g~~.. then because of constant returns to capital and labour.Y.. because of socalled 'productive externalities'.. per worker will have to cease in the lon g run. the growth rate of income per worker is smaller than the assumed grm>Vth rate of capital per worker.I~~. say at 1 per cent per year..S. Therefore the continued constant growth rate in capital per worker cannot be sustained by savings. as in the steady state of the basic Solow model. Productive externalities and endogenous growth © The McGrawHill Companies..P.. How could there possibly be growth in output per worker? With the production function of the basic Solow model... output will increase by 1 per cent per year.... and hence. gf. this will result in growth in GDP per worker in the long nm without any exogenous technological progress being assumed.1 f::.. and the (approximate) growth rate in output per worker.~~~~~r..vill be growth in output per worker. If capital also increases by 1 per cent per year.<?~.~~. In Chapter 3 we explained why growth in income per worker h ad to vanish in the long run according to the basic Solow model. then each year output will also increase by more than 1 per cent and hence there '.l:. Now. the formula gy= ag ~ already reveals the problem. since . that additional units of capital per worker create less and less additional output per worker. Hence output per worker will be constant. exhibiting constant returns to K 1 and L 1 and diminishing returns to K1 alone (c1 < 1). Hence..r.i~. then longrun growth in GDP per worker would be possible without exogenous technological progress. less and less additional savings per worker.. Indeed. Consequently there were also diminishing returns to capital per worker in the production of output per worker: y 1 = k.... If both capital and labour were increasing at a rate of 1 per cent per year. or of output. If capital increases at a given constant rate of more than 1 per cent per year. The models in this category are referred to as endogenous growth models based on productive externalities.
Under constant returns to scale and given input prices.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3. say. The argument here is intended to show that increasing returns can work as a source of longrun growth in income per worl<er. Rather capital inc reases at a greater rate than labour.Endogenous Growth 8. C(Y. profit maximization does not imply welldefined behaviour of the individual Hrm. The Hrm therefore takes aggregate magnitudes such as GDP or the aggregate use of capital as given. goes to infinity. (1) where the firm takes the labour productivity variable A. Constant returns at the firm level and increasing returns at the aggregate level One may think that the idea of an endogenous growth model suggested by the above reasoning is simply the basic Solow model with the production function of the representative firm exhibiting increasing returns. In other words. In our model the aggregate use of capital has to be equal to the use of capital in the single representative firm. with the two exponents on capital and labour summing up to a number greater than 1 rather than exactly 1. First. will be a decreasing function. the representative . Productive externalities and endogenous growth 10 © The McGrawHill Companies. the marginal cost. With a Cobb.firm takes the aggregates as given. is not compatible with increasing returns at the ilrm level. and therefore average and marginal costs will be decreasing in output. In the models in this chapter that formalize this idea. total costs are proportional to total output. If the linn takes the prices of inputs and outputs as given. There is a way to keep the assumption of constant returns at the Hrm level and at the same time h ave increasing returns at the aggregate level. under constant returns average and marginal cost are constant and both equal to the cost of producing one unit. capital and labour do not increase at the same rate in equilib rium. Under increasing returns. to maximize profits it should want to produce an inllnite amount of output.vth and hence diminishing returns would not be a problem. Pricetaking behaviour and perfect competition are not compatible with increasing returns at the flrm level. but we should nevertheless assume that in making its individual decisions.Douglas production fimction with the exponents adding up to a number greater than 1. K~1 and L~'· We assume further that theA. The hypothetical assumption of equal growth rates in capital and labour is made for the sake of the argument. 1 Increasing returns to scale at the aggregate level therefore seems to be a potential source of endogenous growth. Here is why. In our model we only have one representative profitmaximizing Hrm. 2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH 221 it would be unnecessary to build up more and more capital per worker to sustain grm. as given and there are constant returns to the inputs of capital and labour. involving pricetaking behaviour of the individual firms. the idea of competitive markets. of the 1. because of the replication argument. but this firm represents the aggregate behaviour of many firms each of which is small relative to the whole economy. C. since it is too small to have more than a negligible influence on aggregates. < 1. and C(¥ 1) will go to 0 as Y. The reason is simply that an increase in output requires proportional increases in the inputs. . we believe that there should be close to constant returns at the Hrm level to the inputs that can be replicated. an increase in output requires less than proportional increases in the inputs. Second. Hence. that is. We assume that the individual production function of the representative Hnn (in its role as a small individual firm) is: Y t = (K'" La tJ'"(A t Ld\ tl ' 0 < 0.). For two reasons this is not an idea we will pursue.
Furthermore they vary widely across investigations. and our theory of the functional income distribution still has the nice features of constant income shares . 2 Estimates of the sum of our exponents. = aY.Endogenous Growth 8.11. as expressed by the constant elasticity function: (2) The special case¢ = 0 will bring us back to the basic Solow model of Chapter 3. as well as (2) to arrive at: K r. the aggregate production function in (3) has increasing returns because the sum of the exponents is 1 + ¢(1 . is crucial. 96147. pp.(1) > 1. 1 + ¢(1 . Journal of Economic Theory. Since the individual tirm has no inlluence on aggregate capital. (2).=a ( Kf~r )a1 and ( 5) You will easilyveril)rfrom (3) and (5) that r 1K. For thls whole construction the assumption of a productive externality from the aggregate use of capital to labour productivity. and w. = (1 . What could be the motivations tor this? Empirically. The aggregate production results from inserting (2) into (1) and using the facts that clearing of the input markets implies K. are systematically greater than 1. corresponds (with a around ~) to ¢ being around 0 . where L1 is total labour supply in period t: (3) When¢> 0 .75.L. around 1. .' = K1 and L. A doubling of both aggregate inputs implies that aggregate output is multiplied by the factor 2 1 +Q\(1al . There is perhaps a tendency that most estimates (and the most reliable ones) should be found in the (lower) range. 1997. hPin g o . the marginal products entering the 'marginal product equal to real rental rate' conditions for optimal input demands should be those that appear when A 1 is taken as given. A value for the sum of exponents.3 .45 . while a value of 1. 2005 PART 3: ENDO GENOUS G ROWTH i11dividua/ 1irm depends positively on the aggregate stock of capital K.5. 'Comparing Four Models of Aggregat e Fluctuations due to Self·Fulfilling Expectations'.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 222 I Part 3. A survey of empirical estimates of returns to scale in aggregate production functions can be found. Since the individual 11rm takes aggregate capital as given. Taking partial derivatives in (1). say. we therefore have: and (4) in which we can insert the marketclearing conditions Ki = K1 and L~ = L. Productive externalities and endogenous growth © The McGrawHill Companies. c::1pitil l's shilrP. the idea of increasing returns at the aggregate level is not implausible.a)Y1• Hence. it takes A 1 as given. for example.mel n o pnrt~ profits.a). Eq. pointing to¢> 0 .5 would mean that¢ should be 0. we have increasing returns at the aggregate level. This may give an 2. 1. we have achieved what we wanted: we have constant returns at the firm level (so perfect competition can be assumed) . but we will explain below why it may be reasonable to assume¢> 0. ranging trom just above 1 to levels way above 2.a). 1 + ¢ (1 . 12.' = L1. in S tephanie Schmitt·Grohe.
2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH lfD 223 indication of largest possible plausible values for ¢ . 1962. because other Hrms can gain from it by 'looking over their shoulders'.. But why should there be a capability eiTect at all? Here the idea is that one way workers get more skilful and sophisticated is through the installation of new capital in the Hrms. learning by doing. These features should explain why the additional capability effect spills over to firms in general and hence should be modelled as an externality. The benefit from this effect does not only accme to the firm which increases its capital stock. working with new capital. K.173. Equation (6) below is the individual production function of the representative firm taking A 1 as given. L. and are essentially unchanged from the basic Solow model: Y 1 = (Kl'(A. The important idea of productive externalities due to learning by doing comes from a famous article by the economist and Nobel Prize w inner Kenneth J. however.+1 = sY. for instance values close to ~ (1. 29. (8) (9) 3. 'The Economic Implications of Learning by Doing'. but in addition it will have a positive ellect on the capabilities of workers who gain new knowledge by working with the new capital. potentially (if</>> 0) depends on aggregate capital because of learningbydoing productive externalities. + (1 . There may be a bit of hand waving involved here. The two last equations.. Arrow construc ts a model that involves a distinc tion between old and new capital and in which the idea of a capability effect from the use of new capital is perhaps better placed. </>.6 7). since new machinery is a carrier of new technological knowledge. A.. (7) K. but with the equilibrium conditions of the factor markets. respectively.<5)K. The latter corresponds to a</> around 1. '". and L1= L1. 3 The complete model The above explanations have focused on the crucial and only new feature of the growth model we are constructing. and because in the longer run employees may move between firms and thus bring their acquired capabilities with them to new employers. n > .' = K. . We can therefore write down the complete model without further delay. describe capital accumulation and population growth. the capability effect should arise from accumulated gross investment rather than from the stock of capital. (8) and (9).Endogenous 8. Productive externalities and endogenous growth Growth © The McGrawHill Companies. Arrow. but the idea is that the use of(additional) capital in an individual firm will have a direct eiiect on production as expressed by the individual production function (1). inserted. Thus. Taking the great uncertainty of the estimations into account.. o. In all other respects the model is just like the basic Solow model.L1) 1 O<a<l. one cannot completely exclude larger values lor the sum of the exponents. Review of Economic Studies. For the theoretical motivations the key phrase is learning by doing. a possibility that will be of importance below.1. Taking this idea literally. Equation (7) states the assumption that the labour productivity variable. 153. but we may let the latter approximate all the gross investment undertaken in the past (in this connection ignoring depreciation).SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3. workers become more sophisticated. pp.+ 1 = (1 + n)L" ( 6) 0 < s < 1.
/A. As mentioned..... Sometimes it is assumed in similar models that the external learningbydoing eflect on labour productivity really arises from total production rather than from capital use....... L. = 1 (Kt+ l r9\ 1 + 11 ~ • (12) .~~~~g~~~~../(A 1 L1) .......~~... the model will determine the evolutions of all the endogenous variables Y....I...... and the aggregate production function (3) will then have increasing returns to K. We could also have chosen to let the external eflect from capital use (or production) be affecting total factor productivity rather than labouraugmenting productivity.. This would amount to the formulation... K.. the remaining model being the same. of the state variables. K. then there will be diminishing returns to capital alone........t.. but note that A. and hence..~~. From the definition off< 1 we have kt + l = k.. Whether there are diminishing returns or constant returns to capital alone turns out to make an important difference. tl r+t L. An exercise will ask you to verify that this alternative model works qualitatively exactly as the one above. the model above is just the basic Solow model... We will th erefore assume¢> 0 in all that follows.. Instead. At = Yf. Note from (3) that if¢>< 1....... The law of motion Assuming ¢> < 1. From given initial values. Kl+ t K.. while if¢> = 1.~.... and A. on how aggregate capital evolves. We will first consider the case¢>< 1.'Y... =k. of the real factor prices.. L~ no longer growing at an exogenous rate.. It follows straightforwardly from the production function ( 6) that: = = = (10) and (7) implies that: (11) These two equations will be used in the following.. in the special case of¢> = 0. alone....... Y.. the aggregate production function will have constant returns to K..... 2005 PART 3: ENDO GEN O US G ROWTH We have chosen not to restate the expressions for the real factor prices this time. K.... Productive externalities and endogenous growth © The McGrawHill Companies.2 . the evolution of A.... since a + ¢ (1 ... but whenever needed they can be taken from (5) above. is endogenous and depends..Jacobsen: Introducing Advanced Macroeconomics 224 I Part 3. 8. k.&!.'./(A tL1) and Dt y .S.+ t (Kt+l r Lt+l A.. L.. L........r1) will be smaller than 1....... and Lt....Endogenous Growth 8. A t• without any effect on our conclusions. K0 and L0 .... through (2)... Kt+l K./A ..eI SorensenWhitta.. rather than (7) above.S. using (5). we can again analyse the model in terms of the technologyadjusted variables...
gives the transition equation: (13) which we can also write in the form: (11) (You should check this reformulation. Further. +1 = sY.. we get: kt_+ 1 = _ 1_ 1 + . and then derive the condition for this slope being less than 1. converges to a specillc valuek*.(5. 1 b)K. k. insert k* fork.(1 . Productive externalities and endogenous growth Growth Companies. which is exactly the one we have just assumed. To show this one can dillerentiate the transition function given in (13) with respect to k1. . We will establish properties of the transition equation implying that in the long run k. The latter is realistically assumed (note that it is implied by the realistic assumption.) '~'> K.b) )1/(la) • (15) veri(ying a unique positive intersection if (1 + n) l/(1l"l > 1 . for convergence to k* it is important that the transition curve intersects the 45° line at k* trom above.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3.1. n + o> 0).b)K. Inserting th at K. to lind the slope of the transition cunre at k*.Endogenous © The McGrawHill 8.. 1 + n k. !rom (13) follows that it crosses the 45° line for exactly one positive valuek* of k. itis clear that the transition curve passes through (0.) Note that since we have assumed ¢> < 1.(sk. Convergence to steady state The transition equation has the form of a onedimensional firstorder difference equation ink.(5) 1 =. kt = k~ ( S = (1 + n) 1/0l"l. From either of (13) or (14).c5)) 1l".: inserting k. Finally. From (14) it follows that it is everywhere increasing. Multiplying on both sides by k. 8 .". 1+n where we have used (10) for the latter equality. gives _ . = ...1 + (1 . 2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH lfl 225 where we have used (11) and (9) .~+ ) 1" 1 (1 . Also note that if we set¢> = 0.+ 1 = k1 into (13) and solving lor li. as shown in Fig. both of the equations (13) and (14) become the transition equation of the basic Solow model (the model of Chapter 3 with B = 1). 0). the exponents in (14) are all welldefined and positive.b. (sY + (1 1 1 (s[J. You should do this and find the condition (1 + n) 1/( 1 4>) > 1 .. + (1.
could not be constant). Productive externalities and endogenous growth Growth Companies. in steady state. Therefore the righthand side must also equal1: = = 1 (K ) 1 . respectively.__!.Endogenous © The McGrawHill 8. 8. of A 1• This time we do not have an exogenous growth rate ofA1• so to determine the steady state growth rate of y 1 we must determine the endogenous growth rate of A.!. In the Solow model this sulllced for determining the steady state growth rates of k 1 andy. In steady state the lefthand side. both k 1 and y 1 must grow at the same rate as A. . When they are locked at these constant values in steady state. Consider (12) above. y 1/ A 1. is equal to 1. and we have found again that these variables converge to constant steady state values. Consequently.. The auxiliary variables are defined in the same way. 1 + II K. to k* and [J*. This is easy to do. 2005 PART 3: ENDOGENOUS GROWTH k* Figure 8. respectively. k. k. Semiendogenous growth in steady state Our conclusions so tar seem reminiscent of those we arrived at in Chapter 5 for the general Solow model.1 imply that in the long run k1 will converge to k*.Jacobsen: Introducing Advanced Macroeconomics 226 I Part 3. (otherwise k 1/A 1 and yJA. = k~ will converge to the associated: [J* =( (1 + n) )a/(1a) S l /(14>) (1 . g. + n) t/ (1<Pl.1: The transition d iagram of the model of semiendogenous growth The properties just established and illustrated in Fig. This defines the steady state. Both had to be equal to the exogenous growth rate.. k 1/A 1 and !i.eI SorensenWhitta.!.+1 /k 1..c5) (16) · We have thus demonstrated longrun convergence ofk1 and [J. <P = 1 ~ Kt + l = ( 1 K. [J.
From the dellnition of [I 1 the steady state growth path of y. y . and GDP per worker..1 "' g (17) ~ At Hence. must be y'. 2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH lfl 227 Using (11! then gives: At+! = (1 + 11)</1/(1<t>) A.. in steady state the growth rateofA 1 is (1 + n)<P/(l 1>) . which is not farfetched. and we call this growth rate gse· (The constant approximate growth rate.c5) )(a+ [</1/( 1</>)])/( 1a) L<i>/(l<Pi( 1 0 ). *'* At+ l . To exploit the increasing returns in the aggregate production function an increasing labour force is required. but since k. Note that in steady state the in puts of capital and labour are not growing at the same rates. realistic values of ¢ could be up to 0 . then the steady state growth rate of GDP per capita is also 0 . if the growth rate of the labour Ioree is 0. when the economy is in steady state there is a necessary link between K. capital must be growing at approximately the rate n + g.. where we have used that g.: *_ ' S · y .Endogenous © The McGrawHill 8. 5 .. = (1 + n) 1L0 . so y'.. defined as K... Inserting this expression forK. = (k*L.7 5. · (18) This gives the steady state growth path (y '. these parameter values result in steady state growth rates.Cl+11) 1/(l<P) _ (1 . . = fJ*Kf ..In A~' tulfils g~ "" 4>/(1 . so in steady state: K.1. Only i1'11 = 0 are these growth rates equal. For the growth rate of GDP per capita to be positive. Structural policy for steady state Let us now turn to the steady state growth path. This follows because in steady state the variable k.. k .'. g~ = In A1+ 1 .. as you should verily. Labour input grows at rate n. grows at the rate g. = y*(I(*Lr)<P/(1rp) = (k*)"+I<P/(1<Pll Lt/(l<Pl(1 + n) I<P/(1 <t>)lt. = k. Furthermore. is locked at the value k*. into the steady state growth path. . = y* A 1• The evolution of A.) Note that g.. and L. 5 per cent to 1.5 per cent. and from (17) that (1 + n) <l>/(l<P) = 1 +g.0. The corresponding . a positive population growth rate is required. worker gives: = y*Kf.) I/(~<~>>. gse· in the range from 0 . = K1/L. Productive externalities and endogenous growth Growth Companies. In steady state the growth rates ofbotb capital per worker.. We can now finally insert k* !rom (15)..(. ) for output per worker. obeys A. Hence from (17). The term 'semiendogenous' growth refers to the fact that we only have (endogenous) growth in GDP per worker if there is (exogenous) population growth. is determined endogenously and depends on model parameters.L.A t = (1 + I'I)<P/(1</>).. for output per y'. As you will easily verify from (17).cp)n.) = K:"'/L. = Kf .. Is the endogenous growth rate g" of GDP per person derived from this model of a realistic size lor plausible parameter values? Annual population growth rates of around 0 . This should not be surprising in view of the remarks in Section 1.5 per cent are typical for Western countries (see Table A at the end of Book One) and according to the empirical studies mentioned earlier. and L. y'./(A. must then also beg. +g.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3.
is now a+ ¢ (1 .020 0. One way to test this prediction is to plot average annual growth rates in GDP per worker over some period against average annual population growth rates over the same period across countries.00 <1: .1. The count ries considered are countries with a data quality grade of at least C.) The really new feature is that the steady state growth rates of y. "'~ "' ~Ol ~0 ~~ 0> O'l . and adj. which is greater than the (l we have arrived at for the basic Solow model.0. finding a negative. focusing on a longer period and lewer countries.2 does more or less the same. relation.000 0. Productive externalities © The McGrawHill Growth and endogenous growth Companies. Figure 8 .05 ~0 0 o0 0.03. the policy implications are somewhat similar to those in the basic Solow model.03 0. Taking</> to be a given technical parameter (not easily affected by policy) the conclusion from this model's steady state is that. Note. 1. a higher investment rate. R 2 = 0. 2005 Macroeconomics 228 PART 3: E NDO GEN O US G ROWTH steady state growth path (c~) for consumption per worker is obtained by multiplying by 1 .01 0. 55 countries Note: The details of the estimated line are: g' = 0.2000 Figure 8. one should try to pronwte population growth! There are reasons why one would be cautious about such a policy.02 5 0. (As an exercise.eI SorensenWhitta.Jacobsen: Introducing Advanced I Part 3. .3 of Chapter 7. should give higher growth in GDP per capita.0934.son both sides of (18). implying the highest possible position of (c) .>< 0.2: Average annual growth rate of real GDP per worker against average annual growth rate in population. shifts the growth path for y.a). show this and try to explain intuitively why we must now get a higher golden rule savings rate for a given a.02 <CO ~ !: c ~ "' 0> O> Q.035 Average annual growth rate of population. or the population. GBR 0. s.•P l.45n'. 7.010 0. however.030 0. 1951. Structural economic policies for steady state must work to allect the positions of these paths or the growth rates along them (or both) . <110 ~" 0. 0.015 0.Endogenous 8. to promote longrun growth in GDP and consumption per capita. With respect to the positions. Source: Penn World Table 6. For example.0.30. that the golden rule value for s.005 0. just as we did in Fig. = Empirics for semiendogenous growth The model's steady state predicts that a higher growth rate in the labour force.. and c1 depend endogenously on model parameters: both are equal to 9se (1 + n)4>/(l .04 <»"' _.01 0)0. upwards. the standard error of the estimate of the slope is 0. rather than a positive.
Second.1 reports for each of the two subperiods 1870. 2005 lfl Macroeconomics 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH 229 We find (again) a rather clear 11egative relationship between population growth and economic growth across countries in the period considered.Sorensen. rather than labour force. For all countries economic growth increased substantially from the early subperiod to the later one.vth decreased.1990 the average annual population growth rate and the average annual growth rate ofGDP per capita.) What we have found in Fig.1930 and 1930. In th at case the general picture found in Table 8. Table 8. it is not clear exactly what kind of area the model covers. while for all countries but three. First. .4 Still we cannot conclude that the model is wrong. 8. so we should consider the very slow convergence process itself as the model's main prediction. 8. If the firms in one country can 'look over the shoulders' of firms in other countries. 4.Endogenous Growth 8. both numerically and analytically. and the transitory growth depends negatively on population growth. However. growth rates. the evidence shows little sign that growth in GDP per capita should be positively related to population growth. a country. convergence in the model we consider may be quite slow due to the presence of the productive externality. since it is not easy to tell how wide the external ellect of the model reaches.Whitta. These remarks seem to call lor (even) longer run empirical evidence that does not go across countries. It is therefore not clear if crosscountry evidence as reported in Fig. outside steady state behaviour as an approximation of a very longlasting convergence process. however. Furthermore. lor some time. This is exactly the perspective taken in the next section. It could be that convergence according to the model is so slow that it really takes place over periods as long as 120 years. (Exercises 1 and 2 at the end of the chapter will ask you to look in more detail at the convergence process. population grm. or perhaps the (developed) world. (18)). if convergence to steady state is that slow. During the convergence period there will be transitory growth in addition to the underlying (endogenous) steady state growth. but infinite. the model we consider has a steady state and it has convergence to the steady state in the long run . and for the same reason: a decrease in the rate of population growth shifts the steady state growth path up but reduces its slope (see Eq. We will then consider the model's infinite.2 is relevant. The OLSestimated straight line has a signillcantly negative slope (the details of the estimation are reported in the figure legend) . Hence. then perhaps the model should be considered to cover the world. just as in the Solow models. a region. As¢ comes close to 1. during the transition to the new steady state growth path there will. It should be noted that Table 8. convergence '<\Till be very slow. and thus does not take changes in the evolution of participation rates into account.2 may be interpreted as an expression of longlasting transitory growth that should indeed depend negatively on population growth according to our model.1 states population. for many of the countries substantially. This does not necessarily mean that the model of semiendogenous growth is wrong. be a positive transitory contribution to growth arising from the lower population growth.Jacobsen: Introducing Advanced I Part 3. Productive externalities © The McGrawHill and endogenous growth Companies.1 could be compatible with the model's transitory growth. We will apply a speciHc assumption that makes the convergence period not just very long. then the model's steady state is not very descriptive lor periods of interest. If we consider quite long periods and 'stay' within each country.
.a)(1 ... 18701990 Average annual population growth rate (% ) Australia Austria Belgium Canada Denmark Finland France Germany Italy Japan Netherlands New Zealand Norway Sweden Switzerland UK USA Average annual growth rate of GOP per capita (%) 18701930 19301990 18701930 19301990 2..6 1.1: Average annual growth rates in population and real GOP per capita in 17 industrialized countries....9 1.... such as centuries.4 1..2 1...0 1...2 1.....0 1. and several expressions would be meaningless for¢> = 1.5 1. It is only as an approximat ion of the case of a large <fJ that 1> = 1 can be of interest. 2005 Macroeconomics 230 PART 3: ENDO GENOUS G ROWTH Table 8...7 0.4 2.....~~~.5 1..7 1...6 0.5 2...7 0...8 1..Jacobsen: Introducing Advanced I Part 3.4 1.3 2.. including in some denominators.8 0.6 0...1 2..5 2.Endogenous 8. . Monitoring the World Economy 1820.3 2.1 0.1 1..5 0.. 8...1 1....6 0.6 1.2 1.4 1.¢)c5.1 1...0 1....8 1.. In Exercise 2 (which is an important exercise) you are asked to Hnd an analytical expression for the rate of convergence for y.....3 0... but still less than unity..6 1. OEC D.. Productive externalities © The McGrawHill Growth and endogenous growth Companies. This verilles the claim above that a large¢ less than 1 implies very slow convergence... the rate is A"" (1 ..3 ~~~g~.2 0....eI SorensenWhitta.7 0. say....~. population growth is rwt the factor that causes economic growth.9 3.1 2..6 2..1992.6 0. There is no reason to believe that a paramet er value should take one particular knife· edge value..~~~... equal to L..8 0...7 2.9 1. Paris..~. 5 TheAKmodel In the analysis in the previous section the term 1 . The tdea is to consider the growth model of this chapter with ¢ = 1..1 3..... For the case where population growth is smalL rr"" 0.6 2..9 2.......7 0. Hence we 5. according to our model with ¢ < 1.1 1....7 1.. since we now want to investigate the possibility that over periods of interest... Below we will indeed set 11 = 0.&~~...7 0..7 0..2 0.7 1..5 2.0 2.1 1 . 1995.1 1...6 0..... We consider the everlasting convergence resulting from ¢ = 1 to be of interest because it approximates the very longlasting convergence that would result from a¢ that is large.4 1..1 Source: Angus Maddison. This gives a zero rate of convergence and hence an everlasting convergence process. The size of the labour force is then constant...6 0..4 0.4 1......8 1..0 0..¢>appeared in many places..6 1 .
the technology variable A 1 (not to be confused with the constant A= L1 a) must increase at rate g. evolves proportionally to K1• and hence to k. and hence in capital.. gives y 1 = Ak .<5)k.Jacobsen: Introducing Advanced Macroeconomics I Part 3. To understand this result it may be useful to draw some wellknown diagrams. w.. Finally. in order to transform variables into per worker terms. since k 1 = Kr/L. (9). (21) where the first equation is the aggregate production function resulting from (6) and (7). e (24) The latter equation directly gives the constant and endogenous growth rate. Subtracting k 1 from both sides gives the Solow equation: (23) kt+ L . Dividing also both sides of the capital accumulation equation (21) by L. . on the other hand. Dividing both sides of the production function (20) by L. We apply th is notation because the model we consider is so often referred to as the . (19) so the real interest rate. = Land¢ = 1 we get from (5) that: r.o> 0).. = L. g. +1 = sY1 + (1 . since with ¢ = 1. while the second equation is the usual capital accumulation equation repeated from (8). For what we did in Section 1. This is a bit of an abuse of notation. KJL. since y 1 is a constant times ku the growth rate of output per worker is g.a)KjLa. = r =aL 1 " and w. it was not important whether ¢ was smaller than or equal to 1.. or directly from (3). then gives the transition equation: (22) k1+ L =(sA+ 1 .. = K1• It follows that according to the AK model. In addition to the above expressions for the rea! factor prices the model can be condensed to the two equations: = Y. has been replaced by L. is constant and the wage rate.k.Whitta. so the expressions for the real factor prices we found there and the associated theory of income distribution are also relevant when¢ = 1. and hence the growth rate of consumption per worker. = K1L1 a =AK (20) • 1 K.<5. and then dividing both sides by k. = s1\ .. and inserting Ak 1 for y .okl. gives the moditled Solow equation: k . With L. g. = Growth according to the AK model It is easy to see that the model above can result in permanent growth in GDP per worker. but it should cause no confusion if we denote L La by an A without a subscript t. in capital per worker k...f5 is the common endogenous growth rate of all the variables we are interested in (we assume that sA . we haveA. = sAki .o)K.. Hence. c1 = (1 .Sorensen.s)y 1• must beg.AK model'. = (1 . Furthermore. K. r .k kI _ r_+_L_ _r = sA . everlasting economic growth is possible in this model without an assumption of exogenous technological growth. 2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH lfl 231 cannot simply set¢ = 1 in all the equations above.. The model's last equation.. Productive externalities and endogenous growth Growth Companies.Endogenous © The McGrawHill 8.vith ¢ = 1. '. Note that in (20) we have used the definition tl L1 a..(5 =B ..
the Solow diagram associated with (23). which are straight lines in Fig.3. Productive externalities and endogenous growth © The McGrawHill Companies. Rather there are constant returns to capital per worker: y 1 = Ak 1• sA ~n)~k.3: The transition diagram (top). 8.Jacobsen: I Part 3. would in Solow models be true curves due to diminishing returns to capital per worker (Cl < 1).Whitta. and the modified Solow diagram (bottom) of the model of endogenous growth . The 'curves'. 2005 PART 3: ENDOGENOUS GROWTH Figure 8. the Solow diagram (mid:Jie). I Sorensen. 0 ~LL_ __________ Figure 8.+1_k. The diagrams illustrate that in this model there is no steady state and k1 (and hence y 1 and Y1) grow at a constant rate forever. k. and the modilled Solow diagram associated with (24).3 shows the transition diagram associated with (22).Endogenous Introducing Advanced Macroeconomics 232 Growth 8. The linearity of the curves in the present model rellects that there are no diminishing returns to capital at the aggregate level.
Hence the growth rate g. This differs from our earlier conclusions since a higher s no longer just gives a higher level of output per worker in the long run and a temporarily higher transitory growth rate in the intermediate run. Implications for structural policy and the scale effect The main conclusion from the model is that a higher savings (investrnent) rate. Perhaps it is the world. More effective aggregate investment should. and a lower rate of depreciation leads. somehow. This may be most relevant for countries where the government is heavily involved in production so that a large part of investment is not driven by private incentives. Since our model with ¢> = 1 should be seen as an approximation of the semiendogenous growth model with a large if> smaller than 1. but then during the last 200 years the world population has increased rapidly while economic growth rates have not shown a similarly strong increase. and the growth rate would increase if population increased. s. it can perhaps take advantage of this effect. 2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH lfD 233 The 'growth brake' from the Solow models. diminishing returns to capital. In the Solow model a lower depreciation rate results in a higher level of GDP per capita in the long run. but again it is not clear what geographical area the model covers.o is higher the larger the constant population size L is. Note the general tendency that parameter changes that give a longrun level effect on GDP per worker in the Solow models give a longrun ellect on the rate of change in GDP per worker in the model of endogenous growth. but decided upon by government bureaucracies. Policies for more etlective investment are not only relevant in connection with endogenous growth. The source of growth is thus aggregate constant returns to the reproducible factor. to a permanently higher growth rate in GDP per worker. This is the socalled 'scale etlect'. which is also important. lead to a lower rate of depreciation. Historically this method of making investment decisions has often led to inefficient and rapidly worn out (sometimes even useless) investment. gives rise to a permanently higher growth rate in GDP and consumption per worker. By relying more on private. is simply no longer present in the aggregate production function. Assume that the productive externality. however.Endogenous 8. = sA . in the model of endogenous growth. such a country could probably attain higher prosperity and growth in the long run. a government can take actions that make net investment more 'effective'. A decrease in has an effect similar to that of an increased s. which is the driving force of endogenous growth in this chapter. Productive externalities and endogenous growth Growth © The McGrawHill Companies. although there may be transition costs involved in changing from one type of economic system to another. but it may be more difficult to achieve through economic policy. but it results in a permanently higher rate of growth in output per worker. The implications for economic policy are obvious: policies that stimulate savings now give a very longlasting boost to growth. capital.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3. Empirical evidence does not support the hypothesis that larger countries should have larger longrun growth rates. profitmotivated incentives in investment decisions. If. we should remember that strictly speaking the correct statement is that an increase ins gives very longlasting transitory growth in GDP per worker. which is rather controversial. Remember that A= L L a. It is possible to get rid of the scale etlect. arises from capital per o . There is one odd feature of the AK model w hich casts doubt on any policy recommendation derived !rom it.
.._... L...01 ~·~~~=~~. 1. e.4: Average annual growth rate of real GDP per worker against average investment rate in physical capital... Either endogenous growth models have the unattractive scale eflect or they assume.. Source: Penn World Table 6.E. unattractively.b... There is no way around this problem.'"""'++>c"'H"E~ 0...iF+~...~"'ii~~. the A (without subscript t) is equal to 1 and independent of L.07 .. aggregate production (not production in the individual llrm) is now completely insensitive to labour input..~ . s..+:."'Bm::''=:.1082s 1..20 0..2000 Figure 8.03 1. The figure shows a quite good and tight positive relationship.06 1~+BWA 2~ 0. (7) in our model should be replaced by: You will easily verily that with¢ = 1 the aggregate production function becomes Y.00 f'L. I SorensenWhittaJacobsen: I Part 3... 2005 PART 3: ENDOGENOUS GROWTH worker rather than from capital itself.. and the scale effect has been eliminated...40 0. g. at the aggregate level (when¢ = 1). arising at the llrm level is exactly off..00 0. = K...~ (ij ~ ~ RK + + KO R 0.31.. Let us put this prediction to a test by considering crosscountry evidence..:::::. = s . Consequently the endogenous growth rate of the model will be g. The countries considered are countries with a data quality grade of at least C in PWT. g>~ 0.0. :: Q.. The positive effect on production of a higher labour input.. over the same period across countries i.02 1. the standard error of the estimat e of the slope is 0.:oT"'. Empirics for endogenous growth As mentioned. that labour inputs are unproductive at the aggregate level.Endogenous Introducing Advanced Macroeconomics 234 8... R 2 = 0.04 1MidS~ g ~0> ! ::0.... from 1960 to 2000 against average investment rates.. <»a.... 90 countries Note: The estimated line is g 1= 0.45 Average investment rate in physical capital...~+"7 £~ 0.. Productive externalities and endogenous growth Growth © The McGrawHill Companies.0 1 l7a._.:r..0004 + 0.~.. . However. so that Eq. 0. a main prediction of the model of endogenous growth is the positive influence of the savings or investment rate on the longrun growth rate of GDP per worker. 0..:. This is a main reason why economists take the theories of endogenous growth seriously (still remembering that !g 0.02 1..:.4 plots average a1mual growth rates.30 0.. and adj.15 0. Figure 8. that is.01 72.05 ~ 0.25 0.35 0..~et by the negative external eflect of a lower KtfL .''.._.. 1960. 0... .
.. Furthermore. gave rise to a model of semiendogenous growth (Section 2) ........... but there was a scale effect: a larger labour force gave a higher rate of economic growth.... </> = 1. and the longrun growth rate ofGDP per worker was higher the higher the rate of population growth...... In these models perrnmrent growth in GDP per worker wns n reflection of exogenous tedmologicnl growth... resulted in a model of (genuinely) endogenous growth (Section 3). 8.... 2005 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH lfl 235 these should be seen as approximations of theories of very longlasting transitory growth).. in the appropriate model interpretation........... Yet the positive relationship does not prove that the idea of endogenous growth is correct........... The basic mechanism lor endogenous growth in this chapter was productive externalities.. Exogenous versus endogenous growth ... The presence of a positive..... A strong productive externality... so the models did not really explain longrun economic growth. Both the model of semi endogenous growth and the model of endogenous growth delivered explanations of economic growth: the growth rate of GDP per worker depended orr basic structural model parameters... Furthermore.... 0 < </> < 1.... and a fixed population growth would imply an exploding economic growth rate........... semiendogenous and endogenous growth... R. over the period considered typically h ave experienced increases in s....... However. Perhaps the Hgure just shows that countries with a high average s.. In the long run the growth rate of GDP per worker was either independent of or negatively affected by the population growth rate (the latter in the realistic presence of natural resources).... in no way does Fig. very slow convergence to one).Endogenous Growth 8..... and therefore it had the 'convergence property': growth is faster the further below steady state the economy is....Whitta....... Therefore the model did not have the convergence property. All these models had convergence to a steady state and the convergence property.......Sorensen.. so crosscountry evidence may not be the appropriate kind of empirics to study.... and therefore have had relatively high transitory growth rates. No agents were concerned with deliberately producing technological progress... In previous chapters we have studied models of exogenous growth..... but not too strong productive externality.. in steady state economic growth was explained by growth in the labour force..... In this chapter we have studied two distinct and alternative types of endogenous growth .4 as indicating endogenous growth .............. As we have argued several times.... the next section will provide a reason for indeed viewing Fig... both fundamentally caused by productive externalities. The technological growth arising from these came as an unintended byproduct of economic activity.......... it is not clear whether the model covers a country or the world..... This model had convergence to a steady state in the longnm. In the real world we know that a lot of research and development activities intended to create technological progress are undertaken in private companies as .........4 contradict the traditional Solow models according to which an increase in the savings rate will imply a (higher) transitory growth in excess of the underlying exogenous growth... Productive externalities and endogenous growth © The McGrawHill Companies.... Growth in GDP per capita could occur without population growth.... This model had no steady state and hence no convergence to one (or rather.Jacobsen: Introducing Advanced Macroeconomics I Part 3..........
However. and some arguments presented below may be important for deciding this. it could be that at the present stage of knowledge the existing endogenous growth models were not convincing. The knifeedge argument Strictly speaking the model of(genuinely) endogenous growth only pertains to a zero probability. This is again an empirical matter. which does not have zero probability. Explaining growth Everybody agrees that it is good to have explanations oflongrun economic growth. However.Jacobsen: Introducing Advanced Macroeconomics 236 I Part 3.Endogenous Growth 8. which speaks in favour of models of semiendogenous growth. Productive externalities and endogenous growth © The McGrawHill Companies. the simple externalitybased models of endogenous growth considered in this chapter share enough features with the more advanced endogenous growth models to be representative for endogenous growth theory. perhaps the endogenous growth models of the next two chapters in which technological progress is the outcome of an explicit and deliberate production activity will be more convincing. Endogenous growth models are sometimes criticized on such grounds. economic growth and the scale effect If the knifeedge property of the endogenous growth model does not represent a serious objection to it. 2005 PART 3: ENDO GENOUS G ROWTH well as at universities. this is certainly an argument in favour of endogenous growth models. as we have carefully argued. so one would have to rely on exogenous growth models for understanding growth.7 5. Whether the gwwth explanations and other features of the endogenous growth models are sufllciently convincing is mainly an empirical matter.eI SorensenWhitta. According to the model of serniendogenous growth there should be a positive relationship between the (constant) growth rate of GDP per capita and the (constant) . The feature that an increasing population implies an increasing growth rate in GDP per capita is simply not realistic. knifeedge case (¢ = 1) which is uninteresting. Therefore we can already present some main arguments from an ongoing and fascinating debate among growth theorists. For the externality parameter ¢ considered in this chapter. although the understanding would be limited by technological progress being unexplained. The issue in this debate is whether the economic growth we see in the real world is best understood by exogenous growth models or by endogenous growth models. and since the endogenous growth models deliver much more of an explanation than exogenous growth models. However. The presence of the scale effect is the single most important argument against mode!s of (genuinely) endogenous growth. Population growth. On the face of it. Therefore. the scale ellect does. empirical evidence seems to point to positive values of at most up to 0 . The issue is really whether the relevant parameter (here ¢) can realistically be assumed to be so close to a limiting value (here 1) that the model of endogenous growth can be seen as a good approximation. this criticism is therefore not valid. Here are some main arguments frotn the debate. the model should rightly be interpreted as an approximation of a wider case(¢ slightly smaller than 1).
Over the several th ousands of years before that. to 1990'. 'Population G rowth and Technological Change: One Million B. growth becomes slower and slower. As we have seen. we refer to Michael Kremer. 7 In Chapters 5 and 6 we saw that if one controls for structural dillerences between countries in a way suggested by exogenous growth models. one may wonder ifit is really the world's population growth that has caused economic growth. 68 1716.Jacobsen: Introducing Advanced I Part 3. but another fact is. There is thus evidence in support of convergence among countries in the real world. the empirical relationship between population growth and economic growth seems to be mostly in favour of the models of exogenous growth. thousands of years of almost no economic growth have been succeeded by 200 years of rapid economic growth. In the world of models the convergence among countries arises naturally in models with (an appropriate form of) convergence to a steady state. many countries in the West have experienced relatively constant average annual growth rates in GOP per capita of about 2 per cent during the last 200 years. Furthermore.1.Endogenous Growth 8. according to this argument. with the world population being close to constant for hundreds of years up to the eighteenth century and thereafter increasing rapidly. As the steady state is approached (from below). then even for the cow1tries of the world. There is.Whitta. which predict a negative relationship when natural resources are included in the model. This pattern is even more clear between states in the USA. more impressive and more important in a very longrun perspective. For these 'facts' and an interesting discussion of them.C.1 shows that the areas in the world that have experienced the high (and even increasing) economic growth have had decreasing population growth. In the eyes of the authors of this book. USA and (other) parts of Asia.2 and Table 8. or the other way round. 8. The empirically observed convergence between countries is thus most n aturally explained by models of exogenous or semiendogenous growth. however. Based on the evidence presented in Fig. one may have a hard time tlnding this feature convincing. Table 8. such as the OECO countries. a counterargument: what Table 8.1 shows is true. 6. there is a clear tendency for the countries that are initially most behind to grow the fastest and thus catch up to the initially richest. India and China should have caused economic growth in Europe. the average annual growth rates in GOP per capita must have been close to 0 to llt with the level of income per capita around year 1800. Quarterly Journal of Economics. so the country starting out with the lowest GOP per person will grow the fastest. . Population growth has behaved basically the same way. Convergence We saw in Chapter 2 that focusing on countries that can be assumed to be structurally similar. 2005 lfl Macroeconomics 8 PRODU C TIVE EXTERNALIT IES AN D ENDOGEN OU S GRO W TH 237 population growth rate. the initially poorest tend to grow the fastest. Productive externalities © The McGrawHill and endogenous growth Companies. 7. So. 1993. 108. pp. while it seems more plausible that economic growth in Europe and USA has (indirectly) contributed to reduce mortality and create population growth in Africa and parts of Asia.Sorensen. 6 This actually llts with semiendogenous growth. It is bard to believe that population growth in Africa. Two dillerent countries with similar characteristics will converge to the same steady state. Still.
For what it is worth. It must be added.c).4: the clear positive empirical relationship between investment rates and growth rates.. as it stands. 8.. I SorensenWhitta. on the other hand. 'Is Growth Exogenous? Taking Mankiw. Using the data of Bernanke and Giirkaynal< the picture in Fig. 8. that rather natural modifications of the endogenous growth models can take account of convergence. We also think that a serious challenge to endogenous growth theory. What is shown in Fig. Giirkaynak. Productive externalities and endogenous growth © The McGrawHill Companies. There is clearly a positive relationship. 2005 PART 3: ENDO GENOUS G ROWTH The endogenous growth model considered in Section 3. Ben S. 8. and Weil Seriously'. In their article. They then plot these rates against average investment rates across countries. The positive association between investment rates and growth rates could thus be a result of transitory growth. and the existing models. The observed convergence thus contradicts the model and seems to speak against endogenous growth. exactly because this growth rate is exogenous. A. in models of exogenous growth a higher savings or investment rate. s.5 is the strongest empirical argument in favour of endogenous. or perhaps semiendogenous. did not have convergence to a steady state and showed no tendency for the initial position to be important for subsequent growth. . however. but it does so in the intermediate run due to transitory growth. and therefore it does not necessarily contradict the exogenous or semiendogenous growth models. for example. Romer. highly promising. Rather it L~ ongoing. It is thus of interest to llnd out whether investment rates are also positively related to growth rates in A 1• This is indeed an idea pursued in an article by the two economists Ben S. July 2001. 5 emerges. However. The debate from which we have tried to present a few arguments is not settled. The convergence between countries observed in the real world does not deliver a decisive answer to what kind of growth model does the best job. and in the semiendogenous growth model a higher s gives a higher trnnsitory growth in A. comes from the empirical relationship between population growth rates and economic growth rates.. the authors of this textbook llnd that the empirical results linking investment rates to growth rates are sufllciently convincing to make the endogenous growth research programme. The endogenous growth model predicts that a higher savings or investment rate should give a higher growth rate. Growth rates and investment rates The single empirical regularity that speaks most strongly in favour of models of(genuinely) endogenous growth is the one illustrated in Fig. by introducing a gradual technological transmission between countries into the framework. ~One can get an estimate of the average annual growth rate in the productivity variable A 1 by growth accounting. growth models that we know of. Bernanke and Relet S. 8.Endogenous Introducing Advanced Macroeconomics 238 Growth 8. NBER Working paper 8365. could not possibly give a higher growth rate in the technological variable. and indeed high investment rates do go hand in hand with high growth rates. Bemanke and Relet S.Jacobsen: I Part 3. In the models of exogenous or semiendogenous growth a higher savings rate does not give a higher growth rate in the long run. Bernanke and Giirkaynak llrst do growth accow1ting for each of a large number of cow1tries to get estimates of average atmual growth rates of A. Gurkaynak. In the endogenous growth model a higher s will give a higher growth rate in A 1• since the growth rate of A1 is sA .
..023..D <J) 0> I ...... 2.................Douglas production functions where the sum of exponents is often found to be considerably above one.... Productive externalities and endogenous growth Companies......co <O ...... 3...... workers obtain better skills........_ .06 ~~ 0.04 0. if anything........ and adj....Jacobsen: Introducing Advanced Macroeconomics I Part 3....SorensenWhitta............ 2005 8 PROD UC TIVE EXTERNALIT IES AND ENDOGENO US GROWTH 239 0.. then there can be constant returns to capital and labour at the firm level and increasing returns to capital and labour at the aggregate level...05 ... is endogenous: it depends on basic behavioural model parameters and therefore on structural policies that affect these parameters. which underlies the longrun growth in GOP per worker. the individual firm's production function should exhib it constant returns to the inputs of capital and labour at any given techno logical level.. to be a negative one..Endogenous Growth lfl © The McGrawHill 8. W ith productive externalities one can therefore maintain the conven ient assumption of perfect competition and the associated theory of income distribution.. Empirical motivations come..1 and data set for Bernanke and GUrkaynak (200 1).. or with production in general. 1... 84 countries Note: The estimated line is y= 0... The theoretical arguments are associated with learning by doing: by being involved w1th (new) capital... R 2 =0........... while endogenous growth models even predict that a constant positive population growth rate should give an exploding growth rate in GDP per capita....5 $....27. semiendogenous growth models predict a positive ellect of population growth on economic growth... which seems. from attempts at estimating aggregate Cobb .. There are theoretical as well as empirical motivations for productive externalities and aggregate increasing returns .. 8.m o HGK 0.007 + 0........ for instance.03 Average investment rate in physical capital.. 127x.~~Y.. Sources: Penn World Table 6.....':1:~.5: Average annual rate of labouraugmenting technological progress against average investment rate in physical capital......... If there are productive externalities from the aggregate stock of capital (or from aggregate production) to labour productivity or total factor productivity in the individual firm... and at the same time have a potential source of endogenous growth in the model.... ....... 19602000 Figure 8. According to the replication argument. In contrast... In endogenous growth theory the growth rate of technology.. the standard error of the estimate of the slope is 0.......
A 1 = Kf. The intuition is that the increasing returns driving the (semi)endogenous growth can only be exploited if some input increases by itself. ¢ = 1 leads to (truly) endogenous growth. the g rowth rate along the steady state g rowth path was positively influenced by population growth. the growth path for output (and consumption) per worker depend on behavioural model parameters. The negative association between population growth and economic growth found in the data does not contradict the model of semiendogenous growth per se since. In the case ¢< 1. 0 ... The levels of these paths are positively influenced by the rate of saving and investment and negatively influenced by the population growth rate. it took Ku and hence Au as g iven. • Capital accumulated from savings the usual way. Since the firm should be thought of as small relative to the entire economy. The real prices of capital and labour were determined by the marginal products. speaking for polic ies to promote population growth. The feature that growth in the labour force is required for economic growth motivates the term 'semi' in semiendogenous growth. since in the long run the positive growth effect w ill outweigh the negative level effect. equal to. 2005 PART 3 : ENDOGENOUS GROWTH 4. 6. 5. Building on the idea of productive externalities. 8. and the (gross) savings or investment rate was a given constant. G iven that this issue is unsettled. or above 1 : ¢ < 1 leads to semiendogenous growth. and the growth rate along. On the other hand. thereby lowering the level of the basic growth path. we formulated in this chapter a base model w ith the following features: • The representative firm was assumed to produce output (GOP) from the inputs of labour and capital according to a usual CobbDouglas production function with constant returns to capital and labour and a labouraugmenting productivity variable A1• • The labour productivity variable of the representative firm depended on the aggregate stock o f capital as given by a function. Given that the labour force grows. one would be cautious recommending polic ies to promote population growth. 7. and it was strictly positive if and only if both of ¢ and n were strictly positive. This growth rate depended in a specific way on ¢ and n. The labour force was assumed to grow at a given rate. if convergence to steady state is very slow.eI SorensenWhitta. In steady state there was a common constant growth rate for output per worker and for technology. Th s speaks for policies to raise investment rates and dampen population growth. and ¢> 1 (a case we did not consider in the chapter. but which is considered in an exercise) leads to explosive growth and basically an illbehaved model. since it seems relatively certain that higher population growth will erode the stock of capital per worker and increase the pressure on scarce natural resources.Endogenous Growth 8. 9. • n. For the policy implications of the model of semiendogenous growth it is important that both the level of. whereas the positive effect on the growth rate along this path seems doubtful. Empirically it seems hard to find a c lear positive association between population growth and growth in income per capita. It makes a qualitative d ifference in this model if the strength of the external effects is below. Productive externalities and endogenous growth © The McGrawHill Companies. although there are both pros and cons in the debate over whether growth in the labour force really promotes growth in output per worker. the capital stock will grow by a rate higher than n.Jacobsen: Introducing Advanced acroeconomics 240 I Part 3. ¢. the negative level effect of higher population growth w ill dominate . computed for a g iven level of the productivity variable. the model implies convergence of capital per effective worker and of output per effective worker to welldefined steady state levels.
11. and because they have forced growth theorists to think harder about the forces underlying technological change. 2005 8 PRODUC TIVE EXTERNALIT IES AND ENDOGENOUS GROWTH lfl 241 the positive growth effect for a long time. 13. a larger labour force gives higher economic growth and a growing labour force gives explosive economic growth. but close to 1. endogenous growth models have made an interesting and promising contribution to the theory of economic growth. 8.Jacobsen: Introducing Advanced Macroeconomics I Part 3. longrun growth rates in output per worker and in estimates of the productivity variable A. This gave us the socalled AK model. These features are highly counterfactual. and that for </> tending to 1 (here with n = 0). On the other hand these models are plagued by an empirically implausible scale effect: according to endogenous growth models. This is why the model's growth is called 'truly' endogenous. which obtains as if> tends to 1. We took this idea to the extreme and assumed if>= 1. The empirical evidence therefore suggests looking at the model with very slow convergence. Importantly the endogenous growth rate depended positively on the rate of saving and investment. 10. The strength of productive externalities and the speed of convergence: simulations This exercise illustrates by numerical simulations how the speed of convergence in the model of semiendogenous growth (0 <if>< 1) depends on the size of ¢>. . Compared to endogenous growth models. It therefore implies a common and constant positive growth rate of capital per worker and output per worker. The main policy implication of the AK model was therefore that policies to promote saving and investment become even more attractive. The scale effect implied by truly endogenous growth models speak against these models and in favour of models of semiendogenous growth. at the same time assuming a constant labour force. This is in very nice accordance with models of (truly) endogenous growth. because of their ability to account for the observed positive relationship between investment rates and productivity growth. 12. In this case. Nevertheless.SorensenWhitta. convergence becomes very slow so that the model tends to exhibit endogenous growth. semiendogenous and endogenous growth. rather the convergence process will be the relevant thing to study. without any assumption of exogenous technological progress and without growth in the labour force being required.Endogenous Growth 8. on the one hand.6 Exercises Exercise 1. Empirically one finds across countries a rather strong positive relationship between. now not only because of their level effects. the steady state itself will not be very descriptive of the economy.. models of exogenous growth seem easier to reconcile with the empirical evidence on convergence and on the relationship between population growth and economic growth. The everlasting balanced growth described by the model can be seen as approximating the very long lasting transitory growth that would result if </> were smaller than. This model has constant returns to the reproducible factor. We closed this chapter by discussing the arguments for and against the theories of exogenous. investment rates and. capital. but also for their permanent (or very long lasting) effect on economic growth. on the other hand. Productive externalities and endogenous growth © The McGrawHill Companies. however.
o shifts permanently down too'= 0.. 1 . 2005 PART 3: ENDOGENOUS GROWTH We consider the model consi stin~ of Eqs (6) . Comment on the figure with respect to the opening statement of this exercise.Y. 1.lny1) . A= (1. before and after the shift in o.a)]/(( 1. In Y1_ 1) against periods t over the simulated periods for each of the three considered values for¢. Show that the rate of convergence (for tJ> < 1) is: ._. (Of cou rse. After some periods. everything is the same as in Chapter 5 since there is no d ifference in the generic equation studied. Show that the steady state values. Let permanently down to o= 0.91)) s) l/{( 1.a)( 1 .06. of capital and output (also per worker).) Explain in what sense A is a measure of the rate of convergence for ji 1• 2. compute the steady state values for K. In one d iagram p lot the g rowth rate (Y1. 10 say. Hence. . A goes to 0. This is just a matter of defining G appropriately. The strength of productive externalities and the speed of convergence analytically The transition equation.9 . Hence we have excluded longrun g rowth due to population growth from the beginning. Simulate the model over the periods before the shift and 200 periods after to c reate series (K1) and (Y1) . Go through all the operations that lead to th is equation yourself.lnf1 = A(lnf*.)/Y1_ 1 (or the approximate one. Consider three d ifferent values for ¢>: 0 .a)( l . It is important. fall o' = 0 .aj(1 . For each value of ¢>. though. Try to explain intuitively why a large ¢> imp lies s low convergence. For each value of ¢.24 everywhere.. are: s )la+91( 1.( 1 ~ 1 ~1~~ <I>J] . using the same kind of operations that were used in Chapter 5 to arrive at that chapter's equation (3 4). 1 a)(1 Convince you rself that in the special case tJ> = 0.75 and 0 .08. where A= 1 . In Y. Show that as¢> goes to 1.91)1 K* =( and c5 Y* =(0 .(9). 1. this rate is in accordance with the rate found in Chapter 5 . that you can do the full linear approximation you rself. (13) or (14). Productive externalities and endogenous growth Growth Companies. we arrived at in this chapter for tJ> < 1 can be w ritten in the generic form k. a linear approximation around steady state of the transition equation studied in this chapter is: lnf. K* and Y*.. What is the steady state value of A1? In the following let a = ~and s = 0.5.Endogenous Introducing Advanced acroeconomics 242 © The McGrawHill 8. respectively. Exercise 2. and Y.Jacobsen: I Part 3. let the economy be initially in steady state for o= 0 .¢) = (1  ( 1 +n) W 91> (1o) ( + n) W 9~> 1 ¢)[1.06.8 I SorensenWhitta.G'(k* ). but we set n = 0 and we let the constant size of the labour force be L = 1. 2.08 initially and after a shift. 3. 0. 1 = G(k1) just as we did in Section 4 of Chapter 5 for the transition equa tion studied there.
. The model of semiendogenous growth (rp < 1) with the productive externality arising from KtfLt rather than from Kt In our d iscussion in Section 3 of the scale effect in connection w ith the model of endogenous growth (¢ = 1) we said that one can eliminate the effect by assuming a productive externality of the forrr: A.9\)J and * = (s)la+~(l a)J/((1 ")(1 <i>)J. 2005 8 PRODUC TIVE EXTERNALIT IES AND ENDOGENOUS GROWTH lfl 243 Exercise 3. of course.SorensenWhitta. k"" and j *. andy. 2. for instance. . be stronger with ten computers than w ith one.(sk. k. however. Productive externalities and endogenous growth Companies. The size of¢ is.'. at least not for steady state. semiendogenous type of growth in GOP per worker in steady state? The fact that the presence of the external effects does not produce (semi)endogenous growth in the steady state this time may give the impression that¢ is not economically important. 3. The direct {privately appropriated) productive effect will. The model to be considered is thus (6) .1 + (1 b))~ ~. is 0 (also when n> 0). How about the indirect (external) effect? 1. Before turning to the more technical questions. = (1)' ~k.. Find the steady state values. follows from the properties of the transition equation the same way as in this chapter.. What is the aggregate production function in this model? Is labour unproductive at the aggregate level (also when¢ < 1)? Why is it that in this case there is no positive. and Yu respectively.= (~J rather than the form A.(9). 1 +n Convergence to positive steady state values. Show that the constant levels of capital per worker and output per worker in steady state are: k * ( S =  n+O )1/[(1. _l.!il_ A. of a firm w ith ten workers beginning to use computers. = (K.Endogenous Growth © The McGrawHill 8. The assumption on¢ is 0 < ¢ < 1. Show that now: (~\<!> A. important both for the steady state levels and for the convergence process outside steady state.a)(l .jL) <~> most convincing for the productive externality. = Kf or the form A. Let us now consider the model of semiendogenous growth (¢ < 1) w ith the alternative form of the external effect. the transition equation is: . Should the external effect on labour productivity from the use of machinery be expected to come from the amount of capital as such or from capital per worker? Think.. y n+J .. but w ith (7) replaced by the equation above. as the answers to the following two questions should reveal.( 1 + n) ~ ' and that with the usual definitions.Jacobsen: Introducing Advanced Macroeconomics I Part 3. for k. Show that in steady state the common growth rate of A" k .= Kf mainly assumed in the chapter. try to think about whether you find the form A. k"' and f". 4..
2005 PART 3: ENDOGENOUS GROWTH 5. 0 y. that is. This exercise asks you to demonstrate that the uncovered cases lead. when we were setting ¢. of course. = . Show that the (exact) growth rates of k. Write down the transition equation for K.1 .. For instance.(9). in period tare: 1 _.u( 1 + n)<1. Show that (again) A goes to 0 as t/> goes to 1.Jacobsen: I Part 3. andy P respectively. but¢. is constant. (Essentially you are asked to show that all of capital per worker. Is anything puzzling you about this value? Write down the Solow equation forK. 1+n ° and Yt+ 1.a (and since t/> > 1. Then show that the growth path of the model of endogenous growth considered in Section 3 (with n = 0) is one of balanced growth. 1. +(t . is initially greater than K* how does the growth rate of K. to explosive growth. the exponent on K. ( 1 + n)" 2. a very unrealistic feature. _" (1 + n)<t . = 1.y.) Exercise 4.= AK..a)t _ (n + o)] k k k. This is. (The latter result means that the endogenous growth model (¢.'ll. In that model. = 1) without the scale effect approximates the behaviour of the model considered here for large ¢ . according to the linear approximation. has a unique steady state value: K* = [ c5/(sA)] t/ [(t . and sketch the Solow d iagram.a)t + 1 .[sL . Since the models assume a constant growth rate of the labour force (possibly 0) the other requ irements for balanced growth follow. or may lead. = 1 and n > 0 have not been considered.) Exercise 5_ Explosive endogenous growth You may have been wondering about some cases not dealt w ith in this chapter.8 I SorensenWhitta. etc. just below 1) we assumed at the same time n = 0 .? What if K. lett/> = 1 and n > 0 (so that L 1 = (1 + n) 1 L0 increases at a constant rate n). > 1._ +1__.1. Consider the growth model consisting of the equations (6) . Productive externalities and endogenous growth Growth Companies. we have considered ¢. Is K* stable or unstable? Also state the modified Solow equation and draw the modified Solow diagram. Furthermore.1 l. If K. is greater than 1).Endogenous Introducing Advanced acroeconomics 244 © The McGrawHill 8. to a growth rate in GOP per worker that increases over time and eventually goes to infinity. < 1 and ¢. and show that K. This was perhaps for a good reason. evolve over time? How about the growth rates of k . L say). < K* initially? . where A e L t . = 1 in Section 3 (as an approximation of a large ¢. Balanced growth Show that the steady state growth path of the model of semiendogenous growth cons idered in Section 2 is one of balanced growth.a)(~"> . output per worker and wages per worker (the wage rate) grow by one and the same constant rate and that the real rental (interest) rate of capital is constant along both of these growth paths. Show that the aggregate production function can be written Y. find. = __ [sL 1.aXI'> . but not ¢. Now lett/> > 1 (but say also t/> < 2. the rate of convergence A for y. By considering again the linear approximation around steady state. andy. also for th is model.o] . not to be extreme) and n = 0 (so that L.
Show that in longrun equilibrium g is also the constant and endogenous growth rate of Y1 and of y . 1. x1 "" K.c5./H1 tends to its longrun equilibrium value x*. H. these two growth rates tend to the common value: g = s. Define the ratio between physical and human capital. respectively. 2005 8 PRODUC TIVE EXTERNALIT IES AND ENDOGENOUS GROWTH lfl 245 Exercise 6.u . = AK~H) ."' and v = a/(a + f!J). . Show that according to the model above the growth rates of physical and human capital.. where A = L 1 . . but without human capital. K..o)K1 H. respectively. has been put into A and the sum of the exponents on physical and human capital add up to 1. so there is not diminishing. = (K'/)"(H'()'"(A.Endogenous Growth 8.J_'/) ' ""'. Productive externalities and endogenous growth © The McGrawHill Companies. and an external productive effect implies: then if </> = 1. The first equation is the aggregate production function where the influence of the constant labour force. How should we view this model bearing in mind that a knifeedge case such as ¢ = 1 is not of interest in itself? 2.".SorensenWhitta. = g. but constant returns to the reproducible factors taken together.vA. Endogenous growth with both physical and human capital In this chapter we presented endogenous growth models with physical capital. The following three equations make up a model of endogenous growth including human capital: 0 < v < 1.. 1 = sHY1 + ( 1 o)H1• Y.Jacobsen: Introducing Advanced Macroeconomics I Part 3. First show that if the production function of the individual firm is: Y. just as in Chapter6.. K.s~. H. The next two equations describe the accumulation of physical and human capital. = and then show that as x 1= K.. Notation and parameter values are as usual if not explicitly stated. L. the aggregate production function above resu lts. g. in any period tare: K  K. . K1• 1 =sKY1 + (1. H  H. /Hr Show that the model above implies the following dynamic equation for this ratio: and demonstrate that this equation has properties that imply that in the long run x1 converges tox* =s Kfsw 3.
3.</J{1. L 1+ 1 )[</1(1. but also on sH.a)J t (1 +n . 2005 PART 3: ENDOGENOUS GROWTH 4.1 +( 1 0)] (1. growth in GOP per worker depends positively not only on sK as tested in Fig.9'>> > (1 . Comment w ith respect to what c reates growth in output per worker in this model (when</> < 1). Exercise 7. Show that in this model: A 1+ 1 (K . g se• of output per worker in steady state. Productive externalities and endogenous growth Growth Companies. To test th is create a d iagram that plots growth rates in GOP per worker.a)]/(1 .</J{l .a)J/(1 .. a and ({! should be of about equal size. 2.. According to this model. . Semiendogenous growth and endogenous growth if the productive externality arises from Y1 rather than from K1 As mentioned in this chapter.. Assume now that ¢< 1. )a</J/11 1~(1 a)) ( A. Let j 1 . Find the expression for the growth rate. 1 a where the latter equation replaces (7). Show that the transition equation for k1 is: 1 ) 1/(1.a) I k = __ k[s/(a .¢)/ [I . to a test by plotting growth rates Comment. 8. ( l +n = __ Find the steady state values for k1 and j 1. From the empirical observations we described in Chapter 6.o) (which we assume)./(A 1 L1) . gi. Show also that if tJ> = 1 it has constant returns to K1 alone.a)) (sfi". We therefo re consider a growth model consisting of the four equations (6). Show also that the transition equation implies convergence to steady state.4. sometimes models w ith productive externalities assume that the external learning by doing effect arises from workers being involved in production as such rather than (more narrowly) from their being involved w ith the use of capital..q>(1.a)l ./(A 1L1) and j 1 ey. Show that the aggregate production function in th is model is: Why have we assumed </> < 1 /(1 a)? Show that the aggregate production function has increasing returns to (K. Comment.</J{I .against (s~) 1 12 (s~) 112 again taking data from Table A. (8). 4.4 and to the figure plotting g' against s~.</>) +(lO)k(I. s~ across countries i w ith data taken from Table A (avoid countries w ith a quality grade of D). Estimate the line of best fit. L 1) whenever ¢> 0. t 1 ) 1/(1 . .= I<.<P)/(1.</J{I . Compare w ith Fig. . 1.</J{ I.Endogenous © The McGrawHill 8. against investment rates in human capital. and show that these are meaningful whenever (1 + n) 1/(1 . which should here imply v = ! · Put the model's full (endogenous) growth prediction g. (9) and: 1 0< ¢ < .a)] L..11(1. 8 .eI SorensenWhitta. and k1 be defined as usual: k 1e k1/A 1 e K.9'>))<1 .)1/(1 .Jacobsen: Introducing Advanced acroeconomics 246 I Part 3./A 1 e Y. .
9. Show that the aggregate production function (where it is taken into account how A. =rY. Find the growth rate. 2.+ (1. and comment.a )/a 1 K.. Show that the model can be condensed to the two equations: Y.= aG. so government expenditure in period t is: G.. Show that the model is equivalent to an AK model (like (20) and (2 1) in the chapter) with appropriate relabelling of parameters. etc.L.Sorensen. Income is taxed proportionally by the rate r.= L 1. etc.. Taxation and productive government spending: endogenous growth without productive externalities There is a simp le alternative to productive externalities as a mechanism for endogenous growth. Comment. 5.. which yields basically the same models as studied in this chapter: productive government expenditure financed by (proportional) income taxation. Journal of Political Economy. so L. through taxation. pp. . and n = 0.r) Y.) 1 . depends on Y. is equal to some L for all t. Productive externalities and endogenous growth Growth © The McGrawHill Companies. 2005 8 PROD UC TIVE EXTERNALIT IES AND ENDOGENO US GROWTH lfl 247 Now assume ¢ = 1. S103S125.a>faK.(5)K. health care. . a > 0.Endogenous 8. so the tax revenue in period tis r Yl' The budget is balanced. 1 = s( 1 . Government spending (on infrastructure. 0 < r < 1." .) is: Y1 = (arL) <l . Finally we assume a constant labour force: L. Exercise 8. in all periods. Robert J.. 9 and consists of the five equations below.Jacobsen: Introducing Advanced Macroeconomics I Part 3.. Total savings and investment are given as a given fraction of total net income. 1990. 1 = sY. the legal system.= AK~ A "' L( 1. The model studied in this exercise comes from Sarro (1990).) is assumed to be 'productive' in the sense that productivity is influenced positively by expenditure: A.Whitta. ' Government Spending in a Simple Model of Endogenous G rowth'. The production function is as usual with a labour productivity variable A. Find the growth rate of capital per worker and of GOP per worker that the model gives rise to.: Y1 = K'((A. g •' of output per worker. so the capital accumulation equation is: K. + ( 1  o)K. Sarro. 98. for instance. with respect to scale effects.
and find the growth rate of output per worker and of A 1 in steady state. e. if labour supply was an increasing function of the aftertax real wage? Consider a generalization of the model where the equation A 1 = aG.)¢.L1 ). Analyse the model and find the growth rate of GOP per worker. Comment with respect to the scale effect and the aggregate productivity of labour. Discuss the policy implications of your analysis. G. what kind of model do you think g ives the best understanding of economic g rowth? This is not the type of exercise that has one correct answer. 5. do you think./L.g. At this stage. Get together w ith fellow students and line up the most important pros and cons for each type of model. leading growth economists still discuss this issue between them.= a{G. {Hint: Exercise 7 can be of help here.. Exogenous or endogenous growth models? You have now seen some models of exogenous g rowth and some of endogenous growth and you have been presented w ith some pros and cons for both types of models.. verify con  vergence to a steady state. 4. Productive externalities and endogenous growth Companies.. is replaced by A .))? Describe the opposing forces that determine the growthmaximizing tax rate./L. How would policy implications be affected. Find the transition equation for k. and n = 0./A 1 "" KJ (A . if there were traditional d istortionary effects of taxation.e k./L. so that A. above is rep laced by A .. that is productive. as well as for A. Is there a scale effect? Assume alternatively that it is government expenditure per worker. 1.Endogenous Growth © The McGrawHill 8.= a{G. 0 <¢<.= L is replaced by L . so let us now assume ¢ < 1 and n > 0. =aG. and L.eI SorensenWhitta.) Exercise 9. 2005 PART 3: ENDOGENOUS GROWTH 3.. 1 = {1 +n)L 1• The model above corre sponds to ¢ = 1. Analyse the resulting model as we have analysed models of semiendogenous growth in this chapter going through the following steps: find the aggregate production function and describe its returns to scale. . and finally try to make up your own mind on the issue: exogenous or endogenous growth.Jacobsen: Introducing Advanced acroeconomics 248 I Part 3. In fact. What value of r maximizes the rate of economic growth {for the case A 1 = aG.= a{G. Then d iscuss which arguments you find most important and convincing.). Comment with respect to policy implications.
Economic resources in research and development (or in innovative ellort in general) produce ideas. in the real world a lot of production explicitly intended to create technological progress tal<es place.. R&Dbased endogenous growth: Macroeconomic modelling © The McGrawHill Companies. A. The exogenous growth models of Part 2 postulated a fixed growth rate in the technological variable A" so the driving Ioree of economic growth was basically unexplained. No agent was concerned with creating technological progress. A 1• and technological progress '>Viii mean that A 1 increases. so growth in A 1 was endogenous and 'explained' by growth in capital. which is not described as arising from an intentional production process. and on the other hand. We will now begin to study growth models that bring research and development. and it may bring us better tools lor engineering growthpromoting policies. In this chapter and the next we will study growth models in wh ich technological progress is the endogenous and deliberate outcome of a production process which requires productive inputs. In the externalitybased endogenous growth models presented in the previous ch apter. This is a bit of a paradox. In our models technological progress will have a very precise meaning: there will still be a technological variable.rith producing better technology and production methods in private companies as well as at private and public universities and other organizations engaged in the creation of new knowledge.Endogenous Growth 9. on the one hand. We view the production process in the R&D sector as follows . but technological progress was an unintentional byproduct of economic activity in general. the growth models we have considered so far trace economic growth back to tech nological progress. Thus. A 1 was linked to aggregate capital (or income). Macroeconomic modelling .. and no inputs were used lor the purpose of producing increases in A 1• In the real world many people are occupied v. This is an end in itself._ n all our growth models the ultimate source of growth in GDP per worker is technological progress.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3. Up to now technological progress has been either unexplained or unintentional. R&D.+1 . Our purposes are the usual ones: if economic growth is rooted in technological progress.A" from period t to period t + 1. then by better understanding the production process lor technological progress we gain a deeper understanding of growth. 2005 lfl •• • • Chapter :R&Dbased endogenous growth __ . These ideas are I 249 . production processes explicitly into the picture. The ultimate output of the R&D sector in period tis thus the total technological progress.
as will be further explained in the next chapter. so the R&D production process has to be labour intensive. such as Nev. or better 'tools' for existing machinery. Furthermore. Journal of Political Economy. such as the steam engine or a new generation of computers. The model we present in this chapter and the next is known as the Romer model of endogenous growth and was developed by Paul M. We thus view the R&D sector as follows:  Inputs: The innovative ellort of labour and the existing level of technology create as Direct output: new insights and ideas which result in Lhe Ultimate output: a new and higher level of technology.S 102. The fact that ideas are partly or fully n onexcludable implies problems with private production. such as a new or improved piece of software lor existing computers. Romer over several papers. They are 11onrival: the fact that one person uses calculus or (the idea behind) a specific spreadsheet in no way limits another person's ability to use calculus or the same kind of spreadsheet. Ideas can be new insights in basic research. The reason is that ideas. 1 1. The fact that the direct output of the R&D sector essentially consists of ideas makes the associated microeconomics more subtle than in the growth models we have considered earlier. Macro models close to the one we study in this chapter had been suggested before. ' R&D·based Models of Economic Growth'. better tech11ology is created by new ideas. They can also be more 'closetothefactory· inventions of new machinery. although in the real world also some capital is used in R&D activities. So. that Romer has become famous for. Einstein's theory of relativL~m. varying from absolutely nonexcludable (no person can be prevented from using calculus) to partly excludable (in principle people are excluded from using a spreadsheet they h aven't paid for because using it without a licence is a violation of law.Endogenous 9. . Ideas are also impe1jectly excludable. We postpone an explicit treatment of the subtle microeconomics associated with (private) R&D production to the next chapter.rton's mechanics. be it under perfect or imperfect competition. pp. that private production of ideas cannot take place under perfect competition . It should be emphasized. 2005 Macroeconomics 250 modelling PART 3: ENDO GENOUS G ROWTH what ultimately bring better technology. 759784. however. etc. pp. We assume that the only inputs in the R&D sector are the labour power of 'researchers' and the existing stock of knowledge. Ideas come from people..based endogenous © The McGrawHill Growth growth: Macroeconomic Companies. We study the implications of these. such as the assembly line. calculus.Jacobsen: Introducing Advanced I Part 3. In this chapter we present a growth model of a more conventional macroeconomic character. We present a generalized versi3n of the model due to C harles I. they can be ideas for new and more productive ways to organize production. 1995. 103. includin£ the microeconomics of the next chapter. viewed as economic goods. Bohr's atomic theory. 1990.eI SorensenWhitta. 98. Imperfect competition is therefore a necessary ingredient in a model of privately conducted R&D. It is the full model. The nonrival character of ideas implies. Journal of Political Economy. in particular ' Endogenous Technological Change'. increases in A. or they can be ideas for new intermediate or tina! products. S 7 1 . that is. are like public goods. where we interpret ideas in a broad sense. R&D. Jones. that the model is not ad hoc: one conclusion in the next chapter is that the properly microfounded model studied there implies (more or less) the macro model we study in this chapter. but pirate copying takes place). It abstracts from the underlying microeconomics and postulates a number of growth and technology relationships.
... (1) Formally. the amount of which we assume to be proportional to the stock of capital.?. Consequently..1: llrst. the firm is actually to be considered as one of many small firms and therefore it only has a negligible intluence on the aggregates of the R&D sector. According to the replication argument....... K. .~~I?. ..~...for a given inp1. Second..0..~. S... enters as usual. and the inputs K1 and L1.~t::~?. Doubling production by undertaking the same activities twice means doubling the inputs of capital and labour .. A... We will now describe the production processes for each type of output.. can be used in every tirm because the general level of technology is nonrival. From now we '<Viii think ofA 1 more as an input in line with K1 and Ly1 • The variable A 1 is the total productive eflect of the stock of all innovative ideas that have come into existence up to period t. 2005 9 R&DBASED ENDO GENOUS G RO W TH: MACROEC ONO MIC MO DELLI NG 9.... . Ly1 and A 1. and fmal goods that can be used for consumption or for investment in physical capital. the only new feature is that labour input is now denoted by L1. Finally the technological variable. As usual we should distinguish between its two roles.... The production of new technology The R&D sector also has one representative llnn.tt oft. 1 rather th an by L.~~~1?....2 per cent per year in the developed world. A . produced in the sector in period t. a feature that is important for our conclusions.. R&D...Endogenous Growth 9. and thus does not import technology developed elsewhere. A. it produces the entire output.... On the one hand. ~ .r.SorensenWhitta.. .. I?.... LA< . A:+ 1 . it is implicit that the model covers the entire developed world. LAr• and on the total output.~~~~.... The production of final goods A representative firm produces llnal output in the amount Y 1 according to the familiar Cobb.. Another input is capital services.. .. There will be two types of output: 'new technology'..<:>. the production function should have constant returns to the two rival inputs. that is..Jacobsen: Introducing Advanced Macroeconomics I Part 3. On the other hand. the representative llrm cannot adjust its input of technology as it wishes...?.. the production function in (1) has increasing returns in the three inputs.. just lil<e the firm in the final goods sector does...+ 1 ..A. There are important diflerences between the input A..A 1. the representative llrm in the R&D sector tal<es the stock of technology as given at any point in time.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies.. K. the fixed amount of technology... As a consequence.Douglas production function: 0<(1<1..1 lfl 251 ~?:. co11sidering tile developed world as one large eco11omy. Since the economy we conslder will produce all of its technological progress itself.. of new tech nology in period t (since there is only this one firm) . This is because there '<Viii also be a labour input... is given...4. A...eclmology. .. on the total input oflabour.. in the R&D sector and we have to distinguish between the two. The model we study in this chapter is meant to explain the relatively constant economic growth of around 1..... society's stock of ideas or technology... K1 and Ly1 • as assumed in (1).. From the point of view of each individual llrm.
while if A = 1.Whitta.. Rather. Two Hrms creating th e same idea only count for one in the stock of knowledge. Hence.. have constant returns to the input of labour alone. If the combination of the 'standing on shoulders' ellect and the 2. the more difficult it becomes to find a new one. respectively. This ellect speaks lor a positive¢.Jacobsen: Introducing Advanced I Part 3. An explic it modelling of stochastic arrival of ideas would not change our results qualitatively.. they are productive in research and development. LAt• in the R&D sector: p > 0. R&D. then also at the level of the whole R&D sector. As already explained. and L~.. 0 <A. L. each individual firm in the R&D sector takes A. just as it is for the producers of final output. At the aggregate level we will assume that there may be a negative externality from the aggregate use of labour. so the more ideas that have already been harvested. (2) where a. and L~.+ 1 . By d oubling labour inputs.. a doubling of l abour inputs could doubl e t he stochastic arrival intensity of ideas (in a Poisson process describing the production of ideas). We are abst ract1ng from an important aspect of reallife R&D. A. the more probable it is that several individual firms work with the same ideas. If there is no negative 'stepping on toes' ellect. 1. . = L At' so from (2) and (3) the aggregate production function for new technology is: (4) Note some pan icular cases. we mal<e the simpli{ying assumption that capital is not required as an input in the R&D sector. the more difficult it becomes to get new ideas. and the total labour input. This 'fishing out' ellect is based on the view that there is a given. The motivation for the possibility ofsuch an ellect is 'stepping on toes': the more research there is in the R&D sector. Hence the production function should. = A1+ 1 . large (possibly infinite) pool of potential ideas and the easiest ones are discovered first. A . existing technology may promote the production of new technology. according to the replication argument. (3) If A< 1.4 . an R&D firm cannot be sure of doubling the number of ideas it w ill produce in a deterministic sense since ideas arrive in a less predictable way. an ellect sometimes referred to as 'standing on shoulders'. have to equal the total production of ideas.. as given). is an input for the R&D frrms.A. Just as existing insights. For the time being we will not impose any restrictions on the size or sign of¢.Endogenous 9. its output elasticity being¢ (but like the firms in the final goods sector. production methods. namely the stoc1astic nat ure of the arrival of new ideas. there is no such ellect. but generally we will think of ¢ as positive.. A. are productive in the final goods sector. On the other hand. In equilibrium a. = 1. possibly equal to 1.A . the number of ideas produced is proportional to the labour input in the sector. A.based endogenous © The McGrawHill Growth growth: M acroeconomic Companies. etc. 2 The existing stock of knowledge. A 1. is the labour demand of the individual firm in period t .eI Sorensen. This eiTect speaks for a negative value of¢. then there is a negative spillover from the aggregate activity level in the R&D sector to the productivity of the individual finn. . it may be that the more ideas that have already been accumulated in the existing stock of knowledge. computer software. In equilibrium one must have a. is the number of innovative ideas produced by the individual firm. 2005 Macroeconomics 252 modelling PART 3: ENDO GENOUS G ROWTH The production function of the individual firm is: p>O. and L~.
endogenous variables that can adjust in the period in circles.+ 1 =sY. Equation (10) assumes that in all periods a given and exogenous fraction. 0< SR < 1.. (8) is the familiar equation assuming that the labour Ioree. Likewise. 2005 9 R&D BASED ENDOGENOUS GROWTH: MACROECONOMIC MODELLING 253 'fishing out' ellect is such that¢ = 1.A.1.'(A . and A. The first two are the aggregate production functions lor the final goods sector and the R&D sector. as arising from individual behaviour than to think of s and n as representing an aggregation of individual behaviour. K. . You may llnd it more dillkult to think of s R · the share of people working in the R&D sector.based endogenous growth: Macroeconomic modelling Growth Companies. ¢. ¢. then ( 4) directly gives the growth rate of technology as: (A. The state variables of the above model areK" L. (7) (8) (9) n> . b.1: The dynamics of the R&Dbased macro model Note: Predetermined endogenous variables in squares. . A and<~ are technical parameters.<~. A main element of the microeconomic model of the (5) Figure 9.)/A.Jacobsen: Introducing Advanced Macroeconomics I Part 3. . p.. = pL~.<~)K. The third one is the usual capital accumulation equation assuming an exogenous savings (investment) rate. referred to as the R&D share. 1'1. The complete model Our complete model consists of the six equations below. where sR. R&D. s. grows at a fixed exogenous rate. while (1. (5) A 1+ 1 .+(l .Endogenous lfl © The McGrawHill 9. L. respectively.A.) 1 ".+ 1 . 9 . (10) The parameters of the model are a.Ly.SorensenWhitta. and an exogenous depreciation rate. L. p. s R• of all labour is used in the R&D sector. (6) O<s<l.1 Ly.. s. Equation (9) says that the sum of the labour inputs in the two sectors has to be equal to the total amount oflabour available. the model will determine the full dynamic evolutions of all endogenous variables as illustrated in Fig. + LAt = L. = K. nand sR. For given initial values K0 • L0 and A 0 .+1 = (l + n)L" O<b<l. The two last equations are new. Y. should be thought of as a behavioural parameter in line with sand n.A 1 = pAfL~.
Thus.0 0.4 1.4 2.2000 Extrapolated century growth rate 1981 2000 (%) (%) United States European Union Japan 2. reflecting that labour power is a rival good. so should the marginal product in one sector or the other equal the real wage.4 75. Technology and Industry Outlook. R&D shares seem to be around 23 per cent in the year 2000 with some tendency to be increasing over time. For now we will have to treat it as exogenous and look into the consequences. as a percentage of GDP in some major world areas and in some specific countries.7 2. A cmde extrapolation suggests that R&D shares have been increasing at a pace corresponding to a doubling every 100 years.1 2. R&D.6 1. while the stock of knowledge can be used by all frrms in both sectors simultaneously. reflecting that knowledge is nonrival.1: Gross expenditure on research and development (GERD) as a percentage of GOP Average annual 1981.Endogenous Growth © The McGrawHill 9.2 0. taking s R as a given parameter.based endogenous growth: Macroeconomic modelling Companies.1 * 3.2 0. 2005 PART 3: ENDO GENOUS G ROWTH next chapter is exactly an endogenous determination of the R&D share. Table 9.7 1.8 1. the microeconomics and the microeconomic problems of it are visible in certain respects as the model now stands.7* 3.9 3. Furthermore.0 2.0* 1.4 1.8 112. However. and how does it seem to change over time? Table 9 . Labour is used in two sectors.3 1.2* 2. Typically.9 growth rate Main World Areas Countries Denmark Finland Ireland Netherlands Norway Sweden United Kingdom • Data from 1999.2 2. there is no equation for a factor reward rate in the model.7 1.1 reports on the socalled GERD (gross expenditure on research and development). Source : O EC D Science. ofits size. note that labour inputs used in one sector cannot be used in the other. how large should we typically expect it to be.7 107.eI SorensenWhitta. 2002.8 0.4 Total OECD 2. .1 1. It is not clear how the real wage rate should be determined.2 515. or should the two marginal products be equal to each other? The determination of the wage rate will have to await next chapter's microeconomics. On the suppressed microeconomics of the model As mentioned. rather than the causes.8 * 1.Jacobsen: Introducing Advanced Macroeconomics 254 I Part 3. For instance.2 2. we will deal with the microeconomics of our growth model in the next chapter.
. Should it not be this aggregate income rather than just the Y.) for A. +1 A 1)/A 1 + 1 = g . equal to the growth rate plus 1: A 1+1/A 1 = (A. Hence. 2005 9 R&DBASED ENDO GENOUS G RO W TH: MACROEC ONO MIC MO DELLI NG lfl 255 You might also ask what the appropriate measure of GDP and national income in this model should be? There are two production sectors. =pAfl (sRL. we get (with sRconstant) : 9 ~~~ = (A~~TI (L~~ l Here A.Endogenous 9. The R&D sector is nm by the government and its output is not marketed. c1 = (1 .A . As long as just some labour input is used in the research sector there will be positive additions to knowledge all the time. With respect to welfare. the Y1 of the llnal goods sector. =L1 t+l . People working in the R&D sector earn incomes. is the growth factor of A 1.)' . insert s RL.s)y . but the government pays these incomes and llnances its expenditure by taxing all income. so GOP will have to be some aggregate of the productions. but simply made freely available to everybody.)"'. as a means of increasing production of final goods. so the total tax revenue equals the labour cost of running the research sector.+ 1. th at is. we will take the view that production of new technology does not create welfare in itself. = The law of motion We can describe the dynamics oftbe model by two difl'erence equations that we now derive. Y. y.+ 1/A . If we write (11) for period t + 1 as well as period t and divide the first equation by the second.based endogenous growth: Macroeconomic modelling Growth © The McGrawHill Companies. that tech nology creates welfare. r pA f. and insert 1 + n forL. and we can then. respectively.Jacobsen: Introducing Advanced Macroeconomics I Part 3. total income after tax is the sum of the (values of the) productions in the two sectors minus the production in the R&D sector. of the two sectors.. This aftertax income enters into the savings function. ( 11) This shows that g. The dynamics of the rate of growth of technology From (6) we can compute the growth rate in technology in any period by dividing both sides by A:. on the righthand side: g.4 ..A A.SorensenWhitta. using (10). and A 1+ 1 .+ 1/. is strictly positive in all periods.1(1 + n)". Ifweinsert (1 + g. or rather the consumption of final goods per person. It is only indirectly.1L~. for the LA. that entered into the savings function? You will again have to wait until next chapter's microeconomics lor a more explicit treatment of th is issue. of course. R&D. 9t which wecan write as: (12) . Hence our measure of the prosperity of a society should (still) be the production of llnal goods per person.. Yj L. For now you may think of it as follows. +JL" we get: 9 ~+ 1 = (1 + g.
and g1. In this case one gets from (11) the following growth rate of technology: (14) If¢ = 1. since it has the endogenous g.. R&D. = k~(1 . cannot keep pace with the stock. one only needs to know g . (8) and (10) . (A 1 + 1 .. This is easy to see !rom (14): as A 1 goes to inflnity.Jacobsen: Introducing Advanced I Part 3. the absolute change in technology is given directly by (6).+ 1 from (13). but ( 13) is not a law of motion fork. the increases.L. assume that L At is a constant. At. assuming a constant labour input in the research sector.+ 1 L 1 we then get: +" (13) This has been derived in the same way that we derived transition equations in models considered earlier./A. the growth rate ofknowledge.sR)L 1) la. and as A 1 becomes larger and larger. = K. The dynamic system also involving the economic variables Let us define the usual auxiliary variables. the technological growth rate. one must know both k. 111+ 1 . It has been derived exclusively from the three model equations that concern the research sector. on the righthand side.A. 2005 modelling PART 3: ENDO GENOUS G ROWTH This is a flrstorder difference. we can write the production function as Y. However. the stock ofknowledge. A flrst conclusion from our model is thus that the growth rate./A 1• Using (5). '>\Till be constant and positive and equal to pL~ in all periods. The two dillerence equations are only coupled in one direction: to compute k. (7). . has to fall. the system consisting of (121 and (13) is a law of motion in g. A.A1 = pA fL~.+ 1 from (12) . in technology will also become larger and larger... andy. A 1 + 1 . Thus. A 1 + 1 . This implies that A 1 '>Viii be ever increasing and go to inllnity in the long run. There '>\Till be increases in knowledge all the time. (9) and (10). Technological dynamics for tfi = 1 and for 0 < 1/J < 1 It will be high ly illustrative and indicative for what goes on in this chapter to focus for the moment on the independent dynamics of technology. equation in g. k. IfO < ¢ < 1. and k. the increase.sR) 1". (6).A .based endogenous © The McGrawHill growth: Macroeconomic Companies. or transition.. alone.A. LA> 0. A. but to compute g. This time. Hence.1 goes to 0 (remember¢< 1). g. but when¢< 1. = K. . From the capital accumulation equation. = Y. the growth rate of technology will decrease over time and eventually go to 0./(A 1L.Endogenous Growth Macroeconomics 256 9. Dividing both sides by A. gives y. This makes the dynamic system relatively easy to analyse.)/ A. dividing both sides by A. which is positive. because the existing stock of knowledge is not sulllciently productive in the creation of new knowledge..) = k. implying that eventually A 1 goes to in flnity./(A 1L 1) = y . A" '>Viii also be increasing in all periods and go to inflnity: in period t.eI SorensenWhitta. g!' of technology has a dynamic evolution of its own: from any given initial value it develops independently of the other endogenous variables in a way that only depends on the parameters¢.'(A 1(1 . and n (as long ass Ris constant). however.
.... lor any g.. consider the independent dynamic equation (12) for g 1• You will easily verif)r that the transition curve defined by this equation has properties as illustrated in Fig....~.....)..Endogenous © The McGrawHill 9.. > 0. R&D.......)4>.2: The independent dynamics of g.I. 9.... ..<?.+ dFirst.... in this case..~~~~. .. or just an ever increasing stock of knowledge with the growth rate ofthL~ going to 0 (0 < ¢ < 1)........ 9 .+1 = 1 (1 + n)(l + g.....b)k.......2 (1 + ¢g..... Convergence to steady state Let us repeat the dynamic system we have arrived at: (12)  k. &..). is strictly 9t+l 9se Figure 9. also assume that there is some growth in the labour Ioree. its slope.....2 (remember that g..)  (sk~(1  sR) la + (1 .S........ 2005 9 R&D BASED ENDOGENOUS GROWTH: MACROECONOMIC MODELLING lfl 257 It makes an important dillerence whether a fJXed input of labour into the research sector creates a constant positive growth rate in technology (¢ = 1)..?.. . so Fig.... In this section we consider 0 < ¢ < 1.. 9..~I. 0).. n > 0. (1 + n)~(1 + g..Whitta...i..........based endogenous growth: Macroeconomic modelling Growth Companies. ... it passes through (0.Sorensen... (13) First.2 is only drawn lor positive values ofg. g...3 §...J:.~~J?:?.. Second........~~.t.Jacobsen: Introducing Advanced Macroeconomics I Part 3. For reasons that will become obvious we will..r. and g..
and the implied evolution of other endogenous variables). this will also be the implication of the correct dynamic system consisting of (12) and (13).. The new equation would look almost exactly like our transition equations studied earlier. for instance the one in Chapter 5. The slope of the transition c urve (12) is found as follows: dg. because n > 0 . + g. 9.· for g1 in (13) and solving lor the constant solution.2 [1 +g. However..> 0 there is a unique positive intersection with the 45° line: setting g1. we now know that in the long run g 1 converges monotonically to the particular value Us..sR) on the righthand side. and k1 and [J 1 converge to k* andy*. according to this new equation. gradually converges tog . We have shown that in the long run g 1 converges to g. If. for g1• The diagram shows how the technological growth rate. (1 ..2(1 + q. or (15) where indeed g. R&D. The situation where these variables are constant at their longnm values defines steady state. gives: k* = ( 11 s )l/(1a)(1 ....)•.• + (</> dg.. as t goes to infinity.. reflecting that now only this fraction of the labour input is used in the final goods sector. (12). k1 . setting the g1 in (13) constantly equal to g. As mentioned.. (You should note carefully how this conclusion depends on our assumption n > 0 . An exercise will ask you to analyse the case n < 0 with respect to the evolution of g. this is not a real transition equation for k1 because it has g 1 on the righthand side. This is one major conclusion from our model with ¢ < 1 and 11 > 0: in the long nm the growth rate of knowledge converges monotonically to the particular value g" > 0. and would.2..s R) 1a: S [I* = ( 11 )a/(1a) .Endogenous Growth 9.. 3 Third.+(¢ . as illustrated by the staircaseformed iterations in Pig..eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 258 I Part 3. 1 = g1 in (12) gives (1 + g1) 1 <P = (1 + 11)' . respectively. The four features we have stated imply that g1 goes tog.. > 0.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies.)•. 1 ) g~ 1 + g.sR).. because we assume 11 > 0. 2005 PART 3: ENDO GENOUS G ROWTH positive. 9 . because we assume. converges to a speciilc value k* in the long run.] = (1 + n) 1 (1 + g . the slope of the transition curve at g 1 = 0 is (1 + n)~. g1. = ( 1 + n)1 [( 1 + gJ•.. and hence it is strictly larger than 1. imply convergence of k1 to a constant valuek* (provided that n + g. k. Fourth. Next. (13).... > 0. oneway coupled diagrams. (1 7) These expressions have wellknown structures except lor the (1 .g.S R). The ilrst is Fig.·· We could therefore write down a new dynamic equation. (16) We can ilnd the associated y* by using y* = (/(*)"(1 . . by similar reasoning. 1 =k1 = k*.)". consider the other dynamic equation.. which is fulillled here because we assume 11 > 0). 3.1)g. We can illustrate the convergence process to steady state in two. Inserting !I s.2 associated with the independent dynamic equation.+ (5 + ng. +g..) . + c5 + ng 5. + b + ng.2] = (1 + n) 1 (1 + g .
the growth rate in k1 will be proportional to the distance between the decreasing curve.. This will have no impact on Fig. 2005 9 R&DBASED ENDO GENOUS G RO W TH: MACROEC ONO MIC MO DELLI NG lfl 259 The other diagram is much like a modil1ed Solowdiagram and it is associated with a rewriting of the other dynamic equation.. as illustrated.3). 9. and k1 will move towards the intersection between the curve and the line.3 will now be towards a moving target and the convergence time will therefore be t "" k* Figure 9. 9 . the horizontal line in Fig. 9.Jacobsen: Introducing Advanced Macroeconomics I Part 3. These diagrams provide intuition for an important insight. For a given g1 in period t. Assume that the economy starts in an 'old' steady state given by the (11nal) intersection points in Figs 9. This \>\Till initiate adjustments in g 1.. since the horizontal line moves when g1 changes. so the horizontal line gradually comes to rest at the position corresponding tog. where one has the growth rate ink 1 on the lefthand side: (18) For any tixed g. R&D. there will be no further changes of the horizontal line in Fig.2.Endogenous © The McGrawHill 9. (13). say a change in II. Consider now a parameter change that shifts the transition curve in Fig. 9.1 . so the growth rate of A 1 will stay unchanged at its old steady state value.3. become smaller and smaller as g1 eventually converges to g.3. s(l .. 9 . Consider a parameter ch ange that only affects the second of the diagrams.sR) 1 ak~'.3 will sh ift once and for all and thus initiate a process of dynamic adjustments which are like those in the Solow model (since g1 does not change.SorensenWhitta. The horizontal line in Fig.3: The 'mod ified S olow d iagram' t . This time. gradually taking it towards its new steady state value. Convergence in Fig.2. however.2 and 9 .. The speed of convergence will therefore be as in the Solow model..based endogenous growth: Macroeconomic modelling Growth Companies. In the long nm the movements in g. this is like a modil1ed Solow equation and we illustrate it in Fig. 9. 9.3 \>Viii shift. say a change in the depreciation rate b.. the intersection is a 'moving target'. Every time g1 changes. and the horizontal line at n + g 1+ c~ + ng1.
. a parameter change that initiates an adjustment process for g 1. To sustain a positive constant grovvth rate in A. in the long run.•• This L~ in accordance with the requirements for balanced growth.¢)should be in the range of 2 . will be constant and given by: = z* = . annual population growth rates in the range between ~ and 1 per cent per year should be compatible with a steady state growth in GDP per worker of around 2 per cent per year..)n. Hence. a constant labour input into the research sector will imply that the growth rate.Endogenous Growth 9..A. output per worker must.+ (~+ ngs. technology. = (1 + n)'10 <1>1 . Hence. The intuition for population growth being a requirement for economic growth is this: when ¢ < 1. g 1..eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 260 I Part 3. and output per worker.. g 5. k 1.. in steady state the capital..1/>). when k1 = k 1/A 1 and [:1 1 = yrfA 1 have reached their longrun constant values../(1 . . converge towards growing at one and the same constant rate./n "'A/(1 .4 For magnitudes to be reasonable by Western standards. an increasing labour input into the research sector is required.s___ (19) II+ g.l. it follows that steady state output per 4 The formula for the growth rate may be w ritten as 1 +g•• = (1 + n) " 1'•1• Taking logs on both sides and using the approximation ln(1 + x) "'x.1. In steady state the rate of technological growth as well as the rate of economic growth is gsr = (1 + r·z).. Growth in steady state In steady state. z 1 K. This intuition is important for the theory we present in Section 5. This is only positive if n > 0.. Assuming?../(1 . we are more interested in output per worker. [:1 1• From y 1 = y.. In particular./Y1 =k*/y*. than in output per ellective worker...output ratio.4 . Furthermore.q. both k 1 andy 1 must be increasing at the same rate as A 1• In steady state this rate is g5.. develop as Yi = Il A.. y . 2005 PART 3: ENDO GEN O US G ROWTH longer for this latter parameter change. Because of this feature our growth model with ¢>< 1 is referred to as one ofserniendogenous growth (explaining the subscript se on g) . Note that (11) expresses the growth rate.. g. or g. s R• of all labour used in the research sector. In conclusion. /A 1. ?. "'A/( 1 . has the implication that ?. y . A 1. Our formula.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies.1(sRLi =g.. .. that is. will imply slower convergence to the new steady state than a parameter change that does not allect the research sector. as a function of A . and with a constant fraction ./n.¢ sh ould then be somewhere bernreen ! and ~.. As usual. and L1• In steady state g 1 = g. we get g. R&D. g5.¢) :Egs.. capital per worker.'. in steady state there is a necessary link between A 1 and L. of technology will go to 0 in the long run as explained in the previous section. in steady state. In that case (11) reads: pA. a change that somehow atlects the research sector. this can only result if the labour force increases all the time. around 1 (not too much stepping on toes). and therefore in steady state: (20) Using (17) and (20) plus the fact that [J* =y'.Vtl<l>l.
We would be more interested in the convergence process itself.SorensenWhitta. while the latter ellect gives a lower growth rate in the long run. as given also).based endogenous growth: Macroeconomic modelling Growth Companies. one has problems accepLing that the source of longterm economic growth should be growth in the labour force. (21) where we can insert that: (22) thus tracing output per worker in steady state back to initial values and parameters (some of which are in Osrl· The steady state growth path lor consumption per worker may be found asci = (1 .Endogenous © The McGrawHill 9. . its level as well as the growth rate along it. )a/(1 a) ( ) 1/(1<{>) (1 . The Hrst ellect gives a transitory additional growth.. on empirical grounds.. but only that its steady state does not seem to be descriptive. According to our model._ g. In both cases labour input in the research sector will be increasing.11 + ase +c) + 11g. will raise the steady state growth path as given b}' (21) and (22). one should not necessarily conclude that our model with¢ < 1 is wrong. it is hard (at least lor some of us) to reconcile empirical evidence with the feature that economic growth should depend positively on population growth. s R· will. if convergence is so slow that the transitory growth eiiect dominates over time horizons up to 100 years..s)yi .. As we discussed at length in Chapter 8. If convergence in our model is very slow. n. We may therefore pursue an idea similar to that used in Chapter 8. R&D. 2005 9 R&D BASED ENDOGENOUS GROWTH: MACROECONOMIC MOD ELLI NG lfl 261 worker will evolve according to: Yi = /' s 1. or tax deductions for.sR)sU<HJ £.Jacobsen: Introducing Advanced Macroeconomics I Part 3. A decrease in the population growth rate.. Structural policy and empirics for steady state Structural policy aimed at improving the steady state should be concerned with the steady state growth path of consumption per worker. private research activities as well as public investment in universities and other research and learning institutions. have an influence on growth similar to that of an increasing labour force at a constant research share. with a constant labour force a gradual and steady increase in the research share. If it takes a long time to come close to the new steady state growth path. A deeper discussion of how to allect the research share will have to await next chapter's endogenous determination of s R• but obviously some policy instruments to consider are government subsidies to. the llrst effect will dominate for a long time. It will be !eft to you as an exercise to determine the value of s R that will maximize the height of this growth path and to judge whether the result seems reasonable. then we would not be so interested in the model's steady state. Lti<Hl. A policy intended to increase the research share thus seems to be a cautious alternative to one intended to promote population growth. the growth rate along the steady state growth path can only be increased by a policy that manages to increase the population growth rate (taking the technical parameters¢ and A. which creates technological growth. If. since such a policy avoids the erosion of capital and natural resources per worker implied by growth in the labour Ioree. the empirical evidence may be in accordance with the model. However. but also reduce the growth rate along it. Taking s R as given. as long as it can take place.
..... Just look at the transition equation... (e lor endogenous)....... To have a wellbehaved model we will assume n = 0.. a very longlasting convergence of g.. a large¢ (below 1) will have that eflect.....())K1. As an approximation of a large¢ below 1. implying... it follows from an analysis similar to that in Chapter 5 (do this) that [J t con verges to = ..)At lor all periods t. is used in the production of llnal goods.... we have more or less analysed it before. for¢ = 1. We will denote the constant labour force by L.. L.. = (1............ \0\lith zero population growth and with a particular endogenous value for the growth rate of technology......... This model is like a general Solow model with no population growth and an exogenous growth rate of technology g.... the technological growth rate. is now replaced by L. = (1... This is exactly what we will do next.sR) of the constant labour force.. with the one exception that only the fraction (1 .. R&Dbased endogenous growth: Macroeconomic modelling © The McGrawHill Companies....sR)L) la..g. with .... With our usual dellnitions............ [J 1 Y1 /(AtL 1) = y t /A.. = pA..... we get At+ I .. If we insert the latter of these into (6)...(.Ji. > 0 ............ g" will...... AI+ 1 = (1 + g.. a constant labour input s x L into the research sector gives a constant growth rate in technology..sx)L and LA... (12) ... we will consider¢ = 1... In fact.. Kt+ 1 = sYt + (1 ... absurd. of course.......... describing the evolution of the labour force... to a very high value. (7) . Hence...sR)L in the production function for llnal goods..... Inserting Ly... stated in (24).... At the limit.... (23) or equivalently: (24) As stated earlier...... It is easy to analyse the model that results from imposing these parameter restrictions on our base model consisting ofEqs (5)(10)..eI SorensenWhitta. the transition curve will come close to a straight line 'Nith slope ( 1 + ... = sxL lor all t..... = L lor all t. for g 1• As¢ goes to 1.Jacobsen: Introducing Advanced Macroeconomics 262 I Part 3.... etc. be exploding and go to infinity...... this suggests studying the model for parameter values that imply very slow convergence Again ... At+ I = (1 ... since the model is almost identical to the Solow model considered in Chapter 5... Equation (8)..)A.. th e full model reduces to th e three equations Yt = K~'(At1 . (5)... 2005 PART 3: ENDO GENOUS G ROWTH As in Chapter 8....... It follows from (9) and (10) that Ly.Endogenous Growth 9... and this growth rate is now denoted by g... where one should bear in mind the endogenous value ofg. We can eliminate the exploding g1 by considering n = 0 .. Endogenous growth .A.~R L)' . when ¢ = 1.. and repeating the capital accumulation equation........ for n > 0...... which is.
This could. The growth rate g<given by (24) does not depend on the investment rates. Third. The main potential implication lor structural policy is that in order to obtain a high growth rate in output and consumption per worker. s x (whereas. we do not think it is a serious problem in the present model because we only consider¢ = 1 as an approximation of a large¢> below 1. according to an appropriate interpretation of the model. The model can. in contrast to what was found in the externalitybased endogenous growth model of Chapter 8. policies to stimulate saving and investment are growth promoting. S )"/(La) Yi = ( .rill (again) have to await the microeconomics of the next chapter where the R&D share \vill be endogenous.4 plots average annual growth rates in GDP per worker against GERD (recall this variable from Table 9. Hence. will converge towards the growth path: . as can be seen from the expression given in (24). Second. therefore. In particular. R&D. Without assuming exogenous technological growth.Jacobsen: Introducing Advanced Macroeconomics I Part 3.)'. . A main prediction of the model is that the longrun growth rate of GDP per worker and of technology should depend positively on the R&D share. it does not literally have much to say about crosscountry evolutions. s. and in the long nm the growth rate of GDP per worker converges to g•. population growth is not a prerequisite for economic growth. for instance. + o (1 .Endogenous Growth 9. \viii imply a higher growth rate in GDP per worker. Just as we did not th ink that the knifeedge property was a serious drawback of the endogenous growth model of Chapter 8. the government should try to obtain a high R&D share. Still.. The model considered in this section has some obvious merits. be done by subsidizing or undertaking research activities as most governments indeed do.a.sx)Ao(l +g. this share turns out to depend positively on the investment rate. the growth rate of output per worker varies over time as y 1 gradually approaches its longrun growth path given by (25). 2005 9 R&DBASED ENDO GENOUS G RO W TH: MACROEC ONO MIC MO DELLI NG lCD 263 and hence in the long run.1) as a percentage of GDP across countries lor which data were available. Figure 9 . First. This feature is an artifact. be considered to be in accordance with the observed conditional convergence between the countries in the real world. Figure 9. . g•. since the model is explicitly intended to describe the world economy. Another merit of the model L~ that it does have convergence (unlike the externalitybased endogenous growth model we studied in Chapter 8). In our next chapter. we get a constant and positive technology growth rate. Whereas the growth rate of technology is constant.. it is a model of 'truly' endogenous growth. The model is not intended to describe individual countries.SorensenWhitta. under semiendogenous growth. (25) Consumption per worker will consequently converge towards the grmvth path ci = (1 . A deeper analysis of how the R&D share can be intluenced v. y. First. The model also has some less attractive features.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies. a lower initial GDP per worker. there is the common objection th at the model only pertains to a knifeedge case(</> = 1). where s x is endogenous.5 plots average annual . we h ave an expression showing how the growth rate depends on behavioural model parameters. an increasi11g R&D share was needed to promote growth). y 0 . for given values of all the parameters in (2 5). arising from the assumption of an exogenous research share.s)y'. however. but nevertheless it is tempting to look at crosscountry evidence for growth rates and R&D shares.
18 OECD countries Note: The measure of GERD is the average of G ERD in 1981. 17 OECD countries Note: The measure of GERD is the average of G ERD in 1981. and Penn World Tables 6. 2002. the model does not describe individual countries. As mentioned. 1990 and 1998.5 3..050 0.0 40 0. 1981 98 Figure 9.005 0.030 0.0 Average G ERD as a percentage of G DP. Technology and Industry Outlook. ~ • * 0. 8.0 10 0. if anything.015 • . 1981 98 Figure 9. point to a positive connection between growth rates and R&D shares across countries.045 0.. Nevertheless.025 0.5) against GERD relative to GDP. perh aps..~ 0.._ • • • • 0.5 .025 0.0 • 2.Jacobsen: Introducing Advanced Macroeconomics 264 I Part 3.. Both ligures sutler from having lew observations and no clear relationship.5 3. .0 2. Technology and Industry Outlook. Source: OECD Science.. They do not. growth rates of A1 (using the same data as for the construction of Fig.045 • 0.5 1.0 Average G ERD as a percentage of G DP.035 0.040 0. . and data set for B ernanke and GUrkaynak (2001).0 0.5: Average annual rate of labouraugmenting prog ress against average GERD as a fraction of GDP. not relevant.0 15 0. 1.5 2. 2002.. 0. •  • • • 2. 1990 and 1998. so crosscountry evidence is. 2005 PART 3: E NDO GEN O US G ROWTH 0.5 1. seems to be a negative one.0 1.020 0.0 0.__.0 1. R&Dbased endogenous growth: Macroeconomic modelling Growth Companies...__ . however.050 • 0.eI SorensenWhitta.010  • 0.030 • ~ • .4: Average annual growth rate of real GDP per worker against GERD as a fraction of GDP.020 • .000 0.035 0. it is an interesting puzzle that the empirical relationship.Endogenous © The McGrawHill 9. . S ource: O EC D Science.
vth rate in GDP per worker.~~~~)· ························· . L. and also independent longer run evidence on the number of engineers and researchers employed in R&D activities.:. R&D. viewing this as neither semiendogenous nor truly endogenous growth. but grm.vth rate. It is (plausibly) assumed that ¢ < 1. Jones thinks that the very long·run evidence we discussed in C hapter 8 is supportive of the idea of semiendogenous growth.~. to 1 is not realistic. Empirical evidence suggests that indeed labour inputs into the research sectors in Western countries have been growing. 2002.vth at a small positive rate can take place for a long time. certain parameter changes can have very longlasting ellects. lor a given sR' give a higher growth rate in technology and then in the longer run also a higher grm. If a parameter ch ange does not aflect the (independent) dynamics of the technological grm. Jones. Equation (24) shows that a larger constant labour force. should.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3. leading to semiendogenous growth. 220239. in particular in the US. Jones mentioned in Footnote 1. However. Table 9 . This will eventually lead to an increasing growth rate in output per worker.~?.vth in the developed world. This should. this increase in labour input does not primarily result from growth in the labour Ioree. Iflabour input in the research sector grows over time.~~J?. but mostly from steady ongoing growth in the research share.. as this share has a maximum of 1. American Economic Review.Endogenous Growth 9.. The possibility of the hemiendogenous growth interpretation is linked to a certain property of the model we have studied in this chapter: even if no parameter is initially at an extreme value (such as¢ = 1). ································································· The model we consider in this chapter paves the way lor a third interpretation of the last 200 years of economic gro.i. 5 . Reading the section is hopefully rewarding. The evidence points to constant longterm growth rates. This is the opening observation in the article by C harles I. However. Economic Growth in a World of Ideas'. The model considered is like the model of semiendogenous growth with¢ < 1..1 points to increasing R&D shares. (] 1• the 5. 6. 'Sources of U.~~?:~?. Of course.~~?:?. 92. The interpretation is related to the one suggested in Charles I. According to the hemiendogenous growth interpretation. .~~. but not required for further reading of the book. steady growth in the research share cannot go on forever. the view tal<en here is that the real world is always on its way to steady state in response to new increases in the research share.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies. either because L1 grows or because s R grows or both . this L~ not observed.'. or close. pp.S. 2005 9 R&D B ASED ENDO GENOUS G RO W TH: MACROEC ONO M IC MO DELLI NG lCD 265 The main problem of the model '>\lith ¢ = 1 is the severe scale 4fect it implies. the result is a growing technological growth rate. For this reason he studies the model with if> (considerably) below 1. according to the model '>\lith ¢ = 1. We have already given the intuition lor this in Section 3 . have resulted in increasing growth rates in GDP per worker. point to substantial increases. We will use the term 'hemiendogenous' growth for this third interpretation. This section discusses an alternative interpretation of growth in the Western world during the last 200 years. Jones concludes that a value of¢ equal.. so it does have a steady state with a positive relationship between population growth and economic growth. particularly if the share is initially small. g1• as follows from (11) with¢ = 1. because it may be seen as an attempt to explain the process of economic growth in the western hemisphere. so an increasing labour input in the research sector is required to sustain economic growth lor long periods.
while g)1 (the approximate growth rate of !. and s = 0. also with respect to its duration.output ratio. Finally. it will stay at that value during all 200 years. and K0 and Y0 must also llt in the production function.1 it cou ld be reasonable to let sR = 0. It also gives an impression of the duration of the adjustments.3 (do the drm"ling yourseU). since k1 = kJA" the growth rate of k 1 will jump up from g. Figs 9.99. and to a change in the R&D share. For the parameters relating to the research sector we set p = 1 as a pure normalization. at L0 = 1. and we choose A. KJ Y0 = z* . R&D. An annual population growth rate of 1 per cent also seems reasonable in a 200 year perspective. We can then simulate the model consisting ofEqs (5)(1 0) over a number of years. so we set n = 0. which are normal values. = 1 and¢ = f.viii be true lor the growth rate of y 1• Figure 9 .It) jumps to around 0 . while the other does affect the research sector and therefore triggers adjustments in g1• Let us speci(y nonextreme in itial parameter values that are reasonable for a period length of one year.<5 = 0. From Table 9.. We llrst consider an increase in the investment rate from the initial values = 0.2 and 9. and then gradually return tog s··· The same '. let us think about the qualitative ellects using our diagrams. which seems to be a reasonable anchor for our simulations. Consequently.eI Sorensen.Whitta. First. A 0 should obey the necessary steady state link between A 0 and L0 given in (20). Since g1 is initially at its steady state value g. 2005 PART 3: ENDO GENOUS G ROWTH response to it will be as in the Solow model. and set the initial value of the state variable L.. K0 and A0 .02 all the time. s.3 . s w The llrst of these changes does not afiect the research sector and hence does not in itiate an adjustment process lor g1. are unaffected by s. assuming a permanent parameter change in period 10.2 up to s' = 0.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies. but note that assuming n = 0. is around 0. as shown by the dynamic equation (12). Let us also say that in year 0 the economy is in steady state at these parameter values. Note that g1 stays at the unchanged steady state value gse just above 0. This implies that the steady state annual growth rate. If a parameter change does affect the dynamics of g 1• the economy's response to it will potentially be longerlasting since it implies all the types of adjustments found in the general Solow model plus the adjustments of g1• The latter kind of adjustments give the first ones the character of 'convergence towards a moving target'. The initial values.output ratio in steady state is given by (19) which imposes a restriction.06..sR)L0 ) l a . of the other state variables follow from the requirement that the economy is in steady state. To illustrate. the capital. solving the two equations in K0 and Y0 gives K0 "" 3.. the flrst of the diagrams is unallected. With our assumed parameter values and the already determined values for L0 and A0 .01. g.02.0 35 and then starts to fall towards gse· After .02.005 would not change our conclusions substantially. In the second diagram the shift in the investment rate implies that the decreasing curve shifts upwards once and for all. Y0 = Kr~(A 0(1 .Endogenous Growth 9. which initiates a process of gradual increases ink 1• the growth rate of k1 being largest at the beginning and then falling back towards 0 as k1 approaches its new steady state value.. here implying A0 "" 0. Before showing the results of the simulations.2.6 shows some output from the simulation and conflrms the qualitative findings. on the initial capital.Jacobsen: Introducing Advanced Macroeconomics 266 I Part 3.20 (verify this). we will conduct simulations with our model showing the economy's responses to a change in the investment rate. We will do so for 200 years.. = (1 + nf/11 <~>>.. Since the dynamics for g. First we seta = f. We now have values for all parameters and initial values for the state variables.
9../Y. a 50 per cent increase as with the savings rate. since in period 9 we are in steady state. and K. but now there is a change in this parameter in one period. you may flrst think that it does not affect it at all. In Fig.5 w 1\ 0.. Here..6: Simulation.5 9t 0. 2005 9 R&D BASED ENDOGENOUS GROWTH: MACROECONOMIC MODELLING 0.015 1 0 12 23 34 34 45 67 78 89 100 111112133144155166 177 188 199 Period t Figure 9. in period 10 the technological growth rate. After that the higher growth in A 1 will pull up the growth in y :·The phase where A 1 grows faster = . it follows from {12) that (1 + n) J. evolution of g" of. g1• jumps by the factor s~/sw After that. Now turn to a permanent change in the R&D share.02 0. Hence. In the first period where the new s~ works. .3 this means that the horiwntalline first jumps up and then gradually falls down towards its original position. (12) does not hold. The growth rate of final output per worker initially falls.. However. Dividing the first by the second.03 in year 10. assuming that it increases from the values R = 0.02 .05 0.045 0. 9 . since s R does not enter into (1 2). after an increase in s Note: Growth rates left scale.5 \gr 0. eventually the growth rates of both A 1 and y 1 return to gsr· Figure 9. reflecting that a flrst ellect of the increased s R is that relatively fewer people will be working in the final goods sector.04 lCD 3 . since the R&D share is again constant. 15 years it passes through the value 0.03 1. In this one period. R&D. the technological grmoVth rate (up to the next period) as given by (11) is g 10 (A 11 . g9 = pA~(s R L9 )J.SorensenWhitta. Note how g 1 jumps up once and then falls gradually back towards 0. that is._. That is. = g. gives: g 10 = g9 (A u/A9 )"' (s~/sR)(L w/L 9 )' = g9 (1 + g 9 )"'(s~/sx)(1 + n)J.025 ~ 0.~~Lu/• while in period 9.02.2 is unchanged and the associated dynamics gradually take g1 back towards the steady state value gse around 0 .<!Yr 2.2 (you should draw the diagram yourself)? Looking at ( 12 ). How does this affect a diagram like Fig. in the derivation of (12) we assumed that s R was constant. the transition curve in Fig.1110)/ A10 = pA f0(.02 5. so at this point one half of a percentage point of the additional growth remains. in period 10. g9 (1 + g9) "' = 9se· Inserting this gives g 10 = (s~/sR)g_.02 to s~ = 0 .035 267 35 ~ 0. 9. etc.based endogenous growth: Macroeconomic modelling Growth Companies.Jacobsen: Introducing Advanced Macroeconomics I Part 3.Endogenous © The McGrawHill 9. 7 reports on the simulation and gives an impression of the duration of the adjustments. Capital share of output right scale.
the duration of the adjustments depends on the parameter values. . this time remembering that s R depends on time. For instance. Of course. can sustain growth at reasonable levels for a long time? This is exactly the possibility of hemiendogenous growth that we now turn to. write it down (once again) for period t + 1 as well as lor period t. for instance by s R increasing from 0 . ~ gY '  0.7 5 per centto 1.035 0. and the later phase where y 1 grows faster than A. As stated earlier.eI SorensenWhitta. You will easily verif)r that this gives: 9t+ l = (1 + g. although the jump upwards in gf in the beginning is not as large as after the increase in s.Jacobsen: Introducing Advanced Macroeconomics 268 I Part 3. the exogenous R&D share is no longer a constant.04 0. g>. and divide the llrst equation by the second./Y. It grows by a constant annual rate. and t K. falls below 2 .based endogenous growth: Macroeconomic modelling © The McGrawHill Companies.03 0.025 0.1 1. so that population growth is not the factor th at keeps economic growth alive. can the assumed growth in sR in itself sustain? Return to (11). If a once and for all change in the R&D share has longlasting ellects on growth. Consequently. a••. it takes almost SO years before the growth rate of y . begins to increase again.02 K1/Y1 "' g.015 0.1 ~ 0.8 1.1 . Capital share of output right scale.0 1 2. then perhaps a slow and gradual increase ins R· which is not unrealistic by empirical standards. 7 is that the adjustments following the increase in the research share are long lasting.2 2. than y 1 corresponds to the phase where k1 is decreasing in Fig.045 0.05 0. 9. How much growth in A. Let us also assume that the R&D share increases at a pace earlier suggested as reasonable. while L 1 is constant. but these were deliberately chosen to avoid extreme values. evolution of g. 9. Let us llrst assume th at n = 0 .5 per cent to 3 per cent over the next one. but an increasing sequence. this cannot go on forever. but it could take place for 200 years.5 per cent. 5 per cent overthe llrst century and from 1. after an increase in s. implying a doubling every 100 years: (1 + 9s) 100 = 2. 7 per cent.. or g'• = 2 I/HXJ . Note: Growth rates left scale.y g.) <~>.(1 + g. is whenf.Endogenous Growth 9.9 1. 2005 PART 3: ENDO GENOUS G ROWTH 0.7: Simulation.3 2. The main conclusion from Fig.7 12 23 34 45 56 67 78 89 100 11 1 122 133 144 155 166 177 188199 Period Figure 9.3. R&D.
006 . This means that A.82 per cent per year. of the technological parameter is such that already from year 0 .012 0.. s = 0.. For all other parameters we assume the same values as the initial ones above.. which gives A0 "" 0. For the Us.7 5 per cent. will be growing at an annual rate of around 1.. assumed above.1. 'explain ' economic growth rates close to the observed ones. R&Dbased endogenous growth: Macroeconomic modelling Companies. As mentioned. . ." pure normalization.4 per cent over the entire 200 years. From (11) th is requires g =Ag. this is g = 2 2/HJO..4 per cent..(sRconstant)  0.¢>) = 2... . etc.... 2005 9 R&DBASED ENDO GENOUS G RO W TH: MACROEC ONO MIC MO DELLI NG 269 The constant growth rate that can be sustained (inserting g.... Let us also assume that the initial value. over time spans like 200 years... A.1sRo• or 2 21100 . .. Interestingly..0075A 0112 .) is thus: g = (1 + Ds/ /(1•i\) .. we fix the constant labour force at L = 1. = 1..2. Finally we calibrate the initial value Ku of capital so that the growth rate of output per worker also starts at around 1. and then grows for 200 years at an annual rate of Us = 2 I/wo .Endogenous Growth lCD © The McGrawHill 9. that is not far from the observed relatively constant growth rates in GDP per capita over long periods. and J. we consider n = 0 and. .0075 ../(1 .1 = 0 .._g{ 0.8: Simulation of the possible evolution of g~ and gj associated with a gradual increase in S R . The lower curve shows how the growth rate of output per worker would evolve if s R stayed constant at the initial 0 .016 0... If we had assumed a slightly faster increase in s R · we could also have rationalized technological and economic growth rates of 1..4 per cent over all 200 years.... as a . Hence......4 per cent. g. g{ . We then simulate the model over 200 years...004 0. The assumed plausiblelooking gradual increase in the research share can thus. ¢> = 1.002 0 1 12 23 34 45 56 67 78 89 100 111 122 1331 44 155166 177 188 199 Period t Figure 9. Let us assume th at in year zero sR indeed has the Yalue 0 .0 1 0. The annual growth rate of output per worker stays very close to 1. one possible interpretation of the growth in the developed world over the last 200 years is that it is really a 0.1. Figure 9. a plausible looking gradual increase in the R&D share can sustain continuing growth in the technological parameter..8 shows the results.Jacobsen: Introducing Advanced Macroeconomics I Part 3.1"" 1.008 g.014 ' ' .+ 1 = g. without any population growth at all. takes the constant value g.SorensenWhitta.. 0. A0 . p = J. .29 .
..... constant g rowth rate of the labour input into the research sector (when ¢ < 1).. and this sulllces for creating economic growth rates of realistic magnitudes for long times... 1 ....... There is a specific elasticity..of the output of new technology...... labour and (labouraugmenting) technology.. The macro model of R&Dbased endogenous growth studied in this chapter had the following features: • Final goods are produced from the inputs of capital. some day growth will cease...)/A..........Douglas production function.. • In other respects the model is standard: capital accumulates from savings and investment (minus depreciation)....Au with respect to the input of existing technology..... At+1 A 1 .... 2005 PART 3: ENDOGENOUS GROWTH transitory growth that is being maintained all the time by gradual increases in the research share......... but the growth rate.... Qualitatively hemiendogenous grO\vth is not so different from the growth one could have in a basic Solow model if the investment rate were increasing gradually.... goes to 0...... A..... constant growth rate of technology requires a positive. One fascinating consequence of the hemiendogenous growth interpretation is that it cannot go on forever.. ¢... according to a standard Cobb.....Endogenous Growth 9......... According to the hemiendogenous growth interpretation............ 2..... a constant input of labour in the R&D sector creates a constant and positive growth rate of technology... 1..eI SorensenWhitta. A.... 1 ..... because firms in the R&D sector may be 'stepping on each other's toes' by working on the same ideas............ reflecting that it becomes harder to generate new ideas as the more obvious ideas have already been discovered.. reflecting that ideas are nonrival.... • New technology is produced in the R&D sector from the inputs of labour and existing technology according to a production function with constant returns to labour at the firm level but possibly diminishing returns at the aggregate level............ We may expect a few more centuries of growth in GDP per person based on a doubling oft he research share every 100 years... If¢ is equal to 1..A...... buts R cannot keep on doubling forever. (A. • The labour force can be used in either production of final goods or in production of new technology.. 9... whereas existing technology can be used in both sectors............ To maintain a positive.. It is this kind of interpretation of the growth process we refer to as hemiendogenous growth.. The fraction of labour used in the R&D sector is an exogenous parameter called the research (or R&D) share. and the labour force grows at an exogenous rate.... If¢ is less than 1.. This L~ neither semiendogenous nor endogenous growth . ... A...based endogenous growth: Macroeconomic modelling © The McGrawHill Companies. a constant labour input in research will imply that the absolute production of technology. and the impact ofincreases potentially so strong and long lasting that the research share can keep on growing at a small constant rate for a very long time...6 Summary .... R&D.. The dillerence is that the research share is initially so small...Jacobsen: Introducing Advanced acroeconomics 270 I Part 3..... (A . but at that time the constant income per capita may be very high. 1 A 1 )/A 1 .... increases and goes to infinity over time...
. The fu ll model g ives convergence to a steady state where capital and income per worker both grow at the same rate as technology. Mainly because of empirical evidence concerning the links between population growth and economic growth. 4. The reason is that given a constant research share.tc. 6. the growth ro. and in a simple way. the labour input in the research sector can only grow at a constant rate if population grows at a constant rate. 2005 9 R&D BASED ENDOGENOUS GROWTH: MACROECONOMIC MODELLI NG lCD 271 3. According to the model studied in th is chapter there could be an inbetween explanation of economic growth during the last 200 years.scs at o. Of course. on the size of the research share.SorensenWhitta. In the case 0 < 4> < 1.tc of income per worker should explode. 5.52 per cent over periods of 200 years and more. and ¢ < 1) or as truly endogenous growth (where economic growth can occur for a constant labour force because 4> = 1. but an increasing labour force would imply increasing economic growth). so one day hemiendogenous growth is over. The semiendogenous growth arising in this case suggests that a policy to promote economic growth should be one that stimulates population growth. a realistic slow and gradual increase in the research share could. the model implies conver~ence to a steady state where capital and output per worker both grow at the same constant rate as technology. This makes the hemiendogenous growth interpretation look plausible. but of a gradually increasing research share possibly for a constant labour force. R&D. according to our model. When the research share reaches its maximum. We showed that under reasonable parameter specifications. not of population growth given the research share. Hemiendogenous growth. The model of truly endogenous growth arising when 4> = 1 implies a highly imp lausible scale effect: a larger labour force should imply a higher rate of growth of GDP per worker. explain annual growth rates of GDP per capita in the range 1. is c losely related to semiendogenous growth (assuming 4> < 1). the research share cannot grow at a constant rate forever.Jacobsen: Introducing Advanced Macroeconomics I Part 3.nt ro. as we called it. A growthpromoting structural policy is one that increases the research share. In the alternative case where ¢ = 1.Endogenous Growth 9.based endogenous growth: Macroeconomic modelling © The McGrawHill Companies. but the model is tacit about how th is can be achieved since the research share is exogenous. but explains the required growth of the labour input in R&D as the result. a constant labour force and a constant research share imply a constant positive growth rate of technology that depends positively. it can be hard to believe that the observed longlasting economic growth in Western countries should be understood as either purely semiendogenous growth (where growth in the labour force is needed for economic growth because the research share is constant. consto. and if the labour force incrco. increased labour input in the R&D sector (required for economic growth when 4> < 1) can only be achieved by increases in the labour force. and the growth rate of technology depends positively on the rate of growth of the labour force (or population) in such a way that there can only be economic growth if there is population growth.
4.oK.1 )g~ + ) ngA..7 Exercises Exercise 1. . 1 . g A"" A/A. respectively. Still using standard notation and definitions. . if you wish to. K/(AL) = k/A and Y"' Y/(AL ) = y/ A. in the long run. l. 3. the technological growth rate. 2. L=n.. We assume </> > 0 and A..Jacobsen: I Part 3. Show that the technological growth rate is constantly g. also in a 'modified Solowtype d iagram') and show that k and y converge to specific constant values.a . 1. Explain for each of the two cases if> = 1 and </>< 1 how the output of the research sector A. Now consider the case </> = 1 and n = 0. k* and J* . A = pA ~' L~.g.1 L~ = pA1' \sRL)' . and the technological growth rate g A evolve over time when the labour input into the research sector LA= s RL is constant. Show that g A converges monotonically to: Jn g •• "" 1 .based endogenous growth: Macroeconomic modelling Companies. f<. show k: that the model implies the two oneway coupled differential equations in gAand gA= (</> . the technological level A. Show that. Ly+LA= L LA= sRL. Show that the growth rates of capital per worker and of output per worker both converge to g •• in the long run.Endogenous Introducing Advanced acroeconomics 272 Growth © The McGrawHill 9. R&D.¢ . according to this model is: gA= pA ~" . Both notation and parameter restrictions are obvious and standard.= p(sRf'/ and that the growth rates o f capital per worker and of output per worker both converge to g •. 5.a . k=sk"( 1 ..S R) 1. at any time. The R&Dbased macro model of endogenous growth in continuous time In continuous time the model we have studied in this chapter would be: y = Ka(AL) 1 . Illustrate the second equation in a 'Solowtype diagram' (and.(n + gA+ o)k. 2005 PART 3 : ENDOGENOUS GROWTH 9. and let the constant labour force be denoted by r. (Hint: for the first one use the log d ifference trick on the formula for g A) Assume first that </>< 1 and n >O. e.0 I SorensenWhitta. K=sY. Illustrate the first of the above d ifferential equations in a d iagram with g A along the horizontal axis and gA along the vertical one.
(described in Eqs (21) and (22)) and for consumption per worker./( 1 .. y.ular. 2005 9 R&D BASED ENDOGENOUS GROWTH: MACROECONOMIC MODELLI NG 10 273 Exercise 2. 1. Explain this effect by stating how a larger level of the labour force. Show that this would mean that A. Consider now the model of semiendogenous growth presented in Section 3 . d ivide the first by the second and proceed as in Section 2 remembering that g~ = g.1 in all periods up to and including period 0.A 1)/A 1 = (p'fp)g••. permanent increase ir the productivity parameter p of the research sector. as measured by L0 . wh ich is empirically implausible. arising. sR. 9. not on growth rates. among them p > 0.) 3. n > 0 does not imply increasing (but rather positive and constant) growth rates in the long run . Say that reasonable values of A. The effects of an increase in research productivity This exerc se asks you to analyse qualitatively the effects of a once and for all. (Note that the transition curve does not move. Explain for yourself the basic intuition for this. Consider the model of semiendogenous growth (0 < </> < 1 and n > 0). describe how g 1 evolves over time. on the levels of the steady state growth paths. sand s R. Do you find that this scale effect is as imp lausible as the one connected to the model of endogenous growth? Exercise 4. What is the golden ru le value for the R&D share under this assumption and how does it relate to the empirical GERD shares reported in Table 9 . In this model.e (A 1• 1 A 1)/A 1 and g{"" (y 1• 1  y 1)/y 1 have been equal to g se = (1 + n)'/( 1 . nevertheless. R&D. for example. In partit.Endogenous Growth © The McGrawHill 9. be said to involve a 'scale effect'.¢) should be around 2.. c . Assume that in all periods up to and including period 0 the economy has been in steady state at 'old' parameter values. the growth rates g . from population growth (n > 0).~l . First show that this implies the technological growth rate from period 1 to 2: g 1 = (A 2 . A scale effect also in the model of semiendogenous growth The model of endogenous growth studied in Section 4 was criticized for the scale effect it implied: an increasing labour input into the research sector. but there is the initial jump in g 1 you found in Question 1. 2.Jacobsen: Introducing Advanced Macroeconomics I Part 3.. (1 1) of Section 2 for period 1 and for period 0.s)y1 • What are the (golden ru le) values of the fundamental rates.SorensenWhitta. and the steady state growth paths for output per worker. that w ill elevate the steady state growth path for consumption per worker to the highest position? Describe in words the two opposite effects of the R&D share. but on levels. The model can.= (1. wou ld imply ever increasing technological and economic growth rates over time.based endogenous growth: Macroeconomic modelling Companies.2 for the independent dynamics of g 1. and </> w ith a population growth rate of 1 per cent per year should imply a steady state growth rate g •• of 2 per cent per year. (Hint: Write down Eq. Golden rule in the model of semiendogenous growth Consider the model studied in Section 3 where 0 < </> < 1 and n > 0. Working first time in period 1 the productivity parameter jumps permanently up top' > p. Does the longrun growth rate of output per worker change as a result of the change in the research productivity parameter? What are the longrun effects on output and consumption per worker of the change? . Using a d iagram like Fig.1? Exercise 3. affects the level of the steady state growth path for output per worker g iven by (2 1) and (22).
7 is much slower than corre sponding to eliminating one half of the distance to steady state every year? (Hint: In a diagram like the one you made in Question 1 above.1. Argue that (1 . Same questions for ¢< 1 and n < 0.In g 1 £!! ( 1 . draw both the real transition curve and the transition curve corresponding to the linear approximation around steady state and do staircase formed iterations w ith both. what does the transition curve look like and what is the evolution of the technological growth rate g 1 in the long run? (Once again this is really just a description of the scale effect in the model of endogenous growth. 3. In this exercise we are interested in the speed of convergence.01)? How can it be that the convergence of g 1 in Fig.) 2.J) can be viewed as the rate of convergence for the technological growth rate g 1 according to the linear approximation of the transition equation for g 1• Show that: 1 . what is the lonQ·run effect on the Qrowth rate of GOP per worker of a permanent increase in p? Exercise 5.o) = (1 1>)( 1  ( + 1 n.based endogenous growth: Macroeconomic modelling Growth Companies. what is the longrun growth rate in output per worker according to the full model? Exercise 6.G'(g ••))(ln g •• .). If tP = ~. Show that linearizing this around the steady state value g •• and changing into logs yields: In g 1+1 . and for¢ = 1 and n < 0. The rate of convergence of the technological growth rate Consider the transition equation (1 2) for g 1 repeated here for convenience: gl+l = (1 +n)' g. Hence.In g. computed and .f/1) . Hence. 9. for example for the real Solow model and for the Solow model w ith human capital.then 1 . according to this rate of convergence and assuming ¢ = ~ and n close to 0. Consider the generic form g 1+1 = G(g1) of the transition equa1ion.' "" G(gl).w_ 1.Endogenous Introducing Advanced acroeconomics 274 © The McGrawHill 9. every period about one half of the remaining d istance between g 1 and its steady state value g •• should be eliminated. Sketch the transition curve in a d iagram and explain how it shifts as tP increases and eventually goes to 1.( 1 +g~) "' .7 above (showing output from a simulation where indeed ¢=~ and n = 0.G '(a. in the long run g 1 converges tog •• = (1 + n)~/( l . How does this accord with Fig. R&D. 2005 PART 3: ENDOGENOUS GROWTH 4.) What tends to make the error (in the evaluation of the speed of convergence) arising from considering the linear approximation large? We have earlier.G'(g ••) E!! ~ as long as n is not too far from 0. 9.) )· and that the rate of convergence goes to 0 as tP ~ 1. Assume tjJ < 1 and n > 0 . 1. In the model of endoQenous Qrowth (tP = 1. n = 0). 3. How w ill this affect the speed of convergence? 2. Same questions for ¢ > 1 and n = 0. Some special cases Consider the transition equation (12) fo r g 1 • 1. For these ones.0 I SorensenWhitta.Jacobsen: I Part 3.G' (g . If tjJ = 1 and n > 0.
Endogenous Growth 9. consider the model of semiendogenous growth {0 < ¢ < 1 and n > 0).10 the economy is in steady state at the old parameter values. among them n > 0. this may be what makes the model compatib le with the empirical observation {Table 8. Simulate the model over 200 years and illustrate in the same type of diagram. If the positive impact can last for a long time {for reason· able parameter values).3 are affected by the parameter change and explain how g 1 and k. but due to transitory growth effects it may have a positive influence for some time. adjust over time./Yr Comment with respect to how long lasting the positive effect on g>. We know from the chapter's analysis of the model that in the long run a lower n has a negative impact on economic growth. Therefore.03 and initially n = 0. g>. 3.g. so the change from 0. 2.is.005 takes place over the 120 years following period 10 with an equal relative change each year. before and after the parameter change. Starting between period 10 and 11. implied by the economy being in steady state in period 0. Assume parameter values a=~. in the long run. The effects of a decrease in the growth rate of the labour force Consider a decrease in the growth rate n of the labour force in the model of semiendogenous growth. and find the initial values.. L0 = 1. normalize.(10) over periods 0 to 200 assuming that from period 0 . Do you think we made errors of the same magnitude then as the error revealed by this exercise? Exercise 7. again starting from the old steady state in period 0 . Simulate the model consisting of Eqs (5). 4. Working for the first time between the initial period and the next one. s = 0.1) that over quite long periods lower population growth rates seem to go hand in hand with higher growth rates in GDP per capita within countries in the developed world. To describe the qualitative effects. g>. explain how diagrams like Figs 9.. the growth rate of y . the growth rate of the labour force falls to a lower value n' < n.) Create figures like in Section 5 showing the evolution of g.2 and 9.0 15 to 0. but still n' > 0. Explain how there can be a positive contribution to growth in y.0 15. You may have found above that decreases in the growth rate of the labour force. Compute the steady state values for g 1.06. sR= 0. n drops permanently ton' = 0. R&D. 2005 9 R&D BASED ENDOGENOUS GROWTH 275 evaluated rates of convergence by considering linear approximations of the appropriate dynamic equations. A0 and K0 . as those reported in Table 8.2. and K. whether they are once and for all or gradual. b = 0. do not really (in the model of semiendogenous growth with reasonable parameters) by themselves produce additional growth for long enough periods to provide an explanation for the empirical relationship./Y. Assume that in all periods up to and including an 'initial' one the economy has been in steady state at 'old' parameter values. in the beginning and why. ¢=~. and K.based endogenous growth: Macroeconomic modelling 10 © The McGrawHill Companies. 1.Jacobsen: Introducing Advanced Macroeconomics I Part 3.10. Comment again with respect to your simulations being able to ' explain' empirical observations on population growth rates and economic growth rates. (As we did in Section 5. has to fall. Does this mean that we should dispose of the model or could there be other explanations? .SorensenWhitta. p =)= 1. e. Assume that the decrease in the growth rate of the labour force comes gradually..005 while the other parameters stay unchanged.1. For some points in the chapter the duration of this positive effect can be considered important.
In the real world. R&Dbased © The McGrawHill Growth endogenous growth: Microfoundations Companies. certainly a lot of research and development takes place in government and nonprofit organizations like universities and hospitals. though: in this chapter. in this chapter we will consider private. or improve upon existing. types of medicine. In capitalist market economies the profit motive is used as a device for the allocation of resources. The motive is to earn increased profits from the selling of the new or improved medicine. A main conclusion will be that the macro model of Chapter 9 was an appropriate short cut in the sense that its steady state coincides with the steady state of the microfounded model set up in this chapter. profitmotivated production of new technology. however. This is a very important further step.0 I SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Chapter I Part 3. and that ideas are the intended output of research and development (R&D). For instance. thereby working to eliminate the shortage. given the crucial impact of the R&D share on the growth process that 'Ne found in Chapter 9 . This is particularly true for most basic research. Profitseeking producers will therefore produce more.Endogenous 10. There will be one dillerence. the R&D share will be endogenous. Each firm is motivated by its own pront. the chapter should be both easy and useful reading. Reading it is not required for understanding other chapters. there is also a Jot of R&D activities in private firms where the motive lor creating new ideas is the pront that can be earned from the ideas being marketed and used as new technology. but I • This chapter is the most abstract and d ifficult in Book One. This should lead to a better understanding of the sources of technological progress and economic growth. rr there is a shortage of potatoes the price goes up. The public goods nature of ideas gives reasons why (at least some of) the production of improved technology should be done by government or other nonprofit enterprises. private medical companies have large and very expensive R&D departments struggling to create new. The new feature in the present chapter is that we will face the public goods nature of ideas explicitly and develop the somewhat subtle microeconomics associated with the production of ideas. which makes it more profitable to produce potatoes. Nevertheless. Until Section 2. However. th at is. 2005 •• • • ~ R&Dbased • • ~ endogenous growth •• :Micro foundations* • n the previous chapter we argued that technological progress results from ideas. we sidestepped an explicit formulation of tile economics of ideas. However. it '>\Till be explained by the model's basic behavioural and technical parameters. 276 .
. And in any case.. 10.. In relation to public goods... In the real world income creation from ideas takes different forms....... R&D....... The nonrival nature of ideas Once a new idea has been invented.....Jacobsen: Introducing Advanced Macroeconomics I Part 3...... the main content in a new improved piece of software... This will most often take place by the selling of a patent... Private R&D llrms create new ideas motivated by the pronts th at can be earned from selling the ideas.... In any case.. This feature gives ideas the nature of public goods........ abuse of market power.. Therefore it is of interest to analyse private............ a manual. In other cases ideas are created in research departments of firms that want to use the idea themselves in the production of commodities. but this is sold to the customer in a box containing a C'D.. but the idea is sold packaged with a traditional good consisting of the razor itself and some blades......based endogenous growth: Microfoundations © The McGrawHill Companies. The inventing llrm will then want to sell the idea to another linn. society may have an interest in having private R&D activities in so far as the prollt motive is considered an ellective engine for technological development... we may view the invention of an idea and its use in production as two separate activities... 'as if guided by an invisible hand' in Adam Smith 's famous words. is the idea of the computer code..Endogenous Growth 10. public goods............... but thereby they come to act as technological developers to the benent of everybody.. say...... If the market mechanism is useful in allocating resources between the production of potatoes and bread. etc... like a new spreadsheet....... th e real novelty in a new generation of razors is the idea for an improvement (now with four blades and ball bearings).SorensenWhitta........ one important concept is rivalry. This L~ of relevance for the reasoning in this chapter... patents seem to p!ay a crucial role in ensuring that ideas can generate income for those who developed them... Most economists think that markets and the pront motive constitute a mechanism lor the distribution ofscarce resources over competing ends that is so ellective that it should be widely used in actual resource allocation... Sometimes an idea is invented by a pure R&D llrm th at does not intend to use the idea itself............. However. or by licensing (renting) out a patent in return for a royalty income paid period by period.1 Ideas as economic goods . it can perhaps also be effective in allocating resources to R&D activities.. Most economists th ink as well that the market mechanism is not perfect and should be 'corrected' by various forms of government intervention to address imperfections such as externalities... profitmotivated production of ideas and new technology. etc... For instance. the use of the idea in one activity does not in anyway limit the extent to which it can be used in other activities.......... This section explains why this is the case.. This is what we do in this chapter.. the public goods nature of ideas implies dill1culties lor the working of this engine. What we just said means that ideas are 1'1011riva/.... 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS 10 277 comes............... and the dill1culties associated with such private production.... .... Likewise...... to serve society.. through its attempts to maximize prollts. Thus.... A private production of ideas requires that ideas can somehow create an income.. The idea is then patented by the producing llrm itself and creates an income through the selling of a good containing the idea..........
The idea behind it has been developed either by an independent R&D ftrm or in the research department of the producing 11rm. access to fresh air. can be used by many diflerent persons in many diflerent places at the same time. of producing razors. c. involves almost no additional cost (again provided it is not already congested). but falls short of total cost by the amount (AC(qc). but it does not require additional input of ideas or t echnology. The total revenue pcqc exactly covers the variable cost cqc. supply 'viii be zero.. intermediate products. B uilding a new factory identical to the existing one requires a doubling ol labour. 1 . the use of the idea in one 'activity'.based © The McGrawHill Growth endogenous growth: Companies. exceeds the price. For instance. Let us assume that in addition to this fixed cost there is a (small) constant unit cost. Ideas are nonrival. and the amount produced is given by demand. pc =c. since the new factory can simply use already existing technology.0 I SorensenWhitta. 1 If we denote the ftrm's production of razors by q. AC(qc).Endogenous 10. respectively. pc. 10. the development cost is 'sunk'. the production of a specific razor unit. restaurant meals and cars are rival goods. are MC(q) =c. Assume now that the producer is a small perfect competitor who takes the price of razors as given. acquiring the idea has given rise to a cost consisting mainly of the incomes of the developers. etc. for instance. and therefore optimal behaviour only depends on the variable cost component. Potatoes and bread. and the extent to which it is used in one place at one time has no implication for the extent to which it can be used in other places at the same or at other times. Viewing technology as an input we therefore get a so·called 'non·convexity' in the production function. F say. is equal to c. It is rival if the consumption or use of the good by one person or firm prevents other persons or ftrms h·om using it. the services of bridges and parks (in so far as they are not overcrowded). In these circumstances market forces should imply a competitive equilibrium where the price. Hence. Once the idea has been fully developed. and hence the marginal and average cost functions. Classic examples of nonrival goods are national defence. the big discoveries in basic research . like calculus. the supply will be inflnite: if the output price is smaller than c. These are illustrated in Fig. for instance in the production of a traditional good that embodies the idea. In th is equilibrium the firm will earn a negative protlt since average cost. Once the idea has been fillly developed. In any case. the production of another razor unit. does not prevent the idea from being used in another activity.Jacobsen: Introducing Advanced I Part 3. A good is nonrival if its use by one agent does not limit the ability of other agents to use the same good.c)t/ = F. capital. . the services of the legal system ('law and order'). The replication argument says that there should be (close to) constant retlJ'ns to the rival inputs. The horizontal marginal cost curve will then also be the firm's supply curve: if the given razor price exceeds c.1. and if' the razor price is exactly c. as would follow from the replication argument and given input prices. The nonrival character of ideas has an important implication for the cost functions associated with the use of ideas. R&D. Let us return to the example of a new generation of razors. andAC(q) = F/q+ c. One additional passage of a bridge. for instance for the small 11rm implying a production ofqc as indicated in Fig.1. any output level is an optimal supply. although the bridge may originally have been extremely costly to construct. etc. in the production of the new razors that carry the idea. great scenic views. the cost of acquiring the idea is to be considered as a fixed cost. 10. The tlxed cost is uncovered and the production of razors involves a loss ofF. essentially just meaning that it raquires less than a doubling of all inputs to double output. the cost function is C(q) = F + cq. 2005 Macroeconomics 278 Microfoundations PA RT 3 : ENDOG ENOUS G ROWTH An economic good can be rival or nonrival. c. One aspect of a nonrival good is that its marginal cost of production is (close to) zero.
10. perfect competition However. but rather is the only producer of the particular new razor. and patent rights and other legal ways of protecting intellectual property rights serve as instruments for creating such monopoly power. where the marginal cost curve. When the firm is the sole supplier. Let the demand curve for this razor be as illustrated by Din Fig.Endogenous Growth 10. R&D.based endogenous growth: Microfoundations © The McGrawHill Companies. The conclusion is that under the conditimzs of perfect competition (price til king in the output market). we should emphasize that this way of ensuring a . The conclusion is that in order to have a private production of ideabased technological progress it must somehow be ensured that the inventor ofmt idea holds monopoly over the use of the idea. Assume now instead that the llrm is not small in the output market. Under conditions of impe1ject competition (monopoly power in the output market) it is possible to have a private production of ideabased technological progress. however. This leads to the optimal point M in Fig.Jacobsen: Introducing Advanced Macroeconomics I Part 3. MR . At the monopoly optimum. the price. Before turning to that. and the marginal revenue curve. one cmznot have a private production o. This is indeed an important conclusion. before the linn decides to develop or buy the new razor it would be able to foresee what we have just explained: that the whole project can only end up yielding negative prollts. With this demand curve the competitive equilibrium would be given by the point C involving the price c and a specific total output (of which we imagined thesmall flrm above to supply qc). It is well known from microeconomics that the monopoly llrm should reduce output as long as marginal cost is larger than marginal revenue. pM. 2005 10 R&D BASED ENDOGENOUS GROWTH: MICRO FOUNDATIONS 10 279 p AC(qc) q Figure 10. 10. at least for some period. could well be above average cost as illustrated. it can take advantage of the fact that reducing output from the competitive level will imply a higher price (while the competitive tlrm's supply was assumed to be so small relative to the market that it had no significant influence on the price) .2.1: Production of a research intensive product. The firm would therefore never decide to start such a project in the llrst place.SorensenWhitta.2. or in the production of a good that embodies the idea. which has thesamecostcurvesas in Fig.1. and the project of developing the idea and producing the new type of razor could thus imply positive prollts. MC. 10./'ideabased technological progress. intersect.
Furthermore.2. and taxes have negative welfare ellects too because they imply distortions.based © The McGrawHill Growth endogenous growth: Companies. The social cost of having a less than optimal utilization of ideas under private R&D would have to be weighed against the social cost of distortionary taxation under public R&D. These could supply ideas. and to the right of t he vertical line at qM.vpoint. at marginal cost. or other nonprollt..Jacobsen: Introducing Advanced I Part 3. then tile resulti11g ideas will be used too little as compared to a social optimum. the price is above marginal cost (othenvise there would not be positive prollts). as mentioned earlier.Endogenous 10.2: Production of a research intensive product. in so far as we believe that sometimes the profit motive is an ellective engine for technological development. R&D. too few resources are devoted to the utilization of the idea behind the new type of razors. Monopoly implies a dead weight loss in the form of a forgone consumers' surplus. enterprises. above the MC curve. or blue jeans that people really want. Hence. have to be fmanced by taxes. toothbrush . When it comes to developing the kind of razor.0 I SorensenWhitta. 1 0. and be willing to live with the monopoly power necessarily associated with it. however. .2 In other words. or the kind of spreadsheet or catalyser that 2. from a social vie. In the monopoly optimum given by point Min Fig. The experiences of the formerly planned economies of Eastern Europe suggest that the more applied part of research and development is done most effectively in private enterprises. 2005 Macroeconomics 280 Microfoundations PA RT 3 : ENDOG ENOUS G ROWTH p AC(q~ 0 q Figure 10. exactly because it works through giving monopoly power to the firm. in the figure represented by t he area of the triangle below the 0 curve. thus ensuring an optimal utilization of the ideas and simply accept the implied negative profits. monopoly profitmotivated development of ideas is highly imperfect. !f one ensures a private production of ideabased teclmological progress by giving monopoly in tile use of ideas to the creators. This might seem to suggest th at research and development should mainly be conducted by public. or products embodying ideas. This means that the consumers are ready to pay more for additional units of the new razor than the additional cost that would result from producing additional units. we (the general public) may wish to have private R&D activities. The losses would.
excludability can occur to diflerent degrees. however. The llrst are the traditional goods like bread. proiHmotivated llrms. but the exact recipe will not be directly visible and hence not completely open lor use by everybody. but during most of history free access to catching 11sh in the open sea has been considered a right for everybody. as we h ave seen. more eflective than public bureaucracies exactly because they are governed by market signals and the prollt motive. or both nonrival and nonexcludable. etc. fresh air. restaurant meals and travels. Each package of medicine will someh ow contain the idea. restaurant meals. The patent gives Novo Nordisk the right to be the sole user of the design in production of insulin for some period. R&D. While a good is either rival or nonrival (again disregarding congestion). 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS 281 other ftrms can really use. Most economic goods are either both rival and excludable. The importance of excludability A good is excludable if the owner of the good can exclude other persons or frrms from using it (if they do not pay lor it). However. etc. When Novo Nordisk. and the latter are the pure public goods like national defence.based endogenous growth: Microfoundations 10 © The McGrawHill Companies. potatoes. An example of a good that is rival. One person's passage will (under normal circumstances) not limit other persons' possibilities of crossing the same bridge. increased excludability can to some extent be ensured by law and by technology. Consider again the idea lor a new way of producing an improved form of insulin . for instance crossing a (not congested) bridge. A good is nonexcludable if persons or firms cannot be excluded from the use of it.. namely excludability. but nonexcludable could be 'llsh in the sea'. Very often patents serve the purpose of ensuring such a monopoly. like bread and potatoes. somewhat excludable. etc. probably. are excludable.Endogenous Growth 10. haircuts. by nature. Classic examples of nonexcludable goods are. and relatedly. national defence. nonexcludable goods cannot be produced and marketed by private. again. law and order. etc.SorensenWhitta. bas designed a new way of producing an improved form of insulin. be excluded from crossing a bridge if they do not pay. a competitor with a . 'law and order'. Only public enterprL~es not motivated by proftts can produce completely nonexcludable goods. haircuts. private developers are. What the patent really does is thus to exclude other persons and llrms from the use of the idea lor a specific purpose. Furthermore.Jacobsen: Introducing Advanced Macroeconomics I Part 3. so the good is nonrival. This state of affairs requires. If another firm uses the same design in the production of insulin. Most traditional goods. clean air. This leads us naturally to the discussion of another essential concept in relation to public goods. it can protect its 'design' by taking out a patent on it. On the other hand. Since the access to a nonexcludable good cannot be made conditional on payment. so the good is excludable. These two features are important in connection with ideas. People can. rivalry and excludability are conceptually distinct. it will violate the law. Private R&D activities may therefore be socially benellcial.. There are goods that are nonrival and excludable. that somehow the inventor of an idea obtains monopoly power over the use of the idea. They are only handed over the desk upon payment. The idea or design is therefore. and certainly in the real world we do observe a lot of R&D taking place in private linns. under huge development costs. since a fish caught by one person cannot be caught by another person too.
It does not prevent a competitor from analysing the product or from studying the application for the patent.eI SorensenWhitta. but at some cost. Hence the design is not perfectly excludable. a lot of R&D directed at creating new products and improved production methods takes place in private firms. thereby getting full insight into the design. 2005 PA RT 3 : ENDOG ENOUS G ROWTH research department of its own could probably. At least some degree of excludability is needed in order to have private production. Discoveries in basic research are nonexcludable. lor instance Nobel Prize winner Douglass C. while over the several thousands of years before that. The role of intellectual property rights in economic development Some economic historians. consider this point to be absolutely essential lor understanding the unprecedented economic growth over the last 200 years. creates the monopoly power in the use of the idea that is necessary to overcome the problem associated with the nonrival character of ideas. R&D. by giving (through a patent) the exclusive right to use a certain idea lor a certain purpose to the inventor. At the same time. ideas vary between being absolutely nonexcludable and being imperfectly excludable. On the other hand. A legal system of protection ofintellectual property rights thus serves the double purpose of msuring some excludability a11d ensuring some monopoly power in the use of new ideas.Jacobsen: Introducing Advanced Macroeconomics 282 I Part 3. the average annual growth rate ofGDP per capita must have been close to zero to fit '<\lith the level of income . As we have touched upon before. just like a TV transmission in the air (another nonrival good) can be made more excludable by encoding the signal. the degree of excludability can often be much increased by legal or technological arrangements. The latter requires some degree of excludability. the Western world has experienced average annual growth rates of GDP per capita of about 2 per cent during the last 200 years. The knowledge Hrst created by Novo Nordisk can be useful for other developers even though the patent prevents them lr01n using this knowledge lor exactly the same purpose as Novo Nordisk. While nmrrivalry prevents private ideabased productio11 under perfect competition. break or read the code and in that way technically become able to produce the same medicine. Note carefully wh at a patent does. The legal institution of patent rights can to some extent prevent direct copying and thus mal<e the design more excludable than it was by nature. for example in connection with a new design for insulin: it prevents a competitor wh o can break the code from using the same design for the production of insulin. If one cannot prevent people who do not pay for a good from using it. North .based endogenous growth: Microfoundations © The McGrawHill Companies. which is often ensured by the presence of a legal system protecting intellectual property rights. no money can be made from producing the good.Endogenous Growth 10. In the latter case. For these reasons almost all basic research takes place at public universities or at other types of nonprofit research institutions. There may also be ways to produce the new medicine that can mal<e it more dilllcult to detect the invention behind it. Both legal and technological devices can serve to mal<e ideas more excludable. this legal system. By nature. by analysing the product. complete liOnexcludability prevents private productio11 under any market structure. In this way a competitor may obtain increased knowledge that can be used lor a related idea not protected by the patent. while more downtoearth ideas lor new products or production methods most often by nature are partly excludable: they can be copied.
Another feature was the abolition of rules and regulations that bad hitherto prevented free enterprise. and the revolution that took place was the transition from the feudal to the capitalist mode of production. competing '<\lith other firms. The important feature missing from the Marxian story. This is what we do now by going through a seminal contribution to growth theory made by economist Paul M. During centuries of feudalism a very slow technological evolution had taken place. Among other features this transition implied that labour power was set free (or kicked out) from rural estates. 1990. but here it is again: Paul M.Jacobsen: Introducing Advanced Macroeconomics I Part 3. 'Endogenous Technologic al Change'. If new inventions were in no way protected from being copied. a private return on the efl'ort exerted in developing ideas was ensured. during this period. The output price would therefore. pp. With these remarks in mind it seems worthwhile to set up a formal model which focuses on the economic incentives w1derlying private activities in research and development. R&D. Gradually during this process. By protecting through patents new ideas from being copied. This state of affairs would prevent inventing in the first place (note that this argument is basically the same as our argument th at there can be no private ideabased technological progress under perfect competition) . be determined by the higher costs of the competing firms. Journal of Political Economy. Th is story certainly sounds appealing. This may have improved greatly the incentives for a private production of ideas. but according to the economic historians mentioned above one element is missing. Romer. After any such invention there would be a period where the competitors had not yet adopted the new technique. thus obtaining 'freedom' to move to the cities looking for work in the growing manufacturing sector.Endogenous Growth 10. would h ave an individual incentive to create new production methods that could increase productivity and reduce costs. What kind of an event could that be? According to some (lvfarxian. . In this way the prollt motive would continuously work as an incentive for further technological progress. then the time that an inventor would be alone with a new production method would be short. 3 3.Sorensen. still according to these economic historians. The market mechanism thus came to work as a driving force for the creation of new technologies. for instance) historians the big event was the change in the mode of production from feudalism to capitalism. Romer. We already gave the reference to Romer's paper in the introduction to Chapter 9 . Apparently big changes must have occurred around the late eighteenth and early nineteenth centuries. technology came in increasing contradiction to the social relations of feudalism characterized by labour power being tied to rural estates and lots of rules and regulations preventing the free initiative to establish new firms in trade and industry. so there would not be much extra prollt to reap.based endogenous growth: Microfoundations © The McGrawHill Companies. In this way the proflt motive and the allocation of resources through markets became much more important than it had been earlier. and the inventing finn would temporarily earn abovenormal proflts. was an improvement in the legal protection of intellectual property rights that bad occurred gradually up to the time we are talking about. Each flrm. 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS lfD 283 per capita around year 1800. that in turn contributed to the technological progress '<Vitnessed over the last 200 years.Whitta. S7 1S 10 2. 98. changes that have worked to initiate and maintain technological progress and economic growth. at that time reaching the 'critical' level. A social revolution had to occur.
a formula. a new chemical process. and will alter some time drive out older systems.Jacobsen: Introducing Advanced Macroeconomics 284 I Part 3. and there is a theory of endogenous growth that explicitly incorporates the idea of creative destruction. Ideas as designs and designs as commodities In everything that follows an idea will be a 'design' (a recipe. or license. a new material. We find it most convenient to assume the flrst case. or the level of technology. The right to use the patent for a specific intermediate good therelore h as to be transferred from the inventing 11rm in the R&D sector to a 11rm in the intermediate goods sector. Finally. 323352.. which would then have the right to sell it further. Econometrica. + 1 .based endogenous growth: Microfoundations © The McGrawHill Companies. for example. t. For instance. Such ideas often displace older inventions. we will have to be very specific about the concept of an 'idea'. and one can identifY technological progress in period t with the number. so that licences for designs. A firm that has developed a new design obtains an everlasting patent on the use of the design in the production of the intermediate product th at the design supports. An idea can thus be thought of as the design for a new catalyser. or rather for the patents for designs. etc.. 1992. and there are separate intermediate goods producers th at produce the intermediate products that the ideas support.Whitta. A. a new piece of software. 60. 4 Romer's model presented here abstracts from creative destruction and assumes that ideas lead to new intermediate products wit/rout reducing the productivity ofalready existing ones. commodities and markets: an overview of the economy The economic agents are firms and households (consumers). of intermediate products exi~ting at that time.A. Economically it will make no difference which of these two alternatives we consider. 'A Model of Growth through C reative Destruc tion'. Agents. Or the inventor could sell the patent once and for all to an intermediate product firm. will be one of the commodities of our model. We are going to assume that society has dealt with these problems through a wellfunctioning patent system. that is. we will assume th at there are pure R&D 11rms that only produce ideas. For this reason one can identify the stock of knowledge. the right to use the patent to an intermediate product firm . a new spreadsheet or word processing program is really an improved system. This could take different forms. for a good that is used in the production of 11nal goods. with the number. R&D. The process by which new inventions displace older ones has been called 'creative destruction'. at the beginning of any period.Endogenous Growth 10. of ideas for new intermediate products created in the period. 2005 PA RT 3 : ENDOG ENOUS G ROWTH To present Romer's model of endogenous technological progress. There are three different production sectors distinguished by what they produce: the 11nal goods sector. a computer code or the like) lor a new intermediate product. We have mentioned two problems lor private production of ideas associated with nonrivalry and nonexcludability. . a new computer. See. The inventor could keep the patent and rent. A . pp. the 4 . Philippe Aghion and Peter Howitt.eI Sorensen. respectively.
' ' I I I I I I :t A 1 monopoly (license) markets __T.Jacobsen: Introducing Advanced Macroeconomics I Part 3. 2005 10 R&D B AS ED ENDOG ENO US G RO W TH: MICRO FO UNDA TIO NS lfl 285 intermediate goods sector.. The output can be used either lor consumption or for investment in capital. Although only A1 monopoly markets for intermediate products INTERMEDIATE G O ODS SECTOR Capital + one design j t Intermediate prod uct j ...SorensenWhitta. using capital and a design as inputs. so compensation for labour is new designs Figure 10.. . Each firm in the intermediate goods sector produces one particular intermediate product.... Technology... in the research sector produce new designs from the inputs of labour nnd the existing stock of knowledge.::: .' '.__ . To simplify it is assumed that labour is not required as an input._f_ o_ r designs FINAL GOODS SECTOR Labour + intermed iate products t Final goods Competitive capital market R&D SECTOR Labour + technology t New designs HOUSEHOLD SECTOR No formal market here: R&D firms are oneman firms.> are licensed : \ ' . ... ' =. > . is thus assumed to be productive in two distinct ways: in the final goods sector as the number of productive intermediate goods. The representative firm in the final goods sector takes the number of intermediate products {the level of technology) as given.3: Flow chart of the economy considered . The tirm in the final goods sector does not directly use capital as an input. The model thus explicitly accounts for the nonrival character of ideas. only the intermediate goods that are produced from capital (see below). while the design is a nonrival input: only one design is required for the production of any amount of the intermediate good in question.. .! . and in the research sector as a stock of knowledge. and the research (or R&D) sector... Capital is a traditional rival good. out d irectly by ...e '. Firms in the tinal goods sector produce output from labour and intermediate products.3. It will be helpful to follow our explanations below in the flow chart of Fig.. R&Dbased endogenous growth: Microfoundations Companies. __\ \_\. because designs \ ~ ... Firms. . .. ' ' \ households . /' No flows h:. or persons. 10. A1 in period t.Endogenous Growth © The McGrawHill 10. . \ ___....
. rather than a usual wage payment. Consumers sell their labour services to the flrms in the 11nal goods sector or work in the research sector. pp. K. such that the compensation for the labour exerted in research is (the value of) the patents produced. It facilitates our explanations if we think of the 11rms in the research sector as oneperson 11rms. we could equally have presented the model such that researchers were wage earners and patents were owned by the R&D 11rms. the commodities are: final goods. patents also to flrms in the intermediate goods sector. as given at any time. On the other hand. as well as the patents for the A. no one can make or sell widgets without the agreement of the inventor. thus increasing supply in the labour market and reducing the real wage rate. On the other hand. who all take w 1 as given. Paul M.. The consumers can alternatively choose to work in the research sector. if more could be earned by doing R&D than by working in the 11nal goods sector. A. pi" of intermediate product j in period t is set by the monopoly supplier. w . The rental market lor capital is perfectly competitive. Demand comes from the firms of the final goods sector while supply comes from the consumers. that one unit of labour can earn in the competitive labour market must therefore equal the return that one unit of labour can obtain from working with R&D. If the former were larger than the latter. R&D. 1990. and licences lor existing patents.eI SorensenWhitta. .. . workers would move from that sector into the R&D sector until the earnings dillerential is eliminated. and the price in this market is normalized to 1. so the real rental price of one unit of capital. However.If an inventor has a patented design for widgets. labour would move away from the research sector towards the final goods sector.' 5 Each individual in the research sector takes the stock of knowledge.Jacobsen: Introducing Advanced Macroeconomics 286 I Part 3. 98.based endogenous growth: Microfoundations © The McGrawHill Companies. each of the latter two in the number A. the knowledge embodied in the design may be detectable and of use in the production of new designs. The labour market is also perfectly competitive and the real wage rate is denoted by w. other inventors are free to spend time studying the patent application for the widget and learn knowledge that helps in th e design of a wodget. they lease out capital to the firms in the intermediate goods sector. Journal of Political Economy. capital services.frictions'. We assume that workers can shift between the flnal goods and the research sectors without . while pi< is 5. the consumers own the capital stock. existing designs. For each intermediate good.. p. and they rent (license) their . The price. 'Endogenous Technological Change'. the wage. The market for final goods is perfectly competitive. intermediate goods. In period t.Endogenous Growth 10. while supply comes from the single llrm that holds the design required for producing the particular intermediate good in question. Romer. and all researchers can use all of A.1. in period t. In equilibrium. 2005 PA RT 3 : ENDOG ENOUS G ROWTH the 11rm that has licensed the patent lor design j is allowed to produce intermediate product j. demand comes from the firms of the final goods sector. labour services. denoted by r 1. Hence this market is one of monopoly. Capital services are supplied by the consumers and demanded by the llrms in the intermediate goods sector. reflecting the nonrival nature of knowledge. S71S1 02. In Romer's words: . S84. Final goods are supplied by the flrms of the final goods sector and demanded by the consumers for consumption and investment. At the beginning of period t. is taken as given by everybody.
.based endogenous growth: Microfoundations Growth Companies. R&D.a. This is indeed a doubling of inputs. In equilibrium the prices of all intermediate products turn out to be the same. < 1. When we allow the level of technology to change.t 1a _ L yt  A1 L (X . 0 < Cl. then each term. each of which is again used in the amount x. so there are now 2A. is doubled.. the patent is licensed to an intermediate product firm at a rental rate (or royalty) ofr. The production function exhibits constant returns to scale in the rival inputs of labour. The resulting production is 2A. ~ t1. Hence. and xw taking the real input prices. (1) J= l where Y1 is output.1 = x 1. which will turn out to be the same for all patents. since labour input is doubled. Finally there is a market for licences. 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS lfl 287 taken as given by the demanding fums. Ly.1 are doubled. Hence the production function changes from period to period as A 1 increases.tXjt' (2) i= l The firm seeks to maximize this profit through its decisions onLy.Jacobsen: Introducing Advanced Macroeconomics I Part 3. The final goods sector The production function in period t of the representative firm in the final goods sector is assumed to be: tl. X. w 1 andP. intermediate products. First there are A 1 intermediate products.LP. In each period after the one where the new design was invented. Ly. and the sum itself doubles. .1• j = 1. the market for each design is also one of monopoly where the supplier of a patent sets its rental price. and X. There is one consumer who holds the patent.Endogenous © The McGrawHill 10. The profit of the firm in period tis: rrt= Y. This is essential for the model's ability to generate endogenous growth .t . and all are used in equal amotmts. . so the total input of intermediate products has also doubled.11 in period t. but there are twice as many intermediate products. for j = 1. X .r• and the number of intermediate products. We assume that the inventor of a new design. and A. so output is A. Consider an alternative way of doubling the inputs. and each intermediate product is used in the same amount as before. is kept fixed and Ly. We have now described the 'micro world' considered in general terms and will go on to describe each of the agents and sectors in more detail. A u as it should according to the replication argument. This is most easily seen from the second equality in (1): if A . the conswner who created it in a oneman firm. . _ yt  ( ) L X.• t. and intermediate products.SorensenWhitta. the production function for final goods thus has increasing returns to scale.. 1a ). The rental market for design j is not competitive either. Then LYI is doubled. A 1..x~'(2L 1. wiLy.. xj1L ~.1 ) 1 " . is labour input.. of period tenters as the number of intermediate products available in period t. as given. in the sum doubles. The technological level. but there are many firms in the intermediate goods sector that could potentially rent the design. but for the moment it is convenient to have a notational distinction between them as expressed by subscriptj.1 is the input of intermediate productj..tLl't . keeps the patent for the design forever. and each x. A 1.=1 . which is more than t\>Vice the former output.  A.." .. . x~'L~. A 1.
t· The production function of each firm in the intermediate goods sector is thus described by the output of :x~ it requiring a capital input of K.. "v\le will call this dillerence the 'gross prollt' of the llrm and denote it by n . r1 xw of producing X. as given.r· arising from producing and selling X.Jx . Otherwise. An individual consumer who decides to spend one unit of work time in the research sector in period twill produce the number pA. Average cost. The llm1 takes r.1 = x i< and an input of the right to use design j .eI SorensenWhitta. = 1 (no 'steppingontoes') and¢ = 1 (again approximating a¢ in the range just below 1). to maximize the gross prollt :nil' since this will also maximize the net profit. but tbe distinction between¢ = 1 or¢< 1 has the same effect as in Chapter 9. More generally.1: (4) The firm 's net profit is :n. but in equilibrium these two magnitudes h ave to coincide). r At• oflicensing one patent (for design j) for one period. AC(x..1) = r.r· In addition there is a fixed cost. At the beginning of period t + 1 the consumer will therefore be equipped with additional patents . Once a llrm In tbe Intermediate goods sector bas rented design J. Having decided to rent design j. the fum would not decide to rent design j in the first place. In equilibrium the resulting :nit should cover the llxed cost.based endogenous growth: Microfoundations © The McGrawHill Companies.1 + r 1. the former case leading to endogenous growth and the latter to semiendogenous growth. of new ideas. the flrm can do no better than trying to maximize the difference between the revenue.Jacobsen: Introducing Advanced Macroeconomics 288 I Part 3.. (It is a bit of an abuse of notation to let x . You will recognize the type of cost function we considered in Section 1 for the razor example.1 units at price Pw minus the variable cost. 1 = x . we could have assumed 0 < ¢. for simplicity . as well as rAt as given..r denote both the input of intermediate product j in the ilnal goods sector and the output of a ilrm in the intermediate goods sector. taking A. so variable cost is r1 K. and taking into account how demand from the flrms in the final goods sector depend~ on pi'' the Hrm will set P.1 units.1) = r.. rA. R&D.Endogenous Growth 10. is constant. Capital is rented in the capital market at the rental price r.t X. The research sector The production function lor flrms in the research sector is exactly as assumed in Chapter 9.. Hence the cost function of the linn producing intermediate product j is: (3) where r 41 is a fixed cost. while marginal cost.1 = r 1 X. consider the particular case where A. so the production function is (5) where LA. 1 as in Chapter 9. is decreasing. 2005 PA RT 3 : ENDOG ENOUS G ROWTH The intermediate goods sector A firm in the intermediate goods sector that bas acquired the right to use design j in period t can produce xit units of intermediate product j from a capital input of K. r 41.. (An exercise will ask you to analyze the case¢< 1). P. In this chapter we will.rAt• which also takes account of the llxed cost.. MC(x. is the total labour input in the sector.
but we w ill ignc·re int eger problems and allow the number of designs and patents to be any positive real number. so a patent is like an asset.based endogenous growth: Microfoundations © The McGrawHill Companies. be an integer for all I.. Instead of assuming a rental market for licence rights to patents.<~)PAt. The household sector We assume that there is a constant number L of consumers.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3. Let us call this amount PAt· A capital of PA. fulfils or equivalently: (r. r At· and the patent price. but had to buy the patent once and for all at price PAt· Then.41 + 1 .. ... which is distributed over the two sectors using labour power: (8) 6.Endogenous Growth 10. The total wealth that the patent gives rise to at the end of period t is thus rAt+ P. r Au rA<+1. Each consumer is endowed with one unit of labour in each period.<~)P41 that could have been earned if the amount PA1 had instead been invested in capital..1. 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS lfl 289 in the number pA . as we shall see below. and at the end of the period the consumer still has the patent to rent out in periods t + 1. The number A1 should. The value of a patent is the amount of capital (the number of capital units) which would be just as good to own at the beginning of period t as one patent.PAt· (7) TherA< on the lefthand side is the rent (royalty) that an intermediate product ftrm must pay for using one design in one period.. would yield an income in period t ofr 1 PA<· and at the end of the period t he owner would still have the capital net of depreciation. Assume that the Hrm did not rent the right to use a patent period by period. The value of this opportunity is PA1+ 1 .PAl + (1.. that is. as the righthand side shows. the latter including the capital gain t. Th e two formulations are equivalent as long as the necessary link between the rental rate. minus the capital gain on the patent from the beginning of period t to the beginning of period t . The capital value of a patent is important for an essentia l consideration th at consumers have to make.o)PAl +PAt· One patent held at the beginning of period t generates a royalty income of rA< in period t. the value of a patent is therefore essential. So w hat is this value? One patent held by a consumer at the beginning of period t gives rise to a stream of payments.(~)PAt = (r.'"At+ c. t he perperiod opportunity cost of holding the patent would be the netincome (r. PAt· A rewriting of interest is: r. The total endowment of labour is thus L in all periods. equals the income earned from a patent in period t. we could have assumed that patents were traded once and for all.ll = (rt. so his wealth at the end of period t would be r. Hence. t + 2 . of course. the 'user cost' of a design..p AI' (6) Equation !6) says that the (net) income earned in period ton the capital PA. capital. PA<· expressed in ( 6) or ( 7) is l'uliHied. 6 For a household's decision whether to work with R&D or sell labour in the labour market.t... . . Our model has one other asset.})PAt .. R&D. by deHnition the value of a patent PA..
gives rise to an additional w 1 of capital at the beginning of period t + 1. or income. to ease notation. What is the valueadded. Finally. We consider the economic agents in turn and describe their behaviour. GDP. Only part of this income comes as a payment. C1. It is time to analyse the model. patents at the beginning of period t + 1.vo sectors as they wish. production functions and endowments. and the objectives that govern the behaviour of agents. from renting out designs the consumers earn a total cash income ofr At At in period t. R&Dbased endogenous growth: Microfoundations © The McGrawHill Companies. 1 . or income. Lyp paid as a cash !low in period t. Whenever it is clear that the variables considered pertain to period t. as primary inputs in the three production sectors.)PAt+ 1• As for consumption and savings decisions.+ 1 . in period t isY1 + (At +l .s. or total income. if it is added to the capital stock. . also paid out in period t. Y.. but the income does not accrue in cash. we will make a simplif)ting assumption: Total consumption. and A. in period t. and used in the final goods sector. (not of the total income or GDP). we will drop the time subscript t. which. One unit of work time spent in the research sector in period t gives an additional pA .Endogenous Growth 10. which give rise to a stream of payments from period t + 1 and onwards. Denoting the total cash llow of income in period t by Y. From renting out capital the consumers earn rtKt . is a constant fraction . The essential 'arbitrage' condition for an optimal allocation oflabour is thus: (10) The valueadded. 2005 PA RT 3 : ENDOG ENOUS G ROWTH The consumers earn income in period t from the use of their resources.. LAI' spent in the research sector in period t creates valueadded and income as well. we have: (9) The labour.A 1)PAt+ 1 • where we have used (5) . in period t arising from LA? The consumers can shift between working in the a. LAt = (pA 1L 4t)PAt+ 1 = (A. arising from the labour used in the research sector is w. Then we describe the aggregate equilibrium resulting from this behaviour. of the cl!st1. Hence an optimal distribution of labour power requires that the return from supplying one unit of labour in the labour market is the same as from working one unit of time in the research sector. One unit oflabour supplied in the labour market gives an income of w. L.f1ow of income. tv. K. where 0 < s < 1: (11) We h ave now finished our detailed description of the elements of the economy with respect to agents and markets.eI SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 290 I Part 3. The total value of these additional patents at the beginning of period t + 1 is pAtPAt• J· where PAt+ 1 is the capital value of a patent at the beginning of period t + 1.A . the consumers earn an income. in period t: from the labour sold in the labour market. The input L At results in additional patents at the beginning of period t + 1. or a cash llow.
but the inputs.).. It follows that: 111 = (1 . must fulfil the firstorder conditions to be compatible with profit maximization. arising from profit maximization by the firms in the final goods sector.· the final goods firms will be able to substitute more away from intermediate product j (and towards other intermediate products). If we denote the demand function in (15) by x(p. Ly and X. (An exercise will ask you to show that the elasticity of substitution is exactly 1/(1 . the larger the value of a is. for instance. A larger a means that the intermediate products substitute more easily for each other in the production function for flnal goods. the demand elasticity. A. is given by the constant exponent in (15).)..p. 111 and each P.x'(p. (~)1/(1 a) .a)Y/Ly. because all these products enter into the production function for final goods in the same way. Eq. so do.c1) (check this).x.. The input demands must fullll necessary llrstorder conditions. after an increase in P. (13) The firstorder condition for costminimizing inputs of the intermediate products are: j = 1.X _L. and each x i in order to maximize profits: IT = (~ L x")L .) Hence. (14) Again.a) . 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS 291 The final goods sector From (1) and (2).· as given. This elasticity i~ larger. ./x(p.Endogenous 10.)v. the representative firm in the final goods sector intends to choose L1. Note that this demand function is the same for all intermediate products j. R&Dbased endogenous growth: Microfoundations Growth lfl © The McGrawHill Companies. v la  i= l ~ wLY .L. 2L!1 and 2x . . A. so total labour income created in the llnal goods sector is: wLy = (1 . The firstorder conditions (12) and (14) do not determine L y and xi' If some L y and X.. flt in the conditions.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics I Part 3.r i=l taking the number of intermediate products. . as well as the input prices.a)Y. ... t = 1/(1 . In (14) we may isolate xi to get: . the larger the magnitude of a. ' a (15) Given Ly. Setting the derivative of IT with respect to Lv equal to 0 gives the condition for a costminimizing labour demand: (12) This is a usual 'real wage equal to marginal product' condition. (15) is a demand curve for intermediate product j..· The scale of production is not determined by pront maxim ization alone. these are 'real input price equal to marginal product' conditions.
as explained above. end up setting the same price. Hence. Thus. (17) The larger the production (labour input) in the fmal goods sector. so the larger a is. Thellrm will set pi to maximize the gross proflt: :rei = (pi. . the common price set for all existing intermediate products is a markup over the constant marginal cost.a).1/t). 1/( 1 . and the larger will the production.)· This maximization problem is the same for all the llrms. From the production function ( 5). a is a measure of the in tensity of competition in the intermediate goods sector.p . Consequently they will also end up with the same gross prollt. The optimal p must fulflll the llrstorder condition: x'(p)p)] x(p) +(p . of each intennediate product llrm appears if we insert the price we have just found into the demand function ( 15). R&D.r ( . P. r. p Inserting 1/(1. gives: 1 p = r.a) fort. ::rc.r)x'(p) = O*'*x(p) 1 .Jacobsen: Introducing Advanced Macroeconomics 292 I Part 3. this production is pAL A' The issue L~ therefore how much the households decide to work in the research sector..)· Therefore all llrms must. the price p goes down to the marginal cost r. .based endogenous growth: Microfoundations Growth Companies. This gives: . a = 1 corresponds to perfect competition. the more to the 'right' will the demand curves given by (15) be situated.eI SorensenWhitta.r 0~1 .e = O . the lower is the prolltmaximizing price. 2005 PA RT 3 : ENDOG ENOUS G ROWTH The intermediate goods sector The firm that has acquired the right to produce intermediate product j (through a licence) is faced with the demand curve given by (15). = p. and therefore in the selling of ideas because these are embodied in the intermediate products. A larger a implies a smaller markup. x. means that it is more costly in terms of losses of sales to increase the price. You may recall from microeconomics that this markup factor should be equal to e/(E. of each intermediate product be. x. goes to infinity)._ ( r x . and the markup factor is 1/a. If a is larger. The x that each intermediate product llrm produces is also the firm's demand for capital because of the simple production function for intermediate products we have assumed. since they are all laced with the same demand curve x(p. The research sector and the households The production of new designs in the research sector is fully determined by the amount of labour. A more elastic demand. from the conunon demand function.1) = 1/(1 . LA· supplied to the sector. end up producing the same amount. (16) a Hence. The common production and capital demand. where Ly is taken as given.= [ p x(p) p.Ly 2 (l ')1 /(1a} .Endogenous © The McGrawHill 10.r)x(p. as indeed it is. as a result of profit maximization. As a~ 1 (the elasticity of substitution. x. and then. the demand for each intermediate product will be more elastic. in turn..
. The rental rate for capital. the gross pront of each intermediate product firm. so he can set rA such as to extract all of the surplus generated by the use of design j . K/A . and hence the production in each intermediate product firm is x = K/ A. 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS lfD 293 We have already described essential elements in this decision above. total demand for capital is xA. is pushed up until the total amount of capital demanded. ) 1 (19) " . K.. Aggregate equilibrium The stock of capital in the hands of consumers at the beginning of any period is predetermined. r (given Ly): 1 y a. xA. We found in (17) the demand for capital. Clearing of the competitive capital market thus requires that K = xA. Hence: !_ = L A (_!_)1/(1 a). We have seen that this surplus is n. A. has reappeared. r adjusts to make this demand equal to the predetermined value. Since the number of intermediate product Hrms is equal to the predetermined number of designs in the period considered. . r A• of a design. the production function for final goods we have assumed throughout. x i' of intermediate product j used by the representative firm in the tina! goods sector is equal to x. so x i = K/ A for all j = 1. This is the result of integrating the Hnal goods sector \vith the intermediate goods sector. We now add the equilibrium condition for the rental rate. r . and in particular in the macro model of Chapter 9. Hence from (6): (18) Combined with the arbitrage condition (10) . equilibrium in the market for intermediate products implies that the amount. This determines the scale of production in the intermediate goods sector since both KandA are predetermined. This is the supply of capital. As already noted. r. Ly. Since x is also the production of each intermediate good.based endogenous growth: Microfoundations Companies. taking into account appropriately the equilibrium conditions for the capital market and for the intermediate goods markets. R&D. of each intermediate product llrm as depending on the rental rate. this gives the condition for an optimal allocation oflabour. Inserting this value for xi into the production function for final goods gives: Y= (t xj} i. K (20) .SorensenWhitta. and the capital value was then used to express the arbitrage condition (10).. has adjusted exactly to K. v az from w hich we can compute the equilibrium value for the rental rate.Jacobsen: Introducing Advanced Macroeconomics I Part 3. of capital supply per llrm. x. and we found above that each tlrm wlll demand the same amount. The consumer who holds the patent for design j at the beginning of period t is a monopoly supplier. x. Miraculously.a = A(~r L~a = K«(AL 1. in equilibrium r A = :rt. and on the scale of production of Hnal goods. The demand comes from the intermediate product flrms. Th is implies that the pront net offlxed cost in each intermediate product firm is 0 . The rental rate for designs was used to define the capital value of a design in ( 6).Endogenous Growth © The McGrawHill 10. So.
(24) Hence. the stock of capital develops as: K... (21) and (24). 2005 PA RT 3 : ENDOG ENOUS G ROWTH so total capital income is: (21) rK = a 2 Y.= a(1 . Cl. rK and r 4 A (see also (9) above). R&D. and the equilibrium condition from the capital market. Into this expression for n we insert the equilibrium value for r stated in (20).A. r AA. capital and patents in the cash flow of income are 1 . wLy. If we insert thatp = r/a.based endogenous growth: Microfoundations © The McGrawHill Companies.a)K A A (23) so the total licence income from patents is: nA = a(1 . but also has implications for the income distribution. The output of final goods is used for consumption or investment. respectively. (26) The parameter s thus becomes a 'ploughback' rate (the fraction of the production of llnal goods accumulated as capital). 1 . from (13). the shares oflabour. x = K/ A. + (1 .eI SorensenWhitta.a).rx.:..C = sY.Endogenous Growth 10. Y.. The llrst block consists of the four equations: Y1 = K~'(A 1 Ly1) l a.a (22) n =. This gives: 1a Y)(K) a (a. we find that the total cash flow of income of the consumers is: (25) The cash flow of income equals the production of llnal goods. How about the last component. The complete Romer model of endogenous technological progress We will present the complete model in three blocks. 1= sY. Therefore. A. the licence income.. JT = (p . We have the expressions in (13) and (21). Y..a)Y.s)Y. stating how the first two components relate to the production of llnal goods. The consumers' total cash flow of income. so we return to the expression for the gross profit of each intermediate product llrm. Y. We are now ready to collect elements from our description of individual behaviour and the aggregate equilibrium conditions into a complete model. respectively. Hence. 2 and a( 1 .Jacobsen: Introducing Advanced Macroeconomics 294 I Part 3. so total gross investment in capital L~ Y . The parameter a thus measures different things here.r):~.a. which comes from (16). consists of the three income components. from existing patents? We know that r A = n. A larger (1 implies intensilled competition in the selling of intermediate products and ideas. = pA 1L 41• (27) (28) . total consumption is C = (1 . Furthermore. from (11). n =. a. we get: 1.b)K.
and L. The first four equations.+ 1 = sY. Lv. = pAtPAt+l' (35) LA. sRt' (r. . The second restates the production function lor the R&D sector. (29) Ly. 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS lfl 295 K. We consider th is balanced growth path as the steady state prediction of the model. (1 a) . We will not conduct a full analysis of the dynamics arising from the model. respectively: 1\11 = Y. (27)(36). found in (12).Endogenous © The McGrawHill 10. while the third simply defines the R&D share in period t.c~)P At = n . Y. r 1 and rAt ( = n:... (36) The ten equations above contain th e ten endogenous variables: Y 1. = L. nt. s R• must be . In veri(ying this claim. These three equations are needed because the rental rates enter into the endogenous determination of the R&D share.' (32) K. (33) n.SorensenWhitta. PAt ands xt· For time paths (Y. combined with (36) with a constant research share. = s x L1• Here this last equation is replaced by an endogenous determination of the R&D sh are.. Y. The first equation in our last block of three equations repeats the dellnition of the capital value of patents from (18). Th is formula brings us our main results. + t. etc.). The second block simply restates our expressions lor the rental rates w . = L.PAt = P At + )  PAt• (34) w. sR. = a(1 . (Kt). The fourth is a repetition of the resource constraint for total employment.a). the second is a restatement of the arbitrage condition (10). constant research share. as it depends on the model's parameters. (31) r =a. Comparing with the macro model studied in Chapter 9. = sRtL. they must obey all of the equations in all periods. A.). the steady state balanced growth path of the model of this chapter with constant. with the exceptions that here we have assumed rp = A= 1. collected in Eqs (5)(10) in Section 1 of Chapter 9.. Thus. Therefore we get an explicit formu la for the endogenous. we see that the lour equations above correspond exactly to (5)(9). +LA. K1• A" Lyp L AP I~'" r 1. (27)(30). The last equation in the model of Chapter 9 expressed the assumption of an exogenous R&D share: LA. (20) and (23) . of these variables to form an equilibrium. we get additional insight because there is only one particular constant value for the R&D share that is compatible with balanced growth. Instead we show below that the model has exactly one lmlanced growth path along which the endogenous research share is a constant. + (1 . s Rt = s R for all t. endogenous research share. s R' make up exactly the model of Chapter 9..PAt• !!. (30) The 11rst equation is the integrated production function lor 11nal goods derived above. The third is the capital accumulation equation restated from (26 ). R&Dbased endogenous growth: Microfoundations Growth Companies.Jacobsen: Introducing Advanced Macroeconomics I Part 3.c5)K1. which is described in the following two blocks.
. we set the exogenous research share to be equal to the same s R' This verifies our claim that the endogenous growth model of Chapter 9 is a correct short cut ........ the labour inputs in the final goods and the research sector are both constant........sR)L) 1'"... = yJA. constant interest rate..etc. This shows convergence of k......... the sub model has a growth path where output (of tina! goods) per worker. A ..... etc..)............ = Kr'(ArC1 .. = _ k* = ( S ) 1/(1a)  g. !Yom equations (36) and (30)... g1 (A 1+ 1 . and the balanced growth path for the Solow submodel .4 Endogenous growth and the R&D share ...b)K1.. + (5 __! =  Y. and[J 1• respectively.L) and YJA.... respectively. The technological growth rate..eI SorensenWhitta.....+ t = (1 + g..Jacobsen: Introducing Advanced Macroeconomics 296 I Part 3.......)A........... sR' assumed for the research share..... for any constant value...... Ka t = sY1 + (1 .Endogenous © The McGrawHill 10. (K.. L1......... and the wage rate all growing at one and the same constant rate. = The submodel can be analysed in the usual way.) Hence.......... and the capitaloutput ratio.....output ratio.. towards constant levels: [J ... (40) Next tum to the block consisting of the equations for the rental rates. A balanced growth path consists of time paths (Y.. still considering a constant sR...+b (1 sR).for the appropriately micro~founded model.... (28) and (29) then boil dm<Vn to the Solowtype submodel we also considered in Section 4 of Chapter 9: Y... output per worker.. in the latter. K1/Y1• is constant: K k* = sy* 11. Balanced growth and the endogenous R&D share Assume that indeed s Rr is constant and equal to s R in all periods.... and capital per worker.... Then... (38) (39) (It will be good practice lor you to show this.. (32) and (3 3).. lor all the endogenous variables that fulfil the equations (2 7)(3 6) as well as the requirements for balanced growth: constant capital............ 10.... defining k1 Kj(A 1L) = kj(A .... = (1 ........... 2005 PA RT 3 : ENDOG ENOUS G ROWTH identical to the steady state growth path of the model of Chapter 9. is then a constant.. R&D.. y 1• and capital per worker..A 1i/A .... both grow at the same constant rate as A I' namely at the rate g. if.. given by (28) : = (37) The model equations (27).... g1 = g... k..sR)L and LA= sRL.... (31)......)..........based endogenous growth: Microfoundations Growth Companies..
(42) and (43).Endogenous Growth Macroeconomics 10.b. L = a(1 . gr. is a constant: n = a(1 . = . If we insert the constant capitaloutput ratio from (40) into (32). s (42) where g.y p 1 . . y. Start from the arbitrage condition. = . and if so. Inserting Y.. and L\'. Again. This shows that the real interest rate. (34) and (3 5). . inserting v.a)y*L. or the rental rate for designs..Jacobsen: Introducing Advanced I Part 3. what that sh are is.L. n 1 and PAt are constants... and A. along which the wage rate and the rental rates fulfil the three equations (41). Hence. we get: 1 1. If we insert the expression for w.based © The McGrawHill endogenous growth: Companies. is constant. = y .a) _ w. = y L WP.. the lower the capital value h as to be. = a(1 .L. The remaining issue is whether the last two model equations. g. = n 1 + ~ PA. along the path considered. the model equations (27)(33) and (36) imply a balanced growth path.a) (1 . r . A.SorensenWhitta.a) y. of (41) into (44) .. are compatible with a constant research share. r. (35). (33 ). given w 1• The more productive labour is in the research sector (the larger pA 1 is). . R&D. defining the capital value of patents. + c) r.a * PA =.. grow at the common rate..t: t n.sx (45) Next consider equation (34). along the balanced growth path we consider. P41+ 1 has to be constant along this path. along this path the gross profit in each intermediate product tlrm.sn)L.a)y. is constant and equal toy*: (41) This shows that w. will be growing at the same rate as y 1 and AI' along the path considered.. along the balanced growth path considered. 2005 lfl Microfoundations 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS 297 associated with it.y 1 A~' (1 . restated here: w. in the Hrst of these three equations then gives: (1 . =r= a. (r. r. we get: 1 g. If we insert that along the considered path. so PA1 = PA for all t. The arbitrage condition t h en becomes w 1 = pA 1 PA. From the block's third equation.. is given by (3 7). or equivalently: (44) This shows what the capital value of a patent must be in order to ensure that working in the research sector and selling labour on the labour market are equally attractive options.y. = pA 1 PAt+1 • Since w.c})PA.Sx) (1 . in accordance with balanced gro'Wth. n and PA.Sx) where. = (1 . [J 1 = y*. (43) We h ave shown thatjor any assumed constant values R (between 0 and 1) for the research share.
sR . in this macro model. a smaller gross pront earned on each intermediate product.sR) r . In particular.. Our conclusion is that there is exactly one constant value for the R&D share for which our model has a balanced growth path. R&D. Solving for sR (which you should do) gives: (~ 1+ apL SR = . if. One may wonder why the research share. In this way more competition in the markets for research intensive products weakens the incentive for doing research and development. we fix (the exogenous) sR at the value of (48).Jacobsen: Introducing Advanced Macroeconomics 298 I Part 3.based endogenous growth: Microfoundations Growth © The McGrawHill Companies. where discounting is done by the constant interest rate. we get: PA= a(1 . Equating the two expressions for PAin (45) and (47) yields: 1 1 a .~ from (42).'.o)PA = n.o (47) y · L. {J. Since the return to a patent in each period is exactly this gross prollt.eI SorensenWhitta. Equation (48) is the main result from our model and the remainder of our discussion here more or less follows from this formula . we get (r .o.= L. 2005 PA RT 3 : ENDOG ENOUS G ROWTH respectively. <~/(c)+ pL) Competition. n .a) _"' s r.. + o)j.. The balanced growth path resulting from this particular sR is identical to the one we studied in Section 4 of Chapter 9 for the macro model. = psR L > 0. other things being equal. a (48) 1+ s We assume that the sR of this formula is positive and smaller than 1 (it su!Ilces that < c1 2/s < 1. Using (43) to eliminate n from (46). and this is the value stated in (48). . and furthermore g. structural policy and the R&D share Inspection of (48) shows that a larger value of a implies a lowers Rand therefore less R&D. p (1 . or: (46) This shows that the capital value of a patent is the discounted value of the stream of constant gross prolits. from two opposing efJects of an increased a. .<~ Into this equation we can insert r = a 2(g. as evidenced by (2 2). along the balanced growth path there is endogenous growth: A 1 and y 1 grow at the same rate. as given by (48). = psRLfrom (3 7) . which you should show). with sR given by (48). Having done that we are left \>\lith one equation containing only parameters and the unknown. This is an important conclusion from our model and it follows. The reason is that a larger a implies stronger competition in the markets for the intermediate products that embody ideas and therefore. (implying ~pAt = 0). as we explain now. does not necessarily go to 0 as a goes to 1 (for a while assuming that parameters are such that s R > 0 even for a = 1 ). really.Endogenous 10.. r .' . each patent gives rise to less royalty income.
With a separate parameter for the intensity of competition. the phenomenon of creative destmction that we have . In our model. The reason is that a has a double content.11 1• goes to 0 (see (12) or (41)) . a lower a implies that the markup of the price. Such a formulation is indeed possible. research goes to 0. \. R&Dbased endogenous growth: Microfoundations © The McGrawHill Companies. One would like to disentangle its two roles. the gross profit.) Subsidizing research thus seems to be an attractive alternative to weakening competition . as we explained in Section 1. This is the exact opposite policy to wh at most competition authorities try to carry out. so the return to labour in the final goods sector. Increasing a really has two opposite ellects on sw The fall in gross profits pulls downwards. This would raise the return to research above the PA of our model leading to resources being transferred from the final goods sector to the research sector. However. but at the same time it exacerbates the underutilization of existing ideas. (An exercise will ask you to perform this analysis explicitly. our discussion here of appropriate policies in relation to private R&D is far from complete. The result that competition is bad for R&D. as well as in the research sector. and more so the more market power there is. goes to 0 (while 'capital's share'. having two different parameters determining the degree of competition and 'labour's share' respectively. a 2 • goes to 1). and that market power enhances R&D. Less intense competition may work to stimulate the production of new ideas. so nothing can be earned from patents. goes to 0 (see (22) or (43)). n. market power tends to cause underutilization of the ideas that already exist. As a goes to 1. Seemingly. implying a larger forgone opportunity for welfare that could be harvested by using more of each intermediate product. the wage rate. Hence. gives substance to an often heard argument concerning competition policy. There is an important tradeoil' here. but has been avoided here for simplicity. a technological progress to the benefit of everybody. However. Therefore the arbitrage condition is compatible with labour being distributed both to research and to production of final goods as a becomes large. p = (1/a)r. This is w hy we said that Bill Gates' argument was justified 'to some extent'. at least to some extent. What can policy makers do? An obvious suggestion is to subsidize R&D activities by public funds. not Jess. and the first ellect happens to be the strongest in our formulation . and as perfect competition is approached. When Microsoft was accused by the American competition authorities of having a too dominant position in the software market. The parameter a is in a way overburdened here. stronger competition \viii imply less research without any counteracting effect. Our model suggests that Bill Gates was justified in this claim. the 'direction' of the result we have obtained here is correct in the sense that it is the same as in a more satisfactory formulation.Jacobsen: Introducing Advanced Macroeconomics I Part 3. but it does not want to weaken competition lor the reasons just explained.SorensenWhitta. Assume that society wants to stimulate privately conducted R&D. For instance. over marginal cost increases. 2005 10 R&D BAS ED ENDOG ENO US G RO W TH: MICRO FO UNDATIO NS lfl 299 As a goes to 1. in the markets where research intensive products are sold. an appropriate policy for stimulating R&D and growth is to create more market power.Endogenous Growth 10. Bill Gates countered that the market power of Microsoft was exactly what gave the incentive for the development of new software. the fall in the wage rate pulls upwards. In our model we could imagine that government pays a certain amount to households for each unit of time they work in the research sector and (for the sal<e of argument) that government fmances its expenditure by nondistortionary lump sum taxation.
g. a larger s also implies a larger gross profit. sR. capital becomes more abundant. 2005 PA RT 3 : ENDOG ENOUS G ROWTH abstracted from (as explained at the beginning of Section 2) complicates matters. R&D.b. along the balanced growth path. but it is probably better to somehow subsidize research than to weaken competition. Investment and research Another conclusion following from inspection of (48) is that a larger savings and investment rate. Empirics We have found that the endogenous research share. increases with the investmen t rate. as can be seen from (39) and ( 43). Furthermore.. an increase ins increases the capital value of patents thereby stimulating the incentive for doing R&D. affects the growth rate. intended to stimulate savings. This means that for sulficiently large¢ our model accords with the empirical crosscountry evidence presented in Chapter 8. r .Jacobsen: Introducing Advanced Macroeconomics 300 I Part 3. the wish to stimulate savings came from the model property that an increased savings rate would increase the level of the economy's longrun growth path and therefore imply more growth during a transition period and a higher level of GDP (and possibly of consumption) per worker permane11tly. The negative externality would have to be weighed against the positive one... should be seen as an approximation of a¢ in the range just below 1. In all our growth models. where an increase ins does not give perma11e11tly higher growth . since r is a cost in the production of intermediate goods. a larger research share implies that the growth rate is higher for a long time (permanently). when society decides u·it should tax or subsidize R&D. but very longlasting additional growth. g. In most of the growth models we have considered. g. and for¢ close (equal) to 1. in turn. Obviously this makes policies to stimulate savings and investment of more interest. in output per worker. implies. r. see (46). implies a higher level of the balanced growth path. beside the positive one arising from increased knowledge. and hence a lower interest rate.Endogenous Growth 10. and it is not clear exactly how it should do so. that is. a lower rental rate.Whitta. The new feature in this chapter is that the existing level of . s. In the model ofthL~ chapter an increase in the savings and investment rate. a higher propensity to save implies a higher growth rate of output per worker and of At' This has consequences lor the desirability of structural policies.eI Sorensen. lor a given gross pront n. a11d a higher growth rate. tax policies.). as can be seen from (39). e. s.based endogenous growth: Microfoundations © The McGrawHill Companies. since the invention of one firm displaces earlier inventions. but the permanent growthpromoting effect of a larger s depends on the assumption¢ = 1. s. ¢ = 1. sR. One can see from (42) that for a given sR (and g. r. implying a larger sR' Since the research share. a larger s implies a lower rental rate. implies a larger sR' The reason is that if the propensity to save increases. and a larger supply of capital implies a lower price for capital services. We repeat that the knifeedge case. longrun growth of income per worker requires longrun technological progress. A lower interest rate. The conclusion is that it is not clear to what extent society should subsidize R&D. Creative destruction really means that there is a negative externality associated with R&D activities. All in all. that the capital value of a patent increases. indicating a clear positive relationship between investment rates and growth rates of both GDP per worker and labouraugmenting technology.
so sh ould A.4 does not point to a constant growth rate of A. I. the number of patents granted up to the beginning of year t. Many patents turn out to be of no value since the ideas behind them are useless. of the stock of patents from 1800 to 2000. A. and A . useful or not.. Let us take the crude view that the number. = (1 + g)1 AO' In % 5 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 Figure 10..Endogenous Growth l fl) © The McGrawHill 10. but rather to a gradually decreasing one that perhaps flattens out over the last 50 years at a level around 2 per cent. A 1 = uA.)/ A. as we shall now explain . g. that is. Fig. R&D.. A. for our model to accord with the Western experience of 200 years of constant annual growth rates of GDP per capita around 2 per cent.. Let us denote by A. . or shoes for horses are important inventions that have never been patented. the stock of ideas should have grown at a constant annual rate around 2 per cent as well. so we will consider these. We may get some idea of how the stock of ideas resulting from private research eflorts has evolved over time by studying how the (cumulative) number of patents has evolved.. plus a llxed number. However. of (useful) ideas that were never patented.based endogenous growth: Microfoundations Companies. 10. A. grows at a constant rate (which is what we want to test).. For each year since 1790 we know how many patents were granted by the US patent authorities. At llrst glance.. Hence.4: Annual growth rate of the cumulative number of patents granted in the USA Source: Tec1nology Assessment and Forecast {TAF) Branch. Figure 10.. (A.. 2005 10 R&D B AS ED EN DOG ENO US G RO W TH: MICRO FO UNDA TIO NS 301 technology is narrowly identified with the stock of ideas that has resulted from private R&D activities. When there were no patent authorities. so that A.SorensenWhitta.A. indeed grows at a constant rate.1l ideas. For instance. nevertheless. of useful ideas at the begin ning of year t is a certain fraction u of the stock of patents. for instance. Assume further that the number of usefi. if a relatively constant fraction of the patents granted turn out to be useful. if A. so A.Jacobsen: Introducing Advanced Macroeconomics I Part 3.+ 1 . The USA has good patent data. Sails for boats. is not an accurate measure of our model's A. there should be a close relationship between A. According to our model the longrun growth rate of output per worker is equal to the growth rate of this stock of ideas. compatible with a constant growth rate of A. ideas could not be patented.4 shows the annual growth rate. US Pat ent and Trademark O ffice (USPTO). + I. The figure is.
the artillcial smooth curve follows the actual curve rather closely except for some 'arbitrary' fluctuations in the latter..A. g (49) I 1 1A0 (1 +g)' where I/A0 is the initial fraction of unpatented useful ideas..5 sh ows in th e left part the (natural) log or GDP per worker in the USA. Therefore.eI SorensenWhitta. US Pat ent and Trademark O ffice (USPTO). have grown at constant rates over the period considered and the average annual growth rates are remarkably similar.. g At+l .::...based endogenous growth: Microfoundations Companies...I I 1 A. R&D.4 shows the growth rate offl 1 as it should evolve according to (49) with the initial year being 1800 and fore = 0. as pointed to by the model of endogenous growth we have studied in this chapter.. shown in Fig. e=a"' nn. 10. 10..Jacobsen: Introducing Advanced Macroeconomics 302 I Part 3. The smooth curve in Fig.:. and In A... In y I ' with the linear trend indicated in grey. profitmotivated research..5 1. Both In y. (since the denominator in (49) goes to 1 as t gets large)..2 ~~l 0._:. Figure 10.4 ~~7~ g 0 0.Endogenous Growth © The McGrawHill 10.1 1""7l(J..:u:.....9 Average annual growth rate: 1....9 5.8 ~ 0..8'  r / / 7 /  0. the actual curve for the growth rate in the cumulative number of patents in no way contradicts that the stock of knowledge arising from private..=.9 0.. and hence under our assumptions A.1 0 1950 1960 1970 1980 1990 2000 v Figure 10.:..:.4 evolve? You will easily compute: At+ l.AF.2 / 0..:. Clearly.....7 ~~~ m0. The figure gives a strong indication that both y 1 and A1. 10.8 % ~ 0..5: Log of GOP per worker (left) and log of cumulated patents (right) for the USA Sources: Penn World Table 6.. A.. how should the growth rate of A. again with the linear trend indicated in grey.:.::a'' 1 .At =. A..:. The empirical evidence thus indicates a close long run relationship between economic growth and growth in the stock of ideas resulting from private research and development. directly with the growth rate of A. 0...8 [ 0.. a. ~ ~ 1 .4 flattens out and we may identify the growth rate of A.1. . 0.3 _. 1950 1960 1970 1980 1990 2000 1/) 0..:..7 "0 0. 0.___ _ _ _ _ _ _ ___ 0+..4 0 0.02 and the in itial fraction of unpatented ideas equal to 0 ... In A. 2005 PA RT 3 : ENDOG ENOUS G ROWTH these circumstances. should have grown at a constant annual rate of 2 per cent over the last 200 years.._:_'7/ 0.. Technology A ssessment and Forecast (TAF) 3ranch...5 *5 0 0.6~~~q a. Let us now focus on the period 19 50 to 2000 where the curve ofFig.0. have been normalized to 0 in 19 SO.= A1 . and in the right part the log of the cumulative number of patents..6 :.3 Ol 3 1 Averagc::.
....... This way of obtaining private R&D activities is...... 2... it is effective in the sense that much more private R&D will take place in the presence than in the absence of laws that protect intellectual property rights.. The fact that ideas are nonrival implies that to induce private R&D.. and by which means society should try to stimulate private R&D is far from settled..... from a social viewpoint....... 10... It therefore seemed better....... 4... R&D... imperfect.. if one wants to stimulate private research. 5........ However.. and a higher propensity to save and invest tends to increase it.... Otherwise the fixed development cost cannot be covered.. 2005 10 R&D BASED ENDOGENOUS GROWTH: MICRO FO UNDATIONS 303 Final remarks One conclusion that stands out from this chapter is that the last 20 years of 'new growth theory' (the R&Dbased theories of endogenous growth) have highlighted the economic incentive mechanisms that underlie private research and development and economic growth..5 Summary ...... This chapter has studied the microeconomics of private idea (R&D) production. 3... 1.............. Suddenly the canteen belongs to the research department... but would also imply a stronger over· pricing and underutilization of ideas........... the potential eflects of structural policies meant to iniluence and improve the incentives for research and development have come to the forefront of the debate over economic policies........Jacobsen: Introducing Advanced Macroeconomics I Part 3........... market power is needed in the selling of ideas and ideabased products.......... subsidies or tax privileges to specific activities always raise the issue of abuse and the problem of targeting the incentives at socially beneficial activities which would not have been undertaken in the absence of the subsidies. to subsidize research somehow........ Consequently.. We considered private idea production in a fo rmal model explicitly accounting for the nonrival character of ideas and the economic problems associated with nonrivalry. underutilized.... 6. Even though a patent system is not a perfect way of ensuring private research and develop· ment.. to what extent............ According to some economic historians. the unprecedented economic growth during the last 200 years is partly due to the fact that the legal protection of intellectual property rights reached a critical level some two centuries ago........... according to our model. Our main con· elusions were: stronger competition in the selling of ideabased products tends to lower the R&D share. The debate over whether.....SorensenWhitta... increase R&D and therefore result in more ideas being produced.... Ensuring monopoly power to the sellers of ideas implies that ideas are priced above marginal cost and therefore... however.. A policy to ensure more market power would.. The fact that ideas are only imperfectly excludable necessitates some arrangements to increase excludability and prevent copying in order for private production to take place.... The second conclusion (with respect to the influence of the propensity to save) had the implication that policies meant to increase savings and investment rates became (even) more ... The policy implications of the first conclusion (with respect to competitiveness) were not too clear.....based endogenous growth: Microfoundations © The McGrawHill Companies.Endogenous Growth 10.... Patent laws as well as other legal ru les for the protection of intellectual property rights can serve the double purpose of ensuring both excludability and monopoly in the production and selling of dea·based products........
DEGREE OF RIVALRY RIVAL > l= ::::! NON.eI SorensenWhitta. in Proceedings of the World Bank Annual Conference on Development Economics 1992. not just on the level of the longrun growth path.EXCLUDABLE (OR CLOSE) Your task is to discuss and suggest an appropriate placement of the following commodities: • Drinking a beer in a bar • Enjoying the atmosphere in a bar • Bohr's atomic theory • Travels • Great scenic views • The concept for a TV show 7. . supplement to the World Bank Economic Review. beers Passage of a bridge Fish in the sea? (think of quotas. bread.RIVAL Potatoes. national defence FULLY EXCLUDABLE (OR CLOSE) Ill <( 0 ::J _J 0 X UJ INTERMEDIATELY EXCLUDABLE Ll. While rivalry is a zeroone concept. excludability occurs more continuously in various degrees. March 1993. 2005 PART 3: ENDOGENOUS GROWTH interesting than when based on models of exogenous growth. We have inserted a few commodities in the figure. Romer.based endogenous growth: Microfoundations Companies. pp. R&D.Jacobsen: Introducing Advanced lacroeconomics 304 I Part 3. 6391.Endogenous Growth © The McGrawHill 10. that the assumption of ¢ close to 1 is requ ired for having very long lasting effects on growth of increased savings. One should remember.. Paul M. since such a policy would have a permanent positive effect on growth. 0 UJ UJ 0:: C) UJ 0 NON. however. etc.6 Exercises Exercise 1.) Computer code for spreadsheet Fish in the sea Fresh air. according to their characteristics. 10. 'Two S trategies for Economic Development: Using Ideas and Producing Ideas'. Rivalry and excludability The figure below is taken w ith modifications from an article by Paul Romer. in the figure reflected by three degrees. 7 It d ivides the plane of commodity characteristics according to two criteria: degree of excludability and degree of rivalry.
Why? (Hint: You should use (41 ) and (43) to see that. (46). R&D. for a given A" an increased (5 affects the wage w. p. sR' Surprised this time? . n .Endogenous Growth 10. and then how it affects the capital value. when the PA of (47) is set equal to the PA of (45). implies a smaller research share. Exercise 2.based endogenous growth: Microfoundations © The McGrawHill Companies.o. first hold sR fixed. 3. a larger productivity in the R&D sector. a larger rate of depreciation on capital. r. However. Now show that according to (48). it is good exercise to study how s R depends on other parameters as well. However. implies a larger research share. PA. What is the elasticity of the composition of demand. and the interest rate. To understand the dependence found in Question 1. of a patent for a given fixed 'profit'. 4. (Hint: Use Eqs (42) and (46)). this latter effect is neutralized in the complete determination of sR.) Now try to explain the basic mechanism directing more resources towards R&D in response to an increased rate of depre· ciation on capital (when a 2 < s). and in particular to explain the economic mechanisms behind. You can see from (39) and (43) that an increased oimplies a lower gross profit n . ct. Exercise 3. and the gross p rofit n in the same way. at least one for each of the cells. We assume that a 2 < s. which itself pulls in the other direction on PA than does r(5. 1.SorensenWhitta. xj xi. like discoveries in basic research • Cable TV signals • The idea for a new type of motor for cars (discuss legal and technical ways to ensure more excludability) • • Taxi rides Parking in a car park at a shopping centre (discuss the implication of the size of the cost of enforcing a payment system) • Pythagoras' theorem Now use your own imagination and place other examples of commodities in the figure. We said that perhaps the most interesting determinants of sR are a and s. s R ' Does this surprise you or is it more or less as you would have expected? 2. The elasticity of substitution between intermediate products Consider h e demand fo r intermediate products from the final goods sector as given by (15). between two intermediate products PIP? Explain that this elasticity must measure the ease by with respect to the relative price which the intermediate products substitute for each other in the production of final goods. Show that according to (48). that is. The dependence of the research share on the rate of depreciation and on productivity in the research sector In the model we have studied in this chapter the R&D share depends on parameters as given by (48). Investigate how an increased oaffects the rental rate. o. 2005 10 R&D BASED ENDOGENOUS GROWTH: MICRO FO UNDATIONS 305 • TV transmissions 'in the air' (discuss encoded versus nonencoded and the extent to which excludability of nonencoded signals can be obtained by legal arrangements) • • Biscuits The idea of the assembly line • Mozart's 'Eine Kl eine Nachtmusik' • The services of a dentist • Basic R&D. r.Jacobsen: Introducing Advanced Macroeconomics I Part 3.
The compensation for one unit of labour used in the research sector is the capital value of the patents it creates. (37). R&D. while the household receives (1 + r)r At including government's subsidy.eI SorensenWhitta. Show that one still has 3. A way to subsidize research is therefore to levy a negative (or reduced) tax on licence income. In our model. according to the model considered in this chapter (assume a 2 < s)? By which mechanism does a larger L affect s R? (Hint: As in the former exercise you should go through howL affects g. Net of subsidies and taxes this cash flow is: Y= wL r + rK + (1 + r)rA A Y= Y. How does an increase in p affect the interest rate r. affected by an increase in the labour force. Go further and convince yourself that the next time something should be revised is at the point where the consumers' income cash flow is computed. Hence. (42)? Now try to explain 5. the government pays another rrA.based endogenous growth: Microfoundations Companies.sc. affects r. = rrAcA .a)PA= (1 + r)n + !:iPA. 2005 PART 3: ENDOGENOUS GROWTH a~ain s R fixed for a while.o .Endogenous Growth © The McGrawHill 10. L. Total government spending (on research subsidies) in period t is T. in tho.). (18). which is financed by lump sum taxes on the households. We assume that whenever a household rents out a design (patent) to a firm in the intermediate goods sector. g • • affected overall by an increase in p? Exercise 4. Government subsidies to private R&D For the reasons we have discussed. Go through all the arguments and formulae we have presented from the beginning of Section 3. which in turn reflects the licence payment for patents. y L( 1 +r). and then rfor a given sR.t co. Hold the mechanism through which an increased productivity p in R&D implies that a smaller fraction of resou rces is directed towards research. g. and the endogenous growth rate. How is the growth rate. where r is the (constant) real rate of subsidizing income earned from patents. when it is taken into account how p affects g. sR. a government may want to stimulate private research without affecting the competitiveness in the markets for research intensive products. g ive an amount to each unit of work done in the R&D sector.. 2. is to subsidize R&D activities. Exercise 5. . and convince yourself that the first time something should be revised because of the government intervention described is in the formula for the capital value of a patent. Use the above to convince yourself that in the complete model (Eqs (27)(36)). An obvious ideo. etc. government should. T (where T is the lump sum tax in the period considered).Jacobsen: Introducing Advanced lacroeconomics 306 I Part 3. There are no other changes made to the model considered in this chapter. and how g. receiving the rental payment rAc from the firm. . the intermediate product firm pays rAc for using the design in period t. 1. to the household. 6. which should be changed to: (r.a. The scale effect How are the research share. r. Show that the only implication this has for the balanced growth path is that (47) should be replaced by: PA= a(1 a)_* . the only change required is that (34) is replaced appropriately as found in Question 1 above.
We revise the model to consider </> < 1 and positive population growth. and then use (38) and g • =p s RL). To do the analysis below is very good practice. Assume that a similar increase in s R had instead been obtained by a lowering of a. x= K. Define the technological growth rate.1.. Show that it follows from (28') that g. not </> < 1 leading to semiendogenous growth. sR.Jacobsen: Introducing Advanced Macroeconomics I Part 3. PAr• of a patent will no longer be constant. (with the usual definitions. converges towards: g se e (1 + n) 11<1 <P) _ 1 > 0. 5.. 2005 10 R&D BASED ENDOGENOUS GROWTH: MICRO FO UNDATIONS 307 4. s R. growth rate. n > 0. This exercise asks you to analyse the model with semiendogenous growth.= L. an increase in the rate. L.(36) this has the following implications: Equation (28) is replaced by: (28') 0 < </> < 1.+ o)? Comment with respect to the desirability of obtaining an increased research share by weakening competition versus subsidizing research activities. (35). (30") The arbitrage condition. s R. by which research is subsidized implies an increase in the research share. e (A t+1 A . )/A. (more market power). Now show that for g 1 = g se in all periods. 1 = ( 1 + n)L1 . to which each existing idea is used along the balanced growth path affected by an increase in r ? (Hint: Use that in equilibriurT. We will be looking for a balanced growth path with a constant research share. must be replaced by: (35') (explain why).Endogenous Growth © The McGrawHill 10. Show that under the obvious restriction. and finally (36) must be replaced by: (36') Analysing the resulting model is a bit more complicated than analysing the model with </> = 1 .L. (We do not bother to consider a . less than 1).based endogenous growth: Microfoundations Companies. Show that the (endogenous) value for the R&D share compatible with balanced growth is: (5 apL S R= ~~~~ a. 1. For the model (27) . (This is just as in Chapter 9). . g. and find what that s R must be. x. How would that affect x in comparison (realistically assuming s > g.+ LA. just to check which conclusions are changed and which are not. because now the capital value.. one must have. g •. and in the r. R&D./A.= k. sR< 1. s (1 +r) +  and compare to (48). Equation (30) is replaced by the two equations: (30') Ly. along a balanced growth path with a constant research share. Exercise 6.SorensenWhitta. The Romer model with semiendogenous growth In this chapter we have only considered the case </> = 1 leading to endogenous growth. How is the degree.
Verify that along the balanced growth path the real wage must be growing at the same rate.. PAt• must be growing at rate n. "' Y. show that along a balanced growth path with a constant research share.. 2005 PART 3: ENDOGENOUS GROWTH k. . 9se• as technology. (PAt+1 ./A. show that along a balanced growth path with a constant research share. as the labour force. . and gross profit in each intermediate product firm must be growing at the same rate. sR. and verify that along a balanced growth path with a constant research share./(A . this research share must fu lfil: SR= fc5.= k* = ( S )1 /(1. the capital value of a patent. Combine the e)pression you have just found for PAt with the one you found above for :n: t to find an expression for PA1+ 1 • Insert your expression.= n or PAt+J PAt = 1 + n.eI SorensenWhitta.1.y*At. 0 < s R < 1. = r=a 2 n + 9se + c5 + ngse ::.)) : _ _ k. n +g •• +o+ng•• Yt =j* = ( S ) u/(1 . Using your resu lts from Question 2 and the arbitrage condition {35'). :n:t= a( 1 a)f*Lt.PAt )/PA./(A . s R• one must have: :n:t PA. as well as the expression you found above for w. . Now show that along a balanced growth path with a constant research share. 3.=. Using your resu lts above and the definition of the capital value of a patent. 1. in the arbitrage condition {35').fwt_1 = A. s Fi' one must have for the factor rewards {entering into the endogenous determination of the research share): 1a wt= . (rto)PA.= :n: t+!!. w.based endogenous growth: Microfoundations Growth © The McGrawHill Companies. n. 2.Endogenous 10.a) n +g•• +c5+ ng•• (1  SR). that the interest rate is larger than the growth rate of the labour force.L. Kt = s Y.L. K.)._ 1 = 1 + g se etc. sR. that is.) andy. divide the first equation by the latter. 4. rc5n 5. and then use that from Question 2.Jacobsen: Introducing Advanced acroeconomics 308 I Part 3. PA. {Hint: Write {35') for periods t and t..a) ( 1sR).n 1 +ag••( 1 + n) Show that under the assumption. R&D. n+g•• +a+ng•• (This again is as in Chapter 9).S R r.
Show that the endogenous s R is decreasing in a and increasing ins. 2005 10 R&D BASED ENDOGENOUS GROWTH: MICRO FO UNDATIONS 309 6.based endogenous growth: Microfoundations © The McGrawHill Companies. (It is okay to keep 9 se in the formula.Whitta. we know how g se relates to n and ¢and that it only depends on these two parameters). in order to find the endogenous s R along the balanced growth path traced back to parameters. Compare with our results in the chapter and comment.Sorensen.Endogenous Growth 10. Insert the expression you found above for r into the formula of Ouestion 5 for s R• and possibly rearrange for a manageable expression.Jacobsen: Introducing Advanced Macroeconomics I Part 3. How does a change in s R caused by a change in a or s affect the long· term growth rate of output per capita in this model? What good does such an increase in s R possibly do? . R&D. 7.
............................ 311 ...........i....~ ......~&..............~~Y.......~.~~... ~~?.........~..............~~.......Y.......................................~............ 12 ............ ..............Y...l?........l?.................. 1 3 .Stnctural Unemployment I 11.P..~...~~.l?......~...~l?:~....~?...eI SorensenWhittaJacobsen: Introducing Advanced I Part 4..............~~..~.t..... .~......... ..T~..................~P:~~I?......~~l?:~.....~?..................Y!:?.~.............. ~........... ~~~~l?...~?..~~J?:.................................~~. .......~...~J?:~........ 2005 Structural Unemployment • • • • • • • •• • • 11 Some facts and introductory theory about ........... Some facts and introductory theory about I © The McGrawHill Companies.~......~............................
Stnctural Unemployment 11. 2005 . Some facts and introductory theory about unemployment © The McGrawHill Companies.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics Part 4.
313 . However.0 I SorensenWhittaJacobsen: I Part 4 Introducing Advanced Macroeconomics Stnctural Unemployment 11. 11 . Figure 11.1 are not completely comparable between the two countries. For society it is important how high the longrun rate of unemployment is. Our growth models were just silent about what determines the natural rate of utilization. and cyclical fluctuations around the trend. adjustments in the real wage ensured that labour demand was always equal to labour supply. rate. Some facts and introductory theory about unemployment © The McGrawHill Companies. As we explained in Chapter 1. It makes a di!ference if the average rate of utilization of the most important economic resource. for both countries the rates represent all unemployed as a percentage of all worl<ers rather than all insured unemployed as a percentage of all insured workers. 1 The fact that the long run 'gravity' rate of unemployment is not zero but is rather in the range of 5. The fluctuations are sometimes large.7 per cent over very long periods.7 per cent is not a problem for the growth theory presented in th is book.1 shows average annual rates of unemployment for the USA and Denmark over a long historical period. because markets were assumed to be perfectly competitive. and perhaps so large that one would really not call them fluctuations. appearing in the growth models as total employment when the labour supply is used at a normal. and over the 1970s and early 1980s. We can simply reinterpret the L. For instance. the movements in the rate of unemployment can be viewed as consisting of two components: a Iongnm trend which could be described as a tendency towards a constant unemployment rate around 5. Occasionally unemployment has gone to very high levels in both countries. labour. This chapter. In the real world there is always at least some unemployment. Literally speaking we assumed th at there was no unemployment at any point in time. are focused on the forces that determine the longnm or socalled natural rate of unemployment. or natural. I 1. 2005 Chapter Some facts and introductory theory about unemployment n the growth models studied in the previous parts of this book. is 96 per cent or 92 per cent. in the 19 30s the rate of unemployment went up to 25 per cent in the USA and above 15 per cent in Denmark. The data underlying Fig. and the next two. unemployment rose to rates of around or above 10 per cent in both countries.
.... 0konom1Sk V~kst i Danmark..SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 314 I Part 4. we w ill go into more detail w ith the reasons why unemployment insurance typically cannot and does not fully cover the income losses arising from unem· ployment.. 191483. Macmillan.......~~g~~~~~g~~g. They therefore suffer substantial income losses that may force them and their families to move from their homes....1: Rates of unemployment in the USA and Denmark during the twentieth century Sources: R. 1984.S 11. . and it only covers a limited period... 2 In addition to the loss of income. At the level of society.. Thus the income loss of an unemployed person is partly suffered by the unemployed himself and partly by other members of society .... International Historical Statistics... 2005 PART 4: STRUCTURAL U NEMPLOYMENT 30 . Unemployment thus gives rise to serious individual problems..... longer periods of involuntary unemployment may inllict serious psychological costs on the jobless..... High unemployment is a major economic and social problem and a concern of economic policy makers mainly for two reasons. because income that could have been earned to the benefit of the individual as well as society is forgone . When unemployment is high.... Akademisk Fori<g.Stnctural Unemployment 11.. and the tax revenue could have been used to the benefit of other citizens.. etc. S.f. it does not insure a person's income fully..1 o~r unarl..... Mitchell. 2.~~~~p~gy~~!?:~...A. where we d iscuss the costs of the cyclical part of unemployment. 1998.. unemployment is a social waste. At the individual level..... In any case it is a waste of resources to have people who are both qualified and willing to work at the going wage rates... when unemployment is high many persons will have to go through long spells of joblessness.D "' . bad they bad jobs. B ind 2.. 0 ... Hansen. the associated loss of tax revenues will be substantial..B.. take a cut in their standard ofliving.. Bureau of Labor Statistics (US).. S la lislit..... Although unemployment insurance is available in most countries..... T~. because an unemployed person may feel excluded from society and therefore lose self esteem..2 15 :.USA 25 . would have been taxed. change the children's schools..... In Chapter 20. Some facts and © The McGrawHill introductory theory about unemployment Companies..... unable to do so..Denmark ~ 20 Q) 2 . Statistisk Tiarsoversigt.... The income that unemployed persons would have earned....J 10 5 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Year Figure 11.
since otherwise the firm could increase its pro!Hs by tiring the marginal worker. W/P... or because the most productive workers are hired first. An involuntarily unemployed person would like to work at the going real wage rate. that is. then there must be unemployed workers for whom MRS< W/P ~ MPL.. so th is wage rate must be greater than his or her disutility of work. employment and unemployment The rate of unemployment is a stock variable that can be measured at a given point in time.... If a worker's marginal product is larger than his or her MRS... A person will obviously want to work ifotl'ered a real wage above their MRS. Collecting what we h ave found.. that is.. say on a specitlc day.~.. Some facts and introductory theory about unemployment © The McGrawHill Companies.. T. Let the total number of persons who would like to work on day 3. MRS. The marginal product of labour must be held up against the marginal rate of substitution between consumption and leisure. if there is involuntary unemployment and firms maximize prollts... Now.. A worker's MRS is the amount of goods needed to compensate him for the leisure and the value of the home production he must give up when taking a job.~plgY.~......... there will indeed be jobless workers with marginal products exceeding their disutilities of work. e. Typically dillerent workers will have a dill'erent MRS... there '<Viii be unemployed workers for whom the condition W/P ~ MPL is (very close to being) fullilled.... On the plausible assumption that the marginal product of labour decreases gradually '<\lith employment..~!... say.. implying MRS< MPL.....~...~. or the decrease that will result if one individual is taken out of work...g.. W/P ~ MPL.. This chapter lays the basis for answering these questions by going through some empirical regularities concerning w1employment and by presenting some introductory theories..l.... These are exactly the questions to which we now turn. The next two chapters '<Viti present two main theories of structuraltmemployment.... since putting him or her to work gives an additional amount of output sufilcient to compensate the worker and to leave a surplus that can be shared with other members of society (through taxation) to the benellt of everybody.. Hence their joblessness involves a social loss. then a social loss occurs if he or she does not work..... MPL is the increase in production that will arise if one more worker L~ put to work..~. ?.....r.... Usually we think of the marginal product oflabour as being decreasing in employment because of the presence of fixed factors.... ~.. 8 per cent to 4 per cent.....Jacobsen: I Part 4 Introducing Advanced Macroeconomics Stnctural Unemployment 11. there will be unemployed workers with marginal products close to that of the last worker hired.. In other words... if there is involuntary unemployment. An exercise will ask you to restate the explanation of the social cost of unemployment using a standard diagram.. capital or land.. A situation where MRS< MPL thus implies inetllciency in the use of labour resources. also referred to as the marginal disutility of work.....f..... MRS< W/P.. Let the marginal product of (a particular type of) labour be MPL.~~~... 3 For these reasons it is of great interest to know what determines the rate of structural unemployment and what policy measures could bring the natural rate of unemployment down from...... Labour force....0 I SorensenWhitta.. ..~. 2005 11 SOME FACT S AND I NTRODU C TORY THEO RY ABO UT UNEM PLOYM ENT 315 Economists explain the social waste associated with involuntary unemployment in more detail as follows.. Profit maximization implies that the last worker hired h as a marginal product at least as great as the going real wage rate... so the total labour supply induced by a given real wage is the number of workers having an MRS below that wage rate..
etc. Table 11. L1 = n 1 P1• One can further subdivide into various groups such as women and men. A third way is to let L.1= .. is the number of persons actually having a job on that day.1: Labour market data.782 Unemployed (U = L . that is. respectively.). Total employment on day t. namely the number of persons..775 69% 83% 78% 87% 64. Bureau of Labor Statistics (US). n .. One can then let L... Another way is to let L. E. so we can infer that indeed these people want to work. L.735 1.1% Population in working ages* Women Men Participation rates Women Men Labour force Women Men * Person between 16 and 64 years of age. be the sum of all those who have a job (E. and let E. = L. Some facts and © The McGrawHill introductory theory about Companies. The number of unemployed is then simply U. P1.186 88.1 shows an estimate of the labour forces in the USA and Denmark lor the month of January 2002.893 Employed (E) 128. The U. Out of these theE. employed actually have a job." are the participation rates of women and men.. .E.Stnctural Unemployment Macroeconomics 316 11. E.774 169 Rate of unemployment (U/L) 6. '' where we have denoted the rate of employment bye. and it also shows the decomposition of the labour forces into employed and unemployed. u = .. be the number of insured people who have a job.) and all those who are registered at some kind of (public) institution as looking lor a job but not having one (U. inflation.043 72.Jacobsen: Introducing Advanced I Part 4.) or receive unemployment benellts or social assistance benellts motivated by joblessness (U.013 2. How do we measure the number of such people? One way is to limit attention to those covered by unemployment insurance and to assume that individuals who plan to be active in the labour market will want to insure themselves against unemployment.= 1 . The rate of unemployment contains an element that is not as readily measurable as GDP. This is the labour force.E) 8.537 Total Labour force (L) 136. times the participation rate. L1 = n:.e ' L."P. be the total number of people holding unemployment insurance. which can be viewed as the size of the population of working age. L. consist of all persons who either have a job (E. are persons who would like to work and yet do not do so. L.154 1. however.E. January 2002 (in 1000 persons) USA Denmark 92. Statistics Denmark.787 2. employment."r.SorensenWhitta.4% 6."+ n:. and the rate of unemployment is U L. who would like to have a job. where nt and n. 2005 unemployment PART 4: STRUCTURAL U NEMPLOYMENT t be L.". unemployed.) . Sources: Statistisk Tiarsoversigt. A fourth way is to estimate the number of unemployed by survey techniques asking a Table 11.355 1.744 1. ..= 1 ..
4 per cent of all people in the labour force. In the USA the annual rate of unemployment was substantially lower than in Denmark in 1993. but most of these people were unemployed for considerably less than all of the year. shows aruma! rates. Fig. 10. the data for unemployment most often shown are annual (or quarterly or monthly) rates.0 I SorensenWhitta. 20 per cent were unemployed for more than half the year.1. % 7 12 centage of the labour force Unemployment rate % Sources: Bureau of Labor Statistics (US).Jacobsen: I Part 4. namely 7 per cent. who were unemployed for Less than 5 weeks 514 weeks 1526 weeks 27 or more weeks USA Denmark 42 31 12 14 39 25 16 20 100 100 16 23 5 7 Persons having experienced unemployment as a per centage of the labour force 20 29 Unemployment rate. lor instance. and all the data for unemployment you will see in this chapter are of these types. A considerable 36 per cent of those allected by unemployment. who were USA unemployed for Less than 5 weeks 514 weeks 1526 weeks 27 or more weeks Denmark 37 29 15 20 21 23 19 36 100 100 Persons having experienced unemployment as a per 1998 Percentage of people experiencing unemployment during the year. Of these people. . that is. Indeed. Although the rate of unemployment by dellnition pertains to a specific day. Statistisk Tiarsoversigt. The three latter ways are the most common. many more people than 12 per cent of the labour Ioree experienced some unemployment during 199 3. Some facts and introductory theory about unemployment © The McGrawHill Companies. you migh t think that 12 percent of the people in the labour Ioree were unemployed during all of 199 3. This was not the case. implying that only 4 per cent of the people in the labour Ioree were unemployed for more than half the year. Statistics Denmark.2 shows that in Denmark in 1993. Table 11. the fraction of people in the labour force who experienced some unemployment was 29 per cent. while 64 per cent were unemployed for less than half the year.2: Duration of unemployment 1993 Percentage of people experiencing unemployment during the year. 2005 11 SOME FACT S AND I NTRODU C TORY THEO RY ABO UT UNEM PLOYM ENT 317 representative sample of individuals whether they are available for work. were unemployed for more than half the year. and the fraction of people in the workforce who experienced some unemployment was 20 per cent.Stnctural Introducing Advanced Macroeconomics Unemployment 11. The duration of unemployment When you hear that the annual rate of unemployment was 12 per cent in Derunark in 1993. Table 11. but 21 per cent of these people were unemployed lor less th an live weeks. The annual rate of unemployment is simply the average over all the working days in the year of the daily rates of unemployment as defined above. 44 per cent were unemployed less than 14 weeks. 11.
and now only 20 per cent of those affected by unemployment were unemployed for more than half the year. This would correspond to a perfect matching in the labour market between people who want jobs and the employers who oller jobs. and if there are vacancies there should be no unemployment. Let J. one would perhaps say no.2 shows how the points are actually situated in the lS lor the period 1960. . Fig. where the help wanted index measures the number of job advertisements in major US newspapers. A close positive relationship between the overall rate of unemployment and the rate of longterm unemployment provides one of the main reasons lor trying to avoid high unemployment in general. say. When it comes to the individual costs of unemployment. However. 2005 PART 4: STRU CTU RAL U NEMPLO YMENT This illustrates a general fact that we will return to below. a greater share of total unemployment is longterm unemployment. 11. they should take the vacant jobs until there are no more vacancies. at least for some years. the matching is never that perfect. be total employment on that day.. This curve is called a Beveridge curve. If no people were ever unemployed for more than.. The vacancy index equals a 'help wanted' index divided by employment. and the vacancy rate is v . Just as one can deHne a rate of unemployment at a given time. When overall unemployment is high. However. This involves even more practical problems than measuring the number of unemployed people. v . along the horizontal axis and v 1 along the vertical axis. but if the measurement error is the same at different dates. one can deHne a rate of vacancies. but one can try in dillerent ways. a time series for a vacancy rate or index constructed this way will give a correct indication of the direction of changes in labour market conditions. In Denmark a vacancy rate can be computed. The further out to the northeast the curve is situated. If there are unemployed people truly seeking jobs. U.19R8. Some facts and introductory theory about unemployment © The McGrawHill Companies. based on the number of vacant jobs reported by the employers to the public employment agency. The points for Denmark seem to be situated on a fairly stable downwardsloping curve.. it is longterm unemployment that is most severe. The fact that the curve does not just follow the axes is an indication of' mismatch' in the labour markets..E. only points (u .SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 318 I Part 4. In Denmark the annual rate of unemployment had come down to 7 per cent. as opposed to 36 per cent in 1993 when the overall unemployment rate was 12 per cent. can there be vacant jobs at the same time? At Hrst. The perfect matching between 'people wanting jobs' and 'jobs wanting people' described above would imply that in a diagram with u. The number of vacancies is then V.Stnctural Unemployment 11. VJJ. The Beveridge curve If there is (involuntary) unemployment. The Hgures lor 199R also conHrm this. The position of = . and let E. the more severe are the problems of mismatch . If there is unemployment there should be no vacancies. V.. be the total number of jobs ollered by employers on day t. those who are unemployed for long periods may be badly hurt by unemployment. In the USA a 'vacancy index' is computed as a proxy for the rate of vacancies. To obtain data for vacancy rates requires that we can measure the number of vacant jobs. one month during a year. the individual loss would not be severe.2000.) on one of the two axes would be observed. = ]. and in Denmark for the period 1974. None of these ways are perfect.
Working Paper. Look at the Beveridge curve lor the USA based on the vacancy index.16 1974 0.12 )( Q) 0.05 1961 4 2 6 8 10 12 Unemployment rate (%) Denm ark. Then there is a shift outwards. Some facts and © The McGrawHill introductory theory about unemployment Companies.06 0.2: Beveridge c urves Sources: OECO. 19741988. Annu al data 0.1970. but the curve has shifted considerably to the northeast. University of Copenhagen.04 0.00 1983 2 4 6 8 10 12 Unemployment rate (%) Figure 11. One would expect this to be part of the reason why there was more unemployment in the latter of these decades. 200 1. Between 19 60 and 19 70 or 19 71 the points are located on a nice dm>\111wardsloping curve just as lor Denmark.13 1979 0. If the position of the curve shifts over time.Jacobsen: I Part 4 Introducing Advanced Macroeconomics Stnctural Unemployment 11. Albaek and H.14 0.S 0. 19602000.02 0.09 rl 0.06 0. Since 1985 there seems to have been an almost continually ongoing improvement of the basic structures of the American labour market. 2005 11 SOME FACTS AND INTRODUCTORY THEORY ABO UT UNEMPLOYMEN T 319 USA.0 I SorensenWhitta.12 )( Q) '0 .07 0. Hansen.S (j c >"' 0. Department of Economics.08 (j c "' ~ (j 0. the Beveridge curve can thus be seen as an indicator of structural problems in the labour market.10 >. Bureau of labor Statistics (US). . Annual data 0.08 .10 > 0.1 1 '0 0. The points for 19 7 5 to 19 8 5 are again situated relatively nicely on a downwardsloping curve. This indicates that during the decade 19 75 to 1985 people and jobs had much greater problems llnding each other than during the decade 1960. and data material for K. This is exactly why the Beveridge curve is useful. 'The Rise in Danish Une11ployment: Reallocation of Mismatch?'. this indicates shifts in the matching ell1ciency in the labour market.0.
.rth rate of GDP of two percentage points reduces unemployment by about one percentage point. namely the number of people who do not have a job but would have liked to have one. The figure uses the same data as Fig..f~.3 . of course.~~~!. The rate of unemployment is a fraction and by definition it has to stay between 0 and 1..... 11.. increases and decreases seem to ofJset each other so that there is no upward or downward longrun tendency..1 again and asswne th at lor th e US you only .. Over time spans of 100 years....1 and shows the annual absolute change in the rate of unemployment lor the USA and Denmark. Some facts and introductory theory about unemployment © The McGrawHill Companies.~~~.... Figure 11.o/..~.... since they indeed took jobs at these wages when they had the chance... The relationship reported in Fig. do not observe that real wages increase sharply as soon as GDP goes up. then thi~ suggests that the offlcially reported rates of unemployment indeed contain true information. if we observe a clear relationship between increases in GDP and decreases in the rate of unemployment and..4 (as well as Fig. Therefore it cannot increase by a certain percentage each year over many years.... for example it could tend to increase from one level to another higher level.. There is a clear negative relationship between the rate of growth in GDP and the absolute change in the rate of unemploument. and in particular there is no strong tendency that increases in GDP are associated with increases in real wages. a one percentage point reduction in unemployment seems to require only an additional one percentage point of economic growth. If this reduces unemployment without driving up rea l wages too much. we may start by asking if there is a way to test if those who report that they are involuntarily out of work would actually take a job at the going wages if they were ollered one? Assume th at due to 'good times' there is a sudden increase in the demand for labour. have an increasing trend in other ways. It shows.. however. For Denmark..Y. at the same time. it can be taken as a sign that those w ho reported being unemployed actually did want jobs at the going wages...3 is both important and useful. for instance.. Stylized fact I (Okun's Law).. In other words..... that at a constant rate of unemployment. economic growth in the US is around 3 per cent per year. We mentioned that the rate of unemployment contains an element that is not easy to measure.... Consider Pig... that over quite substantial periods there can be a tendency for unemployment to rise or fall...~?...Stnctural Unemployment 11. 11... At t he same time there is a clear tendency that increases in GDP are accompanied by decreases in unemployment. 11 . 11. In Chapter 14 we return to the empirical facts of the business cycle.SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 320 I Part 4.....~~~..~~..... It could..~.. One fact reported is that real wages are not strongly correlated with GDP over the cycle...~?. For this reason there is some disagreement among economists about how much to trust statistics on unemployment rates.. since this would eventually mal<e it larger than 1.. 11..3 has a name: Okun'sLaw. The relationship illustrated by Fig.... or should one only include series of 'truly measurable' variables? Yet most economL~ts believe that one can safely trust reported statistics of unemployment to give at least some indication of the extent of involuntary unemployment. 11.. Should these be included among the time series that theories are held up against.... To see w hy.3 for USA and Denmark...1) indicates that nothing like that is true. Stylized fact 2 There is no up1vards or downwards tre11d in the rate of unemploument in the (veru) long run.~~~P.. as shown in Fig..~.. Note. while a onetime increase in t he grov. 2005 PART 4: STRU CTU RAL U NEMPLO YMENT 11.~~~0..
Jacobsen: Introducing Advanced I Part 4. Economic Outlook. had the data from the early 1940s to the early 1980s. Figures 11 .!::: 5 Q) 0> c "' ~ 0 3 Q) 0> c"' Q) 0 cf 1 3 3 2 1 0 2 3 2 3 Absolute change in the unemployment rate Denmark Q. Q.1 and 11. Something similar is true for Denmark lor the period from the early 1970s to the early 1990s. 7 4 and t = . then it . For t he statistically proficient.4 also bring evidence of another fact which is well known: Stylized fact 3 There is a Jot of variability in the rate of unemployment in the short rz111.15.45 and t= 5. t he regression line lor t he USA has R 2 = 0. Source : OECD. 7 0 ~ . a period of about 40 years.Stnctural 11. while the line l or Denmark has R 2 =0. Some facts and © The McGrawHill Unemployment introductory theory about Companies. Each dot corresponds to a single year.10. 2005 Macroeconomics unemployment 11 SOME FACTS AND INTRODU C TORY THEORY ABO UT UNEMPLOYMEN T 321 USA Q.1 and 11.0 I SorensenWhitta.!::: 5 Q) 0> c "' ~ 0 3 Q) 0> c"' Q) 0 Q.4 is the persistence in the rate of unemployment: ifthe rate of unemployment is high in a specific year. For this long period there seems to be a sustained upward movement in the rate of unemployment.1 3 3 2 1 0 Absolute change in the unemployment rate Figure 11. 7 0 ~ . . A fact that is perhaps not so directly visible from Figs 11.47.3: Okun's Law The scatterplots show the annual percentage change in GOP (1995 PPP US dollars) against the annual absolute change in the rat e of unemployment in the period from 1965 to 2002.
1' ( 1= 1 u 1  x) )2 .....bi\IIIHH 11 1. 1' u . One can show that . 1= 1 ( x1 . The coefficient of correlation between two time series. and x = } 1 x is: I i= p = v. . u 1 = a + 1Jx 1..1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Year Denmark E 4 +r~~11~ ·o ~ 2 0> "' ~ 0 .6 +.4: An nual absolute change in the rate of unemployment Source: As Fig..1 . was probably also high the year before (and will be the year after).v.:.. Ifu.. ....) x 7. respectively.~~LH~~~~L .. p...1 correspond to complete linear covariation along a straight line.. running through t = 1. T and with averages IT = } I {= 1 u.{= 1 (u...SorensenWhittaJacobsen: Introducing Advanced Macroeconomics 322 I Part 4....:.4 +~HH ..1 0 ~ 2 +1~~~~..1 . I.Stnctural Unemployment 11. that is. u 1 and x 1. and p = 1 and p = .1...  .. We can measure this persistence by the coefflcient of correlation between the current unemployment rate and its 0\>Vll past values..... JH'. Some facts and introductory theory about unemployment © The McGrawHill Companies... 1. 11 . roughly the degree to which one series is above average when the other is above average... It should be visible from the formula that the coefllcient p measures the degree of (linear) covariation between the two series. 2005 PART 4: STRUCTURAL U NEMPLOYMENT USA 10 8 +~ . with positive and negative b..6 +.u)(x.. is the annual rate of unemployment and one lets . 1.1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Year Figure 11.
0
I SorensenWhittaJacobsen: I Part 4 Stnctural
Introducing Advanced
Macroeconomics
11. Some facts and
introductory theory about
unemployment
Unemployment
© The McGrawHill
Companies. 2005
11 SOME FACTS AND INTRODU CTORY THEORY ABO UT UNEMPLOYMEN T
323
x , = u, _ 1 , and computes p lor the two series u, and x, (lor t = 2 ..... Tl. one gets the coefllcient of correlation between the rate of unemployment in one year and the year before.
One can then let x, = u 1_ 2 to get the correlation with unemployment two years before, etc.
Figure 11.5 shows such coefl'icients of correlation in unemployment for the USA and
Denmark.
There is a clear positive correlation between the current rate of unemployment and
the rates of unemployment in several previous years, and the degree of correlation
decreases as one goes back in time. The tendency for high unemployment in one year to
imply high unemployment in several previous and succeeding years is what we mean by
the persistence in unemployment.
USA, 19002002
1.00
1.00
0.88
0.80
c
1

1
1

f
1
1




1

0.69
......
0
· ; 0.60
~
0() 0.40
~ 0.20
'(}
~0
0
000
.
 0.20
 0.40
r
0.38

0 .08
O~OB_
Trtt
r
r
1

0 .1 9

0.27
r






1




1

1

1.00
Denmark, 19032002
1.00
0.86
c
0
r""
0.80
O.GG
~
0.62 ..0.55r r 1
~
0 0 .60
()
0
c
Q)
0.40
=
'(}
0.39
0.27 U~;j
0

1


1
1

0.46


1
f
1

1
1
1

1






,..,
Q)
0
0 .69 U_2:l
0.20



0.00
Figure 11.5: Correlation in unemployment over time
The columns illustrat e the coefficient of correlation between the unemployment rate at time t and time I k. Thus,
the column u1_ 3 shows the degree of correlation between t he unemployment rate and the unemployment rate three
years before.
Source: As Fig. 11 .1.
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
324
I Part 4 Stnctural
11. Some facts and
introductory theory about
unemployment
Unemployment
© The McGrawHill
Companies. 2005
PART 4: STRUCTURAL U NEMPLOYMENT
Stylized fact 4
There is a lot ofpersistence ill mmual rates of rmemployment.
In connection with Table 11.2 we examined the relationship between overall unemployment and longterm unemployment. We can compute the annual rate oflongterm
unemployment over several years, defining this rate as. lor example. the percentage of all
people in the labour force who were unemployed for more than half of the year. We can
then plot the rate of longterm unemployment against the overall rate of unemployment.
This is done for the US and lor Denmark in Fig. 11.6.
What is remarkable is not just that the relationship is positive, but also the strength of
the positive correlation. As unemployment increases, longterm unemployment increases
USA, 19702000. Annual data
7
(I)
"' 0
(/) 0
.,_
6
>.0
5
g::;
o..O
<11
a.
E
<I>
1983
4
<1>£
c 0
:::J
E
3
<I>
~ Cl
(I)"'
~c
Cl(l)
2
c 0
0
....J
~
(I)
Q.
0
3
5
4
6
7
8
9
10
Unemployment rate
Denmark, 19791998. Annual data
8
(I)
"' ~
.,_
(/) 0
"0
~
~5

7
6
<11 5
a.o..O
E
<I>
<1>£
c 
:::J
0
~
<I>
Cl
E
(I) " '
~c
Cl(l)
c
0
....J
0
4
3
1979
2
~
(I)
Q.
0
3
5
7
9
11
13
Unemployment rate
Figure 11.6: Long·term unemployment against overall unemployment
These figures plot the percentage of people in the labour force being unemployed for more than half the year,
against the overall unemployment rat e. Each dot corresponds t o a single year.
Sources: Bureau of Labor Statistics (US); Statistisk Tiarsoversigt, S tatistics Denmark.
0
I SorensenWhitta Jacobsen: I Part 4 Stnctural
Introducing Advanced
Macroeconomics
Unemployment
© The McGrawHill
11. Some facts and
introductory theory about
unemployment
Companies. 2005
11 SOME FACTS AND INTRODU C TORY T HEORY ABO UT UNEMPLOYMEN T
325
relatively more. From the figure it seems that lor both countries a 100 per cent increase in
the rate of unemployment implies an approximately 200 per cent increase in longterm
unemployment (perhaps somewhat less for the US), pointing to an elasticity oflong term
unemployment (as defined here) with respect to overall unemployment of around 2 .
Stylized fact 5
Longterm u11employment varies positively and more tlw11 proportio11ally with overall
llllemployment.
As noted above the most severe consequences of unemployment are felt by the
longterm unemployed. Since higher unemployment means even higher longterm
unemployment. this may be one of the main reasons lor fighting high unemployment.
Does unemployment tend to be of equal size in different regions of the world? Perhaps
it does in the very long run. as suggested by Fig. 11.1, but over substantial periods there
can be considerable dillerences between major regions, as shown in Fig. 11.7. For a long
time until the early 1980s, unemployment in Europe was lower than in America. but
since then joblessness has been substantially higher in the EU than in the US.
Stylized fact 6
There can be large differellces in unemployment across geographical arens for long periods of
time.
Other systematic variations are also important. Fig. 11.8 shows rates of unemployment across educational groups for the US and Denmark. Although dillerent educational
categorizations have been used for the 1:\ovo countries, the general picture is clear:
Stylized fact 7
There are consideralJle and lo11glived differellces in rates of unemployment across educational
groups with a broad te11dency for !rig her education to mean lower unemployment.
Variations across other categories are of importance in dillerent com1ections, e.g.
across race or sex. However, a relatively high rate of unemployment lor a particular
12
10
8
% 6
4
2
0
1970
1980
1990
Year
Figure 11.7: Rates of unemployment in the EU and USA
Sources: Statistics Denmark and O ECD, Main Economic Indicators.
2000
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
326
11. Some facts and
I Part 4 Stnctural
Unemployment
© The McGrawHill
Companies. 2005
introductory theory about
unemployment
PART 4: STRU C TU RAL U NEMPLOYMENT
USA
14.12  
:
.._________Less than high school
1 0 l~=
%
8+~~~~
/~
2
~
1!!gh school graduates
, ____________________________, , , ~~~,~·
+__========C=o=lle~g~e~g~r~ad~u:a:te:s~=~~~==~~~~~~
O+...,,.~
1991
1993
1995
1997
1999
Year
2001
2003
Denmark
16.No vocational training
___ _ _ _ _ _ _ __
14 1_/
~='\"
~
12
"
1 0+~~~~~"~~~~,,,, ~
',,,
~
% 8 ~·~ Vocational ed ucation _ , _____

6t~~~~~======~............. ___________ .
4 +~~~=~"~~ikdium·cycle higher ed uc:rli
oo;;n2
o +......~
1989
1991
1993
1997
1995
1999
2001
2003
Year
Figure 11.8: Unemployment by educational attainment
Sources: Bureau cf Labor Statistics (US), Statistisk Tiarsoversigt, Statistics Denmark.
section of the population may, at least partly, reflect that this group contains relatively
many unskilled people compared to the general population.
The Hnal 'law' we will focus on here is an important one, and the remainder of
this chapter is more or less centred around it. Figure 11.9 shows annual rates of unemployment for four countries for the period from 1965 to 2000. For the USA and
Denmark. Fig. 11.9 is thus just a closeup of the most recent part of Fig. 11.1.
We see that (except for a special period in the 1960s and early 1970s). even when the
rate of unemployment is lowest, there is still a considerable amount of unemployment
with annual rates above 4 per cent. Furthermore, if the rate of unemployment fluctuates
around a certain 'gravity level' , as Fig. 11.9 and in particular Fig. 11.1 could indicate,
this level seems to be somewhere between 5 per cent and 7 per cent in the countries
considered. The period in the 19 60s and early 19 70s, where unemployment rates in some
of the countries went all the way down to below 2 per cent. showed many signs that this
0
I SorensenWhitta Jacobsen: I Part 4 Stnctural
Introducing Advanced
Macroeconomics
Unemployment
© The McGrawHill
11. Some facts and
introductory theory about
unemployment
Companies. 2005
11 SOME FACTS AND INTRODU C TORY THEORY ABO UT UNEMPLOYMEN T
USA

Great Britain
Denmark
327
....... Netherlands
12
10
%
8
6
4
2
1975
1985
1995
Year
Figure 11.9: Rates of unemployment in Great Britain, Denmark, the Netherlands and the USA
Source: OECO, Economic Outlook.
was not a sustainable or equilibrium situation: inflation rose sharply in the countries in
question. Accounting for this, the figures indicate that even in the best of times the lowest
possible rates of unemployment are considerable, perhaps around 4 per cent, and gravity
rates of unemployment are around 57 per cent.
Figure 11.1 indicates a strong similarity in the (very) longrun behaviour of the
rate of unemployment between the USA and Denmark. In both countries the rate of
unemployment seems to fluctuate around a common 'natural' level of. say, 5 to 7 per
cent. However, there may also be some indication, particularly from Fig. 11. 9, that the
natural rate of unemployment can shift over time. and Fig. 11.7 indicates that the natural
rate can diller between countries or regions.
Stylized fact 8
When rates of unemployment are at their lowest. there i.> still a substantial amount of unemployment. seemingly around 4 per cent. and tile natural ttnemployment rate tl!at tile a11m1al
rates .fluctuate around is higher. around 57 per cent. Tile natural rate ofunemployment seems
to shift over time and cmz be dij)immt in different regio11s.
Shortrun
cyclical and longrun structural unemployment
................................................................................................................................................................................
This section links the various types of unemployment. shortrun cyclical and longrun
structural, to the diflerent types of wage and price rigidity that economists think are
associated with them. The section will therefore contain some repetition of material from
Chapter 1.
Unemployment and excess supply of labour
Figure 11.10 illustrates a situation of unemployment in a labour market (disregarding
mismatching). The llgure assumes that the individual suppliers oflabour take the nominal
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
328
I Part 4 
Stnctural
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
PART 4: S TRU C TU RAL U NEMPLOYMENT
L
Figure 11.10: A labour market
wage rate. W, as well as the nominal price level. P, as given. The labour supply curve. L ' . is
drawn as an increasing function of the real wage rate. W /P. but nothing in our arguments
would change if the labour supply curve were vertical. The ftrms that demand labour are
also assumed to take the nominal wage rate as given. and the demand curve for labour. L'1,
is decreasing in the real wage rate.
If this labour market were perfectly competitive. the market forces (working on the
nominal wage rate, W) would imply a real wage at the level w*, where supply equals
demand and there is no unemployment. A situation of unemployment, as illustrated in
the figure. means that the real wage rate is at some level '~' at wh ich the supply oflabour
exceeds the demand for labour by the amount U.
Should the excess supply of labour not imply downward pressure on the nominal wage
rate, forcing the real wage down to w*, thus eliminating unemployment? This is a key
question that any theory of unemployment has to face. Let us first just state two trivial. but
nevertheless important. and highly related points in connection to this question:
• In a lafJour market characterized by full competitive wage.flexibility. there cnn be 110 unemployment, since the nominal wage rate adjusts itnmediately to generate a real wage that equates the
competitive (wagetaking) supply of and demmzd.for labour.
• For unemployment to be possible, there must be factors preventing the (real) wage from
adjusting to a level that would equate the competitive supply of labour and the competitive
demmzd for labour.
Wage rigidity in the short run and in the long run
If we define 'wage rigidity' as any factor that prevents a full, competitive equilibration
of the labour market, we can also summarize the two above statements like this: for
0
I SorensenWhitta Jacobsen: I Part 4 Stnctural
Introducing Advanced
Macroeconomics
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
11 SOME FACTS AND INTRODU C TORY THEORY ABO UT UNEMPLOYMEN T
329
lll!employment to exist there must be some wage rigidity. Usually economists distinguish
between two types of wage rigidity and two associated types of unemployment.
Shortrun JWminal wage rigidity a11d cyclical u11employme11t
If a situation like the one illustrated in Fig. 11.10 prevails (for some time) lor the mere
reason that it takes time for the nominal wage rate to adjust to circumstances, we say
that a shortrun nominal wage rigidity is at work. The part of unemployment caused by
shortrun nominal wage rigidity we call shortrun, or cyclical. unemployment
Shortrun nominal wage rigidity can take different forms. One possibility is that the
adjustments of the nominal wage rate are really governed by competitive market forces,
that is. by discrepancies between wagetal<ing supply and demand. but it simply takes
some time for an excess supply oflabour to bring about a full downward adjustment of the
nominal wagerate. The reason is that an excess supply of labour has to be there for some
time before it can invoke a downward pressure on the nominal wage rate. There will
therefore be a period where the real wage rate can be at a level like 1\1 in Fig. 11.10 simply
because the nominal wage adjustments that should bring it down have not yet occurred.
We would expect cyclical unemployment caused by such a nominal wage rigidity to be
eliminated by market forces alone in the longer run when nominal wages have had
time to adjust fully. However, if the economy is hit by new shocks all the time. the
nominal wage rate may be on a neverending job of adjustment, and during this continual
adjustment, there will occasionally be unemployment and occasionally excess demand for
labour or 'overemployment'.
Another possibility that we introduced in Chapter 1 and will return to in Chapter 19
is that the nominal wage rate is really set by an economic agent, let us say a trade union
that organizes the wagetaking individual workers. As we explained in Chapter 1, if
the nominal wage rate is set optimally at the outset and there are small 'menu costs'
associated with nominal wage adjustments, it will be optimal for the trade union not to
react to shocks as long as these are not too large and of a temporary nature.
The real wage, W/P, may be flexible under shortnm nominal wage rigidity because
the nominal price level, P, may adjust in response to conditions in the product markets. The
implied adjustments in the real wage should not be expected to eliminate unemployment
since they do not come in response to circumstances in the labour market.
The dynamics of cyclical unemployment arising from shortrun nominal wage
rigidity is one of the main subjects of Book Two and will not be dealt with further here.
Lo11gnm real wage rigidity and structural unemployment
Consider again the situation in Fig. 11.10 where the wagetaking labour supply exceeds
employment. and assume that the real wage rate has already fully adjusted in accordance
with the real wage .flexibility that the economy possesses. In that case we say that a longrun
real wage rigidity is at work, and the part of unemployment that is caused by longrun real
wage rigidity we call longrun. or structural, or natural unemployment.
Note that this is diflerent from shortrun nominal wage rigidity. It is not that the
nominal wage rate has not yet adjusted to circumstances or that wage setting agents have
preferred not to react to temporary and opposing shocks. Here we assume (artificially) that
the economy has not been exposed to new shocks for a long time so that all relative prices
have fully adjusted to circumstances in accordance with the economy's longrun price
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
330
I Part 4 Stnctural
Unemployment
© The McGrawHill
11. Some facts and
introductory theory about
unemployment
Companies. 2005
PART 4: STRU C TU RAL U NEMPLOYMENT
and wage flexibility, and yet excess supply of labour prevails. We should, of course,
be interested in the exact reasons why an excess supply of labour would not imply a
downward adjustment of the real wage rate in the long run. We brielly consider such
reasons below, while Chapters 12 and 13 deal in depth with two particular and important
reasons lor real wage rigidity and their macroeconomic implications.
The division of all unemployment into shortrun and longrun unemployment is
closely related to the Stylized fact 8 above, and to the nature of the unemployment fluctuations illustrated in Figs 11.1 and 11.9. From time to time there is a sudden increase in
the rate of unemployment. for instance by two percentage points over just one year. It is
natural to see this as caused by a negative shock to which the economy reacts not only with
lower production, but also (temporarily) with higher unemployment because of shortrun
nominal wage rigidity. This can explain the tluctuations in the rate of unemployment, but
not the trend level of S7 per cent that the unemployment rate lluctuates around. Over
time spans such as the 3 5 years in Fig. 11.9. or the 100 years in Fig. 11.1. we catmot
possibly view the economy as being all the time reacting to negative shocks. Sometimes
the shocks must have been mainly positive. sometimes the labour market must have been
ALL UNEMPLOYMENT
Shortrun or cyclical
unemployment
Longrun or structural
or natural unemployment
REASON:
REASON:
Shortrun nominal wage rigidity
Longrun real wage rigidity
SUBDIVISION by cause of rigidity:
•
FRICTIONAL unemployment
(involuntary)
•
SEARCH unemployment
•
Unemployment due to
{voluntary)
EFFICIENCY WAGES
(involuntary)
•
POLICY:
STABILIZATION POLICIES,
e.g. fiscal and monetary policy
Unemployment due to
TRADE UNIONS (involuntary/
voluntary)
POLICY:
STRUCTURAL (LABOUR MARKET)
POLICIES,
e.g. reforms of the unemployment
benefit system, the educational
system, or the tax system.
Figure 11.11: Different types of unemployment
0
I SorensenWhitta Jacobsen: I Part 4 Stnctural
Introducing Advanced
Macroeconomics
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
11 SOME FACT S AND I NTRODU C TORY THEO RY ABO UT UNEM PLOYM ENT
331
close to equilibrium where wage rates are adjusted in accordance with the economy's wage
llexibility. The permanent part of unemployment must therefore be due to longrun real
wage rigidity, i.e .. to the fact that the price and wage llexibility of the economy is simply
not such that it tends to eliminate all unemployment, not even after a long time without
new shod<s. Fig. 11.11 illustrates the categorization of unemployment that we are
explain ing.
One of the main reasons why economists lind it useful to distinguish between the
lluctuating or cyclical part of unemployment associated with shortnm nominal wage
rigidity, and the permanent or 'natural' part of unemployment associated with longrun
real wage rigidity is that the kind of policies that can be expected to work in fighting
the two types of unemployment are dill'erent. Longrun unemployment can only be
affected by policy measures that have an impact on the price and wage llexibility of the
economy. We call such policies structural policies. One example could be reforms of the
unemployment insurance system designed to make the unemployed exert a stronger
downward pressure on wage rates, but as we saw in Chapter 1, a policy which increases
the degree of product market competition can also help to reduce structural unemployment. Shortrun unemployment. on the other hand. can be affected by stabilization
policies, such as traditional fiscal and monetary policy.
Causes of longrun structural unemployment
For longrun real wage rigidity and structural unemployment to exist, there must be
some reason why unemployment does not impose a downward pressure on wages. The
remaining part of this chapter, and Chapters 12 and 13. are focused on this issue.
Why should the presence of unemployed people not give rise to downward pressure
on wage rates, not even in the long run? Let us go through some possible reasons.
Abuse
One potential reason is that some people registered as unemployed may not really be
looking lor jobs but may only be interested in collecting the publicly subsidized unemployment benellts. Hence they do not contribute any downward pressure on wages. There is
hardly any doubt that such abuse occurs to some extent, but we have argued above that
probably most of registered unemployment is not of this type. and in the following we will
disregard it.
Frictionaltmemployrnent
Another reason is that some of the unemployed have already found a job, but have not yet
started in it. Usually people are not hired immediately, but, say, from the first day of the
next month. There will then be a period of time during which they are unemployed. and
involuntarily so, but they do not look for jobs since they already have a job starting a lew
weeks later. Due to the natural development of the economy there will all the time be some
people who lose their jobs. For instance, as time goes by the products of some llrms become
outdated, the firms must close. and the employees lose their jobs. Even at times when these
people do not have dilllculty llnding new jobs, there will be some unemployment due to a
number of people being on their way from one job to another, since the transition takes
some time for technical and formal reasons beyond people's control. We call this frictional
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
332
I Part 4 Stnctural
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
PART 4: STRU CTU RAL U NEMPLO YMENT
unemployment, and frictional unemployment is one part of structural unemployment. It
is of interest to try to estimate how much structural unemployment is frictional, and
we will return to that in the next section. The conclusion will be that probably only a
minor part of longrun unemployment is frictional. Hence there is still a lot of structural
unemployment to explain, and indeed some view the nonfrictional part of longrun
unemployment as the real structural unemployment.
Search unemployment
If the people who lose their jobs spend some time before taking a new job. not for reasons
beyond their control, but as a matter of choice (they may turn down job oilers for some
time). we call the associated unemployment search unemployment. Just like frictional
unemployment, search unemployment cannot be expected to exert downward pressure
on wages. The two types of unemployment are closely related, diilering only in the reason
why the unemployed turn down job offers. In the case of frictional unemployment it is
because they have already accepted a job oiler. In the case of search unemployment it is
because the unemployed choose to wait, or search, for better job oilers. Who says th at the
11rst oiler received is the best one the unemployed worker can get? Perhaps it is wise to
spend some time searching. Search unemployment is thus closely related to uncertainty
about available job oilers: it is caused by imperfect information about job possibilities.
Furthermore. search unemployment is voluntary.
The level of search unemployment depends on a basic tradeoff for the unemployed
person. On the cost side, longer search for a new job implies a larger income loss during
the waiting time. On the benellt side. it implies a higher expected future income from the
day employment starts, assuming that more search ultimately leads to a better job ofi'er.
The more important the benent side is relative to the cost side, the more (voluntary) search
unemployment there will be. Generous unemployment benellts help to keep the cost side
down; they work as a subsidy to search. Therefore. lower benellts would tend to reduce
search unemployment. and this is one example of a structural policy th at could be used to
light a type of structural unemployment.
Sh ould policy makers try to reduce search unemployment as much as possible? No.
there are some positive aspects of search unemployment. It is in the general interest of
society that the qualillcations of the labour Ioree are used optimally, that each individual
is employed where his or her productivity is the highest, since this means higher incomes.
The search by individuals may work to ensure that people are placed in jobs where they
earn well because they have high productivity. So. lor society there is also a basic tradeoil
related to search unemployment. When people are 'separated' from their jobs, we want
them to search lor some time in order for them to lind new jobs where they have high
productivity. but we do not want them to search for too long. because then the output loss
during the search period will outweigh the productivity gain from further search.
The job for policy makers is thus not to reduce search unemployment to a minimum,
but to create incentives through the tax and benellt systems lor an appropriate level ofsearch
unemploynw1t.
Search unemployment and structural policies intended to atl'ect it are important L~sues
in macroeconomics. However, the formal theory of search unemployment is quite complex,
and lor that reason we will not have (much) more to say about search unemployment in this
book.
eI
Sorensen Whitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 4 Stnctural
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
11 SOME FACT S AND I NTRODU C TORY THEO RY ABO UT UNEM PLOYM ENT
333
Ejficiency wages
The efficiency wage theory of the coexistence of unemployment and fullyadjusted wages
represents a minimum departure from the fully competitive model of wage adjustment.
Since it is the subject of Chapter 12, we will be brief about it here. Just note that there
is a hidden assumption behind a supply and demand diagram like Fig. 11.10, or more
precisely behind the labour demand curve in it. The labour demand curve is given by the
decreasing marginal product of labour, since labour is demanded up to the point where
the marginal cost of an additional unit of labour  the real wage  equals the marginal
revenue product. However, behind this construction is the tacit assumption that the wage
paid to the workers does not allect their productivity. In Chapter 12 we argue that if we
keep all other characteristics of the competitive labour market, but add the feature that
productivity may depend positively on real wages, then the traditional labour demand
curve 'disappears', and 11rms may choose a combination of a real wage and a number of
people demanded inside the labour supply curve. The result may well be unemployment
without any tendency for falling wages. Wages fail to adjust because firms do not want
lower wages, and unemployment is involuntary since individual unemployed workers
would really like to work at the going wage rate.
Trade unions
Unionization of the labour market represents a more 'brutal', but realistic and therefore
important, departure from the competitive model. It is simply assumed that the supply side
of the labour market (and sometimes also the demand side) is not atomistic, since the
workers are organized in a trade union that takes care of their interests. The trade union
sets the wage rate and the labour market therefore does not work competitively, but more
like a monopoly market. As we shall see in Chapter 13, this may well imply that the wage
rate chosen by the trade union will be so high that the wagetaking labour supply of all the
workers on the labour market exceeds the labour demand of the firms. Hence there is
unemployment, but there is no tendency lor falling wages, since the wage rate is what
th ose who set it want it to be. The unemployment will be involuntary at the individual
level, since the unemployed workers would really like to work at the wage rate set by the
union. At the collective level, however, the unemployment bas something voluntary
about it, since it is due to the fact that workers h ave decided to organize to exert influence
on the wage rate a n d, h aving obtained market power through a union, let this union
choose a wage rate that is so high that unemployment results.
In this book we focus on ell'iciency wage setting and trade union behaviour to explain
why real wages do not adjust to eliminate lon grun unemployment. To motivate this
focus, we will now argue that a large part of longrun unemployment cannot be explained
by frictional or search unemployment.
11.5 ?.?.~. ~.?~.~~. ~.~. f.~~~~.~S>.J?:~.l. ~.~.~. ~~.~,<;,~ ..~.~.I?.P~?.Y.J?.~J?:~? ............................ .
Let us now return to the simple formula that relates the number of unemployed workers
to the labour force and the number of employed workers, say in a speciHc month ,
U1 = L1  E1• For simplicity assume that the labour force is constant, L1 = Lin all months t,
so that the formula is U, = L E1• How can the number of unemployed workers possibly
change from month t to month t + 1? In exactly two ways.
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
334
I Part 4 Stnctural
Unemployment
© The McGrawHill
11. Some facts and
introductory theory about
unemployment
Companies. 2005
PART 4: STRU CTU RAL U NEMPLO YMENT
Er
Employed
Employed
separated from
jobs, s1E1
Unemployed
finding
jobs, f1U1
Ur
Unemployed
Figure 11.12: The flows in and out of unemployment
Some people who are employed in month t may lose or quit their jobs from t tot+ 1,
causing an increase in unemployment. We denote by st the 'separation rate' of month t,
that is, the fraction of the employed in period t w ho lose their jobs from t tot+ 1. Then the
increase in the number of unemployed coming from job separation can be written ass 1 Et.
Some people who are unemployed in month t may find a job from t to t + 1, which gives
a reduction in unemployment. If we denote the 'job finding rate' of month t by ft. this
reduction in unemployment is.f1 U, . The flow diagram in Fig. 11.12 shows the two ways
the number of unemployed workers can change.
The total change in the number of unemployed workers from month t tot+ 1 is thus:
ut+ I

ul =stEt ft u,.
If we divide both sides by Land remember that in any month the rate of unemployment is
u, = U,/L. we get:
u,+1 
lit =
E
l
f ' u,.
st L
Now useE, = I  U1 to get:
ut+1 = (1  s,;;)ut + s,.
Given the evolution of s, and f, and an initial rate of unemployment, this equation
describes how the rate of unemployment evolves over time. This law h as nothing in
particular to do with frictional or search unemployment. It is simply an identity th at has
to hold for unemployment in general (given a constant labour Ioree).
We will now locus on a 'stationary environment' where both s1 and.f, are constant
over time. that is, st = s andf1 =fin all months, t. You may lind this assumption strange.
Shouldn't the possibility of finding a job, the job finding rate f t, depend on how much
unemployment there is? We make this assumption deliberately since we want to focus
entirely on frictional and search unemployment. We do not want it to be particularly
difficult to find a job. We want to lind out how much unemployment there will be if the
only reasons for unemployment are that when people lose their job for some reason. they
do not start in new jobs immediately, either for purely technical/practical/lormal reasons,
or because they decide to search for an attractive job offer. The dynamics of the rate of
unemployment now simply become:
u,+ 1 = (1  s .f)u 1 + s.
(1)
eI
SorensenWhittaJacobsen:
Introducing Advanced
Macroeconomics
I Part 4 Stnctural
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
11 SOME FACTS AND INTRODUCTORY THEORY ABOU T UNEM PLOYMEN T
335
Here we should think of s as a small positive fraction. perhaps between 0.01 and 0 .02,
corresponding to 1 or 2 per cent of the employed workers losing their jobs each month. 4 If
everyone who loses a job starts in a new one 'the first day of the next month', the
job finding rate would be 1, but since voluntary job search will pull.f down. we should
probably think off as a traction that is somewhat smaller than 1, e.g.f = 0.8 . This will imply
that the parenthesis (1  s}) in trout of u, in (1) is a positive number smaller than one, so
the dynamics given by (1) are such that u, converges monotonically to a constant level u
(draw a diagram with u, + 1 as a function of u, and iterate) given by u = ( 1  s j)u + s, or:
s
!1 =   .
(2)
S+j'
One can compute the 11 given by (2) lor various values of s and .f From the above, a
value lor s in the interval 0 .010.02 could be realistic. If the value off is to be explained
only by frictions or search. a value of0.8 seems small. A larges and a smallftend to make
the u in (2) large. so inserting s = 0.02 andf= 0.8 should give an indication of a large rate
of trictional and search unemployment. With these parameters the computed u is around
2.4 percent. Even if we send fa ll the way down to 0 .5. which is surely low, the computed
u will still be lower than 1 per cent (3.8 per cent to be precise).
We conclude that there seems to be important elements of structural unemployment
that cannot be explained by frictions or search. There seems, therefore. to be good reasons
to investigate other explanations of structural unemplorment.
11.6 §.~l!l.l!l.~~Y................................................................................................................................................... .
1. Unemployment implies individual costs in the form of income losses that cannot be fully insured
as well as psychological stress. Unemployment implies a social cost as well. For an involuntarily unemployed worker the real value of the goods and services needed to compensate for
the sacrifice of leisure (MRS) is smaller than the going real wage. When firms maximize profits,
the real wage equals or is smaller than the marginal product of labour (MPL) . Hence involuntary unemployment implies MPL >MRS. This means that a social gain could be achieved if
some unemployed workers got jobs, since the extra output generated by additional
employment would more than suffice to compensate the newly employed workers. Because
unemployment is a social waste, it is a main concern of economic policy makers.
2. The rate of unemployment is the fraction of unemployed in the total labour force, measured at
a given point in time. The annual rate of unemployment is the simple average of the rates of
unemployment in each day or month of the year. A large number of workers are affected by
unemployment in each year, typically around 20 per cent of the labour force in countries like
the USA and Denmark, but only some of the unemployed are out of work all of the year,
whereas the rest have unemployment spells of varying but shorter duration.
3. The longterm unemployed during a given year are those who were unemployed more than a
given fraction of the year, for example more than half of the year. Longterm unemployment
4. In their article 'The Cyclical Behavior ol the Gross Flows of US Workers', Brooking Papers on Economic Activity, 2,
1990, O livier B lanchard and Peter Diamond find a monthly job separation rate for the US over the period January
1968 to May 1986 of 2.9 per cent. In the period from 1968 to 1986 there was an increasing trend in the rate of
unemployment in the USA (see Fig. 11.9), so the separation rate under the stationary conditions we want to focus
on here is probably somewhat smaller.
SorensenWhitta Jacobsen:
Introducing Advanced
acroeconomics
336
I Part 4 Stnctural
Unemployment
11. Some facts and
introductory theory about
unemployment
© The McGrawHill
Companies. 2005
PART 4: STRUCTURAL UNEMPLOYMENT
is the part of overall unemployment carried by the longterrr unemployed. It is mainly the
longterm part of unemployment that causes serious individual costs, but for the social cost
the overall (average) unemployment rate is also important.
4. Unemployment and vacancies can coexist because the labour market's match between 'people
wanting jobs' and 'jobs wanting people' is not perfect. The Beveridge curve is a plot over time
of the vacancy rate against the unemployment rate and gives an indication of the degree of mismatch in the labour market. For given structural mismatch problems, the Beveridge curve should
be a stable decreasing relationship. If it shifts inwards or outwards, it can be an indication of
decreased or increased structural problems in the labour market, respectively.
5. The empirical evidence shows that (i) there is a clear negative relationship between the rate
of growth in GOP and the change in the rate of unemployment, (ii) there is no upward or
downward trend in the rate of unemployment in the very long run, (iii) there is a lot of variability
in the rate of unemployment in the short run, (iv) there is a lot of persistence in annual rates of
unemployment, and (v) longterm unemployment varies positively and more than proportionally
with overall unemployment.
6. When annual rates of unemployment are at their lowest, there is still a substantial amount
ot unemployment, seemingly around 4 per cent, and annual unemployment rates seem to
fluctuate around a natural unemployment rate of about 5 7 per cent. The natural rate of
unemployment seems to shift over time and can be different in different reg ions of the world.
7. The presence of unemployment reflects that wage rigidities prevent the real wage from
adjusting to equalize supplies and demands for labour. Economists find it useful to distinguish
between shortrun nominal wage rigidities, which are fundamental for explaining the cyclical
fluctuations in unemployment, and longrun real wage rigidities, which cause the positive and
constant trend level of structural unemployment. This part of the book is focused on the
natural rate of unemployment and longrun real wage rigidities.
8. This chapter briefly presented the following four causes of longrun real wage rigidity and
structural unemployment: (i) labour market frictions, (ii) job search, (iii) efficiency wages, and
(iv) market power of trade unions. The last two of these potential causes for structural
unemployment are dealt with in the following two chapters.
9. Studying a model of the flows in and out of unemployment, we derived a simple formula linking
the rate of unemployment to the separation rate (the fraction of employed who lose their jobs
in a month) and the job finding rate (the fraction of unemployed finding a job from one month
to the next). Assuming a stationary environment with constant separation and job finding
rates, and assuming realistic values for these rates, we found that only a minor part of structural unemployment seems to be due to labour market frictions or search. This gives good
reason to study the explanation for structural unemployment offered by the theories of
efficiency wages and trade unions presented in Chapters 12 and 13.
11.7
Exercises
Exercise 1. The definition of unemployment
Exactly how is the rate of unemployment defined and measured in your own country according
to the (most) official statistics? In particular, how is the number of unemployed counted? How
is the labour force measured?
eI
SorensenWhitta Jacobsen:
Introducing Advanced
Macroeconomics
I Part 4 
Stnctural
Unemployment
11. Some facts and
© The McGrawHill
introductory theory about
unemployment
Companies. 2005
11 SOME FACTS AND INTRODUCTORY THEORY ABOU T UNEM PLOYMEN T
337
Exercise 2. The social cost of unemployment
This exercise asks you to explain the social cost of unemployment more or less as it was done
in this chapter, but to do so around a standard diagram that you will see again in Chapter 20.
In the figure below, illustrating a particular labour market, the curve MPL depicts the
marginal product of labour as this depends on the total number of workers employed. The curve
is decreasing because of the presence of fixed factors or because of productivity differences
among workers or both. For each worker there is a certain marginal rate of substitution of
output for leisure, MRS, the amount of goods it takes to compensate the worker for the loss of
leisure and home production resu lting from having a j ob. The MRS curve in the figure shows
how this marginal rate of substitution varies with employment when workers are lined up in
order of increasing MRS.
1. Explain that if the individual workers are wage and price takers then the MRS curve is identical
to the labour supply curve, such that if the going real wage rate is the
w indicated, total labour
supply should be the L' indicated, that is, the labour supply coming from workers with an MRS
below w.
2. Explain that if the firms that buy labour in the market considered are wage and price takers
then their labour demand curve is given directly by the MPL curve, such that the labour
demand arising from a wage rate of w is the L" indicated in the figure. Now assume that the
firms take the nominal wage rate, W, as g iven, but, due to some kind of imperfect competition
in the output market, profit maximization leads to the nominal price, P, being set as a markup
over nominal marginal cost, that is, P = m(W/MPL), where m > 1. Explain that to be in accordance with the profitmaximizing behaviour of firms, a comb ination of a real wage,
w, and
employment, L, must be such that at this combination, W/ P= MPL/m, meaning that the labour
demand curve is MPL/ m. Indicate this labour demand curve in the figure, and explain that
under the circumstances described labour demand g iven
w must be less than L".
MRS
MPL
L
L"
L'
L
SorensenWhitta Jacobsen:
Introducing Advanced
acroeconomics
338
I Part 4 Stnctural
Unemployment
11. Some facts and
© The McGrawHill
introductory theory about
unemployment
Companies. 2005
PART 4: STRUCTURAL UNEMPLOYMENT
3. Explain that if there is involuntary unemployment (some workers, who would like to work at the
going real wage, are without jobs) then the prevailing combination of real wage,
w, and
employment, L, must be strictly to the left of the MRS (labour supply) curve. Then explain that
if firms are wage takers and maximize profits, the combination of wand L must be (weakly) to
the left of the MPL curve.
4. Explain that if there is involuntary unemployment and the firms maximize profits then at the
prevailing level of employment one must have MPL >MRS. Explain why the situation involves a
social loss. Assume that the combination of real wage and employment is as indicated by
w
and L in the figure, a situation of involuntary unemployment compatible w ith profit maximization.
Explain that the area of the triangle ABC represents the total social loss in the situation.
Assume that the combination is instead wand L'', compatible with the firms being perfectly
competitive, but still a situation of involuntary unemployment. Is there still a social loss and if so
how large is it?
5. Return to the initial situation with real wage w and employment L, indicated in the figu re.
Assume that some marginal, hitherto unemployed workers are hired at the wage rate
w, so
that employment increases to a level above L, but below L'. Indicate the social gain that this
gives rise to as an area in the figure. How much of the gain accrues to the hired workers and
how much to the firms? Is there a way to make everybody enjoy part of the gain?
Exercise 3. The duration of unemployment
Assume that you are told that in some year the (average) unemployment rate was 8.3 per cent.
Does this mean that 8.3 per cent of the people in the labour force were w ithout jobs the whole
year, or does it mean that all people in the labour force were without jobs for one month during
the year? Or does it mean something else, and if so what does it (typically) mean? We said
in this chapter that w ith respect to the individual costs of unemployment it is longterm
unemployment that is most severe. What is meant by longterm unemployment and why is this
a particularly bad kind of unemployment?
Exercise 4. Coefficients of correlation
Below is a table showing the rate of unemployment in the USA for the period from 1965 to 1975.
u,
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
4.51
3.78
3.84
3.55
3.51
4.98
5.95
5.60
4.88
5.62
8.47
Source: O ECD Economic Outlook
Compute the coefficient of correlation between u1 and u 1_ I' using the formula for the coefficient of correlation. Types of unemployment: reasons and policies Figure 11. 1 (u 1 .1 presented a certain categorization of unemployment according to type. compute the coefficient of correlation between u1 and u1 + 1 • How does it relate to the one between u 1 and u1_ 1 ? Explain your findings.9 per cent of all workers are separated from their jobs each month and only 90 per cent of those who are unemployed in one month find j obs in