Dynamic Comparative Advantage and

the Welfare Effects of trade

Stephen Redding,
New College, Oxford.
and CEPR

First version: 28th May, 1996
This version: 24th November, 1997
Abstract
This paper argues that developing economies may face a trade-off be-
tween specializing according to existing comparative advantage (in low-
technology goods), and entering sectors in which they currently lack a
comparative advantage, but may acquire such an advantage in the future
as a result of the potential for productivity growth (in high-technology
goods). Comparative advantage is endogenously determined by past tech-
nological change, while simultaneously shaping current rates of innovation.
As a result, it is possible that specialization according to current compar-
ative advantage under free trade is welfare reducing, while protectionist
measures are welfare increasing.
J.E.L. CLASSIFICATION: F10, F43, O41
KEYWORDS: Endogenous Growth, Dynamic Comparative Advantage, Learn-
ing by Doing, Productivity Growth, International Trade

This paper is a revised version of Chapter 6 of my D. Phil thesis and was completed while
I was a student at Nuffield College, Oxford. The research was funded by the ESRC. I am
very grateful to my supervisors Philippe Aghion and Steve Nickell for their helpful comments
and suggestions. I am also very grateful to Christopher Bliss, Andrea Boltho, Gavin Cameron,
Cecilia Garcia-Penalosa, Dan Maldoom, James Proudman, Danny Quah, Robert Reed, Pe-
ter Sinclair, Jon Temple, two anonymous referees and seminar participants at the European
Economic Association Conference 1996, International Economics Study Group 1996, Royal Eco-
nomic Society 1997 and Oxford for their helpful comments. The views expressed in this paper,
together with the results and any errors are the responsibility of the author alone.

Correspondence address: New College, Oxford. OX1 3BN. Tel: 01865 279482. Fax: 01865
279590. E-mail: stephen.redding@economics.ox.ac.uk.
1
1 Introduction
A study by the World Bank in the 1960s “expressed the view that an integrated
steel mill in Korea was a premature proposition without economic feasibility.”
(Pohang Iron and Steel Co. Ltd (1984), p. 23, cited in Amsden (1989)). A
number of factors, including Korea’s deficiency in the required raw materials
and its small domestic market for such a scale-intensive industry, suggested that
steel making was an industry in which Korea was unlikely to have a comparative
advantage.
1
Nonetheless, in 1973, the Korean government founded the Pohang
Iron and Steel Company Ltd. (POSCO) with an initial investment of $ 3.6bn.
Government assistance in a wide variety of forms, including subsidisation of the
cost of capital and investments in infrastructure has been central to POSCO’s
development. The company soon became one of the lowest cost steel-producers
in the world so that, in 1985, Korea unit costs of production were less than those
of Japan and around 2/3 of those in the United States (Amsden (1989), Table
12.2). By 1988, POSCO had become the eleventh largest steel company in the
world, operating 80 individual plants (Enos and Park (1988)).
Although at the time POSCO was founded Korea did not appear to have a
comparative advantage in the iron and steel industry, it seems incontrovertible
that it now does and that the Korean government has played a central role in
its acquiring one. This paper investigates the idea that developing economies
may face a trade-off between specialising according to an existing pattern of
comparative advantage (often in low-technology industries) and entering sectors
in which they currently lack a comparative advantage, but may acquire such
an advantage in the future as a result of the potential for productivity growth
(e.g. high-technology industries). We analyse the circumstances under which
the actions of private sector agents will resolve this trade-off between current
and future patterns of comparative advantage optimally. If the trade-off is not
resolved optimally, then it becomes possible for free trade to be welfare reducing.
Moreover, protectionist measures that induce specialisation in sectors where one
1
See, for example, (Amsden, 1989, Chapter 12) on which the first two paragraphs of this
Section draw.
2
does not currently have a comparative advantage may be welfare increasing.
This paper investigates these ideas within a general equilibrium model of
endogenous growth, in which an economy’s pattern of international trade and
rate of economic growth are jointly and endogenously determined. The paper
is part of a wider literature concerned with relationship been trade and growth.
For example, Krugman (1981) examines the effect of international trade upon
the world distribution of income when there are external economies to physical
capital accumulation in the manufacturing sector. An early formalisation of the
interrelationship between patterns of international trade and rates of technologi-
cal change (although there is no welfare analysis) is provided by Krugman (1987).
As a result of this interaction, initial patterns of international trade become in-
creasingly “locked-in” over time.
More recently, (Grossman and Helpman, 1990, 1991), (Rivera-Batiz and Romer
1991a, 1991b) and (Taylor 1991, 1994) have examined the relationship between
trade and growth, when endogenous growth is the result of profit-seeking in-
vestments in Research and Development (R & D).
2
Young (1991) analyses the
links between trade and growth, when bounded learning by doing leads to the
adoption of new varieties of goods; while Stokey (1991) examines the interaction
between trade and human capital accumulation. In small open economy mod-
els, Matsuyama (1992) and Sachs and Warner (1995) respectively consider the
effects of levels of agricultural productivity and endowments of natural resources
on international trade and growth.
In fact, the existing literature suggests a number of channels through which
trade may affect an economy’s rate of growth. In this paper, motivated by the
empirical discussion above, we focus upon the relationship between endogenous
comparative advantage, economic growth and economic welfare. In contrast to
much of the literature, which has emphasised the beneficial effects of trade on
growth, the analysis suggests that specialisation according to initial comparative
advantage may have negative effects upon both rates of growth and economic
2
Where these investments may either yield new varieties or (as in Aghion and Howitt (1992))
successively higher qualities of intermediate inputs.
3
welfare. The analysis builds upon a number of existing studies of endogenous
comparative advantage and growth (see, in particular, Krugman (1987), (Gross-
man and Helpman, 1990, 1991), and Young (1991)). This paper makes two main
contributions.
First, we examine the endogeneity of comparative advantage within a particu-
larly tractable, general equilibrium model of endogenous growth and international
trade between two, large economies. The tractability of this framework enables
us to undertake a complete analysis of the welfare effects of international trade
and the potential case for selective trade and industrial policies. We are able to
derive necessary and sufficient conditions for free trade, by inducing specialisation
according to current patterns of comparative advantage, to be welfare reducing.
Furthermore, we establish the circumstances under which selective trade and in-
dustrial policies, that induce specialisation in sectors where an economy does not
currently have a comparative advantage, may be welfare improving. Through-
out the analysis, the role of endogenous comparative advantage is made clear.
Motivated by the earlier empirical discussion of the East Asian development ex-
perience (see also Amsden (1989) and Wade (1990)), the paper emphasises the
potential trade-off an economy may face between specialising according to an ex-
isting pattern of comparative advantage, and entering sectors where it currently
lacks a comparative advantage, but may acquire such an advantage as a result of
the potential for productivity growth.
Second, the endogeneity of comparative advantage in models of growth and
trade has led a number of authors in the theoretical literature (see, for example,
Krugman (1987) and Grossman and Helpman (1991)) to speak in terms of ‘dy-
namic comparative advantage.’ This very same term appears in more informal
discussions of the East Asian development experience (see, for example, Amsden
(1989)). This paper’s second objective is therefore to see whether, on the basis of
the theoretical analysis of the relationship between international trade and eco-
nomic growth, any substantiative content can be given to this often-used, but so
far ill-defined concept. The paper suggests that, when comparative advantage is
endogenous in dynamic trade models, the traditional (or ‘static’) notion of com-
4
parative advantage may be usefully augmented with a second ‘dynamic’ concept.
This dynamic concept explains the evolution of patterns of international trade
over time and sheds light upon the circumstances under which welfare improving
selective trade and industrial policies exist. Interestingly, if such policies exist,
they need only be temporary.
The paper is structured as follows. Section 2 introduces the model, while
section 3 solves for static equilibrium under both autarky and free trade. Section
4 is concerned with the relationship between trade and productivity growth, and
shows how comparative advantage is endogenously determined. Section 5 consid-
ers the implications of endogenous comparative advantage for the welfare effects
of international trade. The standard static gains from trade are augmented with
dynamic effects, which may either increase or decrease the intertemporal welfare
of the representative agent. Section 6 addresses the related, but distinct question
whether selective trade and industrial policies to induce entry into a sector where
an economy currently lacks a comparative advantage may be welfare improving.
Section 7 moves on to consider the popular notion of ‘dynamic comparative ad-
vantage.’ The popular notion is formalised and its relationship to the preceding
analysis discussed. Finally, section 8 concludes.
2 A dynamic Ricardian model
In this section, a standard Ricardian model of international trade (see, for exam-
ple, Krugman and Obstfeld (1994)) is augmented with a specification for produc-
tivity dynamics. Productivity in each sector is assumed to evolve endogenously
over time as learning by doing occurs. We consider international trade between
two economies (‘home’ and ‘foreign’), where all foreign variables are denoted by
an asterisk.
Each economy may produce two final goods, a low-technology, traditional
good z (e.g. agriculture, textiles) and a high-technology, frontier good h (e.g.
manufacturing, electronics).
3
Labour is the sole factor of production, and the
3
See Dornbusch et al. (1977) for an exposition of the static Ricardian model with a contin-
uum of goods.
5
two economies are populated with a number of representative consumers (
¯
L and
¯
L

). Time is continuous and is indexed by t.
2.1 The static model
Consumer preferences are assumed to be identical in the two economies, with
instantaneous utility a Cobb-Douglas function of consumption of the low- and
high-tech goods: u(c
z
, c
h
) = c
β
z
c
1−β
h
where 0 < β < 1.
4
Intertemporal utility is the
sum of instantaneous utilities, discounted at the subjective rate of time preference
ρ. For simplicity, we assume that there is no storage or savings technology so
that, at each point in time, expenditure equals income for the representative
consumer. Each consumer is endowed with one unit of labour, which is supplied
inelastically with zero disutility.
Low- and high-tech goods are produced with labour L
j
according to constant
returns to scale technologies, whose productivity we index by A
j
, for j = z, h.
Aggregate output in each sector is thus,
Y
z
= A
z
.L
z
, Y
h
= A
h
.L
h
(1)
Production is assumed to occur under conditions of perfect competition, and
we make the standard assumption that labour is perfectly mobile between sectors
and immobile across countries. Home labour market clearing requires L
z
+L
h
=
¯
L.
2.2 Productivity dynamics
A wide range of empirical evidence suggests that learning by doing is an important
source of productivity improvements. For example, Lucas (1993) cites evidence
that each doubling of cumulative output of “Liberty Ships” in 14 U.S. shipyards
during World War II was associated with a reduction of man-hours required
per ship by between 12 and 24 per cent. Hence, following Krugman (1987), we
assume that productivity in each sector A
j
depends upon a stock of sector-specific
4
In general, lower case letters are used for per capita variables. In order to simplify notation,
we suppress an implicit dependence upon time, except where it is important.
6
production experience K
j
, as well as exogenous factors ψ
j
such as climate, culture,
political institutions and laws,
A
z
(t) = ψ
z
.K
z
(t), A
h
(t) = ψ
h
.K
h
(t) (2)
where ψ
j
> 0 for j = z, h.
While producing output in a given sector, agents acquire productivity-enhancing
production experience K
j
(or learn by doing): for example, through trial and er-
ror, new methods of manufacture or new ways of organising existing processes are
discovered. The rate at which this production experience is acquired is assumed
to depend upon the flow of labour employed in producing a sector’s output, so
that K
j
evolves according to,
5
˙
K
z
(t) = µ
z
.L
z
(t).K
z
(t), µ
z
> 0 (3)
˙
K
h
(t) = µ
h
.L
h
(t).K
h
(t), µ
h
> 0 (4)
where µ
j
parameterises the rate at which knowledge is acquired as part of the
production process in sector j. Learning by doing is assumed to be a pure exter-
nality of the production process: in particular, learning by doing is assumed to
be external to individual firms but specific to a sector and to an economy.
6
5
According to the specification in (3) and (4), learning by doing is technologically un-
bounded. It is possible to extend the analysis to the case of bounded learning by doing (see
Redding (1996)), in which case comparative advantage is not only endogenous but the rate at
which it evolves over time is a function of cumulative production experience. However, this
only complicates the analysis without adding additional insight.
6
There is considerable empirical evidence that international knowledge spillovers are imper-
fect (see, for example, Coe and Helpman (1995)) and that levels of Total Factor Productivity
differ substantially across economies (see, for example, Islam (1995)). It is possible to intro-
duce international knowledge spillovers. However, as long as these are imperfect, the analysis
remains essentially unchanged.
7
3 Static equilibrium
3.1 Autarky
Autarkic equilibrium is fully characterised by the requirement that the relative
price of the low-tech good equals both minus the Marginal Rate of Substitution
(MRS) and minus the Marginal Rate of Transformation (MRT) between low-
and high-tech goods. From the expression for instantaneous utility and (1), we
require β/(1 −β).C
h
/C
z
= p
z
/p
h
= A
h
/A
z
.
Consumer preferences over the low- and high-tech goods exhibit a demand
for variety. Autarkic equilibrium is characterised by incomplete specialisation,
with labour allocated in the constant proportions β and (1 −β) to the low- and
high-tech sectors respectively,
L
z
= β.
¯
L, L
h
= (1 −β).
¯
L (5)
3.2 Free Trade
In this subsection, we allow two previously autarkic economies to engage in free
trade from some arbitrary point in time t
1
onwards. With free trade and zero
transport costs, the price of the low- and high-tech goods must be the same in each
economy. Perfect competition implies that the home wage in the low- and high-
tech sectors will equal w
z
(t) = A
z
(t).p
z
(t) and w
h
(t) = A
h
(t).p
h
(t) respectively;
where, if specialisation in home is incomplete, we require w
z
= w
h
= w.
Throughout the following, we will be largely concerned with equilibria char-
acterised by complete specialisation in both economies.
7
For home to specialise
completely in the low-tech sector and foreign in the high-tech, we require w
z
> w
h
and w

h
> w

z
. That is,
A

h
(t)
A

z
(t)
>
p
z
(t)
p
h
(t)
>
A
h
(t)
A
z
(t)
(6)
7
This is merely for simplicity. It is straightforward to extend the analysis to cases of incom-
plete specialisation.
8
In such an equilibrium, the entire of home’s supply of labour is employed in
the low-tech sector and the entire of foreign’s in the high-tech sector. Equation
(6) and the associated allocation of labour to the two sectors defines the world
supply of the low- relative to the high-tech good: RS
zh
= (Y
z
+ Y

z
)/(Y
h
+ Y

h
).
With Cobb-Douglas instantaneous utility, each representative consumer allo-
cates expenditure to the low- and high-tech sectors in the constant proportions
β and (1 −β) at each point in time t. World demand for the low-tech relative to
the high-tech good is thus,
RD
zh
=
C
z
+ C

z
C
h
+ C

h
=
β
(1 −β)
.
p
h
p
z
(7)
General equilibrium of the static model may be fully characterised in relative
supply, relative demand space, as shown in Figure 1.
Figure 1 about here
4 Endogenous comparative advantage
The pattern of international trade in the static Ricardian model is determined
by the traditional or ‘static’ notion of comparative advantage. Thus, an economy
is said to have a ‘static comparative advantage’ in the low-tech sector at time t
if the opportunity cost of producing the low-tech good at home is lower than in
the other economy,
A
h
(t)
A
z
(t)
<
A

h
(t)
A

z
(t)
(8)
where, from (6), this is a necessary condition for home to specialise in the low-
tech sector in the free trade equilibrium. Throughout the paper, ‘comparative
advantage’ will be used in this traditional, ‘static’ sense unless otherwise speci-
fied. We will largely be concerned with equilibria in which home has an initial
comparative advantage in the low-tech sector.
From equation (8), it is clear that the pattern of comparative advantage at any
point in time depends upon productivity in each sector of the two economies (A
j
,
9
A

j
, j = z, h). This, in turn, is a determined by a combination of exogenous factors
on the one hand (such as climate, political institutions and laws as parameterised
by ψ
j
and ψ

j
, j = z, h) and past technological change on the other (as manifested
in the stocks of cumulative production experience K
j
and K

j
, j = z, h). Thus,
comparative advantage depends upon past technological change in each sector;
while, at the same time, determining the free trade allocation of labour across
sectors and hence (from (3) and (4)) rates of productivity growth in each sector.
Under autarky, home is incompletely specialised in both sectors and, from
(3), (4) and (5), accumulates production experience at the rates g
n
z
= µ
z
.β.
¯
L and
g
n
h
= µ
h
.(1 − β).
¯
L in the low- and high-tech sectors respectively.
8
Similarly, for
foreign we have g
n∗
z
= µ

z
.β.
¯
L

and g
n∗
h
= µ

h
.(1 − β).
¯
L

. In contrast, in the free
trade equilibrium, home’s comparative advantage in the low-tech sector means
that it specialises completely in the production of this good. Thus, home learns
by doing in the low-tech sector alone (at the rate g
f
z
= µ
z
.
¯
L), while foreign learns
by doing in the high-tech sector (at the rate g
f∗
h
= µ

h
.
¯
L

).
9
Specialisation according to comparative advantage under free trade changes
the (endogenous) rate of productivity growth in each sector of the two economies.
Productivity levels dictate comparative advantage, which affects the allocation
of labour between sectors. This in turn determines relative rates of productivity
growth, and thereby feeds back to shape the evolution of productivity levels over
time. In this way, current comparative advantage is endogenously determined.
The endogeneity of comparative advantage in models of growth and trade has
led a number of authors to speak in terms of ‘dynamic comparative advantage’;
although, as yet, this concept has remained ill-defined. A later section comes
back to discuss this idea. However, first, we move on to consider the implications
of endogenous productivity growth and endogenous comparative advantage for
the welfare effects of trade.
8
Where the superscript n (‘no trade’) indexes the value of a variable under autarky.
9
Where the superscript f (‘free trade’) indexes the value of a variable under free trade.
10
5 Trade and welfare
In this section, we compare the representative consumer’s intertemporal welfare
under the alternative regimes of remaining autarkic from time t
1
onwards and
engaging in free trade. In each case, intertemporal welfare is given by the dis-
counted sum of instantaneous utilities. Furthermore, since instantaneous utility
is Cobb-Douglas, it follows that, under both autarky and free trade, the repre-
sentative consumer will allocate the constant proportions β and (1−β) of his/her
expenditure at each point in time t to the low- and high-tech goods respectively.
5.1 Welfare under autarky
Beginning with autarky, specialisation is incomplete and the representative con-
sumer’s income is given by the wage w(t) = p
n
z
(t).A
n
z
(t) = p
n
h
(t).A
n
h
(t). Using
the fact that constant proportions of income are allocated to expenditure on the
low- and high-tech goods, we obtain the following expression for intertemporal
welfare,
U
n
t
1
=


t
1
e
−ρ(t−t
1
)
[β.A
n
z
(t)]
β
[(1 −β).A
n
h
(t)]
1−β
dt (9)
where A
n
j
(t) denotes the level of productivity in each sector j = z, h under autarky
at each point in time t ≥ t
1
.
5.2 Welfare under free trade
In contrast, under free trade, specialisation is complete and the representative
consumer’s income is equal to the wage in the low-tech sector w(t) = p
f
z
(t).A
f
z
(t)
for all t ≥ t
1
. Again using the fact that consumer expenditure is allocated
in constant proportions, we arrive at an analogous expression for intertemporal
welfare,
U
f
t
1
=


t
1
e
−ρ(t−t
1
)

β.A
f
z
(t)

β

(1 −β).A
f
z
(t).p
f
z
(t)/p
f
h
(t)

1−β
dt (10)
where the relative price of the low-tech good is determined on world markets.
11
5.3 Static gains from trade
The right-hand sides of equations (9) and (10) contain information about levels
of instantaneous utility at all points in time t ≥ t
1
. From these two equa-
tions, it is immediately clear that instantaneous utility must be lower under au-
tarky than under free trade at time t
1
when the choice between the two regimes
must be made. Instantaneous utility will be lower under autarky if and only if
p
f
z
(t
1
)/p
f
h
(t
1
) > A
h
(t
1
)/A
z
(t
1
), where we use the fact that A
n
z
(t
1
) = A
f
z
(t
1
) and
A
n
h
(t
1
) = A
f
h
(t
1
) (clearly, these equalities will not, in general, hold for t > t
1
).
This condition must be satisfied in a free trade equilibrium in which home has
a static comparative advantage in and specialises in low-tech production. The
existence of the standard static gains from trade (from specialisation according to
comparative advantage) implies that instantaneous utility must be initially lower
under a regime of autarky than one of free trade.
5.4 Dynamic effects and intertemporal welfare
However, the fact that technological change is endogenous means that a move
from autarky to free trade has additional, dynamic welfare effects. Here, two
aspects of the analysis are important. First, as discussed in the previous section,
specialisation according to comparative advantage leads to reallocations of re-
sources between the low- and high-tech sectors. These reallocations of resources
affect rates of learning by doing and productivity growth in each sector of the
two economies, and hence have dynamic effects on economic welfare.
Second, the literatures on both the microeconomics of technological change
and endogenous growth suggest a number of reasons why the laissez-faire rate of
technological change may be less than the socially optimal rate.
10
In the present
case, technological change takes the form of serendipitous learning by doing. Be-
10
Of course, there are also reasons why the laissez-faire rate of technological change may be
higher than is socially optimal (e.g. the ‘business stealing effect’ and the ‘monopoly distortion
effect’ in Aghion and Howitt (1992)). However, using a theoretical framework drawing on the
endogenous growth literature, Jones and Williams (1997) review the empirical evidence and
find that social rates to Research and Development (R & D) typically exceed the corresponding
private rates of return.
12
cause technological change is a positive externality of current production, private
sector agents do not fully take into account the potential for productivity growth
in each sector.
11
As a result, agents fail to internalise the changes in rates of
productivity induced by international specialisation and the consequent dynamic
effects on economic welfare.
The effects of specialisation according to comparative advantage on produc-
tivity growth rates were the subject of the previous section. This section takes
the analysis one stage further to consider the implications of changes in produc-
tivity growth for economic welfare. Combining the dynamic welfare effects of
international trade with the standard static gains from trade enumerated above,
we evaluate relative levels of intertemporal welfare for the representative agent
under autarky and free trade.
Beginning with autarky, incomplete specialisation implies that home expe-
riences learning by doing in both the low- and high-tech sectors at the rates
g
n
z
= µ
z
.β.
¯
L and g
n
h
= µ
h
.(1 − β).
¯
L respectively. Hence, productivity levels in
the two sectors equal A
n
z
(t) = e
g
n
z
(t−t
1
)
.A
z
(t
1
) and A
n
h
(t) = e
g
n
h
(t−t
1
)
.A
h
(t
1
) respec-
tively for all t ≥ t
1
. Substituting for levels of productivity in the two sectors in (9)
and evaluating the integral, we obtain the following expression for intertemporal
welfare under autarky,
12
U
n
t
1
=
β
β
(1 −β)
1−β
. [A
z
(t
1
)]
β
. [A
h
(t
1
)]
1−β
ρ −βg
n
z
−(1 −β)g
n
h
(11)
Turning now to free trade, complete specialisation implies that home only
experiences learning by doing in the low-tech sector (at the rate g
f
z
= µ
z
.
¯
L),
while foreign only enjoys learning by doing in the high-tech sector (at the rate
g
f∗
h
= µ

h
.
¯
L

). Hence, home productivity in the low-tech sector in (10) may be
expressed as A
f
z
(t) = e
g
f
z
(t−t
1
)
.A
z
(t
1
). However, from (10), before we can solve
explicitly for intertemporal welfare, we require an expression for the equilibrium
11
More generally, the same will be true in models of endogenous technological change through
profit-seeking Research and Development (R & D), as long as the social rate of return to R &
D exceeds the private rate of return (see an appendix available from the author on request and
Redding (1996)).
12
Where for the integral in (9) to converge and intertemporal utility to be finite, we require
ρ > β.g
n
z
+ (1 −β).g
n
h
.
13
free trade relative price of the low-tech good.
With home and foreign specialising completely in the low- and high-tech sec-
tors respectively, the relative supply of the low-tech good is simply A
f
z
(t)
¯
L/A
f∗
h
(t)
¯
L

;
while relative demand is determined according to (7). Hence, the free trade equi-
librium relative price of the low-tech good equals,
p
f
z
(t)
p
f
h
(t)
=
β
1 −β
A
f∗
h
(t)
A
f
z (t)
¯
L

¯
L
(12)
where, from the above, A
f∗
h
(t) = e
g
f∗
h
(t−t
1
)
.A

h
(t
1
) and A
f
z
(t) = e
g
f
z
(t−t
1
)
.A
z
(t
1
) for
all t ≥ t
1
. Using this equation for equilibrium relative prices in (10), substituting
for the productivity levels A
f
z
(t) and A
f∗
h
(t), and evaluating the integral, we obtain
the following expression for intertemporal welfare under free trade,
13
U
f
t
1
=
β. [A
z
(t
1
)]
β
[A

h
(t
1
)]
1−β

¯
L

/
¯
L

1−β
ρ −βg
f
z −(1 −β)g
f∗
h
(13)
From equations (13) and (11), intertemporal welfare under free trade will be
lower than under autarky if and only if,
[A

h
(t
1
)]
1−β

¯
L

1−β
ρ−βg
f
z −(1 −β)g
f∗
h
<

1−β
β

1−β
[A
h
(t
1
)]
1−β
¯
L
1−β
ρ−βg
n
z
−(1 −β)g
n
h
(14)
As we saw in the last subsection, the existence of the standard static gains
from trade means that instantaneous welfare at the time t
1
, when the choice be-
tween the two regimes is made, must be higher under free trade. The numerators
in equations (13) and (11) are, in fact, simply instantaneous utility at time t
1
under the two regimes. Hence, it follows immediately that the numerator on the
left-hand side of the inequality (14) must exceed the numerator on the right-hand
side.
14
However, whether intertemporal welfare will be higher under free trade than
under autarky will depend, not only upon levels of instantaneous utility at time
13
Where, for intertemporal utility to be finite, we require ρ > β.g
f
z
+ (1 −β).g
f∗
h
.
14
To confirm this, note that, by assumption, home has a comparative advantage in low-tech
production at time t
1
. Hence, A
z
(t
1
).p
z
(t
1
)/p
h
(t
1
) > A
h
(t
1
). Substituting for p
z
(t
1
)/p
h
(t
1
)
from (12), we obtain [A

h
(t
1
)] .
¯
L

>

1−β
β

. [A
h
(t
1
)] .
¯
L.
14
t
1
, but also upon the rate of growth of instantaneous utility from time t
1
onwards.
Here, specialisation according to comparative advantage induces three dynamic
effects upon intertemporal welfare.
First, the reallocation of labour to home’s low-tech sector induced by speciali-
sation under free trade raises home’s rate of learning by doing in this sector (from
g
n
z
= µ
z

¯
L to g
f
z
= µ
z
.
¯
L). Second, specialisation in the low-tech sector under
free trade means that home forgoes its own potential to learn by doing in the
high-tech sector (where, as a result, the domestic rate of productivity growth falls
from g
n
h
= µ
h
.(1 − β).
¯
L to g
f
h
= 0). Third, although forgoing its own potential
to learn by doing in the high-tech sector under free trade, home experiences the
benefits of foreign learning by doing in this sector in the form (see equation (12))
of a terms of trade gain (where, under free trade, foreign’s rate of learning by
doing in the high-tech sector is g
f∗
h
= µ

h
.
¯
L

).
The effect of free trade on the rate of growth of instantaneous utility will de-
pend upon the net outcome of these three dynamic effects. In terms of inequality
(14), the rate of growth of instantaneous utility will be lower under free trade if
and only if βg
f
z
+(1 −β)g
f∗
h
< βg
n
z
+(1 −β)g
n
h
. Substituting for the equilibrium
rate of productivity growth in each sector under the two regimes, we obtain (from
the above) the condition,
β.µ
z
.
¯
L +

µ

h
.
¯
L

−(1 −β).µ
h
.
¯
L

< 0 (15)
The first of the three dynamic effects identified above is unambiguously posi-
tive (g
f
z
> g
n
z
), as reflected in the strictly positive first term on the left-hand side
of the inequality (15). The increase in home’s rate of learning by doing in the
low-tech sector brought about by free trade (reflecting increased employment in
this sector) raises intertemporal welfare relative to that under autarky. This is
essentially a ‘scale effect’ of international trade, whereby trade expands the size
of the market for home low-tech goods.
However, home’s rate of learning by doing in the high-tech sector under au-
tarky (g
n
h
= µ
h
.(1−β).
¯
L ) either may or may not exceed foreign’s under free trade
(g
f∗
h
= µ

h
.
¯
L

). Foreign allocates its entire labour force to high-tech production
15
under free trade, while home only allocates a proportion (1 − β) under autarky.
Nonetheless, the relative magnitude of the two rates of learning by doing also
depends upon the size of the two economies (as measured by the labour forces
¯
L
and
¯
L

) and the potential for learning by doing in the high-tech sector in each
of the two economies (as determined by µ
h
and µ

h
). Therefore, the net impact
of the second and third dynamic effects on intertemporal welfare may be either
positive or negative, and this is reflected in the ambiguous term in parentheses
in inequality (15).
The existence of the dynamic effects of international trade on economic wel-
fare (due to the change in rates of productivity growth induced by specialisation
according to comparative advantage) means that free trade is no longer neces-
sarily welfare increasing, as in standard static theories of international trade. A
necessary condition for free trade to be welfare reducing is that the rate of learn-
ing by doing in the high-tech sector is lower under free trade than under autarky,
and that the effect of this on the rate of growth of instantaneous utility exceeds
that of the increase in the rate of learning by doing in the low-tech sector. That
is, we require inequality (15) to be satisfied.
From (15), this necessary condition for free trade to be welfare reducing is
more likely to be satisfied, the larger home’s potential (µ
h
) for learning by do-
ing in the high-tech sector relative to foreign’s (µ

h
) and the smaller the foreign
economy (as measured by its labour force
¯
L

). The effect of the size of the home
economy (as measured by its labour force
¯
L) and the share of consumer expen-
diture devoted to low-tech goods (β) is ambiguous, and depends, for example,
upon the relative values of home’s potential to learn by doing in the low- and
high-tech sectors (µ
z
and µ
h
respectively).
In order for intertemporal welfare to fall as a result of moving from autarky
to free trade, we require that these dynamic welfare losses from free trade exceed
the standard static gains from trade. From equations (13) and (11), this will
occur whenever inequality (14) is satisfied (which thus provides a necessary and
sufficient condition for free trade to be welfare reducing).
16
6 Policy intervention
The previous section has shown that, when technological change is endogenous,
the static gains from trade are augmented with a number of dynamic welfare ef-
fects and international trade is no longer necessarily welfare increasing. Free trade
will reduce the intertemporal welfare of the representative agent if an economy’s
initial pattern of (static) comparative advantage means that it fails to specialise
in a sector in which its potential to learn by doing is large relative to its trading
partner’s.
This immediately raises a further question: could it ever be optimal for a
policy-maker to induce an economy to specialise in the sector where it does not
currently have a comparative advantage but exhibits considerable potential to
learn by doing ? In terms of the analysis of the previous section, the free trade
equilibrium is characterised by home specialising in the low-tech and foreign in
the high-tech sector. Could it ever be optimal for the policy-maker to try to
reverse this initial pattern of international specialisation ? Clearly, the answer
to this further question will depend upon both the economies’ potential rates of
learning by doing in the sector where they specialise under the proposed policy
intervention and the corresponding rates of productivity growth in the sector
where they specialise under the alternative of free trade.
Another way of thinking about the issue is as follows. The pattern of compar-
ative advantage at any one point in time is, as earlier noted, endogenous. In the
free trade equilibrium, the home economy’s initial pattern of comparative advan-
tage leads it to specialise in one way. Given this initial pattern of specialisation,
productivity growth rates and the evolution of comparative advantage over time
are then determined. However, if the initial patterns of comparative advantage
and international specialisation were otherwise (as, for example, the result of
a policy intervention), then rates of productivity growth and the time path for
comparative advantage could be very different indeed. Ascertaining whether the
policy intervention is welfare improving relative to free trade involves an eval-
uation of productivity dynamics and the time path for comparative advantage
17
under each of the alternative regimes.
In this section, we compare intertemporal welfare under free trade (as evalu-
ated above) with intertemporal welfare under a policy of subsidising entry into
the high-tech sector.
15
As before, home is assumed to have an initial compar-
ative advantage in the low-tech sector and we will be concerned with equilibria
characterised by complete specialisation. That is,
A

h
(t
1
)
A

z
(t
1
)
>
p
f
z
(t
1
)
p
f
h
(t
1
)
>
A
h
(t
1
)
A
z
(t
1
)
(16)
In the free trade equilibrium, home and foreign specialise in the low- and
high-tech sectors respectively; and the equilibrium price of the low-tech good at
time t
1
is determined by equation (12).
Consider now a policy intervention of subsidising production in the high-
tech sector, where the home economy does not currently have a comparative
advantage. For each unit of income earned in the high-tech sector, individuals
are assumed to receive a production subsidy of monetary value s > 0. The
subsidy is assumed to be self-financing, being fully funded by a tax ξ, 0 < ξ < 1,
on wage income. The after-tax/after-subsidy wages in the low- and high-tech
sectors are thus,
16
w
s
z
(t) = (1 −ξ) p
s
z
(t)A
s
z
(t), w
s
h
(t) = (1 + s) (1 −ξ) p
s
h
(t)A
s
h
(t) (17)
For a sufficiently large value of the production subsidy s, w
s
h
(t) > w
s
z
(t),
and home will now specialise in the high-tech sector under international trade.
The initial pattern of comparative advantage is reversed, and again we restrict
consideration to equilibria characterised by complete specialisation. Thus,
(1 + s) .A
h
(t
1
)
A
z
(t
1
)
>
p
s
z
(t
1
)
p
s
h
(t
1
)
>
A

h
(t
1
)
A

z
(t
1
)
(18)
where the relative price of the low-tech good at time t
1
under the subsidy is now,
15
The comparison between free trade and the subsidy seems empirically the most relevant.
However, it is straightforward to extend the analysis to compare welfare under the subsidy with
that under autarky.
16
Where the superscript s indexes the value of a variable under the subsidy.
18
p
s
z
(t
1
)
p
s
h
(t
1
)
=
β
1 −β
.
A
h
(t
1
)
A

z
(t
1
)
.
¯
L
¯
L

(19)
Appendix 1 shows that inequalities (16) and (18), and the two equations
for the relative price of the low-tech good, (12) and (19), may be simulta-
neously satisfied for sufficiently large values of s and β ∈ (1/2, 1). The as-
sumption of complete specialisation under both free trade and the subsidy is
thus validated.
17
For the production subsidy to be self-financing, we require
s.p
s
h
(t).A
h
(t) = ξ. (1 + s) .p
s
h
(t).A
h
(t); and the equilibrium tax rate is thus,
ˆ
ξ =
s
1 + s
(20)
General equilibrium under free trade and the subsidy, with the accompanying
change in the pattern of comparative advantage and international specialisation,
is illustrated diagrammatically in Figure 2. Under the production subsidy, home
specialises in the high-tech sector and foreign in the low-tech. The representative
agent’s income in the home economy is given by the after-tax/after-subsidy wage
in the high-tech sector. Replacing the tax rate ξ with its equilibrium value
ˆ
ξ
in (17), we obtain an expression for disposable income. Using this expression
and the fact that, in equilibrium, the representative agent allocates expenditure
in constant proportions to each sector, we may solve for intertemporal welfare
under the subsidy,
U
s
t
1
=


t
1
e
−ρ(t−t
1
)
[β.p
s
h
(t)/p
s
z
(t).A
s
h
(t)]
β
[(1 −β).A
s
h
(t)]
1−β
dt (21)
where the relative price of the high-tech good p
s
h
(t)/p
s
z
(t) is determined according
to equation (19).
Figure 2 about here
The existence of the standard static gains from trade means that instanta-
neous utility at the time t
1
, when the choice between the two regimes is made,
17
For simplicity, we restrict attention to the case of complete specialisation. Again, it is
straightforward to extend the analysis to consider incomplete specialisation.
19
must be lower under the production subsidy (the home economy is choosing to
specialise in a sector in which it has a comparative disadvantage).
18
However, as
in the comparison between autarky and free trade, the change in the pattern of
international specialisation between the two regimes has implications for rates of
productivity growth and hence has dynamic effects upon economic welfare.
Under the production subsidy, complete specialisation implies that home ex-
periences learning by doing in its high-tech sector (at the rate g
s
h
= µ
h
.
¯
L); while
foreign enjoys the fruits of learning by doing in its low-tech sector (at the rate
g
s∗
z
= µ

z
.
¯
L

). Thus, productivity levels in the high-tech and low-tech sectors
in home and foreign respectively may be expressed as A
s
h
(t) = e
g
s
h
(t−t
1
)
.A
h
(t
1
)
and A

z
(t) = e
g
s∗
z
(t−t
1
)
.A
z
(t
1
) for all t ≥ t
1
. Under the subsidy, home forgoes its
own potential to learn by doing in the low-tech sector, and instead benefits from
foreign learning by doing in this sector in the form of a terms of trade gain.
Substituting for the relative price of the high-tech good (from (19)) in the
equation for intertemporal welfare (21), then substituting for productivity levels
in the two sectors and evaluating the integral, we obtain the following expression
for intertemporal welfare under the subsidy,
19
U
s
t
1
=
(1 −β). [A

z
(t
1
)]
β
. [A
h
(t
1
)]
1−β
.

¯
L

/
¯
L

β
ρ −βg
s∗
z
−(1 −β)g
s
h
(22)
From equations (22) and (13), intertemporal welfare will be higher under the
subsidy than under free trade if and only if,
(1−β)
β
[A

z
(t
1
)]
β
[A
h
(t
1
)]
1−β

¯
L

¯
L

2β−1
ρ −βg
s∗
z
−(1 −β)g
s
h
>
[A
z
(t
1
)]
β
[A

h
(t
1
)]
1−β
ρ −βg
f
z −(1 −β)g
f∗
h
(23)
We saw in the discussion above, that the existence of the standard static gains
from trade means that instantaneous welfare at time t
1
must be lower under the
subsidy. The numerators in equations (22) and (13) are simply instantaneous
utility at time t
1
under the two regimes. Hence, it follows immediately that the
18
This is established formally in the proof of a later Proposition (see the proof of Proposition
1 in Appendix 2).
19
Where, for intertemporal utility to be finite, we require ρ > β.g
s∗
z
+ (1 −β).g
s
h
.
20
numerator on the left-hand side of the inequality (23) must be strictly less than
the numerator on the right-hand side.
20
A necessary condition for the production subsidy to the high-tech sector to
be welfare improving is therefore that the rate of growth of instantaneous utility
under the subsidy exceeds the corresponding rate of growth under free trade:
βg
s∗
z
+ (1 −β)g
s
h
> βg
f
z
+ (1 −β)g
f∗
h
. As we saw in the discussion above, home’s
specialisation in the high-tech sector results in it accumulating production ex-
perience at the rate g
s
h
= µ
h
.
¯
L under the subsidy, while foreign accumulates
production experience in the low-tech sector at the rate g
s
z
= µ

z
.
¯
L

. In contrast,
under free trade, the pattern of international specialisation is exactly the reverse.
Home experiences learning by doing in the low-tech sector at the rate g
f
z
= µ
z
.
¯
L,
while foreign learns by doing in the high-tech sector at the rate g
f∗
h
= µ

h
.
¯
L

. A
necessary condition for the production subsidy to be welfare improving is thus,
β.

µ

z
¯
L

−µ
z
¯
L

+ (1 −β).

µ
h
¯
L −µ

h
¯
L

> 0 (24)
That is, we require a weighted average of the change in rates of productivity
growth induced by the reversal of patterns of international specialisation to be
strictly positive. With Cobb-Douglas instantaneous utility, the weights are the
shares of consumer expenditure allocated to each sector. Whether or not this
inequality is satisfied will depend upon the two countries’ relative potentials to
learn by doing in both the low- and high-tech sectors (as parameterised by µ
j
, µ

j
for j = z, h), the share of consumer expenditure allocated to the low-tech sector
(β), and the two economies’ relative sizes (as measured by
¯
L,
¯
L

).
While inequality (24) is a necessary condition for the subsidy to be welfare
improving, it is clearly not sufficient. In order for intertemporal welfare to rise as a
result of implementing the subsidy rather than adopting free trade, we require the
dynamic welfare gains from home specialising in the high-tech sector and foreign
in the low-tech sector to exceed the static welfare losses. From equations (22) and
20
This may be shown formally by substituting for the relative price of the low-tech good
p
z
(t
1
)/p
h
(t
1
) under both the subsidy and free trade (using equations (19) and (12)) in the
proof of Proposition 1 below.
21
(13), this will occur whenever inequality (23) is satisfied (which thus provides a
necessary and sufficient condition for the subsidy to be welfare improving).
In the next section, we consider the relationship between this general equilib-
rium argument for selective trade and industrial policies and the often-discussed
notion of ‘dynamic comparative advantage.’ First, we note that, although one
may establish theoretical conditions for interventionist public policies to be wel-
fare improving, it may be extremely difficult in practice to determine when these
conditions are met. The information requirements to implement these policies
are large - the analysis requires a policy-maker to have information on rates of
productivity growth in either sector of each economy under both the proposed
subsidy and free trade. Furthermore, the literature on the political economy of
trade policy suggests that there may be hidden welfare costs to activist trade
policies in the form of Directly Unproductive Profit-Seeking (DUP) activity (see,
for example, Bhagwati (1982)).
Nonetheless, there may be instances where the potential for productivity
growth in sectors where an economy does not currently exhibit a comparative
advantage is large, and where an active trade policy can be justified in terms of
the theoretical analysis of this section. Developing economies with high levels
of general human capital, which may achieve rapid rates of productivity growth
through imitation, may be a case in point. Indeed, there is empirical evidence
that the development experience of some East Asian economies may be inter-
preted in these terms.
7 Dynamic comparative advantage and the case
for policy intervention
The fact that comparative advantage evolves endogenously over time in theo-
retical models of endogenous growth and trade has led a number of authors to
speak of ‘dynamic comparative advantage’ (see, for example, Krugman (1987)
and Grossman and Helpman (1991)). Somewhat independently, the same con-
cept has been applied in more informal discussions of the East Asian development
22
experience (see, for example, Amsden (1989)). In each case, the concept is left
ill-defined, with its exact usage and meaning unclear.
On the one hand, the use of the concept may reflect a desire to explain the
way in which comparative advantage (as traditionally defined) evolves over time
in dynamic trade models (see, in particular, the discussion in Grossman and
Helpman (1991)). On the other hand, the concept’s use may indicate a concern
with some of the welfare considerations that have been the subject of previous
sections. This seems to be particularly the case with regard to the literature on
the East Asian development experience. Here, the use of the concept seems linked
with the idea that a country’s current pattern of comparative advantage may
work against its long-term interests, and that there may be a trade-off between
specialising according to current comparative advantage and realising dynamic
benefits from specialising in other sectors (see, in particular, the discussion in
Temple (1997)).
This paper has shown, in a model of endogenous technological change and
growth, that specialising according to current patterns of comparative advantage
may not be welfare maximising. Furthermore, in certain circumstances, policy
interventions to induce specialisation in sectors in which an economy does not
currently have a comparative advantage may be welfare increasing. This section
now considers whether there is a concept of ‘dynamic comparative advantage’,
that accords reasonably closely with popular usage, and that sheds light upon
the circumstances under which selective trade and industrial policies are welfare
increasing.
The traditional (or ‘static’) concept of comparative advantage is essentially
concerned with relative levels of opportunity costs of production in different sec-
tors of two economies. Thus, in our case, the home economy is said to have a
(‘static’) comparative advantage in low-tech production at time t if the opportu-
nity cost of producing the low-tech good at time t is lower in the home economy.
In this section, we propose a simple definition of ‘dynamic comparative advan-
tage’, concerned with changes over time in relative levels of opportunity costs.
Thus, the home economy is said to have a ‘dynamic’ comparative advantage in
23
low-tech production at time t if the rate of growth of the opportunity cost of
producing the low-tech good at time t is lower in the home economy. That is,
home will have a dynamic comparative advantage in low-tech production if and
only if,
∂ (A
h
(t)/A
z
(t)) /∂t
A
h
(t)/A
z
(t)
<
∂ (A

h
(t)/A

z
(t)) /∂t
A

h
(t)/A

z
(t)
(25)

˙
A
h
(t)
A
h
(t)

˙
A
z
(t)
A
z
(t)

˙
A

h
(t)
A

h
(t)

˙
A

z
(t)
A

z
(t)

< 0
This formalisation of dynamic comparative advantage is an extremely natural
one, that is the dynamic analogue of the traditional ‘static’ definition. Nonethe-
less, although natural, the definition has (as will be shown below) a surprising
amount of analytical content.
First, while static comparative advantage determines patterns of international
trade at a given point in time, dynamic comparative advantage explains changes
over time. In terms of the analysis of the previous section, the free trade equi-
librium involves home specialising in low-tech production (learning at the rate
g
f
z
> 0) and foreign in high-tech production (accumulating production experience
at the rate g
f∗
h
> 0). Hence, the opportunity cost of low-tech production in home
(A
h
/A
z
) falls over time, while the converse is true for the opportunity cost of
low-tech production in foreign (A

h
/A

z
). In terms of the definition (25), home
will have a dynamic comparative advantage in the low-tech sector, and its initial
static comparative advantage in this sector will be reinforced over time.
However, it is important to realise that dynamic comparative advantage, as
defined above, is not invariant to changes in patterns of international special-
isation. From (25), dynamic comparative advantage is completely determined
by productivity growth rates in each sector of the two economies. However, we
have already seen that these themselves are functions of patterns of international
specialisation. Again, the argument may be illustrated in terms of the analysis
of the previous section.
On the one hand, international specialisation in the free trade equilibrium
implies that home has a dynamic comparative advantage in the low-tech sector
24
(so that the initial pattern of static comparative advantage is reinforced over time,
as described above). On the other hand, under the subsidy, home specialises
in high-tech production (learning at the rate g
s
h
> 0) and foreign in low-tech
production (learning at the rate g
s∗
z
> 0). As a result, the opportunity cost of
low-tech production in home (A
h
/A
z
) rises over time, while the converse is true for
the opportunity cost of low-tech production in foreign (A

h
/A

z
). In terms of the
definition (25), home will actually have a dynamic comparative advantage in the
high-tech sector under the subsidy, and its initial static comparative advantage
in the low-tech sector will be reduced over time. Ultimately, the initial pattern
of static comparative advantage will be reversed; so that, were the subsidy to
be removed at a future point in time, home would continue to specialise in the
high-tech sector.
Thus, patterns of dynamic comparative advantage are very different under
the alternative regimes of free trade and the subsidy (as a result of differences in
international specialisation). This brings us to the link between dynamic compar-
ative advantage, as defined above, and the circumstances under which selective
trade and industrial policies may be welfare improving. As we have seen, an im-
portant element of the popular usage of this concept is the idea that a country’s
current pattern of static comparative advantage may work against its long-term
interests. Instead, it is argued that an economy should specialise in sectors where
it may enjoy various dynamic benefits (where it exhibits a ‘dynamic comparative
advantage’). It turns out that this informal argument may be straightforwardly
related to the formal analysis of the previous section.
Proposition 1 A necessary condition for a subsidy to the high-tech sector (where
the economy does not currently have a static comparative advantage) to be welfare-
improving is that the economy will (under the subsidy) acquire a static compara-
tive advantage in the high-tech sector at some future point in time t

> t
1
.
Proof. A necessary condition for the production subsidy to yield a higher level
of intertemporal welfare than under free trade is that at some future point in time
t

> t
1
home attains a higher level of instantaneous utility under the subsidy than
25
by switching to free trade. Otherwise, a policy-maker could unambiguously raise
intertemporal welfare by abandoning the subsidy.
In Appendix 2, we show that a necessary condition for instantaneous utility at
time t

to be higher under the subsidy is for home to have acquired a static
comparative advantage in the high-tech sector by t

(see Appendix)
In order for the representative agent’s intertemporal welfare to be increased
by a subsidy to a sector where the economy has no current static comparative
advantage, it must be true that (under the subsidy) a static comparative advan-
tage in this sector will be attained at some future point in time. However, a
necessary condition for the initial pattern of static comparative advantage to be
reversed in this way is that (under the subsidy) the home economy has a dynamic
comparative advantage (as defined in (25)) in the very sector in which it initially
has no static comparative advantage.
Thus, as suggested in informal discussions of the East Asian development
experience, one can indeed think of an economy potentially facing a trade-off be-
tween static and dynamic comparative advantage (or between current and future
patterns of static comparative advantage). Furthermore, the fact that the initial
pattern of static comparative advantage must be reversed for the subsidy to be
welfare improving implies that the selective trade or industrial policy need only
be temporary. If the subsidy were removed at time t

, home would (as discussed
above) continue to specialise in the high-tech sector.
However, although the reversal of static comparative advantage (and hence
having a dynamic comparative advantage in the subsidised sector) is a necessary
condition for the subsidy to be welfare improving, it is important to note that
it is not sufficient. Productivity growth rates and the way in which comparative
advantage evolves over time are themselves dependent upon patterns of interna-
tional specialisation (and, therefore, will be different under the subsidy and free
trade).
In Proposition 1, we compare instantaneous utility under the subsidy at each
point in time t

> t
1
with the level that could be achieved by abandoning the
subsidy at t

and engaging in free trade (taking as given productivity growth rates
26
over the interval of time t ∈ [t
1
, t

) - as determined by the pattern of specialisation
under the subsidy). This enables us to establish a necessary condition for the
subsidy to raise intertemporal welfare. However, in order to arrive at a sufficient
condition, two further steps must be taken. First, one must compare productivity
growth rates and the evolution of comparative advantage under the subsidy with
the corresponding values under free trade. In effect, this comparison involves an
evaluation of the (different) patterns of dynamic comparative advantage under
both the subsidy and free trade. Second, one must evaluate any dynamic welfare
gains from implementing the subsidy relative to the standard static welfare losses.
It was precisely such an analysis that was undertaken in the previous section.
8 Conclusion
This paper has considered the idea that developing economies may face a trade-off
between specialising according to an existing pattern of comparative advantage
(often in low-technology industries), and entering sectors where they currently
lack a comparative advantage, but may acquire such an advantage in the future as
a result of the potential for productivity growth (e.g. high-technology industries).
The analysis was undertaken within the context of a general equilibrium model
of endogenous growth, involving international trade between two large economies.
An essentially Ricardian model of international trade was combined with a model
of endogenous technological change in the form of learning by doing. Comparative
advantage - as traditionally defined (‘static’ comparative advantage) - depends
upon past technological advances, while simultaneously determining current rates
of learning by doing and technological change. Thus, comparative advantage itself
becomes endogenous.
Specialisation according to current comparative advantage results in the stan-
dard static gains from trade. However, if individual agents fail to fully internalise
the potential for productivity growth in each sector, it may also mean that an
economy fails to specialise in sectors where its potential for productivity growth
is large relative to its trading partners. As a result, free trade will induce dynamic
27
welfare losses. If sufficiently large, these may outweigh the standard static wel-
fare gains, so that trade reduces the intertemporal welfare of the representative
agent.
Selective trade and industrial policies to induce specialisation in sectors where
an economy currently lacks a comparative advantage, but exhibits a large po-
tential for productivity growth relative to its trading partner, may be welfare
improving. The case for these policies was related to the often-used, but as yet
ill-defined, notion of ‘dynamic’ comparative advantage. A natural formalisation
of this concept was suggested in terms of rates of growth of opportunity costs
of production in each economy. So defined, dynamic comparative advantage
explains the evolution of patterns of international trade over time and proves
informative in evaluating the case for interventionist public policies.
A necessary condition for a selective trade and industrial policy (of the form
suggested above) to be welfare improving is that the initial pattern of ‘static’
comparative advantage is reversed under the policy. However, the initial pattern
of static comparative advantage will only be reversed in this way if the economy
has a dynamic comparative advantage (under the proposed policy intervention) in
precisely the sector in which it initially has no static comparative advantage. This
necessary condition suggests that welfare improving selective trade and industrial
policies, if they exist, need only be temporary.
However, while this yields a necessary condition for interventionist public poli-
cies to be welfare improving, it does not, alone, provide a sufficient condition. In
order to evaluate whether subsidising production in a sector where no current
comparative advantage exists is welfare improving, two further steps are neces-
sary. First, one must compare productivity growth rates and the evolution of
comparative advantage under the subsidy with the corresponding values under
free trade. In effect, this comparison involves an evaluation of the (different) pat-
terns of dynamic comparative advantage under both the subsidy and free trade.
Second, one must evaluate any dynamic welfare gains from implementing the
subsidy relative to the standard static welfare losses.
Once the complete welfare comparison is undertaken, a theoretical case for
28
selective trade and industrial policies exists. Nonetheless, converting this theoret-
ical case into practical policy advice is more difficult. It involves an appreciation
of the large informational requirements of the theoretical argument, alongside po-
tential hidden welfare costs in the form of Directly Unproductive Profit-seeking
(DUP) activities. Nonetheless, developing economies with high levels of general
human capital, which may achieve rapid rates of productivity growth through
imitation, may be examples where a case for intervention exists.
29
References
Aghion, P and Howitt, P (1992) ‘A Model of Growth Through Creative Destruc-
tion’, Econometrica, 60, 323-51
Amsden, A (1989) Asia’s Next Giant: South Korea and Late Industrialisation,
Oxford University Press, New York
Bhagwati, J (1982) ‘Directly Unproductive Profit-Seeking (DUP) Activities’,
Journal of Political Economy, 90, 988-1002
Coe, D and Helpman, E (1995) ‘International R & D Spillovers’, European Eco-
nomic Review, 39, 859-87
Dornbusch, R, Fischer, S, and Samuelson, P (1977) ‘Comparative Advantage,
Trade, and Payments in a Ricardian Model with a Continuum of Goods’, Amer-
ican Economic Review, 67, 823-39
Enos, J and Park, W (1988) The Adoption and Diffusion of Imported Technology:
The Case of Korea, Routlege, London
Grossman, G and Helpman, E (1990) ‘Comparative Advantage and Long-Run
Growth’, American Economic Review, 80, 796-815
Grossman, G and Helpman, E (1991) Innovation and Growth in the Global Econ-
omy, MIT Press
Islam, N (1995) ‘Growth Empirics: A Panel Data Approach’, Quarterly Journal
of Economics, 110, 1127-70
Jones, C and Williams, J (1997) ‘Measuring the Social Return to R & D’, Quar-
terly Journal of Economics, Forthcoming
Krugman, P (1981) ‘Trade, Accumulation and Uneven Development’, Journal of
Development Economics, 8, 149-61
Krugman, P (1987) ‘The Narrow Moving Band, the Dutch Disease, and the
Competitive Consequences of Mrs Thatcher: Notes on Trade in the Presence of
30
Dynamic Scale Economies’, Journal of Development Economics, 27, 41-55
Krugman, P and Obstfeld, M (1994) International Economics, Third Edition,
Harper Collins
Lucas, R (1993) ‘Making a Miracle’, Econometrica, 61, 251-72
Matsuyama, K (1992) ‘Agricultural Productivity, Comparative Advantage and
Economic Growth’, Journal of Economic Theory, 58, 317-34
Redding, S (1996) ‘Endogenous Innovation and Economic Growth’, D. Phil. The-
sis, University of Oxford
Rivera-Batiz, L and Romer, P (1991a) ‘International Trade with Endogenous
Technological Change’, European Economic Review, 35, 971-1001
Rivera-Batiz, L and Romer, P (1991b) ‘Economic Integration and Economic
Growth’, Quarterly Journal of Economics, 106, 531-55
Sachs, J and Warner, A (1995) ‘Natural Resource Abundance and Economic
Growth’, NBER Working Paper, No. 5398
Stokey, N (1991) ‘Human Capital, Product Quality, and Growth’, Quarterly Jour-
nal of Economics, 106, 587-616
Temple, J (1997) ‘St Adam and the Dragons: Neo-classical Economics and the
East Asian Miracle’, Oxford Development Studies, 25, 279-300
Taylor, M (1991) ‘Quality Ladders and Ricardian Trade’, Journal of International
Economics, 34, 225-43
Taylor, M (1994) ‘Once-off and Continuing Gains From Trade’, Review of Eco-
nomic Studies, 61, 589-601
Wade, R (1990) Governing the Market: Economic Theory and the Role of Gov-
ernment in East Asian Industrialisation, Princeton University Press
Young, A (1991) ‘Learning by Doing and the Dynamic Effects of International
Trade’, Quarterly Journal of Economics, 106, 369-405
31
9 Appendix
9.1 Relative prices under free trade and the subsidy
Under free trade, inequality (16) and equation (12) jointly imply that, at time t
1
,
A
z
(t
1
) > β/(1 −β).
¯
L

/
¯
L.A

z
(t
1
) (26)
A

h
(t
1
) > (1 −β)/β.
¯
L/
¯
L

.A
h
(t
1
) (27)
Under the subsidy, inequality (18) and equation (19) jointly imply that, at
time t
1
,
A
z
(t
1
) < (1 + s).(1 −β)/β.
¯
L

/
¯
L.A

z
(t
1
) (28)
A

h
(t
1
) < β/(1 −β).
¯
L/
¯
L

.A
h
(t
1
) (29)
For values of β ∈ (1/2, 1), inequalities (27) and (29) may both be satisfied..
At the same time, for sufficiently large values of s, inequalities (26) is compatible
with (28) and with both (27) and (29). Hence, the assumption of complete
specialisation under both free trade and protectionism is validated.
9.2 Proof of Proposition 1
From equations (21) and (10), instantaneous welfare at any time t

could be
increased by abandoning the subsidy and moving to free trade if and only if,
[βA
h
(t

)]
β
[(1 −β)A
h
(t

)]
1−β
<
¸
β
p
s
z
(t

)
p
s
h
(t

)
A
z
(t

)
¸
β
¸
(1 −β)A
z
(t

)
p
f
z
(t

)
p
f
h
(t

)
¸
1−β
(30)
In any equilibrium, in which home has a static comparative advantage in
low-tech production and a subsidy is required to induce it to specialise in the
high-tech sector, inequalities (16) and (18) jointly imply,
(1 + s).A
h
(t

)
A
z
(t

)
>
p
s
z
(t

)
p
s
h
(t

)
>
A

h
(t

)
A

z
(t

)
>
p
f
z
(t

)
p
f
h
(t

)
>
A
h
(t

)
A
z
(t

)
(31)
32
Thus, A
h
< p
f
z
/p
f
h
.A
z
< p
s
z
/p
s
h
.A
z
. As long as home has a static comparative
advantage in the low-tech sector and the subsidy is required to induce it to
specialise in the high-tech sector, inequality (30) must be satisfied.
Suppose instead, that home acquires a static comparative advantage in the
high-tech sector and a subsidy is no longer required to induce specialisation in
this sector. In this case,
(1 + s).A
h
(t

)
A
z
(t

)
>
A
h
(t

)
A
z
(t

)
>
p
s
z
(t

)
p
s
h
(t

)
=
p
f
z
(t

)
p
f
h
(t

)
>
A

h
(t

)
A

z
(t

)
(32)
and A
h
> p
f
z
/p
f
h
.A
z
= p
s
z
/p
s
h
.A
z
. Hence, inequality (30) is no longer satisfied.
It follows immediately that a necessary condition for the subsidy to yield a
higher level of instantaneous utility at some point in time t

is for home to have
acquired a static comparative advantage in this sector by time t

.
33

1

Introduction

A study by the World Bank in the 1960s “expressed the view that an integrated steel mill in Korea was a premature proposition without economic feasibility.” (Pohang Iron and Steel Co. Ltd (1984), p. 23, cited in Amsden (1989)). A number of factors, including Korea’s deficiency in the required raw materials and its small domestic market for such a scale-intensive industry, suggested that steel making was an industry in which Korea was unlikely to have a comparative advantage.1 Nonetheless, in 1973, the Korean government founded the Pohang Iron and Steel Company Ltd. (POSCO) with an initial investment of $ 3.6bn. Government assistance in a wide variety of forms, including subsidisation of the cost of capital and investments in infrastructure has been central to POSCO’s development. The company soon became one of the lowest cost steel-producers in the world so that, in 1985, Korea unit costs of production were less than those of Japan and around 2/3 of those in the United States (Amsden (1989), Table 12.2). By 1988, POSCO had become the eleventh largest steel company in the world, operating 80 individual plants (Enos and Park (1988)). Although at the time POSCO was founded Korea did not appear to have a comparative advantage in the iron and steel industry, it seems incontrovertible that it now does and that the Korean government has played a central role in its acquiring one. This paper investigates the idea that developing economies may face a trade-off between specialising according to an existing pattern of comparative advantage (often in low-technology industries) and entering sectors in which they currently lack a comparative advantage, but may acquire such an advantage in the future as a result of the potential for productivity growth (e.g. high-technology industries). We analyse the circumstances under which the actions of private sector agents will resolve this trade-off between current and future patterns of comparative advantage optimally. If the trade-off is not resolved optimally, then it becomes possible for free trade to be welfare reducing. Moreover, protectionist measures that induce specialisation in sectors where one
See, for example, (Amsden, 1989, Chapter 12) on which the first two paragraphs of this Section draw.
1

2

does not currently have a comparative advantage may be welfare increasing. This paper investigates these ideas within a general equilibrium model of endogenous growth, in which an economy’s pattern of international trade and rate of economic growth are jointly and endogenously determined. The paper is part of a wider literature concerned with relationship been trade and growth. For example, Krugman (1981) examines the effect of international trade upon the world distribution of income when there are external economies to physical capital accumulation in the manufacturing sector. An early formalisation of the interrelationship between patterns of international trade and rates of technological change (although there is no welfare analysis) is provided by Krugman (1987). As a result of this interaction, initial patterns of international trade become increasingly “locked-in” over time. More recently, (Grossman and Helpman, 1990, 1991), (Rivera-Batiz and Romer 1991a, 1991b) and (Taylor 1991, 1994) have examined the relationship between trade and growth, when endogenous growth is the result of profit-seeking investments in Research and Development (R & D).2 Young (1991) analyses the links between trade and growth, when bounded learning by doing leads to the adoption of new varieties of goods; while Stokey (1991) examines the interaction between trade and human capital accumulation. In small open economy models, Matsuyama (1992) and Sachs and Warner (1995) respectively consider the effects of levels of agricultural productivity and endowments of natural resources on international trade and growth. In fact, the existing literature suggests a number of channels through which trade may affect an economy’s rate of growth. In this paper, motivated by the empirical discussion above, we focus upon the relationship between endogenous comparative advantage, economic growth and economic welfare. In contrast to much of the literature, which has emphasised the beneficial effects of trade on growth, the analysis suggests that specialisation according to initial comparative advantage may have negative effects upon both rates of growth and economic
Where these investments may either yield new varieties or (as in Aghion and Howitt (1992)) successively higher qualities of intermediate inputs.
2

3

Furthermore. Motivated by the earlier empirical discussion of the East Asian development experience (see also Amsden (1989) and Wade (1990)). Amsden (1989)). 1991). First. (Grossman and Helpman. we examine the endogeneity of comparative advantage within a particularly tractable. Second. the role of endogenous comparative advantage is made clear. Krugman (1987). may be welfare improving. and entering sectors where it currently lacks a comparative advantage. but may acquire such an advantage as a result of the potential for productivity growth. We are able to derive necessary and sufficient conditions for free trade. large economies. on the basis of the theoretical analysis of the relationship between international trade and economic growth. the paper emphasises the potential trade-off an economy may face between specialising according to an existing pattern of comparative advantage.’ This very same term appears in more informal discussions of the East Asian development experience (see. in particular. to be welfare reducing. This paper’s second objective is therefore to see whether. any substantiative content can be given to this often-used. the endogeneity of comparative advantage in models of growth and trade has led a number of authors in the theoretical literature (see. This paper makes two main contributions.welfare. general equilibrium model of endogenous growth and international trade between two. 1990. The tractability of this framework enables us to undertake a complete analysis of the welfare effects of international trade and the potential case for selective trade and industrial policies. we establish the circumstances under which selective trade and industrial policies. for example. by inducing specialisation according to current patterns of comparative advantage. Krugman (1987) and Grossman and Helpman (1991)) to speak in terms of ‘dynamic comparative advantage. that induce specialisation in sectors where an economy does not currently have a comparative advantage. The analysis builds upon a number of existing studies of endogenous comparative advantage and growth (see. but so far ill-defined concept. and Young (1991)). for example. when comparative advantage is endogenous in dynamic trade models. Throughout the analysis. the traditional (or ‘static’) notion of com4 . The paper suggests that.

if such policies exist. while section 3 solves for static equilibrium under both autarky and free trade. a standard Ricardian model of international trade (see. where all foreign variables are denoted by an asterisk. The standard static gains from trade are augmented with dynamic effects. Section 2 introduces the model. they need only be temporary. section 8 concludes.’ The popular notion is formalised and its relationship to the preceding analysis discussed. for example. The paper is structured as follows. Section 5 considers the implications of endogenous comparative advantage for the welfare effects of international trade. textiles) and a high-technology. This dynamic concept explains the evolution of patterns of international trade over time and sheds light upon the circumstances under which welfare improving selective trade and industrial policies exist. and shows how comparative advantage is endogenously determined. and the See Dornbusch et al. manufacturing. agriculture. traditional good z (e. 3 5 .3 Labour is the sole factor of production. Each economy may produce two final goods. Section 4 is concerned with the relationship between trade and productivity growth. We consider international trade between two economies (‘home’ and ‘foreign’). frontier good h (e. but distinct question whether selective trade and industrial policies to induce entry into a sector where an economy currently lacks a comparative advantage may be welfare improving. Productivity in each sector is assumed to evolve endogenously over time as learning by doing occurs. Section 6 addresses the related. a low-technology. which may either increase or decrease the intertemporal welfare of the representative agent.parative advantage may be usefully augmented with a second ‘dynamic’ concept. Section 7 moves on to consider the popular notion of ‘dynamic comparative advantage. 2 A dynamic Ricardian model In this section. electronics). Finally. Interestingly. Krugman and Obstfeld (1994)) is augmented with a specification for productivity dynamics.g. (1977) for an exposition of the static Ricardian model with a continuum of goods.g.

Aggregate output in each sector is thus. shipyards during World War II was associated with a reduction of man-hours required per ship by between 12 and 24 per cent. and we make the standard assumption that labour is perfectly mobile between sectors and immobile across countries. at each point in time. for j = z. except where it is important. h. expenditure equals income for the representative consumer. Time is continuous and is indexed by t. Hence. Home labour market clearing requires Lz + Lh = ¯ L. In order to simplify notation.4 Intertemporal utility is the z h sum of instantaneous utilities. lower case letters are used for per capita variables. 4 6 .Lh (1) Production is assumed to occur under conditions of perfect competition.¯ two economies are populated with a number of representative consumers (L and ¯ L∗ ).1 The static model Consumer preferences are assumed to be identical in the two economies. ch ) = cβ c1−β where 0 < β < 1. 2. For example. Yz = Az . which is supplied inelastically with zero disutility. Yh = Ah . we assume that productivity in each sector Aj depends upon a stock of sector-specific In general.S. whose productivity we index by Aj . For simplicity. discounted at the subjective rate of time preference ρ.and high-tech goods are produced with labour Lj according to constant returns to scale technologies. Lucas (1993) cites evidence that each doubling of cumulative output of “Liberty Ships” in 14 U. we assume that there is no storage or savings technology so that.Lz . Each consumer is endowed with one unit of labour. we suppress an implicit dependence upon time. 2.and high-tech goods: u(cz . following Krugman (1987). Low.2 Productivity dynamics A wide range of empirical evidence suggests that learning by doing is an important source of productivity improvements. with instantaneous utility a Cobb-Douglas function of consumption of the low.

It is possible to introduce international knowledge spillovers. as long as these are imperfect. agents acquire productivity-enhancing production experience Kj (or learn by doing): for example. The rate at which this production experience is acquired is assumed to depend upon the flow of labour employed in producing a sector’s output. for example. new methods of manufacture or new ways of organising existing processes are discovered. Islam (1995)). political institutions and laws. where ψj > 0 for j = z. as well as exogenous factors ψj such as climate. 5 7 . this only complicates the analysis without adding additional insight. for example.Lh (t).Lz (t).Kh (t).production experience Kj . through trial and error. so that Kj evolves according to. learning by doing is assumed to be external to individual firms but specific to a sector and to an economy. Coe and Helpman (1995)) and that levels of Total Factor Productivity differ substantially across economies (see. While producing output in a given sector.Kh (t) (2) ˙ Kh (t) = µh . µh > 0 (4) where µj parameterises the rate at which knowledge is acquired as part of the production process in sector j. Learning by doing is assumed to be a pure externality of the production process: in particular. in which case comparative advantage is not only endogenous but the rate at which it evolves over time is a function of cumulative production experience. µz > 0 (3) Ah (t) = ψh . the analysis remains essentially unchanged. 6 There is considerable empirical evidence that international knowledge spillovers are imperfect (see. It is possible to extend the analysis to the case of bounded learning by doing (see Redding (1996)). h. However.6 According to the specification in (3) and (4). learning by doing is technologically unbounded.Kz (t).5 ˙ Kz (t) = µz . However. culture. Az (t) = ψz .Kz (t).

Autarkic equilibrium is characterised by incomplete specialisation. With free trade and zero transport costs.and high-tech goods exhibit a demand for variety.1 Static equilibrium Autarky Autarkic equilibrium is fully characterised by the requirement that the relative price of the low-tech good equals both minus the Marginal Rate of Substitution (MRS) and minus the Marginal Rate of Transformation (MRT) between lowand high-tech goods.7 For home to specialise completely in the low-tech sector and foreign in the high-tech.3 3. we require β/(1 − β).2 Free Trade In this subsection.Ch /Cz = pz /ph = Ah /Az . if specialisation in home is incomplete.and high-tech sectors respectively. Throughout the following. we allow two previously autarkic economies to engage in free trade from some arbitrary point in time t1 onwards.L (5) 3. A∗ (t) pz (t) Ah (t) h > > ∗ (t) Az ph (t) Az (t) 7 (6) This is merely for simplicity. 8 .and high-tech goods must be the same in each economy. ¯ Lh = (1 − β). where. with labour allocated in the constant proportions β and (1 − β) to the low. we require wz > wh ∗ ∗ and wh > wz . we will be largely concerned with equilibria characterised by complete specialisation in both economies.pz (t) and wh (t) = Ah (t). From the expression for instantaneous utility and (1).ph (t) respectively. It is straightforward to extend the analysis to cases of incomplete specialisation. the price of the low.L.and hightech sectors will equal wz (t) = Az (t). we require wz = wh = w. Perfect competition implies that the home wage in the low. That is. Consumer preferences over the low. ¯ Lz = β.

RDzh = ∗ ph Cz + Cz β . We will largely be concerned with equilibria in which home has an initial comparative advantage in the low-tech sector. Equation (6) and the associated allocation of labour to the two sectors defines the world supply of the low. it is clear that the pattern of comparative advantage at any point in time depends upon productivity in each sector of the two economies (Aj . = ∗ Ch + Ch (1 − β) pz (7) General equilibrium of the static model may be fully characterised in relative supply. an economy is said to have a ‘static comparative advantage’ in the low-tech sector at time t if the opportunity cost of producing the low-tech good at home is lower than in the other economy.and high-tech sectors in the constant proportions β and (1 − β) at each point in time t. Throughout the paper. the entire of home’s supply of labour is employed in the low-tech sector and the entire of foreign’s in the high-tech sector.relative to the high-tech good: RSzh = (Yz + Yz∗ )/(Yh + Yh∗ ). from (6).In such an equilibrium. as shown in Figure 1. ‘comparative advantage’ will be used in this traditional. Figure 1 about here 4 Endogenous comparative advantage The pattern of international trade in the static Ricardian model is determined by the traditional or ‘static’ notion of comparative advantage. From equation (8). With Cobb-Douglas instantaneous utility. each representative consumer allocates expenditure to the low. Thus. World demand for the low-tech relative to the high-tech good is thus. ‘static’ sense unless otherwise specified. this is a necessary condition for home to specialise in the lowtech sector in the free trade equilibrium. relative demand space. Ah (t) A∗ (t) < h Az (t) A∗ (t) z (8) where. 9 .

The endogeneity of comparative advantage in models of growth and trade has led a number of authors to speak in terms of ‘dynamic comparative advantage’. Thus. 8 9 Where the superscript n (‘no trade’) indexes the value of a variable under autarky. which affects the allocation of labour between sectors. although. as yet. first. This in turn determines relative rates of productivity growth. from ¯ (3).L).β. 10 . and thereby feeds back to shape the evolution of productivity levels over time.(1 − β).L and z ¯ = µh . Thus.A∗ . (4) and (5).9 h Specialisation according to comparative advantage under free trade changes the (endogenous) rate of productivity growth in each sector of the two economies. j = z. A later section comes back to discuss this idea. h). This.8 Similarly. j = z. accumulates production experience at the rates g n = µz . home is incompletely specialised in both sectors and. in the free z z h h n gh trade equilibrium. while foreign learns z by doing in the high-tech sector (at the rate f gh ∗ ¯ = µ∗ . at the same time.L∗ and g n∗ = µ∗ . In contrast. comparative advantage depends upon past technological change in each sector. for ¯ ¯ foreign we have g n∗ = µ∗ . Productivity levels dictate comparative advantage. in turn. home learns ¯ by doing in the low-tech sector alone (at the rate g f = µz .L∗ ).(1 − β). while.and high-tech sectors respectively. determining the free trade allocation of labour across sectors and hence (from (3) and (4)) rates of productivity growth in each sector. Where the superscript f (‘free trade’) indexes the value of a variable under free trade. is a determined by a combination of exogenous factors j on the one hand (such as climate. current comparative advantage is endogenously determined. this concept has remained ill-defined. Under autarky. home’s comparative advantage in the low-tech sector means that it specialises completely in the production of this good. h). In this way. h) and past technological change on the other (as manifested ∗ in the stocks of cumulative production experience Kj and Kj . political institutions and laws as parameterised ∗ by ψj and ψj . we move on to consider the implications of endogenous productivity growth and endogenous comparative advantage for the welfare effects of trade. However.β.L∗ . j = z.L in the low.

5.1 Welfare under autarky Beginning with autarky. Using z z h h the fact that constant proportions of income are allocated to expenditure on the low.An (t) = pn (t). under both autarky and free trade. In each case.5 Trade and welfare In this section.2 Welfare under free trade In contrast. since instantaneous utility is Cobb-Douglas.Af (t).pf (t)/pf (t) z z h 1−β dt (10) where the relative price of the low-tech good is determined on world markets. we obtain the following expression for intertemporal welfare.and high-tech goods respectively.Af (t) z z for all t ≥ t1 . Utf1 = ∞ t1 e−ρ(t−t1 ) β. Utn = 1 ∞ t1 e−ρ(t−t1 ) [β. h under autarky j at each point in time t ≥ t1 . we arrive at an analogous expression for intertemporal welfare.Af (t) z β (1 − β). the representative consumer will allocate the constant proportions β and (1 −β) of his/her expenditure at each point in time t to the low. Furthermore.An (t)]β [(1 − β). under free trade. it follows that. specialisation is complete and the representative consumer’s income is equal to the wage in the low-tech sector w(t) = pf (t). specialisation is incomplete and the representative consumer’s income is given by the wage w(t) = pn (t). 11 . Again using the fact that consumer expenditure is allocated in constant proportions. we compare the representative consumer’s intertemporal welfare under the alternative regimes of remaining autarkic from time t1 onwards and engaging in free trade. 5. intertemporal welfare is given by the discounted sum of instantaneous utilities.An (t).An (t)]1−β dt z h (9) where An (t) denotes the level of productivity in each sector j = z.and high-tech goods.

BeOf course. Instantaneous utility will be lower under autarky if and only if pf (t1 )/pf (t1 ) > Ah (t1 )/Az (t1 ). using a theoretical framework drawing on the endogenous growth literature. These reallocations of resources affect rates of learning by doing and productivity growth in each sector of the two economies. there are also reasons why the laissez-faire rate of technological change may be higher than is socially optimal (e. it is immediately clear that instantaneous utility must be lower under autarky than under free trade at time t1 when the choice between the two regimes must be made. 5. in general. Jones and Williams (1997) review the empirical evidence and find that social rates to Research and Development (R & D) typically exceed the corresponding private rates of return. the literatures on both the microeconomics of technological change and endogenous growth suggest a number of reasons why the laissez-faire rate of technological change may be less than the socially optimal rate. technological change takes the form of serendipitous learning by doing. 10 12 . dynamic welfare effects. From these two equations. these equalities will not. The existence of the standard static gains from trade (from specialisation according to comparative advantage) implies that instantaneous utility must be initially lower under a regime of autarky than one of free trade. Second. hold for t > t1 ).3 Static gains from trade The right-hand sides of equations (9) and (10) contain information about levels of instantaneous utility at all points in time t ≥ t1 .10 In the present case. two aspects of the analysis are important. as discussed in the previous section. h h This condition must be satisfied in a free trade equilibrium in which home has a static comparative advantage in and specialises in low-tech production. where we use the fact that An (t1 ) = Af (t1 ) and z z z h An (t1 ) = Af (t1 ) (clearly.5. specialisation according to comparative advantage leads to reallocations of resources between the low. the ‘business stealing effect’ and the ‘monopoly distortion effect’ in Aghion and Howitt (1992)). Here. However. the fact that technological change is endogenous means that a move from autarky to free trade has additional. First.and high-tech sectors. and hence have dynamic effects on economic welfare.4 Dynamic effects and intertemporal welfare However.g.

L). we obtain the following expression for intertemporal welfare under autarky.L∗ ). complete specialisation implies that home only f ¯ experiences learning by doing in the low-tech sector (at the rate gz = µz . before we can solve z explicitly for intertemporal welfare.Az (t1 ). This section takes the analysis one stage further to consider the implications of changes in productivity growth for economic welfare. Beginning with autarky. Substituting for levels of productivity in the two sectors in (9) and evaluating the integral. we require an expression for the equilibrium More generally.Az (t1 ) and An (t) = egh (t−t1 ) . [Az (t1 )]β .12 Utn 1 β β (1 − β)1−β . [Ah (t1 )]1−β = n n ρ − βgz − (1 − β)gh (11) n n Turning now to free trade. However. Combining the dynamic welfare effects of international trade with the standard static gains from trade enumerated above.β.L and g n = µh .and high-tech sectors at the rates ¯ ¯ g n = µz .11 As a result. the same will be true in models of endogenous technological change through profit-seeking Research and Development (R & D).cause technological change is a positive externality of current production. productivity levels in z h the two sectors equal An (t) = egz (t−t1 ) .gz + (1 − β).L respectively. private sector agents do not fully take into account the potential for productivity growth in each sector. agents fail to internalise the changes in rates of productivity induced by international specialisation and the consequent dynamic effects on economic welfare.Ah (t1 ) respecz h tively for all t ≥ t1 . incomplete specialisation implies that home experiences learning by doing in both the low. Hence.(1 − β).gh . we evaluate relative levels of intertemporal welfare for the representative agent under autarky and free trade. 12 Where for the integral in (9) to converge and intertemporal utility to be finite. The effects of specialisation according to comparative advantage on productivity growth rates were the subject of the previous section. Hence. while foreign only enjoys learning by doing in the high-tech sector (at the rate ¯ g f ∗ = µ∗ . from (10). 11 f 13 . we require n n ρ > β. home productivity in the low-tech sector in (10) may be h h expressed as Af (t) = egz (t−t1 ) . as long as the social rate of return to R & D exceeds the private rate of return (see an appendix available from the author on request and Redding (1996)).

home has a comparative advantage in low-tech production at time t1 . Substituting for pz (t1 )/ph (t1 ) ¯ ¯ from (12).free trade relative price of the low-tech good. not only upon levels of instantaneous utility at time f f Where. we obtain z h the following expression for intertemporal welfare under free trade.pz (t1 )/ph (t1 ) > Ah (t1 ).gh ∗ . To confirm this.L. [Az (t1 )]β [A∗ (t1 )]1−β L∗ /L h f f ρ − βgz − (1 − β)gh ∗ 1−β Utf1 = (13) From equations (13) and (11).A∗ (t1 ) and Af (t) = egz (t−t1 ) . intertemporal welfare under free trade will be lower than under autarky if and only if. Af ∗ (t) = egh h f∗ (12) f (t−t1 ) . we obtain [A∗ (t1 )] .and high-tech sec¯ ¯ tors respectively.Az (t1 ) for h z all t ≥ t1 .gz + (1 − β). Hence. [Ah (t1 )] . must be higher under free trade.14 However. whether intertemporal welfare will be higher under free trade than under autarky will depend. we require ρ > β. The numerators in equations (13) and (11) are. from the above. Az (t1 ). substituting for the productivity levels Af (t) and Af ∗ (t). Hence. when the choice between the two regimes is made. for intertemporal utility to be finite. With home and foreign specialising completely in the low. h β 13 14 14 . ¯ pf (t) β Af ∗ (t) L∗ z h = ¯ 1 − β Af (t) L pf (t) z h where. z h while relative demand is determined according to (7). and evaluating the integral.13 ¯ ¯ β. the free trade equilibrium relative price of the low-tech good equals.L∗ > 1−β . Using this equation for equilibrium relative prices in (10). Hence. simply instantaneous utility at time t1 under the two regimes. ¯ [A∗ (t1 )]1−β L∗ h 1−β 1−β 1−β β f f ρ−βgz − (1 − β)gh ∗ < ¯ [Ah (t1 )]1−β L1−β n n ρ−βgz −(1 − β)gh (14) As we saw in the last subsection. note that. the relative supply of the low-tech good is simply Af (t)L/Af ∗ (t)L∗ . in fact. by assumption. it follows immediately that the numerator on the left-hand side of the inequality (14) must exceed the numerator on the right-hand side. the existence of the standard static gains from trade means that instantaneous welfare at the time t1 .

L∗ ). the domestic rate of productivity growth falls ¯ from g n = µh .L < 0 h (15) The first of the three dynamic effects identified above is unambiguously posif n tive (gz > gz ). but also upon the rate of growth of instantaneous utility from time t1 onwards.µh .µz . Here. home experiences the benefits of foreign learning by doing in this sector in the form (see equation (12)) of a terms of trade gain (where. This is essentially a ‘scale effect’ of international trade. However. Third.L). ¯ ¯ ¯ β. specialisation according to comparative advantage induces three dynamic effects upon intertemporal welfare. The increase in home’s rate of learning by doing in the low-tech sector brought about by free trade (reflecting increased employment in this sector) raises intertemporal welfare relative to that under autarky. First. the rate of growth of instantaneous utility will be lower under free trade if f f n n and only if βgz + (1 − β)gh ∗ < βgz + (1 − β)gh . as reflected in the strictly positive first term on the left-hand side of the inequality (15). although forgoing its own potential h h to learn by doing in the high-tech sector under free trade. foreign’s rate of learning by ¯ doing in the high-tech sector is g f ∗ = µ∗ . whereby trade expands the size of the market for home low-tech goods. Substituting for the equilibrium rate of productivity growth in each sector under the two regimes.L ) either may or may not exceed foreign’s under free trade h f (gh ∗ ¯ = µ∗ . as a result.t1 . the reallocation of labour to home’s low-tech sector induced by specialisation under free trade raises home’s rate of learning by doing in this sector (from ¯ ¯ g n = µz . Foreign allocates its entire labour force to high-tech production h 15 .β L to g f = µz .L to g f = 0). home’s rate of learning by doing in the high-tech sector under au¯ tarky (g n = µh . we obtain (from the above) the condition.(1 − β).L + µ∗ .(1−β). under free trade. h h The effect of free trade on the rate of growth of instantaneous utility will depend upon the net outcome of these three dynamic effects.L∗ ). Second. In terms of inequality (14). specialisation in the low-tech sector under z z free trade means that home forgoes its own potential to learn by doing in the high-tech sector (where.L∗ − (1 − β).

under free trade. In order for intertemporal welfare to fall as a result of moving from autarky to free trade. this will occur whenever inequality (14) is satisfied (which thus provides a necessary and sufficient condition for free trade to be welfare reducing). the larger home’s potential (µh ) for learning by doing in the high-tech sector relative to foreign’s (µ∗ ) and the smaller the foreign h ¯ ∗ ). The existence of the dynamic effects of international trade on economic welfare (due to the change in rates of productivity growth induced by specialisation according to comparative advantage) means that free trade is no longer necessarily welfare increasing. upon the relative values of home’s potential to learn by doing in the low. From (15). Nonetheless. the net impact h of the second and third dynamic effects on intertemporal welfare may be either positive or negative. and this is reflected in the ambiguous term in parentheses in inequality (15). we require inequality (15) to be satisfied. From equations (13) and (11). The effect of the size of the home economy (as measured by its labour force L ¯ economy (as measured by its labour force L) and the share of consumer expenditure devoted to low-tech goods (β) is ambiguous. and that the effect of this on the rate of growth of instantaneous utility exceeds that of the increase in the rate of learning by doing in the low-tech sector. the relative magnitude of the two rates of learning by doing also ¯ depends upon the size of the two economies (as measured by the labour forces L ¯ and L∗ ) and the potential for learning by doing in the high-tech sector in each of the two economies (as determined by µh and µ∗ ). while home only allocates a proportion (1 − β) under autarky. That is. and depends.and high-tech sectors (µz and µh respectively). 16 . Therefore. this necessary condition for free trade to be welfare reducing is more likely to be satisfied. we require that these dynamic welfare losses from free trade exceed the standard static gains from trade. A necessary condition for free trade to be welfare reducing is that the rate of learning by doing in the high-tech sector is lower under free trade than under autarky. for example. as in standard static theories of international trade.

Another way of thinking about the issue is as follows. The pattern of comparative advantage at any one point in time is. Could it ever be optimal for the policy-maker to try to reverse this initial pattern of international specialisation ? Clearly. the static gains from trade are augmented with a number of dynamic welfare effects and international trade is no longer necessarily welfare increasing. Given this initial pattern of specialisation. the answer to this further question will depend upon both the economies’ potential rates of learning by doing in the sector where they specialise under the proposed policy intervention and the corresponding rates of productivity growth in the sector where they specialise under the alternative of free trade. endogenous. as earlier noted. when technological change is endogenous. Ascertaining whether the policy intervention is welfare improving relative to free trade involves an evaluation of productivity dynamics and the time path for comparative advantage 17 . then rates of productivity growth and the time path for comparative advantage could be very different indeed. However. Free trade will reduce the intertemporal welfare of the representative agent if an economy’s initial pattern of (static) comparative advantage means that it fails to specialise in a sector in which its potential to learn by doing is large relative to its trading partner’s. the free trade equilibrium is characterised by home specialising in the low-tech and foreign in the high-tech sector. if the initial patterns of comparative advantage and international specialisation were otherwise (as. productivity growth rates and the evolution of comparative advantage over time are then determined. the result of a policy intervention). This immediately raises a further question: could it ever be optimal for a policy-maker to induce an economy to specialise in the sector where it does not currently have a comparative advantage but exhibits considerable potential to learn by doing ? In terms of the analysis of the previous section. In the free trade equilibrium. for example.6 Policy intervention The previous section has shown that. the home economy’s initial pattern of comparative advantage leads it to specialise in one way.

and the equilibrium price of the low-tech good at time t1 is determined by equation (12). The comparison between free trade and the subsidy seems empirically the most relevant. it is straightforward to extend the analysis to compare welfare under the subsidy with that under autarky. being fully funded by a tax ξ. For each unit of income earned in the high-tech sector. In this section. individuals are assumed to receive a production subsidy of monetary value s > 0. and home will now specialise in the high-tech sector under international trade.15 As before. (1 + s) . However. Consider now a policy intervention of subsidising production in the hightech sector. z z s wh (t) = (1 + s) (1 − ξ) ps (t)As (t) h h (17) s s For a sufficiently large value of the production subsidy s. The initial pattern of comparative advantage is reversed. we compare intertemporal welfare under free trade (as evaluated above) with intertemporal welfare under a policy of subsidising entry into the high-tech sector. A∗ (t1 ) Ah (t1 ) pf (t1 ) h > > z (16) f ∗ (t ) Az 1 Az (t1 ) ph (t1 ) In the free trade equilibrium. Thus. 16 Where the superscript s indexes the value of a variable under the subsidy. home and foreign specialise in the low. home is assumed to have an initial comparative advantage in the low-tech sector and we will be concerned with equilibria characterised by complete specialisation. where the home economy does not currently have a comparative advantage. 15 18 .and high-tech sectors respectively.and high-tech sectors are thus.16 s wz (t) = (1 − ξ) ps (t)As (t).Ah (t1 ) ps (t1 ) A∗ (t1 ) z > s > h (18) Az (t1 ) ph (t1 ) A∗ (t1 ) z where the relative price of the low-tech good at time t1 under the subsidy is now. That is.under each of the alternative regimes. The after-tax/after-subsidy wages in the low. 0 < ξ < 1. and again we restrict consideration to equilibria characterised by complete specialisation. wh (t) > wz (t). The subsidy is assumed to be self-financing. on wage income.

Uts1 = ∞ t1 e−ρ(t−t1 ) [β. 1). h h ˆ ξ= s 1+s (20) General equilibrium under free trade and the subsidy. we restrict attention to the case of complete specialisation.As (t)]β [(1 − β).ps (t). Under the production subsidy. home specialises in the high-tech sector and foreign in the low-tech. the representative agent allocates expenditure in constant proportions to each sector. we require s. ¯∗ ps (t1 ) 1 − β Az (t1 ) L h (19) Appendix 1 shows that inequalities (16) and (18). Again. Figure 2 about here The existence of the standard static gains from trade means that instantaneous utility at the time t1 .As (t)]1−β dt h z h h (21) where the relative price of the high-tech good ps (t)/ps (t) is determined according h z to equation (19). Replacing the tax rate ξ with its equilibrium value ξ in (17). The assumption of complete specialisation under both free trade and the subsidy is thus validated. we may solve for intertemporal welfare under the subsidy. is illustrated diagrammatically in Figure 2.Ah (t) = ξ. (12) and (19). we obtain an expression for disposable income.Ah (t). it is straightforward to extend the analysis to consider incomplete specialisation. (1 + s) . 17 19 . in equilibrium.ps (t). with the accompanying change in the pattern of comparative advantage and international specialisation.17 For the production subsidy to be self-financing. and the equilibrium tax rate is thus. ∗ . The representative agent’s income in the home economy is given by the after-tax/after-subsidy wage ˆ in the high-tech sector. For simplicity.¯ ps (t1 ) β Ah (t1 ) L z = . and the two equations for the relative price of the low-tech good. may be simultaneously satisfied for sufficiently large values of s and β ∈ (1/2. when the choice between the two regimes is made. Using this expression and the fact that.ps (t)/ps (t).

[Ah (t1 )]1−β . complete specialisation implies that home ex¯ periences learning by doing in its high-tech sector (at the rate g s = µh .L∗ ). Thus. Under the subsidy.must be lower under the production subsidy (the home economy is choosing to specialise in a sector in which it has a comparative disadvantage). Hence. L∗ /L z s s∗ ρ − βgz − (1 − β)gh β Uts1 = (22) From equations (22) and (13). it follows immediately that the This is established formally in the proof of a later Proposition (see the proof of Proposition 1 in Appendix 2). 19 s∗ s Where. as in the comparison between autarky and free trade. productivity levels in the high-tech and low-tech sectors z in home and foreign respectively may be expressed as As (t) = egh (t−t1 ) . [A∗ (t1 )]β . The numerators in equations (22) and (13) are simply instantaneous utility at time t1 under the two regimes.L). Under the production subsidy. then substituting for productivity levels in the two sectors and evaluating the integral. the change in the pattern of international specialisation between the two regimes has implications for rates of productivity growth and hence has dynamic effects upon economic welfare.Az (t1 ) for all t ≥ t1 . and instead benefits from foreign learning by doing in this sector in the form of a terms of trade gain.gh . we require ρ > β.Ah (t1 ) h and A∗ (t) = egz z s∗ (t−t ) 1 s . we obtain the following expression for intertemporal welfare under the subsidy. for intertemporal utility to be finite.19 ¯ ¯ (1 − β). Substituting for the relative price of the high-tech good (from (19)) in the equation for intertemporal welfare (21).gz + (1 − β). (1−β) β [A∗ (t1 )]β [Ah (t1 )]1−β z 2β−1 ¯ L∗ ¯ L s s∗ ρ − βgz − (1 − β)gh [Az (t1 )]β [A∗ (t1 )]1−β h > f f ρ − βgz − (1 − β)gh ∗ (23) We saw in the discussion above. that the existence of the standard static gains from trade means that instantaneous welfare at time t1 must be lower under the subsidy. 18 20 . intertemporal welfare will be higher under the subsidy than under free trade if and only if.18 However. home forgoes its own potential to learn by doing in the low-tech sector. while h foreign enjoys the fruits of learning by doing in its low-tech sector (at the rate s∗ ¯ gz = µ∗ .

the weights are the shares of consumer expenditure allocated to each sector. Whether or not this inequality is satisfied will depend upon the two countries’ relative potentials to learn by doing in both the low. z under free trade. ¯ ¯ ¯ ¯ β. µ∗ j for j = z. h). While inequality (24) is a necessary condition for the subsidy to be welfare improving.20 A necessary condition for the production subsidy to the high-tech sector to be welfare improving is therefore that the rate of growth of instantaneous utility under the subsidy exceeds the corresponding rate of growth under free trade: f s∗ s f βgz + (1 − β)gh > βgz + (1 − β)gh ∗.L∗ . µh L − µ∗ L∗ > 0 z h (24) That is. home’s specialisation in the high-tech sector results in it accumulating production ex¯ perience at the rate g s = µh . With Cobb-Douglas instantaneous utility.L. A h necessary condition for the production subsidy to be welfare improving is thus. and the two economies’ relative sizes (as measured by L. z while foreign learns by doing in the high-tech sector at the rate f gh ∗ ¯ = µ∗ . ¯ Home experiences learning by doing in the low-tech sector at the rate g f = µz . we require the dynamic welfare gains from home specialising in the high-tech sector and foreign in the low-tech sector to exceed the static welfare losses.L∗ .L under the subsidy. we require a weighted average of the change in rates of productivity growth induced by the reversal of patterns of international specialisation to be strictly positive. the pattern of international specialisation is exactly the reverse. µ∗ L∗ − µz L + (1 − β). the share of consumer expenditure allocated to the low-tech sector ¯ ¯ (β). it is clearly not sufficient. while foreign accumulates h s ¯ production experience in the low-tech sector at the rate gz = µ∗ .and high-tech sectors (as parameterised by µj . In contrast. In order for intertemporal welfare to rise as a result of implementing the subsidy rather than adopting free trade. L∗ ). As we saw in the discussion above. From equations (22) and This may be shown formally by substituting for the relative price of the low-tech good pz (t1 )/ph (t1 ) under both the subsidy and free trade (using equations (19) and (12)) in the proof of Proposition 1 below. 20 21 .numerator on the left-hand side of the inequality (23) must be strictly less than the numerator on the right-hand side.

In the next section. the same concept has been applied in more informal discussions of the East Asian development 22 . Somewhat independently. we consider the relationship between this general equilibrium argument for selective trade and industrial policies and the often-discussed notion of ‘dynamic comparative advantage. 7 Dynamic comparative advantage and the case for policy intervention The fact that comparative advantage evolves endogenously over time in theoretical models of endogenous growth and trade has led a number of authors to speak of ‘dynamic comparative advantage’ (see. we note that. Indeed. for example. Nonetheless. the literature on the political economy of trade policy suggests that there may be hidden welfare costs to activist trade policies in the form of Directly Unproductive Profit-Seeking (DUP) activity (see. Bhagwati (1982)). may be a case in point.’ First. although one may establish theoretical conditions for interventionist public policies to be welfare improving. this will occur whenever inequality (23) is satisfied (which thus provides a necessary and sufficient condition for the subsidy to be welfare improving).the analysis requires a policy-maker to have information on rates of productivity growth in either sector of each economy under both the proposed subsidy and free trade. there is empirical evidence that the development experience of some East Asian economies may be interpreted in these terms.(13). Krugman (1987) and Grossman and Helpman (1991)). Furthermore. The information requirements to implement these policies are large . for example. and where an active trade policy can be justified in terms of the theoretical analysis of this section. there may be instances where the potential for productivity growth in sectors where an economy does not currently exhibit a comparative advantage is large. Developing economies with high levels of general human capital. which may achieve rapid rates of productivity growth through imitation. it may be extremely difficult in practice to determine when these conditions are met.

In each case. with its exact usage and meaning unclear. for example. In this section. the concept’s use may indicate a concern with some of the welfare considerations that have been the subject of previous sections. the concept is left ill-defined. and that sheds light upon the circumstances under which selective trade and industrial policies are welfare increasing. Furthermore. the discussion in Temple (1997)). in a model of endogenous technological change and growth. the use of the concept seems linked with the idea that a country’s current pattern of comparative advantage may work against its long-term interests. On the one hand. The traditional (or ‘static’) concept of comparative advantage is essentially concerned with relative levels of opportunity costs of production in different sectors of two economies. concerned with changes over time in relative levels of opportunity costs. in particular. This seems to be particularly the case with regard to the literature on the East Asian development experience. in particular. in certain circumstances. that accords reasonably closely with popular usage. This section now considers whether there is a concept of ‘dynamic comparative advantage’. Amsden (1989)). policy interventions to induce specialisation in sectors in which an economy does not currently have a comparative advantage may be welfare increasing. in our case. the home economy is said to have a (‘static’) comparative advantage in low-tech production at time t if the opportunity cost of producing the low-tech good at time t is lower in the home economy. On the other hand. the discussion in Grossman and Helpman (1991)). the home economy is said to have a ‘dynamic’ comparative advantage in 23 . and that there may be a trade-off between specialising according to current comparative advantage and realising dynamic benefits from specialising in other sectors (see.experience (see. the use of the concept may reflect a desire to explain the way in which comparative advantage (as traditionally defined) evolves over time in dynamic trade models (see. we propose a simple definition of ‘dynamic comparative advantage’. Thus. that specialising according to current patterns of comparative advantage may not be welfare maximising. Here. Thus. This paper has shown.

although natural. dynamic comparative advantage explains changes over time. However. the definition has (as will be shown below) a surprising amount of analytical content. dynamic comparative advantage is completely determined by productivity growth rates in each sector of the two economies. However. home h z will have a dynamic comparative advantage in the low-tech sector. is not invariant to changes in patterns of international specialisation. and its initial static comparative advantage in this sector will be reinforced over time. we have already seen that these themselves are functions of patterns of international specialisation. From (25). the argument may be illustrated in terms of the analysis of the previous section. Again. that is the dynamic analogue of the traditional ‘static’ definition. That is. home will have a dynamic comparative advantage in low-tech production if and only if. Nonetheless. as defined above. while the converse is true for the opportunity cost of low-tech production in foreign (A∗ /A∗ ). In terms of the analysis of the previous section. the opportunity cost of low-tech production in home (Ah /Az ) falls over time. the free trade equilibrium involves home specialising in low-tech production (learning at the rate f gz > 0) and foreign in high-tech production (accumulating production experience f at the rate gh ∗ > 0). On the one hand. it is important to realise that dynamic comparative advantage. ∂ (Ah (t)/Az (t)) /∂t ∂ (A∗ (t)/A∗ (t)) /∂t h z < Ah (t)/Az (t) A∗ (t)/A∗ (t) h z ⇔ ˙ ˙ ˙ ˙ Ah (t) Az (t) A∗ (t) A∗ (t) h − − z − ∗ Ah (t) Az (t) Ah (t) A∗ (t) z <0 (25) This formalisation of dynamic comparative advantage is an extremely natural one. while static comparative advantage determines patterns of international trade at a given point in time. In terms of the definition (25). Hence.low-tech production at time t if the rate of growth of the opportunity cost of producing the low-tech good at time t is lower in the home economy. international specialisation in the free trade equilibrium implies that home has a dynamic comparative advantage in the low-tech sector 24 . First.

as described above). home will actually have a dynamic comparative advantage in the high-tech sector under the subsidy. were the subsidy to be removed at a future point in time. home would continue to specialise in the high-tech sector. Thus.(so that the initial pattern of static comparative advantage is reinforced over time. and its initial static comparative advantage in the low-tech sector will be reduced over time. A necessary condition for the production subsidy to yield a higher level of intertemporal welfare than under free trade is that at some future point in time t > t1 home attains a higher level of instantaneous utility under the subsidy than 25 . the opportunity cost of low-tech production in home (Ah /Az ) rises over time. home specialises s in high-tech production (learning at the rate gh > 0) and foreign in low-tech s∗ production (learning at the rate gz > 0). patterns of dynamic comparative advantage are very different under the alternative regimes of free trade and the subsidy (as a result of differences in international specialisation). it is argued that an economy should specialise in sectors where it may enjoy various dynamic benefits (where it exhibits a ‘dynamic comparative advantage’). This brings us to the link between dynamic comparative advantage. Proof. As we have seen. Ultimately. Proposition 1 A necessary condition for a subsidy to the high-tech sector (where the economy does not currently have a static comparative advantage) to be welfareimproving is that the economy will (under the subsidy) acquire a static comparative advantage in the high-tech sector at some future point in time t > t1 . under the subsidy. It turns out that this informal argument may be straightforwardly related to the formal analysis of the previous section. Instead. so that. As a result. and the circumstances under which selective trade and industrial policies may be welfare improving. the initial pattern of static comparative advantage will be reversed. while the converse is true for the opportunity cost of low-tech production in foreign (A∗ /A∗ ). In terms of the h z definition (25). On the other hand. as defined above. an important element of the popular usage of this concept is the idea that a country’s current pattern of static comparative advantage may work against its long-term interests.

In Proposition 1. therefore. Furthermore. the fact that the initial pattern of static comparative advantage must be reversed for the subsidy to be welfare improving implies that the selective trade or industrial policy need only be temporary.by switching to free trade. However. However. although the reversal of static comparative advantage (and hence having a dynamic comparative advantage in the subsidised sector) is a necessary condition for the subsidy to be welfare improving. Otherwise. a necessary condition for the initial pattern of static comparative advantage to be reversed in this way is that (under the subsidy) the home economy has a dynamic comparative advantage (as defined in (25)) in the very sector in which it initially has no static comparative advantage. one can indeed think of an economy potentially facing a trade-off between static and dynamic comparative advantage (or between current and future patterns of static comparative advantage). we show that a necessary condition for instantaneous utility at time t to be higher under the subsidy is for home to have acquired a static comparative advantage in the high-tech sector by t (see Appendix) In order for the representative agent’s intertemporal welfare to be increased by a subsidy to a sector where the economy has no current static comparative advantage. If the subsidy were removed at time t . as suggested in informal discussions of the East Asian development experience. it is important to note that it is not sufficient. In Appendix 2. we compare instantaneous utility under the subsidy at each point in time t > t1 with the level that could be achieved by abandoning the subsidy at t and engaging in free trade (taking as given productivity growth rates 26 . Thus. Productivity growth rates and the way in which comparative advantage evolves over time are themselves dependent upon patterns of international specialisation (and. a policy-maker could unambiguously raise intertemporal welfare by abandoning the subsidy. home would (as discussed above) continue to specialise in the high-tech sector. will be different under the subsidy and free trade). it must be true that (under the subsidy) a static comparative advantage in this sector will be attained at some future point in time.

It was precisely such an analysis that was undertaken in the previous section. In effect. The analysis was undertaken within the context of a general equilibrium model of endogenous growth.as traditionally defined (‘static’ comparative advantage) . This enables us to establish a necessary condition for the subsidy to raise intertemporal welfare. t ) . Thus. However. 8 Conclusion This paper has considered the idea that developing economies may face a trade-off between specialising according to an existing pattern of comparative advantage (often in low-technology industries). but may acquire such an advantage in the future as a result of the potential for productivity growth (e. An essentially Ricardian model of international trade was combined with a model of endogenous technological change in the form of learning by doing.g. However. while simultaneously determining current rates of learning by doing and technological change.depends upon past technological advances. First. one must compare productivity growth rates and the evolution of comparative advantage under the subsidy with the corresponding values under free trade.as determined by the pattern of specialisation under the subsidy). in order to arrive at a sufficient condition. Specialisation according to current comparative advantage results in the standard static gains from trade. if individual agents fail to fully internalise the potential for productivity growth in each sector. two further steps must be taken. it may also mean that an economy fails to specialise in sectors where its potential for productivity growth is large relative to its trading partners.over the interval of time t ∈ [t1 . this comparison involves an evaluation of the (different) patterns of dynamic comparative advantage under both the subsidy and free trade. Comparative advantage . one must evaluate any dynamic welfare gains from implementing the subsidy relative to the standard static welfare losses. free trade will induce dynamic 27 . Second. high-technology industries). involving international trade between two large economies. comparative advantage itself becomes endogenous. and entering sectors where they currently lack a comparative advantage. As a result.

This necessary condition suggests that welfare improving selective trade and industrial policies. but as yet ill-defined. However. provide a sufficient condition. the initial pattern of static comparative advantage will only be reversed in this way if the economy has a dynamic comparative advantage (under the proposed policy intervention) in precisely the sector in which it initially has no static comparative advantage. Second. A natural formalisation of this concept was suggested in terms of rates of growth of opportunity costs of production in each economy. may be welfare improving. these may outweigh the standard static welfare gains. while this yields a necessary condition for interventionist public policies to be welfare improving. The case for these policies was related to the often-used. alone. so that trade reduces the intertemporal welfare of the representative agent. two further steps are necessary. First. one must evaluate any dynamic welfare gains from implementing the subsidy relative to the standard static welfare losses. So defined. if they exist. need only be temporary. dynamic comparative advantage explains the evolution of patterns of international trade over time and proves informative in evaluating the case for interventionist public policies. this comparison involves an evaluation of the (different) patterns of dynamic comparative advantage under both the subsidy and free trade. it does not. Selective trade and industrial policies to induce specialisation in sectors where an economy currently lacks a comparative advantage. but exhibits a large potential for productivity growth relative to its trading partner. a theoretical case for 28 .welfare losses. If sufficiently large. A necessary condition for a selective trade and industrial policy (of the form suggested above) to be welfare improving is that the initial pattern of ‘static’ comparative advantage is reversed under the policy. In effect. However. Once the complete welfare comparison is undertaken. notion of ‘dynamic’ comparative advantage. In order to evaluate whether subsidising production in a sector where no current comparative advantage exists is welfare improving. one must compare productivity growth rates and the evolution of comparative advantage under the subsidy with the corresponding values under free trade.

may be examples where a case for intervention exists. alongside potential hidden welfare costs in the form of Directly Unproductive Profit-seeking (DUP) activities.selective trade and industrial policies exists. Nonetheless. 29 . It involves an appreciation of the large informational requirements of the theoretical argument. converting this theoretical case into practical policy advice is more difficult. developing economies with high levels of general human capital. Nonetheless. which may achieve rapid rates of productivity growth through imitation.

the Dutch Disease. Oxford University Press. E (1990) ‘Comparative Advantage and Long-Run Growth’. Journal of Political Economy. 323-51 Amsden. MIT Press Islam. D and Helpman. European Economic Review. Econometrica. Journal of Development Economics. 149-61 Krugman. and the Competitive Consequences of Mrs Thatcher: Notes on Trade in the Presence of 30 . 8. J (1982) ‘Directly Unproductive Profit-Seeking (DUP) Activities’. P (1987) ‘The Narrow Moving Band. London Grossman.References Aghion. Quarterly Journal of Economics. S. New York Bhagwati. American Economic Review. E (1995) ‘International R & D Spillovers’. 823-39 Enos. G and Helpman. Accumulation and Uneven Development’. 67. Routlege. American Economic Review. 988-1002 Coe. Fischer. P and Howitt. Quarterly Journal of Economics. G and Helpman. E (1991) Innovation and Growth in the Global Economy. 80. 796-815 Grossman. P (1992) ‘A Model of Growth Through Creative Destruction’. J (1997) ‘Measuring the Social Return to R & D’. and Samuelson. W (1988) The Adoption and Diffusion of Imported Technology: The Case of Korea. 60. Forthcoming Krugman. 110. P (1981) ‘Trade. C and Williams. and Payments in a Ricardian Model with a Continuum of Goods’. N (1995) ‘Growth Empirics: A Panel Data Approach’. 859-87 Dornbusch. R. 39. Trade. P (1977) ‘Comparative Advantage. J and Park. A (1989) Asia’s Next Giant: South Korea and Late Industrialisation. 90. 1127-70 Jones.

106. 25. J (1997) ‘St Adam and the Dragons: Neo-classical Economics and the East Asian Miracle’. Quarterly Journal of Economics. 34. R (1990) Governing the Market: Economic Theory and the Role of Government in East Asian Industrialisation. Thesis.Dynamic Scale Economies’. NBER Working Paper. A (1995) ‘Natural Resource Abundance and Economic Growth’. 589-601 Wade. L and Romer. P and Obstfeld. M (1994) ‘Once-off and Continuing Gains From Trade’. 225-43 Taylor. 41-55 Krugman. D. P (1991a) ‘International Trade with Endogenous Technological Change’. Third Edition. 58. Harper Collins Lucas. Princeton University Press Young. Journal of International Economics. 251-72 Matsuyama. M (1991) ‘Quality Ladders and Ricardian Trade’. Review of Economic Studies. University of Oxford Rivera-Batiz. No. J and Warner. P (1991b) ‘Economic Integration and Economic Growth’. Journal of Economic Theory. M (1994) International Economics. 106. Econometrica. 106. L and Romer. European Economic Review. 27. Oxford Development Studies. A (1991) ‘Learning by Doing and the Dynamic Effects of International Trade’. and Growth’. 587-616 Temple. 317-34 Redding. 531-55 Sachs. R (1993) ‘Making a Miracle’. Quarterly Journal of Economics. K (1992) ‘Agricultural Productivity. Phil. 369-405 31 . 5398 Stokey. Journal of Development Economics. 279-300 Taylor. Product Quality. S (1996) ‘Endogenous Innovation and Economic Growth’. 971-1001 Rivera-Batiz. Comparative Advantage and Economic Growth’. 61. Quarterly Journal of Economics. 61. 35. N (1991) ‘Human Capital.

(1 − β)/β. ¯ ¯ Az (t1 ) > β/(1 − β). at time t1 .A∗ (t1 ) z ¯ ¯ A∗ (t1 ) < β/(1 − β). ps (t ) z Az (t ) ps (t ) h β [βAh (t )] [(1 − β)Ah (t )] β 1−β < β (1 − β)Az (t ) pf (t ) z pf (t ) h 1−β (30) In any equilibrium. instantaneous welfare at any time t could be increased by abandoning the subsidy and moving to free trade if and only if. in which home has a static comparative advantage in low-tech production and a subsidy is required to induce it to specialise in the high-tech sector. Hence.2 Proof of Proposition 1 From equations (21) and (10). for sufficiently large values of s. 1).9 9. inequality (16) and equation (12) jointly imply that. 9. inequalities (27) and (29) may both be satisfied. inequalities (16) and (18) jointly imply.L∗ /L..L/L∗ . At the same time.L∗ /L. the assumption of complete specialisation under both free trade and protectionism is validated.Ah (t1 ) h (26) (27) Under the subsidy.Ah (t1 ) h (28) (29) For values of β ∈ (1/2. ¯ ¯ Az (t1 ) < (1 + s).L/L∗ .A∗ (t1 ) z ¯ ¯ A∗ (t1 ) > (1 − β)/β. at time t1 . inequalities (26) is compatible with (28) and with both (27) and (29). inequality (18) and equation (19) jointly imply that.Ah (t ) ps (t ) A∗ (t ) pf (t ) Ah (t ) z > s > h > z > f ∗ (t ) Az (t ) ph (t ) Az Az (t ) ph (t ) 32 (31) . (1 + s).1 Appendix Relative prices under free trade and the subsidy Under free trade.

inequality (30) is no longer satisfied. inequality (30) must be satisfied.Ah (t ) A∗ (t ) Ah (t ) ps (t ) pf (t ) z > h > > s = z Az (t ) Az (t ) ph (t ) A∗ (t ) pf (t ) z h (32) and Ah > pf /pf .Az < ps /ps . Hence.Az . As long as home has a static comparative z z h h advantage in the low-tech sector and the subsidy is required to induce it to specialise in the high-tech sector. 33 . that home acquires a static comparative advantage in the high-tech sector and a subsidy is no longer required to induce specialisation in this sector. In this case.Az . z z h h It follows immediately that a necessary condition for the subsidy to yield a higher level of instantaneous utility at some point in time t is for home to have acquired a static comparative advantage in this sector by time t .Thus. (1 + s). Ah < pf /pf .Az = ps /ps . Suppose instead.