This action might not be possible to undo. Are you sure you want to continue?
No 381
Reassessing the impact of
finance on growth
by Stephen G Cecchetti and Enisse Kharroubi
Monetary and Economic Department
July 2012
JEL classification: D92, E22, E44, O4
Keywords: Growth, financial development, credit booms, R&D
intensity, financial dependence
BIS Working Papers are written by members of the Monetary and Economic Department of
the Bank for International Settlements, and from time to time by other economists, and are
published by the Bank. The papers are on subjects of topical interest and are technical in
character. The views expressed in them are those of their authors and not necessarily the
views of the BIS.
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2012. All rights reserved. Brief excerpts may be
reproduced or translated provided the source is stated.
ISSN 10200959 (print)
ISSN 16827678 (online)
iii
Reassessing the impact of finance on growth
Stephen G Cecchetti and Enisse Kharroubi*
July 2012
Abstract
This paper investigates how financial development affects aggregate productivity growth.
Based on a sample of developed and emerging economies, we first show that the level of
financial development is good only up to a point, after which it becomes a drag on growth.
Second, focusing on advanced economies, we show that a fastgrowing financial sector is
detrimental to aggregate productivity growth.
JEL classification: D92, E22, E44, O4
Keywords: Growth, financial development, credit booms, R&D intensity, financial
dependence
* Cecchetti is Economic Adviser at the Bank for International Settlements (BIS) and Head of its Monetary and
Economic Department, and Kharroubi is Economist at the BIS. This paper was prepared for the Reserve Bank
of India’s Second International Research Conference in Mumbai, India, on 1–2 February 2012. We thank
Claudio Borio, Leonardo Gambacorta, Christian Upper and Fabrizio Zampolli for helpful suggestions; and
Garry Tang for valuable research assistance. The views expressed in this paper are those of the authors and
not necessarily those of the BIS.
1
1. Introduction
One of the principal conclusions of modern economics is that finance is good for growth. The
idea that an economy needs intermediation to match borrowers and lenders, channelling
resources to their most efficient uses, is fundamental to our thinking. And, since the
pioneering work of Goldsmith (1969), McKinnon (1973) and Shaw (1973), we have been able
to point to evidence supporting the view that financial development is good for growth. More
recently, researchers were able to move beyond simple correlations and establish a
convincing causal link running from finance to growth. While there have been dissenting
views, today it is accepted that finance is not simply a byproduct of the development
process, but an engine propelling growth.
1
This, in turn, was one of the key elements
supporting arguments for financial deregulation. If finance is good for growth, shouldn’t we be
working to eliminate barriers to further financial development?
It is fair to say that recent experience has led both academics and policymakers to
reconsider their prior conclusions. Is it true regardless of the size and growth rate of the
financial system? Or, like a person who eats too much, does a bloated financial system
become a drag on the rest of the economy?
In this paper, we address this question by examining the impact of the size and growth of the
financial system on productivity growth at the level of aggregate economies. We present two
very striking conclusions. First, as is the case with many things in life, with finance you can
have too much of a good thing. That is, at low levels, a larger financial system goes hand in
hand with higher productivity growth. But there comes a point – one that many advanced
economies passed long ago – where more banking and more credit are associated with
lower growth.
Our second result comes from looking at the impact of growth in the financial system –
measured as growth in either employment or value added – on real productivity growth. Here
we find evidence that is unambiguous: faster growth in finance is bad for aggregate real
growth. One interpretation of this finding is that financial booms are inherently bad for trend
growth.
At first, these results may seem surprising. After all, a more developed financial system is
supposed to reduce transaction costs, raising investment directly, as well as improving the
distribution of capital and risk across the economy.
2
These two channels, operating through
the level and composition of investment, are the mechanisms by which financial development
improves growth.
3
But the financial industry competes for resources with the rest of the
economy. It requires not only physical capital, in the form of buildings, computers and the
like, but highly skilled workers as well. Finance literally bids rocket scientists away from the
satellite industry. The result is that people who might have become scientists, who in another
1
The view that financial development is simply a byproduct of growth is discussed in Robinson (1952): “Where
enterprise leads, finance follows”. For the more recent work establishing causality, see Levine et al (2000) for
countrylevel evidence and Rajan and Zingales (1998) for industrylevel evidence. For an alternative view, see
Easterly et al (2000), who suggest that financial development may only be good up to a point.
2
See Pagano (1993) for a simple analytical model of financial development as a reduction in transaction costs
in the context of an AK model. A more comprehensive approach is developed in Holmström and Tirole (1997),
who provide a model for why different financial patterns (direct finance vs intermediated finance) may coexist
altogether.
3
Theoretical contributions relating the role of financial intermediaries to the composition of investment include:
Acemoğlu and Zilibotti (1997), who look at how the presence of financial intermediaries affects the risk return
profile of entrepreneurs’ projects; Holmström and Tirole (1998), who examine how financial intermediaries can
help save on liquidity hoarding; and Aghion et al (2010), who show how financial development helps reduce
the growth cost of economic fluctuations.
2
age dreamt of curing cancer or flying to Mars, today dream of becoming hedge fund
managers.
4
There is an important sense in which this description of the consequences of a financial
boom is no different from those of the dotcom boom of the 1990s, or the impact of any other
boom tied to more tangible output. Booming industries draw in resources at a phenomenal
rate. It is only when they crash, after the bust, that we realise the extent of the
overinvestment that occurred. Too many companies were formed, with too much capital
invested and too many people employed. Importantly, after the fact, we can see that many of
these resources should have gone elsewhere. Following the dotcom bust innumerable
computers were scrapped, office buildings vacated and highly trained people laid off.
The remainder of the paper provides the empirical evidence for our conclusions. In Section 2,
we examine the impact of financial system size on productivity growth in a sample of 50
advanced and emerging market economies over the past three decades. To measure
financial sector size, we consider both output measures like private credit to GDP as well as
input measures like the financial sector’s share in total employment (in this latter case, we
restrict our analysis restricted to advanced countries because of limited data availability).
Considering the level of financial development, we find that when private credit grows to the
point where it exceeds GDP, it becomes a drag on productivity growth. Using employment
measures, we find that when the financial sector represents more than 3.5% of total
employment, further increases in financial sector size tend to be detrimental to growth.
In Section 3, we examine the impact of the growth rate of the financial system on aggregate
productivity growth in a sample of advanced countries over the past three decades. Again,
our analysis is restricted to advanced economies due to data limitations. There we find that,
compared with a country where the financial sector’s share in total employment is stable, a
typical financial boom – employment growth of 1.6% per year – reduces growth in aggregate
GDP per worker by roughly one half of 1 percentage point.
2. The inverted Ushaped effect of financial development
We begin by examining the relationship between the size of a country’s financial system and
its productivity growth to see whether there is a point where bigger is no longer better. In
what follows, we examine various measures of financial development, starting with the ratio
of private credit to GDP.
2.1 Private credit and growth
We begin with a simple histogram constructed from a sample of 50 advanced and emerging
countries over the period 1980–2009. Using fiveyear nonoverlapping GDPperworker
growth and private credit to GDP we compute the average growth conditional on the quartiles
of the ratio of private credit to GDP. The resulting histogram in Graph 1, computed from a
total of 300 data points, shows that GDPperworker growth increases from the first to the
4
Philippon and Reshef (2009) provide empirical evidence that, over the past 30 years, the US banking industry
has become relatively skilledlabourintensive. Analytical contributions investigating occupational choices
between producing and financing include Philippon (2007), which provides a model where human capital is
allocated between entrepreneurial and financial careers, and where entrepreneurs can innovate but face
borrowing constraints that financiers can help to alleviate. Cahuc and Challe (2009) also develop an analytical
model focusing on the allocation of workers between financial intermediation and production sectors in the
presence of asset price bubbles.
3
third private credit to GDP quartile, before declining in the final quartile. That is, countries
with the highest level of private credit to GDP have lower trend growth than the rest.
Graph 1
Average GDPperworker growth by private credit to GDP quartiles
Each bar represents the fiveyear average GDPperworker growth conditional on the fiveyear average private
credit to GDP ratio belonging to a specific quartile of the sample distribution. The sample covers 50 countries over
the period 1980–2009. GDPperworker growth and private credit to GDP are averaged for nonoverlapping
periods over five years. For country sample and sources, see data appendix.
Graph 2
Private credit to GDP ratio and growth
Graphical representation of ( ) ( )
t k t k
y
t t k
fd
t t k
fd
k t t k
y
, ,
2
5 , , 1 5 , , 0 , 5 ,
c o ¸ ¸  o + ÷
+
+
+
+ + =
+
A over the period
1980–2009. For country sample and sources, see data appendix.
Of course, this histogram does not imply that private credit is bad for growth at high levels.
Countries with high private credit to GDP are more developed economies, which grow more
slowly for a variety of reasons. Convergence effects are the most obvious. To address this,
we compute deviations from countryspecific means, and control for initial conditions. The
result, based on the sample of data, is in Graph 2. The relationship is clearly not monotonic.
That is, at low levels of credit, more credit raises trend growth. But there comes a point
where the additional lending and a bigger financial system become a drag on growth.
0.0
0.5
1.0
1.5
2.0
first quartile second quartile third quartile fourth quartile
–0.10
–0.05
0.00
0.05
0.10
–0.50 –0.25 0.00 0.25 0.50 0.75 1.00
F
i
v
e

y
e
a
r
a
v
e
r
a
g
e
G
D
P

p
e
r

w
o
r
k
e
r
g
r
o
w
t
h
Fiveyear average private credit to GDP
(Deviation from country mean)
(
D
e
v
i
a
t
i
o
n
f
r
o
m
c
o
u
n
t
r
y
m
e
a
n
)
4
To get a more precise sense of this relationship, and to test our hypothesis that the effects of
finance on growth can go from good to bad, we turn to a panel regression.
5
We regress the
fiveyear average growth in output per worker in a given country on the following variables:
the level of financial development; the squared level of financial development (looking for the
parabola in Graph 2); and a series of control variables known to influence aggregate growth.
To fix ideas and notation, we write this as:
( ) ( )
t k t k t t k t t k t t k k t t k
y X fd fd y
, , , 5 , 2
2
, 5 , 1 , 5 , 0 , 5 ,
c o ¸ ¸ ¸  o + ÷ + + + + = A
+ + + +
(1)
where y
k,t
is the log of output per worker in country k in year t; Ay
k,t+5,t
is the average growth in
output per worker in country k from time t to t+5; fd
k,t+5,t
is the average ratio of private credit to
GDP in country k from time t to t+5, our measure of financial development; X
k,t+5,t
is a set of
control variables averaged from time t to t+5, including working population growth, openness
to trade measured by the ratio of imports and exports to GDP, the share of government
consumption in GDP and CPI inflation; o is a constant and 
k
is a vector of country dummies;
and c is the error term, which we allow for heteroskedasticity. Our hypothesis is that ¸
0
will be
positive and ¸
1
negative.
In the first column of Table 1 we report the result with no controls. Here we see what we
expect – the relationship is parabolic. Continuing across the columns of the table, we
sequentially add controls. The nonlinearity is robust. Regardless of the exact specification,
the coefficient of the level of financial development, ¸
0
, is around 0.035 and that on the
quadratic term, ¸
1
, is always close to –0.018.
6
We can use the estimated coefficients to compute an estimate of the peak of the inverted U –
the vertical line in Graph 2. These are reported near the bottom of the table, together with
95% interval estimates. The point estimates all roughly 100% of GDP, a figure that is quite
close to the threshold of 90% computed in Cecchetti et al (2011).
7
To see what these numbers mean, we can look at a few examples. Starting with New
Zealand, in the first half of the 1990s, private credit was below 90% of GDP. Credit then rose
steadily, reaching nearly 150% of GDP by the time of the crisis. The estimates in Table 1
suggest that this increase created a drag of nearly one half of 1 percentage point on trend
productivity growth.
Thailand is another interesting case. In the runup to the Asian crisis of 1997–98, the ratio of
Thai private credit to GDP reached 150%. More recently, this measure of financial sector
size has fallen to roughly 95%. This time, the result is a benefit of roughly one half of
1 percentage point in trend productivity growth.
5
Our work in this section builds on an extensive body of research, especially the empirical literature relating
growth to finance. Notable contributions on this topic include King and Levine (1993), Islam (1995), Levine and
Zervos (1998), Beck et al (2000) and Cecchetti et al (2011). Comprehensive surveys can be found in Levine
(1997, 2005).
6
We note that the results reported in Table 1 are robust to adding a variety of changes to equation (1). These
include: (i) using GDP per capita instead of GDP per worker as the dependent variable; (ii) using alternative
measures of financial development like private credit by banks to GDP, bank deposits to GDP, financial
system deposits to GDP, or bank assets to GDP; and (iii) dropping certain countries, such as the former
communist countries, from the sample.
7
The difference between the estimates is probably a result of differences in data and methods. The current
study uses a broader set of countries, while the latter employs a somewhat more sophisticated econometric
model.
5
Finally, take the example of the United States, where private credit grew to more than 200%
of GDP by the time of the financial crisis. Reducing this to a level closer to 100% would, by
our estimates, yield a productivity growth gain of more than 150 basis points.
Table 1
GDPperworker growth and private credit to GDP
Dependent variable: five
year average real GDP
perworker growth
(1) (2) (3) (4) (5) (6)
Fiveyear average private
credit to GDP
0.036*** 0.038*** 0.035*** 0.035*** 0.035*** 0.048**
(0.011) (0.011) (0.011) (0.011) (0.011) (0.021)
Fiveyear average private
credit to GDP squared
–0.018*** –0.018*** –0.018*** –0.017*** –0.017*** –0.022***
(0.005) (0.005) (0.005) (0.005) (0.005) (0.008)
Log of real GDP per worker –0.742*** –1.020*** –1.110*** –1.110*** –1.160*** –6.220***
(0.211) (0.210) (0.208) (0.207) (0.204) (1.200)
Fiveyear working
population growth
–0.478*** –0.480*** –0.471*** –0.501*** –0.685***
(0.162) (0.160) (0.163) (0.152) (0.162)
Fiveyear average
openness to trade
0.010*** 0.010*** 0.009*** 0.054***
(0.003) (0.003) (0.003) (0.010)
Fiveyear average
government consumption
share in GDP
0.0106 0.0107 –0.145
(0.046) (0.045) (0.331)
Fiveyear average CPI
inflation
0.0378 0.047
(0.036) (0.037)
Turning point for the effect
of private credit to GDP on
real GDPperworker growth
0.98 1.02 0.99 0.99 1.01 1.08
95% confidence interval [0.97;1.00] [1.01;1.03] [0.98;1.01] [0.98;1.01] [0.99;1.02] [1.06;1.11]
Observations 270 270 270 270 270 270
Rsquared 0.098 0.160 0.190 0.190 0.213 0.424
The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country,
which yields six observations per country. Fiveyear averages for the independent variables are computed over
the same period as the dependent variable. The log of real GDP per worker is the natural logarithm of real
GDP per worker for the initial year of the period over which the averages are computed, divided by 100. All
estimates include a nonreported constant. Column (6) includes country dummies. Robust standard errors are
in parentheses. Significance at the 1/5/10% level is indicated by ***/**/*. The turning point for the effect of
private credit to GDP on real GDPperworker growth is the level for private credit to GDP below (above) which
an increase in private credit to GDP is estimated to raise (reduce) real GDPperworker growth. For country
sample and sources, see data appendix.
We should be very clear that we do not in any way view these peak debt values as targets,
and neither should any readers (or authorities). These are levels of debt that a country
should only approach in extremis. And as discussed in more detail by Cecchetti et al (2011),
under normal circumstances we would expect to see debt at much lower levels than these
thresholds. Keeping debt well below 90% of GDP provides the room needed to respond in
6
the event of a severe shock. Otherwise, should a crisis arise, the additional accumulation of
debt would result in a drag on growth that would make recovery even more difficult than it
already is.
8
2.2 Alternative measures of financial development
In the previous section, financial development was measured using total credit extended to
the private sector. In this section we examine the robustness of this result to the use of
alternative measures of financial sector size. We start by looking at the consequences of
using bank credit rather than total credit, and then move on to study the relationship between
growth and the share of employment accounted for by the financial sector.
2.2.1 Bank credit as a measure of financial development
Differences in financial system structure imply that credit can mean different things in
different countries. And this difference could be the driving force behind our results. For
example, the inverted Ushape could reflect compositional effects, with bankbased financial
systems being one part of the parabola and marketbased financial systems being the other.
To examine this possibility, we replace private credit with private credit by banks (relative to
GDP) in equation (1).
Table 2 presents the results of this exercise. Again, we start with no controls in the first
column and add controls sequentially moving to the right in the table. The results confirm
both the parabolic relationship and its robustness. Furthermore, the point estimates
themselves are very close to those in Table 1.
Looking at the peak of the parabola, we estimate that for private credit extended by banks,
the turning point is closer to 90% of GDP – somewhat lower than for total credit. Many
countries are close to or beyond this level, suggesting that more credit will not translate into
higher trend growth. For example, in Portugal, private credit by banks was 160% of GDP at
the onset of the financial crisis. The corresponding figure for the UK was 180% of GDP and
even reached 200% of GDP in Denmark. By contrast, a country like India, where bank credit
is less than 50% of GDP, can still reap significant benefits from further financial deepening in
terms of increasing productivity growth.
8
This argument is consistent with the results of welfare maximisation, which would imply that, in normal times,
debt should be maintained below the level at which borrowing constraints become binding.
7
Table 2
GDPperworker growth and private credit by banks to GDP
Dependent variable:
fiveyear average real
GDPperworker growth
(1) (2) (3) (4) (5) (6)
Fiveyear average private
credit by banks to GDP
0.0369*** 0.0373*** 0.0336*** 0.0334*** 0.0325*** 0.0477**
(0.0117) (0.0116) (0.0118) (0.0118) (0.0121) (0.0208)
Fiveyear average private
credit by banks to GDP
squared
–0.0196*** –0.0193*** –0.0185*** –0.0184*** –0.0178*** –0.0229***
(0.00520) (0.00522) (0.00521) (0.00519) (0.00543) (0.00836)
Log of real GDP per
worker
–0.732*** –0.979*** –1.046*** –1.049*** –1.086*** –6.279***
(0.195) (0.197) (0.195) (0.195) (0.191) (1.174)
Fiveyear working
population growth
–0.461*** –0.463*** –0.455*** –0.483*** –0.675***
(0.164) (0.163) (0.166) (0.155) (0.164)
Fiveyear average
openness to trade
0.0105*** 0.0105*** 0.00979*** 0.0545***
(0.00295) (0.00297) (0.00296) (0.0108)
Fiveyear average
government consumption
share in GDP
0.00962 0.00977 –0.114
(0.0453) (0.0445) (0.329)
Fiveyear average CPI
inflation
0.0364 0.0456
(0.0368) (0.0368)
Turning point for the effect
of private credit to GDP
on real GDPperworker
growth
0.94 0.96 0.91 0.91 0.92 1.04
95% confidence interval [0.93;0.95] [0.95;0.98] [0.90;0.92] [0.90;0.92] [0.90;0.93] [1.02;1.07]
Observations 269 269 269 269 269 269
Rsquared 0.103 0.161 0.194 0.194 0.215 0.426
The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country,
which yields six observations per country. Fiveyear averages for the independent variables are computed over
the same period as the dependent variable. The log of real GDP per worker is the natural logarithm of real
GDP per worker for the initial year of the period over which the averages are computed, divided by 100. All
estimates include a nonreported constant. Column (6) includes country dummies. Robust standard errors are
in parentheses. Significance at the 1/5/10% level is indicated by ***/**/*. The turning point for the effect of
private credit by banks to GDP on real GDPperworker growth is the level for private credit by banks to GDP
below (above) which an increase in private credit to GDP is estimated to raise (reduce) real GDPperworker
growth. For country sample and sources, see data appendix.
2.2.2 Financial sector employment as a measure of financial development
The use of credit as a measure of financial development means that we are focusing on the
output of the sector. An alternative gauge of financial sector size and financial development,
one based on inputs, is the financial sector’s share in the economy’s total employment. Using
a more limited sample drawn from 21 OECD economies over the period from 1980 to 2009,
we look at the relationship between the financial sector’s use of the economy’s labour
resources and aggregate growth. In addition to providing a different measure of financial
development, the analysis using an inputbased measure of financial development provides
8
an important check that the inverted Ushaped effect on growth is not simply the result of
using a sample which mixes advanced and emerging market economies.
Graph 3
Financial sector share in employment and growth
1
1
Graphical representation of ( ) ( )
t k t k
y
t t k
fs
t t k
fs
k t t k
y
, ,
2
5 , , 1 5 , , 0 , 5 ,
c o ¸ ¸  o + ÷
+
+
+
+ + =
+
A over the period
1980–2009. For country sample and sources, see data appendix.
We start with a scatter plot in Graph 3, which is analogous to Graph 2. The results confirm
our previous results: the relationship between growth and the financial sector’s share in
employment is an inverted U. At low levels, an increase in the financial sector’s share in total
employment is actually associated with higher GDPperworker growth. But there is a
threshold beyond which a larger financial sector becomes a drag on productivity growth.
Turning to the regression analysis, we estimate equation (1) using the fiveyear average
financial sector share in total employment in country as a measure of financial development
(fd). Again, on the lefthand side we have the fiveyear average growth in output per worker
in a given country. And on the righthand side, we have the financial sector’s share in total
employment, the financial sector’s share in total employment squared, and various controls.
Table 3 presents the results from this exercise. Here we see the result that we expect – the
relationship is parabolic.
9
Again, the result is robust to the addition of controls.
9
Introducing country fixed effects changes the results only modestly. In particular, the estimated coefficient for
the linear and the quadratic term are such that the share of financial intermediation in total employment that
maximises growth is lower than the one obtained without country fixed effects. This reinforces the idea that the
estimated turning point should be regarded not as a target but rather as an upper bound.
–0.04
–0.03
–0.02
–0.01
0.00
0.01
0.02
–0.004 –0.002 0.000 0.002 0.004 0.006
F
i
v
e

y
e
a
r
a
v
e
r
a
g
e
G
D
P

p
e
r

w
o
r
k
e
r
g
r
o
w
t
h
Fiveyear average financial intermediation share in total employment
(Deviation from country mean)
(
D
e
v
i
a
t
i
o
n
f
r
o
m
c
o
u
n
t
r
y
m
e
a
n
)
9
Table 3
GDPperworker growth and financial sector share in employment
Dependent variable: five
year average real GDP
perworker growth
(1) (2) (3) (4) (5) (6)
Fiveyear average financial
intermediation share in total
employment
3.345*** 3.335*** 3.354*** 3.347*** 3.341*** 5.574**
(0.690) (0.677) (0.675) (0.706) (0.705) (2.602)
Fiveyear average financial
intermediation share in total
employment squared
–43.35*** –43.31*** –43.48*** –43.37*** –43.30*** –103.6***
(9.025) (9.004) (8.980) (9.516) (9.493) (35.82)
Log of real GDP per worker –3.346*** –3.334*** –3.409*** –3.417*** –3.407*** –6.087***
(0.665) (0.672) (0.708) (0.708) (0.707) (1.537)
Fiveyear working
population growth
0.0243 0.00762 0.00799 0.0129 –0.111
(0.189) (0.174) (0.174) (0.181) (0.237)
Fiveyear average
openness to trade
0.00195 0.00194 0.00193 0.0171
(0.00431) (0.00442) (0.00439) (0.0173)
Fiveyear average
government consumption
share in GDP
0.00260 0.00249 –0.360
(0.0581) (0.0586) (0.269)
Fiveyear average CPI
inflation
–0.00256 0.0181
(0.0236) (0.0250)
Turning point (in %) for the
effect of financial
intermediation share in total
employment on real GDP
perworker growth
3.86 3.85 3.86 3.86 3.86 2.69
95% confidence interval [1.20;6.51] [1.28;6.42] [1.29;6.42] [1.22;6.50] [1.22;6.50] [–6.34;11.7]
Observations 95 95 95 95 95 95
Rsquared 0.331 0.331 0.333 0.333 0.333 0.536
The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country,
which yields six observations per country. Fiveyear averages for the independent variables are computed over
the same period as the dependent variable. The log of real GDP per worker is the natural logarithm of real
GDP per worker for the initial year of the period over which the averages are computed, divided by 100. All
estimates include a nonreported constant. Column (6) includes country dummies. Robust standard errors are
in parentheses. Significance at the 1/5/10% level is indicated by ***/**/*. The turning point for the effect of the
financial sector’s share in total employment on real GDPperworker growth is the level of the financial sector’s
share in total employment below (above) which an increase in the financial sector’s share in total employment
is estimated to raise (reduce) real GDPperworker growth. For country sample and sources, see data
appendix.
To see what these numbers mean, we first look at the recent data for the sample countries
and evaluate them against the estimate for the turning point of 3.9% reported at the bottom
of Table 3. Graph 4 shows that in most countries, the financial sector’s share in total
employment is below or significantly below the threshold beyond which the effect on GDP
perworker growth turns from positive to negative. Indeed, the size of the financial sector is
above this growthmaximising point only in some cases. Examples are Canada, Switzerland,
Ireland and, to a lesser extent, the United States. However, as was stressed above, the
10
growthmaximising size of the financial sector should not be considered as a target, in
particular because it is possible that the negative effect on growth may start materialising for
lower levels. From that point of view, all countries in our sample are considerably above the
lower band of the 95% confidence interval around the estimate for the turning point. This
means that for all countries, further increases in financial sector size are most likely to have
mixed effects on productivity growth. However, this result also owes to the limited sample we
use, which mechanically raises the size of the confidence interval around the estimated
turning point.
Graph 4
Average financial sector share in total employment, 2005–09
1
1
AU = Australia, AT = Austria, BE = Belgium, CA = Canada, CH = Switzerland, DE = Germany, DK = Denmark,
ES = Spain, FI = Finland, FR = France, GB = United Kingdom, IE = Ireland, IT = Italy, JP = Japan, KR = Korea,
NL = Netherlands, NO = Norway, NZ = NewZealand, PT = Portugal, SE = Sweden, US = United States.
Sources: OECD Structural Analysis database; authors’ calculations.
Coming back to the countries where the financial sector’s share in total employment is above
the growthmaximising point, we can compute the gain in GDPperworker growth if their
financial sectors were to shrink back to the growthmaximising point. For Canada, the gain is
1.3 percentage points, for Switzerland 0.7 percentage points and for Ireland 0.2 percentage
points.
The case of Ireland is interesting because over the period 1995–99, the share of the Irish
financial sector’s share in total employment was 3.84% – very close to the growth
maximising value. But over the next 10 years, the share rose to more than 5%. Had the
share been constant between1995–99 and 2005–09, our estimates suggest that Irish trend
GDPperworker growth could have been as much as 0.4 percentage points higher over the
past decade.
3. The real effects of financial sector growth
Having established that there is a point at which financial development switches from
propelling real growth to holding it back, we now turn to an examination of the impact of the
speed of development on productivity growth. Put another way, we examine how financial
0
1
2
3
4
5
6
AU AT BE CA CH DE DK ES FI FR GB IE IT JP KR NL NO NZ PT SE US
Average financial sector share in total employment, 20052009
Growthmaximizing level
95% lower confidence band
11
sector booms – periods when financial development is moving at a particularly fast pace –
can affect growth.
10
Unlike in the earlier exercise, we cannot simply rely on the ratio of private credit to GDP to
measure financial sector growth. If we were to do this, financial sector growth would be
negatively associated with GDP growth by construction. To bypass this problem, we use data
on employment in the financial sector and measure financial sector growth as the growth rate
in the financial sector’s share in total employment.
As noted in the previous section, using employment data comes at a cost since employment
data for the financial sector are available for a limited subset of our previous sample of
countries. Hence we focus on the 21country subset of OECD countries.
Graph 5
Financial sector growth and productivity growth
1
1
Graphical representation of
t k t k
y
t t k
fd
k t t k
y
, , , 5 , 0 , 5 ,
c o ¸  o + ÷
+
A + + =
+
A over the period 1980–2009, where y
k,t
is
the log of output per worker in country k in year t; Ay
k,t+5,t
is the average growth in output per worker in country k
from time t to t+5; Afd
k,t+5,t
is the average growth in financial intermediation employment in country k from time t to
t+5; 
k
is a vector of country dummies; and c
k,t
is a residual. For country sample and sources, see data appendix.
Graph 5 summarises our main finding. Again, on the vertical axis we plot the fiveyear
average GDPperworker growth. On the horizontal axis, we now plot the fiveyear average
growth in the financial sector’s share in total employment.
11
(As in Graphs 2 and 3, both
variables are measured as deviations from their countryspecific means.) The result is quite
10
There is a large and wellknown literature on this financial accelerator and its quantitative implications for the
business cycle (see Bernanke and Gertler (1989) and Bernanke et al (1999), for instance). Likewise, there is a
significant body of research examining credit cycles (from Kiyotaki and Moore (1997) to more recent work by
Caballero et al (2006) on the dotcom bubble or Lorenzoni (2008), who look at the normative implications of
credit booms). We are, however, unaware of empirical studies on the implications of financial booms for long
run growth.
11
The results reported in this section are robust to the use of financial sector value added in place of the
financial sector’s share in total employment.
–0.04
–0.02
0.00
0.02
0.04
–0.04 –0.02 0.00 0.02 0.04
F
i
v
e

y
e
a
r
a
v
e
r
a
g
e
r
e
a
l
G
D
P

p
e
r

w
o
r
k
e
r
g
r
o
w
t
h
Fiveyear average financial intermediation employment growth
(Deviation from country mean)
(
D
e
v
i
a
t
i
o
n
f
r
o
m
c
o
u
n
t
r
y
m
e
a
n
)
12
striking: there is a very clear negative relationship. The faster the financial sector grows, the
slower the economy as a whole grows!In order to verify that this relationship is robust, we
follow the same procedure as before, estimating a panel regression with the fiveyear
average annual growth rate in GDP per worker as the dependent variable. In addition to the
controls used in equation (1), we now introduce financial sector growth. But, unlike in the
earlier exercise, we cannot simply take the change in the ratio of private sector to GDP as
the object of interest. If we were to do this, we would have GDP growth on the lefthand side
of the regression and the inverse of GDP growth on the righthand side, so finding a negative
relationship would be wholly uninformative. It is for this reason that we now measure financial
sector growth using employment growth and estimate:
t k t k t t k t t k k t t k
y X fd y
, , , 5 , 1 , 5 , 0 , 5 ,
c o ¸ ¸  o + ÷ + A + + = A
+ + +
(2)
where all variables are defined as before, with the exception of Afd
k,t+5,t
, which is the average
growth in the financial sector’s share in total employment in country k from time t to t+5. We
note that the vector of controls in equation (2) includes the growth rate of the working
population, trade openness measured as the ratio of imports plus exports to GDP, the share
of government consumption in GDP, CPI inflation and the level of financial development.
Table 4 presents the results of estimating equation (2) using a variety of measures of
financial sector size as controls. Our interest is in the first row of the table, which reports the
estimates of ¸
0
, the coefficient on the financial sector growth. The result evident in Graph 4 is
confirmed by more careful statistical analysis: the faster financial sector employment grows,
the worse it is for productivity growth (measured as fiveyear average growth in GDP per
worker). Moreover, this effect survives regardless of the combination and definition of the
controls.
12
To assess the magnitude of the effects, we start by comparing a country with constant
employment in financial intermediation with one in which employment grows at
1.6 percentage points per year, the sample average for those with positive growth. The
elasticity estimate of –0.33 implies that the first country will grow on average 50 basis points
faster than the second country. Given that the sample average productivity growth rate is
1.3%, this strikes us as sizeable.
Turning to some country examples, we look at Ireland and Spain – admittedly extreme
cases. During the five years beginning in 2005, Irish and Spanish financial sector
employment grew at an average rate of 4.1% and 1.4% per year, while output per worker fell
by 2.7% and 1.4%, respectively. Our estimates imply that if financial sector employment had
been constant in these two countries, it would have shaved 1.4 percentage points from the
decline in Ireland and 0.6 percentage points in Spain. In other words, by our reckoning
financial sector growth accounts for one third of the decline in Irish output per worker and
40% of the drop in Spanish output per worker.
12
Note that the effect of control variables is relatively different from what it was in the previous regression. In
particular, government size now has a significant negative effect on growth at the margin. This could be
related to the fact that here we focus on advanced economies, where high government consumption is more
likely to have a detrimental effect on the private sector. The speed of convergence is also much higher
(between 7 and 8% a year) than in the previous regressions, which is also probably related to sample
difference.
13
Table 4
GDPperworker growth and financial sector growth
Dependent variable: fiveyear
average real GDPperworker
growth
(1) (2) (3) (4) (5)
Fiveyear average financial
intermediation employment growth
–0.471*** –0.327*** –0.325*** –0.328*** –0.331***
(0.083) (0.074) (0.073) (0.073) (0.074)
Fiveyear working population growth
–0.356* –0.275 –0.286 –0.270 –0.259
(0.204) (0.186) (0.183) (0.188) (0.191)
Fiveyear average openness to trade
0.007 0.022 0.023 0.022 0.022
(0.0148) (0.0138) (0.0143) (0.0142) (0.0138)
Fiveyear average government
consumption share in GDP
–0.762*** –0.636*** –0.626*** –0.637*** –0.635***
(0.212) (0.219) (0.220) (0.220) (0.219)
Fiveyear average CPI inflation
0.021 0.011 0.011 0.011 0.011
(0.018) (0.018) (0.018) (0.018) (0.018)
Log of real GDP per worker
–0.083*** –0.073*** –0.072*** –0.074*** –0.076***
(0.014) (0.012) (0.012) (0.012) (0.012)
Financial intermediation share in total
employment
–1.732***
(0.529)
Private credit to GDP
–0.001
(0.005)
Private credit by banks to GDP
–0.002
(0.006)
Financial system assets to GDP
–0.000
(0.006)
Banking system assets to GDP
0.002
(0.005)
Observations
104 110 110 110 110
Rsquared
0.616 0.583 0.584 0.583 0.583
The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country.
Fiveyear averages for the independent variables are computed over the same period as the dependent
variable. The log of real GDP per worker is the natural logarithm of real GDP per worker for the initial year of
the period over which the averages are computed. The financial intermediation share in total employment is
the share of the financial intermediation sector in total employment for the initial year of the period over which
the averages are computed. Private credit (by banks) to GDP is the ratio of private credit (by banks) to GDP for
the initial year of the period over which the averages are computed. Financial (banking) system assets to GDP
are measured as the ratio of financial (banking) system assets to GDP for the initial year of the period over
which the averages are computed. All estimates include country dummies. Robust standard errors are in
parentheses. Significance at the 1/5/10% level is indicated by ***/**/*. For country sample and sources, see
data appendix.
Overall, the lesson is that big and fastgrowing financial sectors can be very costly for the
rest of the economy. They draw in essential resources in a way that is detrimental to growth
at the aggregate level.
14
4. Conclusion
In this paper, we study the complex real effects of financial development and come to two
important conclusions. First, financial sector size has an inverted Ushaped effect on
productivity growth. That is, there comes a point where further enlargement of the financial
system can reduce real growth. Second, financial sector growth is found to be a drag on
productivity growth. Our interpretation is that because the financial sector competes with the
rest of the economy for scarce resources, financial booms are not, in general, growth
enhancing. This evidence, together with recent experience during the financial crisis, leads
us to conclude that there is a pressing need to reassess the relationship of finance and real
growth in modern economic systems. More finance is definitely not always better.
15
Data appendix
Data sources for Graphs 1–2 and Tables 1–2:
Penn World Tables: real GDP per worker, working population, ratio of imports and exports to
GDP, ratio of government consumption to GDP and CPI.
World Bank Financial Structure and Development database: ratio of private credit to GDP.
50 countries: Argentina, Australia, Austria, Bangladesh, Belgium, Brazil, Canada, Chile,
China, Colombia, the Czech Republic, Denmark, Egypt, Estonia, Finland, France, Germany,
Greece, Hungary, Iceland, India, Indonesia, Ireland, Italy, Japan, Korea, Luxembourg,
Mexico, Morocco, the Netherlands, New Zealand, Nigeria, Norway, Pakistan, the Philippines,
Poland, Portugal, Russia, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland,
Thailand, Turkey, the United Kingdom, the United States, Venezuela and Vietnam.
Data sources for Graphs 3–4 and Tables 3–4:
Penn World Tables: real GDP per worker, working population, ratio of imports and exports to
GDP, ratio of government consumption to GDP and CPI.
World Bank Financial Structure and Development database: ratio of private credit to GDP,
ratio of private credit by banks to GDP, financial system assets to GDP and banking system
assets to GDP.
OECD Structural Analysis database: financial sector’s share in total employment.
21 countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany,
Ireland, Italy, Japan, Korea, the Netherlands, New Zealand, Norway, Portugal, Spain,
Sweden, Switzerland, the United Kingdom and the United States.
16
References
Acemoğlu, D and F Zilibotti (1997): “Was Prometheus unbound by chance? Risk,
diversification, and growth”, Journal of Political Economy, 105, pp 709–75.
Aghion, P, M Angeletos, A Banerjee and K Manova (2010): “Volatility and growth: credit
constraints and the composition of investment”, Journal of Monetary Economics, 57,
pp 246–65.
Beck, T, A DemirgüçKunt and R Levine (2000): “A new database on financial development
and structure”, World Bank Economic Review, 14, pp 597–605.
Beck, T and R Levine (2002): “Industry growth and capital allocation: does having a market
or bankbased system matter?”, Journal of Financial Economics, 64, pp 147–180.
Beck, T, R Levine and N Loayza (2000): “Finance and the sources of growth”, Journal of
Financial Economics, 58, pp 261–300.
Bernanke, B and M Gertler (1989): “Agency costs, net worth, and business fluctuations”,
American Economic Review, 79, pp 14–31.
Bernanke, B, M Gertler and S Gilchrist (1999): “The financial accelerator in a quantitative
business cycle framework”, in J Taylor and M Woodford (eds), Handbook of
Macroeconomics, Chapter 21, Elsevier, pp 1341–93.
Caballero, R, E Farhi and M Hammour (2006): “Speculative growth: hints from the US
economy”, American Economic Review, 96, pp 1159–1192.
Cecchetti, S, M Mohanty and F Zampolli (2011): “The real effect of debt”, BIS Working
Papers, no 352.
Cahuc, P and E Challe (2009): “Produce or speculate? Asset bubbles, occupational choice,
and efficiency”, IZA Discussion Paper, no 4630.
Easterly, Islam and Stiglitz (2000): “Shaken and Stirred: Explaining Growth Volatility”,
Discussion paper, The World Bank.
Goldsmith, R (1969): Financial structure and development, Yale University Press.
Heston, A, R Summers and B Aten (2011): “Penn World Table Version 7.0”, Center for
International Comparisons of Production, Income and Prices at the University of
Pennsylvania.
Holmström, B and J Tirole (1997): “Financial intermediation, loanable funds, and the real
sector”, Quarterly Journal of Economics, 112, pp 663–691.
——— (1998): “Private and public supply of liquidity”, Journal of Political Economy, 106,
pp 1–40.
Islam, N (1995): “Growth empirics: a panel data approach”, Quarterly Journal of Economics,
110, pp 1127–70.
King, B and R Levine (1993): “Finance and growth: Schumpeter might be right”, Quarterly
Journal of Economics, 108, pp 713–737.
Kiyotaki, N and J Moore (1997): “Credit cycles” Journal of Political Economy, 105,
pp 211–48.
Levine, R (1997): “Financial development and economic growth: views and agenda”, Journal
of Economic Literature, 35, pp 688–726.
——— (2005): “Finance and growth: theory and evidence”, Handbook of Economic Growth,
1, Elsevier, pp 865–934.
17
Levine, R, N Loayza and T Beck (2000): “Financial intermediation and growth: Causality and
causes”, Journal of Monetary Economics, 46, pp 3177.
Levine, R and S Zervos (1998): “Stock markets, banks and economic growth”, American
Economic Review, 88, pp 537–58.
Lorenzoni, G (2008): “Inefficient credit booms”, Review of Economic Studies, 75, pp 809–33.
McKinnon, R (1973): Money and capital in economic development, Brookings Institution,
Washington DC.
Organisation for Economic Cooperation and Development (2010): Economic Outlook,
2010/2.
——— (2011): OECD Database for Structural Analysis (STAN).
Pagano, M (1993): “Financial markets and growth: an overview”, European Economic
Review, 37, pp 613–22.
Philippon, T (2007): “Financiers vs engineers: should the financial sector be taxed or
subsidized”, NBER Working Papers, no 13560.
Philippon, T and A Reshef (2009): “Wages and human capital in the US financial industry:
1909–2006”, NBER Working Papers, no 14644.
Rajan, R and L Zingales (1998): “Financial dependence and growth”, American Economic
Review, 88, pp 559–86.
Robinson (1952): “The Generalization of the General Theory”, in The Rate of Interest and
Other Essays (London: Mac Millan).
Shaw, E (1973): Financial deepening and economic development, Oxford University Press,
New York.
18
Previous volumes in this series
No Title Author
380
June 2012
Characterising the financial cycle: don’t lose
sight of the medium term!
Mathias Drehmann,
Claudio Borio and
Kostas Tsatsaronis
379
May 2012
When capital adequacy and interest rate
policy are substitutes
(and when they are not)
Stephen G Cecchetti
and Marion Kohler
378
April 2012
Ensuring price stability in postcrisis Asia:
lessons from the recovery
Andrew Filardo
377
April 2012
Rapid credit growth and international credit:
Challenges for Asia
Stefan Avdjiev, Robert
McCauley and Patrick
McGuire
376
April 2012
Systemic Risks in Global Banking: What Can
Available Data Tell Us and What More Data
Are Needed?
Eugenio Cerutti, Stijn
Claessens, and Patrick
McGuire
375
April 2012
Loan loss provisioning practices of Asian
banks
Frank Packer and
Haibin Zhu
374
March 2012
A Comprehensive Look at Financial Volatility
Prediction by Economic Variables
Charlotte Christiansen,
Maik Schmeling and
Andreas Schrimpf
373
March 2012
Collateral requirements for mandatory central
clearing of overthecounter derivatives
Daniel Heller and
Nicholas Vause
372
February 2012
Inflation Dynamics in the Presence of
Informal Labour Markets
Paul Castillo and Carlos
Montoro
371
February 2012
Stochastic Herding in Financial Markets
Evidence from Institutional Investor Equity
Portfolios
Makoto Nirei, Theodoros
Stamatiou and
Vladyslav Sushko
370
February 2012
Credit at times of stress: Latin American
lessons from the global financial crisis
Carlos Montoro and
Liliana RojasSuarez
369
January 2012
Stresstesting macro stress testing: does it
live up to expectations?
Claudio Borio, Mathias
Drehmann and Kostas
Tsatsaronis
368
January 2012
The sustainability of pension schemes Srichander Ramaswamy
367
December 2011
Is the longterm interest rate a policy victim, a
policy variable or a policy lodestar?
Philip Turner
366
December 2011
Currency Momentum Strategies Lukas Menkhoff, Lucio
Sarno, Maik Schmeling
and Andreas Schrimpf
365
December 2011
Was This Time Different?: Fiscal Policy in
Commodity Republics
Luis Felipe Céspedes
and Andrés Velasco
All volumes are available on our website www.bis.org.
BIS Working Papers are written by members of the Monetary and Economic Department of the Bank for International Settlements, and from time to time by other economists, and are published by the Bank. The papers are on subjects of topical interest and are technical in character. The views expressed in them are those of their authors and not necessarily the views of the BIS.
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2012. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.
ISSN 10200959 (print) ISSN 16827678 (online)
Based on a sample of developed and emerging economies. we show that a fastgrowing financial sector is detrimental to aggregate productivity growth. on 1–2 February 2012. This paper was prepared for the Reserve Bank of India’s Second International Research Conference in Mumbai. India. financial * Cecchetti is Economic Adviser at the Bank for International Settlements (BIS) and Head of its Monetary and Economic Department. E44. and Kharroubi is Economist at the BIS. E22. O4 Keywords: Growth. credit booms. We thank Claudio Borio. after which it becomes a drag on growth. focusing on advanced economies. JEL classification: D92. R&D intensity. iii . and Garry Tang for valuable research assistance. Second. dependence financial development. The views expressed in this paper are those of the authors and not necessarily those of the BIS. we first show that the level of financial development is good only up to a point. Christian Upper and Fabrizio Zampolli for helpful suggestions. Leonardo Gambacorta.Reassessing the impact of finance on growth Stephen G Cecchetti and Enisse Kharroubi* July 2012 Abstract This paper investigates how financial development affects aggregate productivity growth.
.
a more developed financial system is supposed to reduce transaction costs. First. A more comprehensive approach is developed in Holmström and Tirole (1997). we address this question by examining the impact of the size and growth of the financial system on productivity growth at the level of aggregate economies. is fundamental to our thinking. like a person who eats too much. More recently. The idea that an economy needs intermediation to match borrowers and lenders. we have been able to point to evidence supporting the view that financial development is good for growth. see Levine et al (2000) for countrylevel evidence and Rajan and Zingales (1998) for industrylevel evidence. but an engine propelling growth. While there have been dissenting views. and Aghion et al (2010). researchers were able to move beyond simple correlations and establish a convincing causal link running from finance to growth. It requires not only physical capital. At first. but highly skilled workers as well. After all. finance follows”. Introduction One of the principal conclusions of modern economics is that finance is good for growth. see Easterly et al (2000). computers and the like. these results may seem surprising. who provide a model for why different financial patterns (direct finance vs intermediated finance) may coexist altogether. as is the case with many things in life. as well as improving the distribution of capital and risk across the economy. 2 3 1 . Here we find evidence that is unambiguous: faster growth in finance is bad for aggregate real growth. who examine how financial intermediaries can help save on liquidity hoarding. Is it true regardless of the size and growth rate of the financial system? Or. channelling resources to their most efficient uses. For an alternative view. operating through the level and composition of investment. since the pioneering work of Goldsmith (1969). at low levels. Finance literally bids rocket scientists away from the satellite industry.3 But the financial industry competes for resources with the rest of the economy. We present two very striking conclusions.1 This. Our second result comes from looking at the impact of growth in the financial system – measured as growth in either employment or value added – on real productivity growth. And. But there comes a point – one that many advanced economies passed long ago – where more banking and more credit are associated with lower growth. For the more recent work establishing causality. One interpretation of this finding is that financial booms are inherently bad for trend growth. a larger financial system goes hand in hand with higher productivity growth. in turn. who suggest that financial development may only be good up to a point. who in another 1 The view that financial development is simply a byproduct of growth is discussed in Robinson (1952): “Where enterprise leads. McKinnon (1973) and Shaw (1973). does a bloated financial system become a drag on the rest of the economy? In this paper. The result is that people who might have become scientists.1. If finance is good for growth. in the form of buildings. today it is accepted that finance is not simply a byproduct of the development process. are the mechanisms by which financial development improves growth. with finance you can have too much of a good thing. who show how financial development helps reduce the growth cost of economic fluctuations. was one of the key elements supporting arguments for financial deregulation.2 These two channels. raising investment directly. who look at how the presence of financial intermediaries affects the risk return profile of entrepreneurs’ projects. shouldn’t we be working to eliminate barriers to further financial development? It is fair to say that recent experience has led both academics and policymakers to reconsider their prior conclusions. Holmström and Tirole (1998). See Pagano (1993) for a simple analytical model of financial development as a reduction in transaction costs in the context of an AK model. That is. Theoretical contributions relating the role of financial intermediaries to the composition of investment include: Acemoğlu and Zilibotti (1997).
2. starting with the ratio of private credit to GDP. a typical financial boom – employment growth of 1. we examine the impact of financial system size on productivity growth in a sample of 50 advanced and emerging market economies over the past three decades. Using employment measures. that we realise the extent of the overinvestment that occurred. after the bust. we consider both output measures like private credit to GDP as well as input measures like the financial sector’s share in total employment (in this latter case. we examine various measures of financial development. Considering the level of financial development. The inverted Ushaped effect of financial development We begin by examining the relationship between the size of a country’s financial system and its productivity growth to see whether there is a point where bigger is no longer better. it becomes a drag on productivity growth. we examine the impact of the growth rate of the financial system on aggregate productivity growth in a sample of advanced countries over the past three decades. further increases in financial sector size tend to be detrimental to growth.age dreamt of curing cancer or flying to Mars.5% of total employment. our analysis is restricted to advanced economies due to data limitations. In what follows. Importantly. There we find that. over the past 30 years. The remainder of the paper provides the empirical evidence for our conclusions.1 Private credit and growth We begin with a simple histogram constructed from a sample of 50 advanced and emerging countries over the period 1980–2009. Analytical contributions investigating occupational choices between producing and financing include Philippon (2007). In Section 2. compared with a country where the financial sector’s share in total employment is stable. In Section 3. To measure financial sector size. the US banking industry has become relatively skilledlabourintensive. It is only when they crash. shows that GDPperworker growth increases from the first to the 4 Philippon and Reshef (2009) provide empirical evidence that. we restrict our analysis restricted to advanced countries because of limited data availability). Again. office buildings vacated and highly trained people laid off. we find that when private credit grows to the point where it exceeds GDP. today dream of becoming hedge fund managers. computed from a total of 300 data points. 2. which provides a model where human capital is allocated between entrepreneurial and financial careers. Too many companies were formed. Following the dotcom bust innumerable computers were scrapped. and where entrepreneurs can innovate but face borrowing constraints that financiers can help to alleviate. 2 . or the impact of any other boom tied to more tangible output. after the fact. Booming industries draw in resources at a phenomenal rate.4 There is an important sense in which this description of the consequences of a financial boom is no different from those of the dotcom boom of the 1990s. Cahuc and Challe (2009) also develop an analytical model focusing on the allocation of workers between financial intermediation and production sectors in the presence of asset price bubbles. The resulting histogram in Graph 1. we can see that many of these resources should have gone elsewhere. we find that when the financial sector represents more than 3.6% per year – reduces growth in aggregate GDP per worker by roughly one half of 1 percentage point. Using fiveyear nonoverlapping GDPperworker growth and private credit to GDP we compute the average growth conditional on the quartiles of the ratio of private credit to GDP. with too much capital invested and too many people employed.
third private credit to GDP quartile. countries with the highest level of private credit to GDP have lower trend growth than the rest.00 Graphical representation of y k .50 Fiveyear average private credit to GDP (Deviation from country mean) 0.75 –0. But there comes a point where the additional lending and a bigger financial system become a drag on growth.t 5 1 fd k . The result.0 0. The sample covers 50 countries over the period 1980–2009. more credit raises trend growth. To address this.10 0.25 0. Of course. is in Graph 2.0 first quartile second quartile third quartile fourth quartile Each bar represents the fiveyear average GDPperworker growth conditional on the fiveyear average private credit to GDP ratio belonging to a specific quartile of the sample distribution.t .05 –0. we compute deviations from countryspecific means. which grow more slowly for a variety of reasons.t k .t 5 2 y k .0 1.00 –0. For country sample and sources.5 0. GDPperworker growth and private credit to GDP are averaged for nonoverlapping periods over five years. Convergence effects are the most obvious. The relationship is clearly not monotonic.25 0. Graph 2 Private credit to GDP ratio and growth 0.t .t (Deviation from country mean) over the period 1980–2009.10 1. see data appendix. based on the sample of data.t k 0 fd k . That is. For country sample and sources. Graph 1 Average GDPperworker growth by private credit to GDP quartiles 2. 3 . see data appendix. this histogram does not imply that private credit is bad for growth at high levels. at low levels of credit.50 –0.05 Fiveyear average GDPperworker growth 0. That is.00 0.5 1.t 5. Countries with high private credit to GDP are more developed economies. before declining in the final quartile. and control for initial conditions.
The estimates in Table 1 suggest that this increase created a drag of nearly one half of 1 percentage point on trend productivity growth.t 5. including working population growth. 6 7 4 . Beck et al (2000) and Cecchetti et al (2011). we can look at a few examples.t is the log of output per worker in country k in year t. is a constant and k is a vector of country dummies. a figure that is quite close to the threshold of 90% computed in Cecchetti et al (2011). These are reported near the bottom of the table. reaching nearly 150% of GDP by the time of the crisis. or bank assets to GDP. Islam (1995). the squared level of financial development (looking for the parabola in Graph 2). More recently.t+5. To fix ideas and notation. yk. the ratio of Thai private credit to GDP reached 150%. Our hypothesis is that 0 will be positive and 1 negative. from the sample. and (iii) dropping certain countries. financial system deposits to GDP. especially the empirical literature relating growth to finance. The nonlinearity is robust. 0. Notable contributions on this topic include King and Levine (1993).t+5. and to test our hypothesis that the effects of finance on growth can go from good to bad.t 5. Xk. our measure of financial development. Credit then rose steadily. (ii) using alternative measures of financial development like private credit by banks to GDP. we turn to a panel regression.t yk . this measure of financial sector size has fallen to roughly 95%. 1. private credit was below 90% of GDP. The point estimates all roughly 100% of GDP.035 and that on the quadratic term. This time.6 We can use the estimated coefficients to compute an estimate of the peak of the inverted U – the vertical line in Graph 2. which we allow for heteroskedasticity. in the first half of the 1990s. the result is a benefit of roughly one half of 1 percentage point in trend productivity growth. the share of government consumption in GDP and CPI inflation. In the first column of Table 1 we report the result with no controls.t 1 fd k .To get a more precise sense of this relationship.018. Here we see what we expect – the relationship is parabolic.7 To see what these numbers mean.t is the average ratio of private credit to GDP in country k from time t to t+5. Regardless of the exact specification. Continuing across the columns of the table.t+5.t 5. These include: (i) using GDP per capita instead of GDP per worker as the dependent variable. we sequentially add controls. 2005). Comprehensive surveys can be found in Levine (1997.t 5. is around 0. The difference between the estimates is probably a result of differences in data and methods.t 2 (1) where yk. while the latter employs a somewhat more sophisticated econometric model. fdk. and a series of control variables known to influence aggregate growth. together with 95% interval estimates.t k . such as the former communist countries. Thailand is another interesting case.t is a set of control variables averaged from time t to t+5.5 We regress the fiveyear average growth in output per worker in a given country on the following variables: the level of financial development. and is the error term. we write this as: yk . the coefficient of the level of financial development. The current study uses a broader set of countries. 5 Our work in this section builds on an extensive body of research. is always close to –0. Starting with New Zealand.t 2 X k . We note that the results reported in Table 1 are robust to adding a variety of changes to equation (1). bank deposits to GDP.t k 0 fd k . In the runup to the Asian crisis of 1997–98.t is the average growth in output per worker in country k from time t to t+5. openness to trade measured by the ratio of imports and exports to GDP. Levine and Zervos (1998).
01.1. and neither should any readers (or authorities).145 (0. We should be very clear that we do not in any way view these peak debt values as targets.047 (0.01] 270 0.742*** (0.Finally.207) –0.480*** (0. And as discussed in more detail by Cecchetti et al (2011).022*** (0.003) 0.98 1. yield a productivity growth gain of more than 150 basis points.162) 0.06.1.1. divided by 100.045) 0.200) –0. where private credit grew to more than 200% of GDP by the time of the financial crisis.021) –0.478*** (0.0378 (0.005) –1. take the example of the United States.011) –0.213 [1.048** (0.018*** (0. Fiveyear averages for the independent variables are computed over the same period as the dependent variable.98.018*** (0. These are levels of debt that a country should only approach in extremis.00] 270 0.01 0.163) 0.005) –1.020*** (0.220*** (1.1. Column (6) includes country dummies.1.190 [0.005) –0. Robust standard errors are in parentheses.160*** (0.331) 0.08 [0.02 0. by our estimates.471*** (0.0107 (0.99 (1) (2) (3) (4) (5) (6) 0.035*** (0.03] 270 0. The log of real GDP per worker is the natural logarithm of real GDP per worker for the initial year of the period over which the averages are computed.036*** (0.99 0. Significance at the 1/5/10% level is indicated by ***/**/*. All estimates include a nonreported constant.204) –0.035*** (0. see data appendix.160 [0.010) –0.003) 0.010*** (0.017*** (0.97.005) –1.098 [1. The turning point for the effect of private credit to GDP on real GDPperworker growth is the level for private credit to GDP below (above) which an increase in private credit to GDP is estimated to raise (reduce) real GDPperworker growth.208) –0.02] 270 0.054*** (0.160) 0.190 [0.501*** (0.0106 (0.009*** (0. Keeping debt well below 90% of GDP provides the room needed to respond in 5 .152) 0.011) –0.98. which yields six observations per country.424 The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country.110*** (0.036) 1.003) 0.011) –0.110*** (0.011) –0.210) –0. under normal circumstances we would expect to see debt at much lower levels than these thresholds.010*** (0.038*** (0.11] 270 0.008) –6. For country sample and sources.211) 0.046) 0.01] 270 0.685*** (0. Table 1 GDPperworker growth and private credit to GDP Dependent variable: fiveyear average real GDPperworker growth Fiveyear average private credit to GDP Fiveyear average private credit to GDP squared Log of real GDP per worker Fiveyear working population growth Fiveyear average openness to trade Fiveyear average government consumption share in GDP Fiveyear average CPI inflation Turning point for the effect of private credit to GDP on real GDPperworker growth 95% confidence interval Observations Rsquared 0.011) –0.018*** (0.037) 1.162) 0.1.017*** (0.99.035*** (0.005) –1. Reducing this to a level closer to 100% would.
private credit by banks was 160% of GDP at the onset of the financial crisis.1 Bank credit as a measure of financial development Differences in financial system structure imply that credit can mean different things in different countries. debt should be maintained below the level at which borrowing constraints become binding. financial development was measured using total credit extended to the private sector. 2. and then move on to study the relationship between growth and the share of employment accounted for by the financial sector. Table 2 presents the results of this exercise. with bankbased financial systems being one part of the parabola and marketbased financial systems being the other.8 2. Many countries are close to or beyond this level. The results confirm both the parabolic relationship and its robustness. 6 . And this difference could be the driving force behind our results. we start with no controls in the first column and add controls sequentially moving to the right in the table. To examine this possibility. the point estimates themselves are very close to those in Table 1. By contrast.2 Alternative measures of financial development In the previous section. Otherwise. The corresponding figure for the UK was 180% of GDP and even reached 200% of GDP in Denmark. a country like India. We start by looking at the consequences of using bank credit rather than total credit. the inverted Ushape could reflect compositional effects. For example. can still reap significant benefits from further financial deepening in terms of increasing productivity growth.2. Looking at the peak of the parabola. in normal times. For example. the additional accumulation of debt would result in a drag on growth that would make recovery even more difficult than it already is. where bank credit is less than 50% of GDP. in Portugal. we estimate that for private credit extended by banks. Furthermore.the event of a severe shock. should a crisis arise. the turning point is closer to 90% of GDP – somewhat lower than for total credit. 8 This argument is consistent with the results of welfare maximisation. which would imply that. In this section we examine the robustness of this result to the use of alternative measures of financial sector size. suggesting that more credit will not translate into higher trend growth. Again. we replace private credit with private credit by banks (relative to GDP) in equation (1).
0116) –0.93] 269 0.0.215 [1.0368) 0.00962 (0.0. we look at the relationship between the financial sector’s use of the economy’s labour resources and aggregate growth.0445) 0.195) –0. Column (6) includes country dummies. Fiveyear averages for the independent variables are computed over the same period as the dependent variable.0325*** (0.174) –0. one based on inputs. For country sample and sources.0208) –0.00977 (0.00522) –0.0477** (0.0456 (0.197) –0. is the financial sector’s share in the economy’s total employment.98] 269 0.114 (0.0105*** (0.0364 (0.194 [0.194 [0.0185*** (0.0.91 0.0336*** (0.2 Financial sector employment as a measure of financial development The use of credit as a measure of financial development means that we are focusing on the output of the sector.279*** (1.163) 0.00979*** (0. Robust standard errors are in parentheses.426 The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country.0105*** (0. Using a more limited sample drawn from 21 OECD economies over the period from 1980 to 2009. An alternative gauge of financial sector size and financial development.675*** (0. Significance at the 1/5/10% level is indicated by ***/**/*.1.463*** (0.94 0.0118) –0.00520) –0.979*** (0.049*** (0.0121) –0. which yields six observations per country. see data appendix.0.732*** (0. In addition to providing a different measure of financial development.164) 0.0184*** (0.0545*** (0.95] 269 0.00836) –6.02.329) 0.0229*** (0. 2.103 [0.00521) –1.164) (3) 0.0368) (6) 0.90.155) 0.0196*** (0.92] 269 0.93.195) (2) 0.046*** (0.0118) –0.00297) 0.461*** (0.0193*** (0.2.455*** (0.91 0.0108) –0.92] 269 0.0178*** (0. divided by 100.00519) –1. The log of real GDP per worker is the natural logarithm of real GDP per worker for the initial year of the period over which the averages are computed.0373*** (0. The turning point for the effect of private credit by banks to GDP on real GDPperworker growth is the level for private credit by banks to GDP below (above) which an increase in private credit to GDP is estimated to raise (reduce) real GDPperworker growth.90.00543) –1.195) –0.483*** (0.0369*** (0.07] 269 0.Table 2 GDPperworker growth and private credit by banks to GDP Dependent variable: fiveyear average real GDPperworker growth Fiveyear average private credit by banks to GDP Fiveyear average private credit by banks to GDP squared Log of real GDP per worker Fiveyear working population growth Fiveyear average openness to trade Fiveyear average government consumption share in GDP Fiveyear average CPI inflation Turning point for the effect of private credit to GDP on real GDPperworker growth 95% confidence interval Observations Rsquared (1) 0.0453) (5) 0.166) 0. All estimates include a nonreported constant.00296) 0.086*** (0.191) –0.96 0.92 1.04 [0.00295) (4) 0.90.95.0117) –0.0. the analysis using an inputbased measure of financial development provides 7 .161 [0.0334*** (0.
For country sample and sources.000 0.9 Again.t k . which is analogous to Graph 2. Turning to the regression analysis. the financial sector’s share in total employment squared.t 5 1 fs k . an increase in the financial sector’s share in total employment is actually associated with higher GDPperworker growth.t (Deviation from country mean) 0.04 0. We start with a scatter plot in Graph 3.t . This reinforces the idea that the estimated turning point should be regarded not as a target but rather as an upper bound.01 over the period 1980–2009. At low levels.02 0. and various controls.00 –0. 9 Introducing country fixed effects changes the results only modestly.an important check that the inverted Ushaped effect on growth is not simply the result of using a sample which mixes advanced and emerging market economies. the estimated coefficient for the linear and the quadratic term are such that the share of financial intermediation in total employment that maximises growth is lower than the one obtained without country fixed effects. Here we see the result that we expect – the relationship is parabolic. Table 3 presents the results from this exercise.002 0.002 0. In particular. The results confirm our previous results: the relationship between growth and the financial sector’s share in employment is an inverted U. 8 .t k 0 fs k . we estimate equation (1) using the fiveyear average financial sector share in total employment in country as a measure of financial development (fd).t 5 2 y k .t .02 –0.01 –0. we have the financial sector’s share in total employment. But there is a threshold beyond which a larger financial sector becomes a drag on productivity growth. And on the righthand side. on the lefthand side we have the fiveyear average growth in output per worker in a given country.t 5.006 1 Graphical representation of y k . the result is robust to the addition of controls. Again. Graph 3 Financial sector share in employment and growth1 Fiveyear average GDPperworker growth 0. see data appendix.03 –0.004 Fiveyear average financial intermediation share in total employment (Deviation from country mean) –0.004 –0.
665) (2) 3.42] 95 0.417*** (0. to a lesser extent.237) 0.42] 95 0.29.7] 95 0.0586) –0.708) 0.37*** (9.004) –3.00442) 0.602) –103.174) 0.69 [1.360 (0.407*** (0.0581) (5) 3.6*** (35.516) –3.675) –43.111 (0. All estimates include a nonreported constant.331 [1.409*** (0.269) 0.30*** (9.86 3.22.706) –43.174) 0.6.708) 0.574** (2.22.48*** (8.705) –43.335*** (0.333 [1.86 (1) 3.86 (6) 5.347*** (0.00439) 0.00193 (0.0250) 2.333 95 0.181) 0. the United States.86 3. the 9 .00762 (0.28.0236) 3.50] 95 0. which yields six observations per country.0129 (0. Graph 4 shows that in most countries.333 [1. the financial sector’s share in total employment is below or significantly below the threshold beyond which the effect on GDPperworker growth turns from positive to negative. Examples are Canada. Column (6) includes country dummies.82) –6.189) (3) 3. Fiveyear averages for the independent variables are computed over the same period as the dependent variable.20. the size of the financial sector is above this growthmaximising point only in some cases.334*** (0.85 3.345*** (0.6.35*** (9.00431) (4) 3.0243 (0. Switzerland.00195 (0.0171 (0.537) –0.00194 (0. Indeed. For country sample and sources.9% reported at the bottom of Table 3.690) –43.354*** (0. Ireland and.707) 0.331 [1.50] [–6.087*** (1.0173) –0. as was stressed above. The log of real GDP per worker is the natural logarithm of real GDP per worker for the initial year of the period over which the averages are computed.Table 3 GDPperworker growth and financial sector share in employment Dependent variable: fiveyear average real GDPperworker growth Fiveyear average financial intermediation share in total employment Fiveyear average financial intermediation share in total employment squared Log of real GDP per worker Fiveyear working population growth Fiveyear average openness to trade Fiveyear average government consumption share in GDP Fiveyear average CPI inflation Turning point (in %) for the effect of financial intermediation share in total employment on real GDPperworker growth 95% confidence interval Observations Rsquared 3.346*** (0.6.672) 0. To see what these numbers mean. divided by 100.34.00260 (0.341*** (0.00256 (0. Significance at the 1/5/10% level is indicated by ***/**/*.0181 (0.025) –3.6. see data appendix.677) –43. Robust standard errors are in parentheses.00799 (0. However.6. The turning point for the effect of the financial sector’s share in total employment on real GDPperworker growth is the level of the financial sector’s share in total employment below (above) which an increase in the financial sector’s share in total employment is estimated to raise (reduce) real GDPperworker growth.00249 (0.31*** (9. we first look at the recent data for the sample countries and evaluate them against the estimate for the turning point of 3.51] 95 0.980) –3.11.536 The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country.493) –3.
DK = Denmark. JP = Japan. The case of Ireland is interesting because over the period 1995–99. our estimates suggest that Irish trend GDPperworker growth could have been as much as 0. AT = Austria. in particular because it is possible that the negative effect on growth may start materialising for lower levels. 2005–091 6 Average financial sector share in total employment. the share of the Irish financial sector’s share in total employment was 3. CH = Switzerland. From that point of view. KR = Korea.growthmaximising size of the financial sector should not be considered as a target. PT = Portugal. IE = Ireland. US = United States. NZ = NewZealand. for Switzerland 0. IT = Italy. However. 20052009 Growthmaximizing level 95% lower confidence band 5 4 3 2 1 0 AU 1 AT BE CA CH DE DK ES FI FR GB IE IT JP KR NL NO NZ PT SE US AU = Australia. 3. For Canada. which mechanically raises the size of the confidence interval around the estimated turning point.7 percentage points and for Ireland 0.2 percentage points. This means that for all countries. BE = Belgium. the share rose to more than 5%. we examine how financial 10 . Put another way. further increases in financial sector size are most likely to have mixed effects on productivity growth. we can compute the gain in GDPperworker growth if their financial sectors were to shrink back to the growthmaximising point. ES = Spain. FR = France. NO = Norway. Sources: OECD Structural Analysis database. FI = Finland. But over the next 10 years. we now turn to an examination of the impact of the speed of development on productivity growth. Coming back to the countries where the financial sector’s share in total employment is above the growthmaximising point. SE = Sweden. authors’ calculations. NL = Netherlands. the gain is 1. Had the share been constant between1995–99 and 2005–09. all countries in our sample are considerably above the lower band of the 95% confidence interval around the estimate for the turning point.3 percentage points. this result also owes to the limited sample we use. GB = United Kingdom.4 percentage points higher over the past decade.84% – very close to the growthmaximising value. The real effects of financial sector growth Having established that there is a point at which financial development switches from propelling real growth to holding it back. Graph 4 Average financial sector share in total employment. DE = Germany. CA = Canada.
04 1 Graphical representation of y k . using employment data comes at a cost since employment data for the financial sector are available for a limited subset of our previous sample of countries. see data appendix.t over the period 1980–2009.t is the average growth in output per worker in country k from time t to t+5. financial sector growth would be negatively associated with GDP growth by construction.) The result is quite 10 There is a large and wellknown literature on this financial accelerator and its quantitative implications for the business cycle (see Bernanke and Gertler (1989) and Bernanke et al (1999).t 5. there is a significant body of research examining credit cycles (from Kiyotaki and Moore (1997) to more recent work by Caballero et al (2006) on the dotcom bubble or Lorenzoni (2008).00 0. 11 (Deviation from country mean) 11 . unaware of empirical studies on the implications of financial booms for longrun growth. both variables are measured as deviations from their countryspecific means. yk.t+5. Hence we focus on the 21country subset of OECD countries. and k.04 0.sector booms – periods when financial development is moving at a particularly fast pace – can affect growth. where yk.00 –0. on the vertical axis we plot the fiveyear average GDPperworker growth. who look at the normative implications of credit booms). however.02 Fiveyear average financial intermediation employment growth (Deviation from country mean) –0. Graph 5 Financial sector growth and productivity growth1 0.04 0. we use data on employment in the financial sector and measure financial sector growth as the growth rate in the financial sector’s share in total employment. The results reported in this section are robust to the use of financial sector value added in place of the financial sector’s share in total employment. If we were to do this.t is the log of output per worker in country k in year t. fdk. Again.10 Unlike in the earlier exercise. As noted in the previous section. k is a vector of country dummies.t 5.t k 0 fd k .02 –0.t is a residual. we now plot the fiveyear average growth in the financial sector’s share in total employment.11 (As in Graphs 2 and 3. For country sample and sources.02 0. Graph 5 summarises our main finding.t is the average growth in financial intermediation employment in country k from time t to t+5.04 –0.02 Fiveyear average real GDPperworker growth 0. On the horizontal axis.t y k .t k . To bypass this problem. We are. for instance).t+5. we cannot simply rely on the ratio of private credit to GDP to measure financial sector growth. Likewise.
which is also probably related to sample difference.t. we would have GDP growth on the lefthand side of the regression and the inverse of GDP growth on the righthand side. 12 Note that the effect of control variables is relatively different from what it was in the previous regression.striking: there is a very clear negative relationship. The faster the financial sector grows. by our reckoning financial sector growth accounts for one third of the decline in Irish output per worker and 40% of the drop in Spanish output per worker. while output per worker fell by 2. this strikes us as sizeable.7% and 1. During the five years beginning in 2005. Our estimates imply that if financial sector employment had been constant in these two countries. unlike in the earlier exercise. The elasticity estimate of –0. the share of government consumption in GDP. The speed of convergence is also much higher (between 7 and 8% a year) than in the previous regressions. If we were to do this. Our interest is in the first row of the table. CPI inflation and the level of financial development. the slower the economy as a whole grows!In order to verify that this relationship is robust. respectively. this effect survives regardless of the combination and definition of the controls.4%. we start by comparing a country with constant employment in financial intermediation with one in which employment grows at 1.4 percentage points from the decline in Ireland and 0. estimating a panel regression with the fiveyear average annual growth rate in GDP per worker as the dependent variable.t y k . where high government consumption is more likely to have a detrimental effect on the private sector. government size now has a significant negative effect on growth at the margin. We note that the vector of controls in equation (2) includes the growth rate of the working population.1% and 1. it would have shaved 1.3%. In particular. Irish and Spanish financial sector employment grew at an average rate of 4. we now introduce financial sector growth. But. Turning to some country examples.t 5. In other words. In addition to the controls used in equation (1). the coefficient on the financial sector growth. This could be related to the fact that here we focus on advanced economies. which is the average growth in the financial sector’s share in total employment in country k from time t to t+5.33 implies that the first country will grow on average 50 basis points faster than the second country. we cannot simply take the change in the ratio of private sector to GDP as the object of interest. the sample average for those with positive growth. the worse it is for productivity growth (measured as fiveyear average growth in GDP per worker). trade openness measured as the ratio of imports plus exports to GDP.4% per year. we follow the same procedure as before.t (2) where all variables are defined as before.t+5.t k 0 fd k . with the exception of fdk. Table 4 presents the results of estimating equation (2) using a variety of measures of financial sector size as controls. Given that the sample average productivity growth rate is 1.t 5. so finding a negative relationship would be wholly uninformative. we look at Ireland and Spain – admittedly extreme cases. The result evident in Graph 4 is confirmed by more careful statistical analysis: the faster financial sector employment grows.6 percentage points per year.t 1 X k . 12 . Moreover.t 5.t k . It is for this reason that we now measure financial sector growth using employment growth and estimate: y k . which reports the estimates of 0.6 percentage points in Spain.12 To assess the magnitude of the effects.
083) –0.356* (0.0138) –0.002 (0.259 (0. Overall.072*** (0. Financial (banking) system assets to GDP are measured as the ratio of financial (banking) system assets to GDP for the initial year of the period over which the averages are computed.0142) –0.529) (2) –0.325*** (0.006) 0.212) 0.635*** (0.626*** (0.014) –1.204) 0.011 (0.616 110 0.011 (0.018) –0.583 110 0.018) –0.001 (0.327*** (0.0138) –0. They draw in essential resources in a way that is detrimental to growth at the aggregate level.011 (0. Significance at the 1/5/10% level is indicated by ***/**/*.074*** (0.018) –0.023 (0.073) –0.637*** (0.022 (0.006) –0.011 (0. the lesson is that big and fastgrowing financial sectors can be very costly for the rest of the economy.183) 0.005) –0. For country sample and sources.471*** (0.022 (0.073*** (0.762*** (0.188) 0. All estimates include country dummies.018) –0.002 (0.012) (3) –0.219) 0.Table 4 GDPperworker growth and financial sector growth Dependent variable: fiveyear average real GDPperworker growth Fiveyear average financial intermediation employment growth Fiveyear working population growth Fiveyear average openness to trade Fiveyear average government consumption share in GDP Fiveyear average CPI inflation Log of real GDP per worker Financial intermediation share in total employment Private credit to GDP Private credit by banks to GDP Financial system assets to GDP Banking system assets to GDP Observations Rsquared 104 0.286 (0.005) 110 0. Private credit (by banks) to GDP is the ratio of private credit (by banks) to GDP for the initial year of the period over which the averages are computed.012) (5) –0.275 (0.022 (0.219) 0.636*** (0. see data appendix.331*** (0.076*** (0.074) –0.000 (0.191) 0.021 (0.220) 0.074) –0.186) 0.220) 0.012) (4) –0.732*** (0.584 110 0.012) –0.073) –0.270 (0.328*** (0.583 The dependent variable is the fiveyear average real GDPperworker growth for 1980–2009 for each country.0143) –0. The log of real GDP per worker is the natural logarithm of real GDP per worker for the initial year of the period over which the averages are computed.583 (1) –0.083*** (0.007 (0.0148) –0. Robust standard errors are in parentheses. Fiveyear averages for the independent variables are computed over the same period as the dependent variable. The financial intermediation share in total employment is the share of the financial intermediation sector in total employment for the initial year of the period over which the averages are computed.018) –0. 13 .
Second. Conclusion In this paper. in general. leads us to conclude that there is a pressing need to reassess the relationship of finance and real growth in modern economic systems. financial sector size has an inverted Ushaped effect on productivity growth. First. That is. we study the complex real effects of financial development and come to two important conclusions. together with recent experience during the financial crisis. More finance is definitely not always better. 14 . financial booms are not. This evidence. there comes a point where further enlargement of the financial system can reduce real growth. financial sector growth is found to be a drag on productivity growth. Our interpretation is that because the financial sector competes with the rest of the economy for scarce resources. growthenhancing.4.
Switzerland. Indonesia. South Africa. Poland. Data sources for Graphs 3–4 and Tables 3–4: Penn World Tables: real GDP per worker. the United States. Morocco. the Netherlands. 21 countries: Australia. Germany. Denmark. ratio of imports and exports to GDP. working population. 50 countries: Argentina. Switzerland. Egypt. World Bank Financial Structure and Development database: ratio of private credit to GDP. Japan. Australia. Finland. 15 . Sweden. New Zealand. ratio of government consumption to GDP and CPI. Spain. ratio of private credit by banks to GDP. Estonia. the Philippines. Colombia. ratio of imports and exports to GDP. Belgium. OECD Structural Analysis database: financial sector’s share in total employment. Russia. Norway. working population. Korea. Hungary.Data appendix Data sources for Graphs 1–2 and Tables 1–2: Penn World Tables: real GDP per worker. Belgium. New Zealand. Luxembourg. the Netherlands. Germany. Ireland. Portugal. Nigeria. Bangladesh. the Czech Republic. Japan. India. Brazil. Ireland. Finland. Chile. Italy. Austria. World Bank Financial Structure and Development database: ratio of private credit to GDP. Thailand. financial system assets to GDP and banking system assets to GDP. Iceland. Norway. Slovenia. China. Spain. Austria. Greece. Italy. ratio of government consumption to GDP and CPI. Turkey. Venezuela and Vietnam. Denmark. Slovakia. France. Mexico. Sweden. the United Kingdom and the United States. the United Kingdom. Pakistan. Canada. France. Korea. Canada. Portugal.
14. B and J Tirole (1997): “Financial intermediation. and business fluctuations”. Bernanke. Elsevier. Journal of Monetary Economics. 108. diversification. The World Bank. pp 688–726. P. and the real sector”. 79. R Summers and B Aten (2011): “Penn World Table Version 7. Quarterly Journal of Economics. S. pp 1159–1192. pp 865–934. Journal of Political Economy. BIS Working Papers. R (1997): “Financial development and economic growth: views and agenda”. in J Taylor and M Woodford (eds). 112. pp 713–737. pp 1127–70. 105. Journal of Political Economy. loanable funds. N and J Moore (1997): “Credit cycles” Journal of Political Economy. net worth. Beck.0”. 106. and efficiency”. Cahuc. Discussion paper. Handbook of Economic Growth. Beck. American Economic Review. 105. pp 709–75. Journal of Economic Literature. R (1969): Financial structure and development. pp 1341–93. D and F Zilibotti (1997): “Was Prometheus unbound by chance? Risk. pp 261–300. 16 . 110. occupational choice. pp 663–691. M Mohanty and F Zampolli (2011): “The real effect of debt”. Cecchetti. pp 211–48. T and R Levine (2002): “Industry growth and capital allocation: does having a marketor bankbased system matter?”. Income and Prices at the University of Pennsylvania. Quarterly Journal of Economics. IZA Discussion Paper. T. 96. Heston. 57. Easterly. Journal of Financial Economics. Aghion. pp 597–605. 1. B. A. A DemirgüçKunt and R Levine (2000): “A new database on financial development and structure”. N (1995): “Growth empirics: a panel data approach”. E Farhi and M Hammour (2006): “Speculative growth: hints from the US economy”. Bernanke. Holmström. no 4630. King. Caballero. American Economic Review. B and R Levine (1993): “Finance and growth: Schumpeter might be right”. pp 14–31. M Gertler and S Gilchrist (1999): “The financial accelerator in a quantitative business cycle framework”. Chapter 21. T. Elsevier. Islam. R. pp 1–40. pp 246–65. P and E Challe (2009): “Produce or speculate? Asset bubbles. pp 147–180. Handbook of Macroeconomics. Quarterly Journal of Economics. and growth”. 64. World Bank Economic Review. Islam and Stiglitz (2000): “Shaken and Stirred: Explaining Growth Volatility”. ——— (1998): “Private and public supply of liquidity”. B and M Gertler (1989): “Agency costs. Goldsmith. no 352. Yale University Press. Beck. ——— (2005): “Finance and growth: theory and evidence”. Kiyotaki. Journal of Financial Economics.References Acemoğlu. Center for International Comparisons of Production. M Angeletos. Levine. 35. R Levine and N Loayza (2000): “Finance and the sources of growth”. 58. A Banerjee and K Manova (2010): “Volatility and growth: credit constraints and the composition of investment”.
pp 809–33. in The Rate of Interest and Other Essays (London: Mac Millan). Levine. ——— (2011): OECD Database for Structural Analysis (STAN). Robinson (1952): “The Generalization of the General Theory”. 17 . R and L Zingales (1998): “Financial dependence and growth”. pp 559–86. Oxford University Press. M (1993): “Financial markets and growth: an overview”. Philippon. no 13560. G (2008): “Inefficient credit booms”. 37. Review of Economic Studies. Shaw. 75. 88. pp 537–58. European Economic Review. American Economic Review. 2010/2. Brookings Institution. Lorenzoni. R and S Zervos (1998): “Stock markets. T and A Reshef (2009): “Wages and human capital in the US financial industry: 1909–2006”. Philippon. Washington DC. American Economic Review. E (1973): Financial deepening and economic development. R (1973): Money and capital in economic development. 88. N Loayza and T Beck (2000): “Financial intermediation and growth: Causality and causes”. pp 3177. R. NBER Working Papers. Journal of Monetary Economics. banks and economic growth”. NBER Working Papers. T (2007): “Financiers vs engineers: should the financial sector be taxed or subsidized”. Organisation for Economic Cooperation and Development (2010): Economic Outlook. Rajan. 46. no 14644. McKinnon.Levine. New York. pp 613–22. Pagano.
and Patrick McGuire Frank Packer and Haibin Zhu Charlotte Christiansen. Maik Schmeling and Andreas Schrimpf Daniel Heller and Nicholas Vause Paul Castillo and Carlos Montoro Makoto Nirei. Maik Schmeling and Andreas Schrimpf Luis Felipe Céspedes and Andrés Velasco Was This Time Different?: Fiscal Policy in Commodity Republics All volumes are available on our website www. Claudio Borio and Kostas Tsatsaronis Stephen G Cecchetti and Marion Kohler Andrew Filardo Stefan Avdjiev. 18 . Mathias Drehmann and Kostas Tsatsaronis Srichander Ramaswamy Philip Turner Lukas Menkhoff. a policy variable or a policy lodestar? Currency Momentum Strategies Author Mathias Drehmann. Theodoros Stamatiou and Vladyslav Sushko Carlos Montoro and Liliana RojasSuarez Claudio Borio. Lucio Sarno. Robert McCauley and Patrick McGuire Eugenio Cerutti.bis.Previous volumes in this series No 380 June 2012 379 May 2012 378 April 2012 377 April 2012 376 April 2012 375 April 2012 374 March 2012 373 March 2012 372 February 2012 371 February 2012 370 February 2012 369 January 2012 368 January 2012 367 December 2011 366 December 2011 365 December 2011 Title Characterising the financial cycle: don’t lose sight of the medium term! When capital adequacy and interest rate policy are substitutes (and when they are not) Ensuring price stability in postcrisis Asia: lessons from the recovery Rapid credit growth and international credit: Challenges for Asia Systemic Risks in Global Banking: What Can Available Data Tell Us and What More Data Are Needed? Loan loss provisioning practices of Asian banks A Comprehensive Look at Financial Volatility Prediction by Economic Variables Collateral requirements for mandatory central clearing of overthecounter derivatives Inflation Dynamics in the Presence of Informal Labour Markets Stochastic Herding in Financial Markets Evidence from Institutional Investor Equity Portfolios Credit at times of stress: Latin American lessons from the global financial crisis Stresstesting macro stress testing: does it live up to expectations? The sustainability of pension schemes Is the longterm interest rate a policy victim.org. Stijn Claessens.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.