The World’s Largest Open Access Agricultural & Applied Economics Digital Library

This document is discoverable and free to researchers across the

globe due to the work of AgEcon Search.

Help ensure our sustainability.

Give to AgEcon Search

AgEcon Search
h p://ageconsearch.umn.edu
[email protected]

Papers downloaded from AgEcon Search may be used for non-commercial purposes and personal study only.
No other use, including pos ng to another Internet site, is permi ed without permission from the copyright
owner (not AgEcon Search), or as allowed under the provisions of Fair Use, U.S. Copyright Act, Title 17 U.S.C.

No endorsement of AgEcon Search or its fundraising ac vi es by the author(s) of the following work or their
employer(s) is intended or implied.
tt
ti
ti
ti
tt
 Global Trade Analysis Project
https://www.gtap.agecon.purdue.edu/

This paper is from the

GTAP Annual Conference on Global Economic Analysis
https://www.gtap.agecon.purdue.edu/events/conferences/default.asp
 INTERNATIONAL TRADE IN FINANCIAL SERVICES,
COMPETITION, AND GROWTH PERFORMANCE†

Joseph F. Francois
Tinbergen Institute and CEPR

Felix Eschenbach
Tinbergen Institute

This version August 2004

Keywords: financial services trade, financial sector openness, service trade and
imperfect competition, trade in services and growth, banking and growth

JEL codes: [F40], [F13], [F43], [G15]

†. This paper represents our opinions, and is not meant to represent the official position or view of any
organization with which we may have ever been affiliated. All remaining errors are due to confusion
on the part of the authors.
 FINANCIAL SECTOR COMPETITION,
INTERNATIONAL TRADE IN FINANCIAL SERVICES,
AND ECONOMIC GROWTH

This version: August 2004

Abstract: We explore dynamic linkages between financial/banking sector openness,

financial sector competition, and growth. We first develop a model highlighting links
between long-run economic performance and services trade, through scale economies
and market and cost structures in the financial services sector. This is followed by an
econometric exercise based on data for 130 countries for the 1990s. Our results point
to a strong positive relationship between financial sector competition /performance
and financial sector openness (meaning foreign bank access to domestic markets),
and between growth and financial sector competition/performance. They also point
to the presence of scale economies in the sector.

Keywords: financial services trade, service trade and imperfect competition, trade
in services and growth, financial competition

JEL codes: [F40], [F13], [F43]. [G15]

I. INTRODUCTION
In the early neoclassical growth literature, financial services played a passive role, simply
funneling household savings to investors. Goldsmith [1969] and McKinnon [1973] were
among the first to make a break from this mold, emphasizing a more active role for financial
services in promoting growth. Since then, a considerable theoretical and empirical literature
has emerged analyzing the role of finance in growth and development.
This paper explores issues that straddle two different literatures. First, they relate
directly to the nascent literature on trade in services, and to the impact of services trade, in
terms of foreign bank/institutional participation in domestic capital markets, on economic
performance.1 Second, they are also closely related, though not identical, to those issues
found in the strand of the finance and growth literature involving capital market liberalization
and financial flows restrictions.2 The mechanisms we highlight are linked to the pro-
competitive effects of openness, and come on top of those emphasized in the current
literature on financial sector development.

1
Traditionally, the formal trade literature has focused on trade in goods, with the literature on services trade
being a relatively limited and recent addition. While there is a sizable empirical literature on service sector
policy and deregulation, this is largely focused on domestic deregulation. A thorough overview is provided by
WTO (1998).
2
See the survey of Edison, Klein, Ricci, and Sløk (2002).

1
 The recent empirical literature emphasizes two ways in which domestic financial
services affect growth -- capital accumulation and technical innovation. Gains in these areas
can result either in temporarily higher growth rates (transitional or bounded growth effects)
or in permanently higher growth rates. In general, the approach involves employing financial
sector development indicators as independent variables in growth regressions. Most of this
literature has looked at indicators of banking sector development and the degree of private
sector involvement in financial services and the allocation of savings, together with distortion
and financial service cost measures. In addition, a few studies have examined the
relationship between stock-market development and growth. No real emphasis is placed on
the role of traded financial services. By this, we mean focus on identification of a possible
causal chain linking financial sector openness, financial sector performance, and growth
performance.
Among an ever-increasing number of authors3 DeGregorio and Guidotti [1995] report
a significant link between private sector credit and economic growth, while Demetriades and
Hussein [1996] and Jung [1986] find that financial sector development/depth and growth
have a bi-directional relationship. Roubini and Sala-i-Martin [1992] and Mattesini [1996]
report a negative relationship between real interest rate distortions and lending-deposit
spreads on the one hand, and growth on the other. Arestis, Demetriades and Luintel (2001, 5
countries using time-series techniques) find only weak evidence for the hypothesis that stock
market activity accelerates growth. The empirical results of the latter attribute more weight to
the role of banks in promoting growth. More recently sectoral and long-run vs. short run
growth effects have been investigated. Fisman and Love [2004], for instance, use US data to
show that in the short run financial development benefits industries with high growth
potential, while it reallocates resources towards sectors with a ‘natural reliance’ on external
finance in the long run.
While the empirical literature has moved us from assumptions of a passive financial
intermediation mechanism to explicit linkages between intermediation and growth, the role of
open markets in general, and financial sector openness (in a trade sense) in particular, has not
been emphasized. In addition, there is apparent confusion about the different transmission
channels linking international finance with growth. In general authors mean the liberalization
of financial flows when they touch upon this issue. This is of particular importance as
substantial controversy exists about the question of whether capital account liberalization

3
See Levine [1997] and Eschenbach [2004] for surveys of the literature.

2
yields significant long-term growth effects. Its proponents argue that it may foster resource
reallocation from capital abundant to capital scarce countries. On top of that it is asserted that
free capital flows promote productivity growth through greater international risk
diversification (see Obstfeld [1994]). Its opponents, however, point out that liberalization of
financial flows in the presence of pre-existing trade distortions may have immiserizing effects
(see Eichengreen [2001]). Edison, Levine, et al. [2002], mention the importance of sound
policies and institutions in this context. With theory giving no straightforward prediction,
empirical results are no more clear cut. While Edison, Levine et al. find no hard cross-country
evidence linking capital account openness with growth, Edison, Klein, et al. [2002]
summarize the empirical literature such that positive effects are found in industrialized
countries. Whatever the outcome of this strand of research may be, the approach taken here is
fundamentally different. We start with the definition of trade in financial service as foreign
institutional participation in the domestic financial system. Accordingly the growth-
promoting mechanism is different. It works through openness, market structure and
competition, reducing the physical resource cost absorbed (and not invested) by financial
intermediaries and ultimately giving a boost to incentives to save and invest. While the
approach is relatively new in a comprehensive sense, a step in its direction is recent work
linking openness with financial development and performance. For example, Claessens and
Glaessner [1998] have shown that barriers to financial services trade have slowed down the
development of financial markets in East Asia. In a more recent study Guiso et al. [2004] use
Italian industry and firm-level data to show that financial integration accelerates growth by
improving the financial infrastructure. Claessens, Demirgüç-Kunt and Huizinga [2001] have
shown that greater foreign presence reduces profit margins for domestic banks in developing
country financial sectors. In our view, this suggests the first link in the causal chain explored
here between financial sector openness, financial sector performance, and economic growth.
It seems worth mentioning here that our approach also differs from the literature on growth
effects induced by domestic financial liberalization. Andersen and Tarp [2003] for instance,
argue that fostering financial sector competition in the presence of (almost) inevitable market
imperfections may have detrimental growth effects. In the case of trade in financial services,
however, foreign presence in the domestic financial market improves the framework within
which intermediaries operate [see Levine, 2001, for instance]. In this context competition
may even reduce market imperfections.
The approach followed here involves a mix of theory and empirics, and is organized
as follows. First, in Section II we use an analytical model to explore possible linkages

3
between banking/financial sector competition, scale, openness, and growth. This motivates
the empirical exercise in Section III, which is based on a sample of 130 countries, covering
their experience in the 1990s. These data include standard cross-country growth indicators
(macroeconomic stability, inflation, etc.), along with finance sector indicators (financial
openness, banking concentration, etc.). We find evidence for an economically and
statistically significant link between financial sector openness, competition in local financial
service markets, and economic growth. We summarize in Section IV.

II. THEORY
A. Basic Structure
We start by exploring linkages between competition, openness and growth analytically. This
helps to motivate the empirical exercise offered in the next section. We are primarily
concerned with the location of finance firms at the nexus of the savings and investment
mechanism, and the implications of services trade and competition for the working of this
mechanism.
We first assume a national GDP function that is Cobb-Douglas.

Q = AK a L1−a (1)

In equation (1), Q is GDP, K is production capital, L is labor and 0<a<1. The composite Q
also serves as the numeraire good. We assume a Ramsey-type long-run macroeconomic
closure, with constant relative risk aversion (CRRA) preferences defined over consumption
of the composite good Q and with consumers engaged in intertemporal optmization. This
means the model has certain well-known properties. In particular, consumers strike a balance
between present and deferred consumption, yielding the following modified version of the
well-known steady-state condition in equation (2).

r = ρ +δ +φ (2)

In equation (2), r is the return earned by capital in the steady-state, ρ is the subjective rate of
time discount, δ is the rate of capital stock depreciation, and φ is the cost of financial
intermediation (i.e. the payment made in units of numeraire Q as discussed below).

4
 The critical assumption at this juncture is that financial service firms provide a
necessary bridge between savings (i.e. the creation of financial capital) and actual investment
(the creation of physical capital available for investment expenditures). The theoretical
literature on financial intermediation is extensive, and offers numerous alternative
explanations for the observation of intermediation activities.4 Investment projects for instance
have a different size than individuals’ savings. So intermediaries pool funds. There are
information asymmetries and monitoring costs between savers and investors. Accordingly
financial institutions screen and evaluate entrepreneurs and investment opportunities. They
eventually allocate resources, monitor management and exert corporate control. Increasing
sophistication of financial systems allows investors to trade, hedge, diversify and pool risk.
Financial intermediaries also provide trade financing. The reasons for the emergence of
financial institutions, however, are not crucial to our analysis. We expect that they exist for
the range of reasons offered in the literature, and take their existence as given here. What we
do require here is that in the reduced form financial intermediation involves a real resource
cost that drives a wedge between the gross returns earned by physical capital and the net
returns realized by financial capital owners as a basic incentive for saving and investment.
Hence, changes in this incentive drive changes in the evolution of the capital stock.
To facilitate simplification of the analytics we employ several normalizations. These
follow from the following assumption. While we have assumed a concave aggregate
production technology in terms of K and L, we also assume a Ricardian (i.e. linear)
transformation technology between the composite Q and each of its alternative uses as (i.)
consumption good C, (ii.) investment good (physical capital) K and (iii) financial services F.
Hence we define units so that one unit of Q yields one unit of C or K, and we assume that
financial intermediation activities are also scaled so that one unit of financial services (at
price φ) is required per physical capital unit per period. The resource cost and pricing of
financial services is discussed below.
The market for Q is competitive, as are factor markets. Capital and Labor both earn
their value of marginal product measured in units of the numeraire Q. Hence, from the first
order conditions, we will have r = αQ /K . Combining this with the steady-state condition in
equation [2] allows us to derive the following steady-state values (for a given price of
financial services):

4
Examples include Diamond (1984,1991), Leland and Pyle (1977), and Williamson (1987a, 1987b).

5
 Q* = AΨ α / 1−α L
K * = Ψ 1 / 1−α L (3)
S * = δΨ 1 / 1−α L

In equations (3) a * denotes a steady-state value while Ψ = αA /( ρ + δ + φ ) and S denotes the

level of financial savings.
To close the system we specify the competitive structure of financial markets so that φ
is determined along with the other variables in equation (3). To do this we assume a
Cournot-Nash equilibrium in the financial services sector, with constant marginal cost in the
financial services sector (measured in units of Q) represented by b. There are n financial
service firms. For now, assume that the value of n is simply set directly by regulatory
authorities. These firms set quantities strategically in the sense that they are engaged in a
game where they exercise market power by limiting the level of services supplied (or
identically they strategically set the size of the investment basket they are willing to service).
We adopt the classic Cournot assumption. Each firm believes that other firms will not adjust
quantities when it does.
What does equilibrium look like in the region of the steady-state? From equations
(3), we can derive the demand elasticity for financial services in the region of the steady-
state:

 φ  (4)
ε* = − [ ρ + δ + φ ]−1 < 0
1− α 

The Nash equilibrium conditions combined with equation [4] then give us the following
relationship between n and φ in the region of the steady-state.

−bn − (1− a) ρ − (1− a)δ (5)

φ* = −
n − (1− a)

b −bn − (1 − a) ρ − (1 − a)δ b(n − (1 − a)) bn

where ∂φ * ∂n = + < 0 since
2
<
2
. Entry
n − (1 − a) n − (1 − a) (n − (1 − a)) (n − (1 − a))

implies lower prices, and hence through equation (2) higher steady-state capital stocks, with
related implications for the transition path from one steady-state to another. Directly

6
lowering prices through trade will have similar effects, as would entry of foreign banks into a
domestic oligopoly. We explore these issues in the next subsection.
Making a substitution into equation (3) yields the steady-state per-capita capital stock.
1
 n (ρ + δ + b) −(1− a )
k* =   (6)
 aA(n − (1− a)) 

So far we have assumed the number of firms is set exogenously. To close the system, we are
now going to add conditions sufficient to determine the number of firms n. In formal terms,
we specify a limit entry condition. If unit profits are below a critical level π, firms exit, and if
they are above this level, we have entry. The critical level could, for example, represent a
regulatory target for long-term financial institution health. (It could also, of course, be zero).
With symmetry across banks we then have:

cn (7)
φ* = π + +b
Lk *

where c represents fixed costs (if any) and b again represents marginal costs. Together,
equations (5), (6), and (7) are sufficient to define φ, n, and k in the region of the steady-state.

B. Trade and Market Size

There are several ways in which trade may affect long-run capital stocks (and hence
transitional and long-run economic performance) in our analytical framework. The simplest
approach is to assume a small country, with directly imported financial services setting a
maximum price in the domestic market. Regulation and related trade barriers can be assumed
to influence the import price, and hence the domestic price level, directly. From equation (3),
we then have:

2− a
∂k * 1  ρ + δ + φ −(1− a )
= <0
∂φ −(1− a)aA  aA  (8)

If cross-border services trade barriers are reduced, and this leads to a reduction in financial
service prices domestically, we then expect k* to rise as well.

7
 Under the WTO, “trade” is actually defined as a mix of cross-border trade and local
establishment (FDI) in the case of services. We are therefore also interested in the case
where foreign banks are allowed to enter the domestic market, where they then act like other
banks in the local market. For simplicity, we assume they are subject to local cost conditions
in the pure FDI case. This scenario effectively increases the size of n. Assuming that n is
allowed to increase (which may require adjustment of any regulatory target for π set by the
government), we then have the following effect related to entry of foreign banks from
equation (6):

1
∂k *  n( ρ + δ + c) −(1−a )
=  >0 (9)
∂n (n − (1− a))aA 

As in cross-border trade, local establishment can also be expected to have positive medium-
and long-run effects related to the evolution of the capital stock.
What happens (as is often the case in developing countries) if the government sets a
quantity limit for the foreign banking sector, leaving the rest of the domestic market to
domestic firms? If we define K as the regulated size of the foreign banking sector, the
demand elasticity for the domestic sector is directly related to the size of the foreign banking
enclave:

K* (10)
ε˜* = ε *
K * −K

Note that K may be set under our trade or FDI scenario. In either case, working through the
rest of the system as defined above, market power is weakened by an expansion of K ,
implying lower prices and a higher value for k*.
Next, consider market size. The reader can verify that, under constant returns (i.e.
when c=0), the size of the market simply does not matter. However, with scale economies in
the banking sector, size plays a pro-competitive role, leading to entry and an increase in the
overall capital stock k* in the steady-state. If we differentiate the system defined by
equations (5), (6), and (7), we have the following:

8
 ∂n * c ( n * − (1 − a )) n *
2

= >0
∂L δ ,a ,b,c ,π , ρ 2
L [(1 − a )b + (1 − a ) ρ + (1 − a )δ + (1 − a ) Lk * + ( n − (1 − a ))( n − ( 2 − a ))c ] (11)

∂k * ck * (n − (1 − a))
= >0
∂L δ ,a,b,c,π ,ρ L[(1 − a)b + (1 − a) ρ + (1 − a)δ + (1 − a)Lk * +(n − (1 − a))(n − (2 − a))c ] (12)

As is the case with econometric industry studies, we can expect scale economies to imply a
link between country size and service pricing across a sample. In the present context, this
will also be manifested by an indirect linkage between country size, concentration, and k*.

C. Transition Dynamics
The same mechanisms that link services trade in our model with long-run incomes also link
financial sector openness with transitional or medium-term economic growth. Consider, for
example, a constant returns world initially characterized by a closed banking sector and
oligopoly pricing. Starting from the steady-state, prices are given by equation (7), and the
steady-state levels of per-capita capital k* and consumption c* are then given by the two
differential equation system:

k& = 0 = f (k ) − (δ )k − c (13)
c& = 0 = −Θ −1 [ f ' (k ) − ( ρ + δ + φ )] (14)

where Θ is the coefficient of relative risk aversion, and f(.) is the Cobb-Douglas production
function defined in equation (1). From the c& curve we can directly solve for the steady-state
level of per-capita capital k*, and obviously, if we introduce trade in services, such that the
price φ is driven below its steady-state oligopoly value (recall the discussion of equation [8]),
then we will have an increase in steady-state capital stocks. The resulting transition path will
then be standard, involve rising consumption after an initial drop to seed the rise in the capital
stock, and rising capital stocks k. In turn, the growth of the capital stock implies a process of
medium-term transitional economic growth, as the new capital accumulated through equation
[13], fed through equation [1] also then yields GDP growth.

9
III. EMPIRICS
Following our discussion in Section II above, we have a number of candidate relationships.
First, banking sectors in smaller countries may be more concentrated due to economies of
scale in the provision of financial services (equation 11). On top of that we want to test
whether open financial systems tend to foster competition in the banking sector. More
competition, i.e. less concentration in the banking sector, would then drive down market
power as reflected in price. (equation 5). The final link in the chain is between market power
and economic growth. As the financial sector becomes more efficient, we may expect to see
higher rates of capital accumulation and a transition to a higher steady state capital stock per
capita (equations 6, 8, 9, 13). In short, we are interested in indications of higher growth rates
in the transition, all other things being equal, for countries with more open financial systems
and comparable income levels.
We follow the approach of the recent empirical literature. This involves cross-country
growth regressions, wherein we include a number of variables that seem to perform robustly
in the literature.5 To this mix of variables, we also add measures of financial sector openness
and the degree of competition in the financial services sector. Our data are drawn from a
number of sources, and provide a set of indicators for 130 countries for the decade spanning
1990-1999, including most of the transition economies. (These data are available from the
authors upon request.) The variables we work with are summarized in Table 1. We are
ultimately interested in economic growth, for which we take the average growth rate for per-
capita income for the period 1990-99 from the IMF’s World Economic Outlook dataset
(PCGDPGR). Based on the literature, we also work with the standard deviation of inflation
(INFLATE, also from WEO) over this period as an indicator of macroeconomic stability, the
degree of trade openness (TRADE, measured by the share of trade in GDP, taken from the
World Bank’s World Development Indicators), political stability, and a dummy for the
transition economies. Our indicator of political stability (POLSTAB) is from a relatively new
dataset provided by the World Bank, the Worldwide Governance Research Indicators. Initial
per-capita GDP (PCGDP90) is taken from WDI and measured in 1995 US Dollars. It serves
as an overall indicator of base period development. Population growth (POPGR, from WDI)
is also the average for the 1990-99 period. Country size is measured by GDP, and scaled by
world GDP (SIZE, from WDI). We also work with an indicator of financial stability

5
The most recent literature has also explored estimation with small panel datasets, where each set of
observations represents a decade in cross-section. Because many of our financial indicators only became
available over the last ten years, this is not possible in the present context but may be in future work.

10
(FINCRIS) which is from the IBCA Bankscope dataset and measures loan-loss provisions
scaled by net interest revenue. Finding general cross-country measures of the degree of
competition in banking is problematic at best. The measures we work with are rough: the
share of domestic banking assets held by the three largest banks (effectively a proxy for
concentration as developed in the previous section), an index of bank profitability, and a
measure of bank markups (see CONCENT, PROFIT, and NIM in Table 1). The data are from
the bankscope dataset, which, in spite of its widespread use, has been questioned because of
potential underreporting of banks from developing countries. More specifically the
concentration index is subject to controversy as it may be biased upward for countries where
fewer small and domestic banks report. Cetorelli and Gambera [2001], however, have tested
for an underreporting bias and found it to be insignificant. Regardless of this finding we
introduce a correction by taking only the largest one hundred banks as a basis for the
concentration index. By doing so we accomodate the assumption that in OECD countries a
larger share of small banks reports compared to developing countries.
For financial sector openness we have three measures. One is a crude estimate of
tariff-equivalents for financial services trade, based on GATS (General Agreement on Trade
in Services) commitments within the WTO, labeled TARREQ. (For details see Hoekman
1995). A second is the Heritage Foundation’s “Bank Freedom” index, that we call
BANKFREE. It is a subindicator of the index of economic freedom. We use an average of the
issues dating from 1995-1999. This is a reasonable approximation for the 1990s as each
year’s issue refers to a period starting a few years before its publication. Among other
distortions it measures the extent to which foreign banks are allowed to operate in the
domestic financial system. The third is the share of foreign banks in the domestic banking
system measured by the number of foreign banks in the total number of banks for the 1990s
(FOREIGN). This variable is also drawn from the Bankscope database. If foreign banks tend
to be overrepresented for developing countries, this measure would be biased upward rather
than downward. This would weaken instead of support our view so that a correction is not
advisable in order to achieve conservative estimates.
Our approach is to estimate a simple simultaneous system of three equations.

CONCENTi = AX1,i + ε1,i (15.1)

NIMi (or PROFITi) = BX2,i + ε2,i (15.2)
PCGRi = CX3,i + ε3,i (15.3)

11
Because we are interested in the linkages between growth, competition, and openness, a
system estimation approach makes sense. We then have a number of estimation options
available. We offer two full sets of estimates in Tables 2, 3, and 4. We make some
comparison with alternative approaches in Table 5.
The first set of estimates provided in Tables 2, 3, and 4 are OLS estimates. These
take no account of the transmission of errors across equations. The second set, indicated as
mixed least squares (MLS) estimates, does adjust for this issue. Our MLS estimates involve
an approach that strips out the error component associated with endogenous right-hand
variables from our estimates, as discussed below. Under this approach we are able to include
observations (countries) for estimation of some equations, even if sufficient information is
not available for including all countries in the sample in all three equations. This improves
the information content on which our estimates are based.
Given the uneven coverage for some of our indicators (especially when we work with
tariff-equivalent data), if we only worked with countries that could be included in all three
equations, we would exclude a non-trivial share of the information available from the full
dataset. In addition, our measures of bank markups are themselves rough, and we expect
them to be prone to error as well (as implied by the structure of equation 15.2, where
CONCENT appears on the right hand side, and equation 15.3, where NIM or PROFIT appears
on the right hand side). This means we will have some correlation between right hand side
variables and error terms, with transmission of error terms across equations. To handle the
issue of a non-square dataset while still filtering the transmission of error terms, we use an
approach similar to iterative staged estimation methods, though we actually solve the system
explicitly as a simultaneous least squares problem. Basically, where possible we substitute
right hand side endogenous variables with model estimates (or more formally the estimated
means conditional on exogenous variables within the system) to then obtain least squares
estimates for the entire system. This substitution is done simultaneously across equations,
with the full system solved at once. Like comparable instrument variable methods, this
allows us to effectively sweep error terms transmitted from right hand side variables out of
the regression equations above. At the same time, we salvage information in observations
dropped under some other approaches because this approach allows the inclusion of
observations (countries) in any one equation, though they may not appear in all. Note that
without the replacement of right-hand side endogenous variables by system estimates, this
approach simply collapses to the ordinary least squares estimates also reported in the Tables.

12
 Formally, to obtain our parameter estimates, the entire system defined by equations
(15) is estimated simultaneously as a constrained minimization problem for the sum of the
system squared errors in GAMS (a non-linear programming language used for large-scale
numerical problems). 6 The constrained minimization problem is

n 3
min ∑ ∑εi,2 j
j=1 i=1 (16)
s.t. 15.1, 15.2, 15.3

Because we solve the system (including right hand side substitutions) simultaneously, while
mixing estimates based on (potentially) different though largely overlapping sets of
observation units (countries), we refer to these as “mixed least squares” or MLS-based
estimators.
We are first interested in the relationship between financial sector openness and our
competition index. These estimates (equation 15.1) are presented in Table 2. In equation
15.1, X1 includes SIZE and one of the openness indicators. SIZE is included because (see
Section II) larger markets imply more scope for competition when scale economies are
present. TARREQ, BANKFREE, and FOREIGN are included as our measures of financial
sector openness. The SIZE variable emerges as consistently significant at the .01 level.
Smaller economies are highly correlated with a greater degree of concentration. Critical to
the present exercise, our financial sector openness variables, TARREQ, BANKFREE, and
FOREIGN, all emerge with coefficients that are significant at the .01 level as well. We will
examine the policy "significance" of these coefficients, in terms of the size of this effect, after
we look at growth.
Consider next the link between our competition index and financial pricing (measured
by NIM and PROFIT). These are included in X2 above, and are reported in Table 3. They
emerge, under all specifications, with significance at the .01 level. Our parameter for
financial crisis is less robust, typically being generally significant at between the .01 and .15
level.
Growth results are reported in Table 4. Generally, the standard cross-country growth
variables emerge with significant coefficients in the .01 to .05 range. Our measures of

6
GAMS code is available from the authors upon request. Solutions to the problem of system estimation under
such conditions can be traced to Theil (1953). The present approach combines instrument variable strategies for

13
financial sector performance, NIM and PROFIT, all emerge with significant coefficients at
the .010 level in both the OLS and MLS estimates. The estimated qualitative effect is highly
robust to the model specification chosen (in terms of openness indicator). We do identify a
stronger effect with the MLS estimates than is suggested by the OLS estimates, though in all
cases this is significant under both approaches.
For further comparison, Table 5 takes the first system of equations, and reports results
for SUR system estimates (from STATA). Because standard SUR and staged estimation
procedures require a square dataset (i.e. observations appear in all 3 equations), these
estimates mean we lose observations in all equations, and end up with 114 observations
across all three equations. In Table 5, these are compared to OLS estimates for the same 114
observations, and MLS estimates for the same 114 observations. We then also report OLS
and MLS estimates for the full sample. Comparable results follow for the SUR and MLS
estimates, while both differ in a similar manner from the reported OLS estimates.
From the results for equations 15, we have identified the following pattern in the data.
Open financial sectors appear to be more competitive, leading ultimately to lower financial
service prices and profits. Those regimes featuring more competitive financial sectors are in
turn strongly linked with higher growth rates. Note also that this effect comes in addition to
the effect of other financial variables, as is highlighted in the established literature. The
coefficients of the standard growth regression variables including the dummy for transition
economies all have the expected signs and generally emerge at high significance levels as
well.
Finally we are interested in how strong the identified effects are. Within our sample,
protection in the financial services sector (or identically closed financial sectors) is
concentrated in the lower income countries. This is illustrated in Figures 2 and 3. The
OECD countries in the sample tend to have the most open financial service sectors, so that
the question of gains from liberalization can also be viewed as one particularly relevant for
developing countries. Figures 1 and 2 present a picture of differences in financial openness
mapping to differences in growth. However, many of these differences also follow from
other factors (hence the need for an approach like the regressions above). Consider our
estimates for the BANKFREE and TAREQ openness indicators. Working from these
estimates, and holding all else constant, we can characterize what happens in our sample
when a “typical” lower income country moves from the openness level characterizing the

non-linear system estimation with sparse information sets methods now being explored for large-scale general
equilibrium system estimation. SUR estimates in Table 5 have been produced with STATA8.

14
average lower income regime to the more liberal openness level characterizing the average
higher income regime (i.e. a roughly 50% liberalization of financial services trade by these
measures). Based on the coefficients in Tables 2-4, this is associated with an increased degree
of competition in the financial services sector and in turn with growth rates that are then
higher (in the range of 0.4 percent to 0.6 percent per year on a per capita basis). This is in
line with other recent estimates of financial development and growth linkages, though the
mechanism is different, being grounded in market structure and competition.

IV. SUMMARY
Recent empirical studies have applied both endogenous and bounded growth frameworks to
assess the effect of financial service sector development on growth rates and per-capita
income levels. In general, the approach involves employing financial sector development
indicators as independent variables in growth regressions. Most of this literature has looked at
indicators of banking sector development and the degree of private sector involvement in
financial services and the allocation of savings, and at distortion and financial service cost
measures. In addition, a few studies have examined the relationship between stock-market
development and growth.
Along these lines, this paper examines the pro-competitive effects of trade in financial
services. We highlight the role of financial services at the nexus of the savings and
accumulation mechanism that drives economic growth. Following a brief review of the
literature, we develop an analytical model in Section II characterized by Ramsey
accumulation and an oligopolistic financial services sector. This model is used to highlight
channels through which financial services trade may lead to dynamic, pro-competitive
effects. This analytical exercise helps motivate the econometric exercise in Section III. In
cross-country regressions on data for 130 countries in the 1990s, we find that there is a strong
positive relationship between financial sector competition and financial sector openness, and
between growth and financial sector competition.

15
REFERENCES

Andersen, T. B., and Finn Tarp, “Financial Liberalization, Financial Development and
Economic Growth in LDCs,” Journal of International Development 15 [2], March
[2003], 189-209.

Arestis, P., Demetriades, P.O., and Luintel, K. B., " Financial Development and Economic
Growth: The Role of Stock Markets," Journal of Money, Credit, and Banking 33 (1),
February [2001], 16-41.

Blanchard, O.J. and Fischer, S., Lectures in Macroeconomics, MIT Press: Cambridge MA
[1989].

Bureau van Dijk / Fitch IBCA , Bankscope Database, Brussels [2000].

Cetorelli, N., and Gambera, M., “Banking Market Structure, Financial Dependence and
Growth: International Evidence from Industry Data,” The Journal of Finance, April
[2001], 617-648.

Claessens, S., Demirguc-Kunt, A., and Harry Huizinga "How Does Foreign Entry Affect the
Domestic Banking Market," Journal of Banking and Finance 25 (5), [May 2001],
891-911.

Claessens, S., and Glaessner, T., "The Internationalization of Financial Services in Asia,"
World Bank PRWP no. 1911 [1998].

De Gregorio, J., Guidotti, P., "Financial Development and Economic Growth," World
Development 23(3), March [1995], 433-48.

Demetriades, P.O., Hussein, K.A., "Does Financial Development Cause Economic Growth?
Time-Series Evidence from Sixteen Countries," Journal of Development Economics
51(2), [December 1996], 387-411

16
Diamond, D.W., "Financial Intermediaries and Delegated Monitoring," Review of Economic
Studies [July 1984], 393-414.

Diamond, D.W., "Monitoring and Reputation: the Choice Between Bank Loans and Directly
Placed Debt," Journal of Political Economy, [August 1991], 689-721.

Edison, H., Klein, M., Ricci, L., and Torsten Sløk, “Capital Account Liberalization and
Economic Performance: A Review of the Literature,” IMF Working Paper 02/120
2002 (Washington: IMF) .

Edison, H., Levine, R. Ricci, L., and Torsten Sløk, “International Financial Integration and
Economic Growth,” Journal of International Money and Finance [November 2002],
749-776.

Eichengreen, B., “Capital Account Liberalization: What Do the Cross-Country Studies Tell
Us ?,” University of California, Mimeo [2001].

Eschenbach, F., “Finance and Growth: A Survey of the Theoretical and Empirical
Literature,” Tinbergen Institute Discussion Paper 2004-039/2.

Fisman, R., and Inessa Love, “Financial Development and Growth in the Short and Long
Run,” NBER Working Paper 10236 [2004].

Goldsmith, R. W., Financial Structure and Development. New Haven, CT: Yale U. Press
[1969].

Guiso, L., Jappelli, T., Padula, M., and Marco Pagano, “Financial Market Integration and
Economic Growth in the EU,” CEPR Discussion Paper 4395 [2004].

Hoekman, B., “Assessing the General Agreement on Trade in Services,” in Will Martin and
L. Alan Winters (eds.) The Uruguay Round and the Developing Economies,
Discussion Papers No. 307, Washington DC.: The World Bank [1995].

17
Levine, R. , “Financial Development and Economic Growth: Views and Agenda,” Journal of
Economic Literature [June 1997], 688-726.

Levine, R., “International Financial Liberalization and Economic Growth,” Review of

International Economics 9[4], [2001], 688-702.

Leland, H. and Pyle, D. "Informational Asymmetries, Financial Structure, and Financial

Intemediation," Journal of Finance [May 1977], 371-387.

Mattesini, F., "Interest Rate Spreads and Endogenous Growth," Economic Notes, 25[1],
[1996], 111-29

McKinnon R.I., Money and Capital in Economic Development, Washington D.C.: Brookings
Institution [1973].

Obstfeld, Maurice (1994). “Risk-Taking, Global Diversification, and Growth,” American

Economic Review [December 1994], 1310-1329.

Roubini, N. and Sala-i-Martin, X., "Financial Repression and Economic Growth," Journal of
Development Economics 39 [1992], 5-30.

Williamson, S.D., "Costly Monitoring, Loan Contracts, and Equilibrium Credit Rationing,"
The Quarterly Journal of Economics [February 1987a], 135-145.

Williamson, S.D., "Financial Intermediation, Business Failures, and Real Business Cycles,"
Journal of Political Economy [December 1987b], 1196-1216.

World Trade Organization (WTO), "Economic Effects Of Services Liberalization: Overview

Of Empirical Studies," Background Note by the WTO Secretariat, document
S/C/W/26/Add.1, 29 [May 1988].

18
Figure 1

Foreign Banking Restrictions and Growth in the 1990s

(Bank Freedom Index as a Proxy for Restrictions on Foreign Banks)

1.00
2.00
3.00
4.00
3.00
5.00

2.00
Low Medium High

1.00
average annual growth rate

0.00

-1.00

-2.00

-3.00 High Income

es
Middle Income
tri

-4.00
un
co

Lower Middle Income

-5.00

Low Lower Income

Medium
High
level of foreign banking restrictions

note: Growth rates are per-capita, while banking sector openness is based on the banking freedom index.
High income countries (24 in total) have 1990 GDP per capita above US $10,000 ; medium income countries (32 total)
have incomes above US $2,500, lower middle income countries have incomes above US $1,000 per capita (27 in total)
and lower income countries constitute the remainer (43 countries).
Low bank restrictions have an index value of 1 to 2.33, medium ranges from 2.4 to 3.66, and high ranges from 3.7 to 5.
The sample of high income countries with high banking restrictions includes only Greece.

19
Figure 2

Foreign Banking Restrictions and Growth in the 1990s

(Tarrif Equivalents as a Proxy for Restrictions on Foreign Banks)

3.00
2.50
2.00
average annual growth rate

1.50
1.00
0.50
0.00
-0.50 High Income
-1.00 Middle Income
-1.50 Lower Middle Income
-2.00 Lower Income
Low to
Medium High

High income countries (23 in total) have 1990 GDP per capita above US $10,000 ; medium income countries (32 total)
have incomes above US $2,500, lower middle income countries have incomes above US $1,000 per capita (27 in total)
and lower income countries constitute the remainer (43 countries).
Low to medium bank restrictions have a tariff equivalent index below 35. High ranges from 35 to 50.
The OECD countries with an index below 20 include Sweden, the United States, Canada, Switzerland, Austria,
Australia, New Zealand, and Finland.

20
Table 1: Overview of dataset

Variable Description Source

1. A Macroeconomie indicators

FINCRIS Financial crisis indicator (based on economy-wide loan loss IBCA Bankscope
provisions over net interest revenue
INFLATE The standard deviation of the inflation rate over the 1990-99 period. IMF WEO

PCGDP90 Per-capita GDP in 1990. World Bank WDI

PCGDPGR The average of per-capita growth over the 1990-99 period. IMF WEO

POLSTAB Political stability indicator from –2.5 to 2.5 (-2.5=most unstable, World Bank
2.5=most stable). Worldwide Gover-
nance Research
Indicators
POPGR Average rate of population growth over the 1990-99 period. World Bank WDI

SIZE Total value of GDP, averaged over 1990-99, and scaled by total value World Bank WDI
of world GDP.
TRADE Exports and imports as a share of GDP, averaged over the 1990-99 World Bank WDI
period.
TRANSEC Transition economy (1=yes, 0=no). 26 countries

1. B Financial sector indicators

BANKFREE Openness of banking sector in terms of restrictions on ability of Heritage Foundation

foreign banks to open branches and subsidiaries, barriers to domestic
bank formation, government influence over credit allocation,
government ownership of banks, government regulations like deposit
insurance, and restrictions on providing all kinds of financial services
from 1 to 5 (1= very low restrictions, 5=very high restrictions).
CONCENT Concentration in the financial sector: the assets of the largest three IBCA Bankscope
banks as a share of total assets in percent (corrected for income by
using only largest 100 banks' assets), averaged over 1990-99 period.
FOREIGN The share of the banking sector accounted for by foreign banks. IBCA Bankscope

NIM Net interest income over total banking assets in percent, IBCA Bankscope
averaged over 1990-99 period.
PROFIT Commercial banks’ gross operating profits over total assets in percent, IBCA Bankscope
averaged over 1990-99 period.
TARREQ Estimated tariff equivalent of trade protection of the domestic banking Hoekman 1995,
and financial services sector in percent as derived from WTO Members' see references
GATS commitments in financial services (excluding insurance);
ranges from 0 (free) to 50 (most protectionist). Extended from the original
set reported by Hoekman (1995) to include transition economies.

21
Table 2. Concentration and financial openness: Dependent variable CONCENT

Market Power Equations OLS/MLS OLS/MLS OLS/MLS

system (1), (4) (2), (5) (3), (6)
SIZE -2.71 -3.28 -1.46
-(4.63)*** -(4.96)*** -(2.92)***

TARREQ 0.69
(3.53)***

FOREIGN -1.20
-(3.87)***

BANKFREE 14.51
(8.23)***

TRANSEC 8.94 12.00 16.57

(2.12)** (2.50)*** (3.07)***

OBS 128 121 77

R-squared 0.483 0.308 0.604

Systems (1) – (3) dependent variable NIM, systems (4) – (6) dependent variable PROFIT
*** denotes significance at the .01 level for a two-tailed test, ** denotes significance at the .05 level, and *
denotes significance at the .10 level.

Table 3. Intermediation costs and concentration: Dependent variables NIM and PROFIT
OLS MLS OLS MLS
system (1),(2),(3) (1) (2) (3) (4),(5),(6) (4) (5) (6)
CONCENT 0.04 0.06 0.05 0.04 0.02 0.03 0.03 0.02
(3.19)*** (5.47)*** (4.06)*** (3.21)*** (3.60)*** (4.65)*** (5.27)*** (3.70)***

FINCRISIS -0.02 -0.02 -0.02 -0.01 -0.02 -0.01 -0.01 -0.01

-(2.41)*** -(1.97)** -(1.58) -(1.54) -(3.57)*** -(1.55) -(1.61) -(1.10)

TRANSEC 2.95 2.53 2.64 2.73 1.75 1.47 1.39 1.47

(3.98)*** (3.32)*** (3.54)*** (3.67)*** (4.78)*** (3.91)*** (3.67)*** (3.92)***

OBS 123 123 123 123 120 120 120 120

R-squared 0.223 0.236 0.18 0.191 0.293 0.22 0.225 0.221

Systems (1) – (3) dependent variable NIM, systems (4) – (6) dependent variable PROFIT.
*** denotes significance at the .01 level for a two-tailed test, ** denotes significance at the .05 level, and *
denotes significance at the .10 level.

22
Table 4. Per-capita GDP growth and market-power: Dependent variable PCGDPGR
OLS MLS OLS MLS
system (1),(2),(3) (1) (2) (3) (4),(5),(6) (4) (5) (6)
PCGDP90 -9.62E-05 -1.05E-04 -1.04E-04 -1.03E-04 -9.17E-05 -1.07E-04 -1.07E-04 -1.07E-04
-(3.11)*** -(3.16)*** -(2.93)*** -(2.75)*** -(2.97)*** -(3.09)*** -(3.19)*** -(2.79)***

POPGR -0.64 -0.42 -0.42 -0.44 -0.42 -0.31 -0.31 -0.31

-(2.58)*** -(1.59) -(1.50) -(1.47) -(1.63) -(1.09) -(1.11) -(0.97)

TRADE 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01

(1.54) (2.26)** (2.10)** (1.96)** (1.89)* (2.19)** (2.27)** (1.99)**

INFLATE -7.16E-04 -8.34E-04 -8.33E-04 -8.32E-04 -1.00E-03 -9.41E-04 -9.42E-04 -9.43E-04

-(2.47)*** -(2.69)*** -(2.51)** -(2.38)** -(4.20)*** -(3.04)*** -(3.13)*** -(2.74)***

NIM -0.20 -0.47 -0.62 -0.71

-(2.12)** -(4.61)*** -(5.72)*** -(6.22)***

PROFIT -0.428 -1.078 -0.97 -1.4

-(3.16)*** -(7.10)*** -(6.58)*** -(8.31)***

POLSTAB 1.25 1.24 1.24 1.27 1.37 1.34 1.33 1.33

(3.62)*** (3.34)*** (3.14)*** (3.04)*** (3.93)*** (3.42)*** (3.50)*** (3.06)***

TRANSEC -4.51 -3.19 -2.76 -2.54 -4.10 -2.93 -3.14 -2.42

-(6.06)*** -(4.00)*** -(3.22)*** -(2.82)*** -(5.13)*** -(3.27)*** -(3.62)*** -(2.44)**

OBS 115 115 115 115 112 112 112 112

R-squared 0.525 0.544 0.543 0.539 0.563 0.553 0.554 0.554

Systems (1)–(3) dependent variable NIM, systems (4)–(6) dependent variable PROFIT.
*** denotes significance at the .01 level for a two-tailed test, ** denotes significance at the .05 level, and *
denotes significance at the .10 level.

23
Table 5. Comparison of unrestricted OLS and MLS estimates with OLS, SUR, and MLS
estimates for a balanced sample, system (1).
unrestricted sample
restricted sample estimates estimates
EQUATIONS OLS SUR MLS OLS MLS
Growth
PCGDP90 -1.01E-04 -1.11E-04 -1.06E-04 -9.62E-05 -1.05E-04
-(3.29)*** -(3.89)*** -(3.10)*** -(3.11)*** -(3.16)***

POPGR -0.66 -0.49 -0.43 -0.64 -0.42

-(2.68)*** -(2.15)** -(1.58) -(2.58)*** -(1.59)

TRADE 0.01 0.01 0.01 0.01 0.01

(1.47) (2.24)** (2.19)** (1.54) (2.26)**

INFLATE -6.68E-04 -6.35E-04 -8.34E-04 -7.16E-04 -8.34E-04

-(2.32)** -(2.37)** -(2.60)*** -(2.47)** -(2.69)***

NIM -0.25 -0.55 -0.61 -0.20 -0.47

-(2.52)** -(6.00)*** -(5.56)*** -(2.12)** -(4.61)***

POLSTAB 1.26 1.13 1.26 1.25 1.24

(3.67)*** (3.54)*** (3.30)*** (3.62)*** (3.34)***

TRANSEC -4.65 -3.87 -3.63 -4.51 -3.19

-(6.26)*** -(5.35)*** -(4.39)*** -(6.06)*** -(4.00)***
Pricing
CONCENT 0.03 0.05 0.05 0.04 0.06
(2.90)*** (5.57)*** (5.07)*** (3.19)*** (5.47)***

FINCRISIS -0.02 -0.02 -0.01 -0.02 -0.02

-(2.71)*** -(2.29)** -(1.89)* -(2.41)** -(1.97)**

TRANSEC 1.88 1.46 1.44 2.95 -2.53

(3.10)*** (2.40)** (2.31)** (3.98)*** (3.32)***
Market Power
SIZE -2.47 -2.38 -2.49 -2.71 -2.71
-(4.09) -(4.12) -(4.08)*** -(4.63)*** -(4.63)***

BANKFREE 13.81 13.44 13.81 14.51 14.51

(6.62)*** (6.78)*** (6.62)*** (8.23)*** (8.23)***

TRANSEC 10.72 10.87 10.72 8.94 8.94

(2.31)** (2.39)** (2.31)** (2.12)** (2.12)**

OBS: PCGR 114 114 114 115 115

OBS: NIM 114 114 114 123 123
OBS: CONCENT 114 114 114 128 128
R-squared: PCGR 0.536 0.480 0.536 0.525 0.544
R-squared: NIM 0.194 0.150 0.184 0.223 0.236
R-squared: CONCENT 0.421 0.421 0.421 0.483 0.483
*** denotes significance at the .01 level for a two-tailed test, ** denotes significance at the .05 level, and *
denotes significance at the .10 level.

24