41 Pages

Topological multiple recurrence for polynomial configurations in ...

Gain access to the library to view online
Learn more


  • exposé
XPEH-TEX Topological multiple recurrence for polynomial configurations in nilpotent groups V. Bergelson, A. Leibman Abstract We establish a general multiple recurrence theorem for an action of a nilpotent group by homeomorphisms of a compact space. This theorem can be viewed as a nilpotent version of our recent polynomial Hales-Jewett theorem ([BL2]) and contains nilpotent extensions of many known “abelian” results as special cases. 0. Introduction 0.1.
  • van
  • accordance with the principles of ramsey theory
  • monomial mappings
  • nilpotent group
  • polynomial mapping
  • finite coloring
  • nilpotent group of self-homeomorphisms
  • metric space
  • theorem



Published by
Reads 15
Language English

Credit Constraints in Trade:

Financial development
and export composition

Kalina Manova
Harvard University

First draft: Feb 23, 2005
This draft: Nov 25, 2005

Abstract. Traditional explanations of export composition focus on comparative advantage arising
from relative factor endowments and production technologies. This paper provides evidence of an
additional comparative advantage channel based on the level of financial development. I argue that
potential exporters face borrowing constraints and that their capacity to enter an industry depends
on the sector’s asset structure and reliance on external financing. I find that countries with better-
developed financial systems tend to export relatively more in highly external capital dependent
industries and in sectors with fewer tangible assets that can serve as collateral. I establish causality
in a panel of 107 countries and 27 industries in 1985-1995, and isolate a financial channel
independent of other institutions, factor endowments, or the overall level of development. I also
find that equity market liberalizations increase exports disproportionately more for sectors more
reliant on outside funding or characterized by softer assets. This effect is more pronounced in
countries with initially less active stock markets, suggesting that foreign equity flows may
substitute for an underdeveloped domestic financial system.

I thank Pol Antras, Elhanan Helpman, and Marc Melitz for their invaluable guidance. I also thank
Philippe Aghion, Davin Chor, Gita Gopinath, Dirk Jenter and Nathan Nunn for insightful
conversations, and the participants of the Harvard International Lunch for helpful comments.
1. Introduction
The standard Heckscher-Ohlin model predicts that a country rich in labor, natural resources,
physical or human capital has a comparative advantage in goods intensive in the abundant input
factors. This view abstracts from market frictions that may arise from agency problems, and
presumes that entrepreneurs can enter any industry regardless of its need for outside finance or
endowment of collaterizable assets. In the presence of financial frictions, however, borrowing
constraints will vary across industries and affect the sectoral composition of a country’s exports by
limiting the investment opportunities open to producers with insufficient private capital.
A small but growing literature on finance and trade has indeed found suggestive evidence of
an additional comparative advantage channel based on the level of financial development. In
particular, a number of papers have argued that financially developed countries export relatively
more in sectors that require more outside finance. However, the cross-sectional approach and
focus on worldwide exports by sector in these studies has made it difficult to establish a causal
link from finance to growth. Moreover, the financial channel has been confounded with the effects
of other institutions, making the results difficult to interpret.
This paper exploits the time dimension in a panel of bilateral exports by sector and examines
the effects of equity market liberalizations on trade to address these concerns. I show that
countries with better-developed financial systems tend to export relatively more in highly external
capital dependent industries and in sectors with few tangible assets that can serve as collateral.
This result is not driven by factor endowments or the overall level of development. I establish
causality by using lagged values of financial development in a country-fixed-effects regression
with an annual panel of 107 countries and 27 industries in 1985-1995. To isolate an independent
financial channel I carefully control for the effects of other institutions in both the exporting and
the importing country. I also consider the impact of an exogenous shock to the availability of
outside finance, which is arguably orthogonal to other institutional developments: equity market
liberalizations. I find that liberalizations increase exports disproportionately more in sectors
intensive in external finance and with softer assets. Moreover, this effect is more pronounced in
countries with initially less active stock markets, suggesting that foreign equity flows may
substitute for an underdeveloped domestic financial system.
These findings contribute to the recent empirical literature on the role of financial institutions
for trade. For example, in a cross-section of 56 countries and 36 industries, Beck (2003) finds that
the average 1980-1990 export share of industries that use more outside funds is higher in
financially developed countries. In another cross-sectional analysis for 1995, Becker and
Greenberg (2004) reach a similar conclusion using different industry measures of fixed upfront
1costs. Similarly, Hur et al. (2004) show that a better financial environment (as well as many other
institutions) is associated with a larger 1980-1990 average share of exports in sectors with fewer
internal funds and hard assets.
A weakness of all prior studies is that they present cross-sectional analyses of worldwide
exports by sector, which exposes the results to reverse causality. To address this concern
researchers have instrumented for private credit with legal origin. (Private credit is the most
commonly used measure of financial development and is also my main measure.) However, legal
origin has been shown to impact institution formation and the economy more broadly, which in
2turn are likely to affect sectors and sectoral exports differentially. It is thus not obvious that this
instrument meets the exclusion restriction. In contrast, my annual panel allows me to exploit the
variation in the level of financial development over time by using lagged values of private credit to
explain current export flows. My results remain unchanged or stronger when I hold private credit
fixed at its value immediately before 1985 (the first year in the sample) or use moving lagged 5-
year averages. Since lagged measures of financial development are not contaminated by current
exports, these findings constitute more conclusive evidence of a causal link from financial
development to export composition than has been previously shown. Moreover, I show that
instrumenting with legal origin and creditor rights protection produces higher coefficient point
estimates, which may reflect the role of other variables that covary with the instruments.
The prior literature has also confounded the effects of financial development with those of
other institutions. Establishing a separate role for financial development is problematic because it
tends to be highly correlated with many desirable institutional features. While some of the above
studies use indices such as accounting standards and creditor and property rights protection, they
never include them in the same regression as financial development. Instead, they interpret the
institutional results as robustness checks for the estimation with private credit. This is justified to
the extent that these characteristics are related to the ability to raise capital.

Svaleryd and Vlachos (2005) find similar results by constructing an index of industrial specialization.
La Porta et al. (1997) find that legal origin is a strong predictor of current financial development, measured by either
total credit to the private sector or stock market capitalization. However, La Porta et al. (1998) show that legal origin
also predicts the level of rule of law, corruption, the efficiency of the judiciary, the risk of expropriation, and the
repudiation of contracts by the government.
Institutions, however, may affect exports through at least two other channels. In particular,
recent papers have highlighted the role of institutions in alleviating hold-up problems in the
production process. For example, Nunn (2005) develops an incomplete-contracts model of
relationship-specific investments and finds that countries with better contract enforcement have a
comparative advantage in industries intensive in such investments. Similarly, Levchenko (2004)
and Claessens and Laeven (2003) show that property rights protection and the rule of law affect
the composition of trade. Secondly, institutions in both the exporting and the importing country
may matter for trade negotiations. For instance, quality control or trade agreements may be more
difficult to establish in some industries, making them more dependent on the level of contract
enforcement in both partner countries.
In contrast to the prior literature, I establish an independent financial channel using three
different approaches. First, I include institutional controls directly in the regressions together with
private credit. This addresses the concern that institutions play a non-financial role in the
production process. The results survive this test. Second, the bilateral nature of my panel allows
me to test whether private credit proxies not only for financial contractibility, but for a contractual
environment in general which is conducive to trade negotiations and more exports. If so, then both
the exporter’s and the importer’s level of financial development should affect sectoral exports.
Once I control for contract enforcement in both countries, though, I find that only the exporter’s
financial development matters. I thus isolate a producer-side-only effect of the financial system,
which can therefore be attributed to credit constraints.
Finally, I consider how liberalizing foreign equity flows affects the sectoral composition of
exports. Liberalizations plausibly increase the availability of external financing for any level of
3domestic financial development. In addition, their exact timing is usually the product of complex
political processes and, hence, exogenous from the perspective of a producer. Thus, liberalizations
provide a source of exogenous variation in the supply of financing that is arguably independent
from developments in the institutional environment in an economy (Bekaert et al. (2005)). My
results on the effects of liberalizations on trade composition provide further evidence of a causal
financial channel that is independent of other institutions.
The results on equity market liberalizations are new to the literature and contribute
significantly to the establishment of financial development as a source of comparative advantage.

For example, Mitton (2005) shows that firms have higher investment rates when they open to foreign stock holding.
They also add importantly to the debate on the economic consequences of liberalizing capital
markets, and equity flows in particular. While some papers have found strong support for growth-
4enhancing effects, Prasad et al. (2004) emphasize that the evidence is best summarized as mixed
and inconclusive. In this context, my findings suggest an alternative mechanism through which
liberalizations can benefit countries, namely by affecting their comparative advantage and
potentially improving their terms of trade. In addition, I find stronger effects of liberalization in
countries with less active stock markets prior to reform, as measured by initial market turnover or
value traded as a share of GDP. This suggests that foreign equity flows may compensate for an
underdeveloped domestic financial system. The effects do not depend on the size of the stock
market, which suggests that market activity better reflects the allocative efficiency of the financial
My results are based on a panel of 107 countries and 27 industries over the 1985-1995
period, and are confirmed with both bilateral exports and total exports to the world. I interact a
country-level measure of financial development with industry-level measures of asset tangibility
and external finance dependence, and include the interactions in a gravity model of export
volumes by sector. As my main measure of financial development I use the amount of credit by
5banks and other financial intermediaries extended to the private sector as a share of GDP. As in
Rajan and Zingales (1998), external finance dependence is calculated as the share of capital
expenditures not funded by cash flow from operations for the median U.S. firm in each industry.
Asset tangibility is similarly defined as the share of net property, plant and equipment in total
6book-value assets for the median U.S. firm in a sector, as in Braun (2003). Variations in these
three measures provide identification for the differential effect of financial development on
exports by industry in a generalized difference-in-difference framework. When I turn to equity
market liberalizations, I repeat the analysis using dummies for before and after liberalization, as
well as measures of its intensity.
My results are highly statistically and economically significant. For example, if the
Philippines, the country at the first quartile of the distribution of private credit, were to improve its
financial system to the level of the third quartile country (Italy), the Philippines could increase its

See, for example, Bekaert et al. (2005) and Gupta and Yuan (2004).
5 This is a standard measure of financial development and financial contractibility in the literature. I also show results
with alternative measures such as liquid liabilities and stock market size and activity.
Both measures are constructed with data from all publicly traded U.S.-based companies over the 1986-1995 period.
rdtextile exports (highly dependent on external finance, 3 quartile) by 24.7 percentage points more
stthan its mineral products exports (intensive in internal funding, 1 quartile). At the same time,
stexports of low tangibility sectors (other chemicals, 1 quartile) would grow by 14.7%, while
rd 7exports of high tangibility sectors (wood products, 3 quartile) would rise by only 4.5%. Thus the
effect of financial development on low-tangibility sectors is almost three times as large as that on
high-tangibility sectors. Equity market liberalizations result in an even more powerful
redistribution of exports. The magnitude and significance of these effects are robust to alternative
The dramatic effects of financial frictions on export composition are perhaps less surprising
in view of the large body of literature showing that financial development promotes growth,
8especially in financially vulnerable industries. In particular, Rajan and Zingales (1998) and Braun
(2003) argue that sectors intensive in external finance and intangible assets grow
disproportionately faster in financially developed economies. Financial development should affect
export composition for the same reasons that it redistributes production across sectors. This
intuition is based on the observation that entrepreneurs with limited private resources have to
borrow to produce and export. When industries differ in their need for outside finance and
endowment of tangible assets, financial development determines the cost of external funds for
each industry and therefore which sector an entrepreneur can enter.
An important aspect of financial underdevelopment is poor financial contractibility, in the
sense that financial contracts are less likely to be enforceable. Asset structure matters in a
financially underdeveloped environment because poor financial contractibility increases creditors’
exposure to the risk of default. Tangible assets such as real estate, machinery and plants that can
9serve as collateral offer the financier protection and increase his willingness to invest. Thus, the
poorer financial contractibility is, the more important asset hardness becomes for accessing
external finance. Because industries differ in asset tangibility for technological reasons (Claessens

7 To focus on one industry characteristic at a time, the comparative statics for external finance dependence (asset
tangibility) assume that both sectors have the cross-industry mean asset tangibility (external finance dependence).
Some of the earliest studies include Cameron et al. (1967), Goldsmith (1969), and McKinnon (1973). See King and
Levine (1993), Levine and Zervos (1998), and Beck et al. (2000) for more recent cross-sectional studies that
instrument for financial development.
9 In contrast, it is more difficult to transfer control over a firm’s human and organizational capital, research and
development, and even accounts receivable, cash and inventories. Grossman and Hart (1986), Hart and Moore
(1988,1990,1994), Hart (1995), and Shleifer and Vishny (1992) model these effects theoretically.
and Laeven (2003), Braun (2003)), in less financially developed countries investment, production
and exports will be disproportionately shifted towards sectors endowed with harder assets.
The extent to which a sector depends on external finance is similarly important. When the
nature of the industry is such that projects can be funded with internal cash flows, the level of
financial development matters little since firms do not need to borrow from creditors. As a sector’s
dependence on external funds increases, the availability of outside capital becomes more relevant
and the extent of financial contractibility more important. In an environment with better
functioning capital markets, an entrepreneur of a given wealth level can secure more funding and
enter a more external capital dependent industry (Rajan and Zingales (1998)).
A number of theoretical models have formalized the above intuition. Kletzer and Bardhan
(1987) and Beck (2002) both study a world with two sectors, one of which depends (more) on
external finance. Matsuyama (2004) instead proposes an exogenous limit on how much
entrepreneurs can borrow in a continuum of sectors, while Chaney (2005) focuses on liquidity
constraints. The results in this paper are consistent with the prediction common across these
models, namely that better financially developed countries will specialize in sectors that require
10more outside funds. In addition, my findings on asset structure suggest a complimentary effect
that remains to be explored theoretically.
My results are also consistent with the dynamic model of finance and trade proposed by
Wynne (2005), who explores the implications of income inequality. Here the driving force is not
exogenously given financial institutions, but wealth distribution, which is endogenous to export
volumes and composition. This approach is complementary to those above and consistent with the
underlying intuition that the availability of both private resources and external finance interacts
with industry characteristics to affect comparative advantage. Indeed, these models highlight
different aspects of the credit constraints problem, and together present a comprehensive picture of
the role of external finance dependence in the determination of trade flows.
The remainder of the paper is organized as follows. The next section outlines the empirical
approach, introduces the data and comments on its advantages and limitations. Section 3 presents
the analysis of the effects of financial development on sectoral exports, while the results on equity
market liberalization are discussed in Section 4. The last section concludes.

10 Financially developed countries may also have a comparative advantage in sectors subject to demand shocks or
when more productive technologies are more expensive to pursue (Baldwin (1989), Beck (2003)). I focus on the
financial dependence of sectors and not their exposure to demand shocks or access to alternative technologies.
2. Data and Empirical Approach
I use a generalized difference-in-difference approach to test for the differential effect of financial
development on export volumes across sectors. I interact a country-level measure of financial
development with industry-level measures of asset tangibility and external finance dependence,
and estimate the following gravity model specification:

T = b Fin Devt + b Fin Devt x Fin Dep + b Fin Devt x Tang + (1) ijkt 0 it 1 it k 2 it k
+ a + a GDP + a GDP + a DIST + X g + h + h + h + h + e , 0 1 it 2 jt 3 ij i j k t ijkt

where T is the natural logarithm of the value of trade from country i to country j (which can be ijkt
the rest of the world) in industry k in year t. GDP and GDP give the log of the exporter’s and it jt
importer’s real gross domestic products in year t respectively, while DIST measures the distance ij
between the two countries. Fin Devt is a measure of the level of financial development in the it
exporting country, which I take to be the amount of credit extended to the private sector in most
specifications. Fin Dep and Tang correspond to the level of external capital dependence and the k k
extent of asset tangibility in sector k. X is a vector of control variables, which I return to below.
The coefficients of interest are b and b , and we expect b > 0 and b < 0, that is financial 1 2 1 2
development favors exports in sectors intensive in external financing or soft assets.
In all specifications with bilateral trade flows I allow for exporter, importer, industry and
year fixed effects; when I consider total exports to the world the importer’s fixed effect, GDP level
and distance from the exporter are naturally excluded. I cluster e by exporter. Note that I cannot ijkt
estimate the main effects of Fin Dep and Tang because they are subsumed by the industry fixed k k
effects, which also control for other sector-specific omitted characteristics that may contribute to
exports equally across all exporters. The time fixed effects control for year-to-year changes in the
trading environment that affect all countries and industries equally, such as technological
improvements or price shifts across the board. Finally, country fixed effects control for intransient
country-level characteristics that might affect exports in all industries, such as remoteness or
institutions that do not change during the sample period. Therefore, the main effect of financial
development b is identified from within-country variation in private credit over time, and does 0
not exploit the cross-country variation in average private credit over the period.
The identification of b and b comes from the combination of cross-sectional and time-1 2
series variation in financial development across countries, and cross-industry variation in external
11capital dependence and asset tangibility. These coefficients can therefore be interpreted in one of
two equally valid ways: (1) b (b ) estimates the comparative (dis)advantage that a country has in 1 2
industries more dependent on credit markets (industries with few tangible assets) relative to
sectors with more internal financing (sectors with harder assets). (2) At the same time, b (b ) also 1 2
compares the volume of exports in a given industry relative to its expected outcome in a more
financially developed environment. Finally, when both the country- and industry-level measures
are allowed to vary, I can use the interaction terms to make statements about export composition
either across countries in a given year or for a specific economy over time.
All robustness checks that I perform can be viewed as modifications of specification (1)
above. When I instrument for financial development, (1) gives the second stage of the estimation.
When I test for different sources of omitted variable bias, I worry about the potential for a country-
level measure (A) correlated with financial development to interact with an industry characteristic
(B or C) correlated with Fin Dep or Tang respectively, with the interaction affecting trade k k
composition. I then include the main and interaction effects of these potentially problematic
variables, and estimate:

T = b Fin Devt + b Fin Devt x Fin Dep + b Fin Devt x Tang + (2) ijkt 0 it 1 it k 2 it k
+ d A + d A x B + d A x C + 0 it 1 it k 2 it k
+ a + a GDP + a GDP + a DIST + h + h + h + h + e , 0 1 it 2 jt 3 ij i j k t ijkt

corr(A , Fin Devt ) 0 , corr(B , Fin Dep ) 0 , corr(C , Tang ) 0 , it it k k k k

where I have now made the controls in X more explicit. In particular, I use (2) to control for factor
endowments as a source of comparative advantage, as well as to allow for institutional features
other than financial development to have a differential effect on trade across sectors. (2) is also the
framework in which I study the role of the importer’s level of financial development and legal
enforcement, with the latter two variables taking the place of A.
When I turn to equity market liberalizations, I use (1) and replace Fin Devt with Liberal , it it
which is either a dummy or a continuous liberalization intensity measure as described below. To

The exporter fixed effects control for the effects of average private credit over the period on exports in a reference
sector. The estimation therefore exploits the interaction of industry variation with the within-country changes in
private credit over time. In addition, in any given year countries vary in how much they deviate from their average
private credit. Hence the estimation also exploits the interaction of these deviations with differences across industries.
test whether the magnitude of the effects of liberalization depends on how active the domestic
equity market is (Mkt Act ), I extend (1) to include triple interactions: it

T = b Liberal + b Liberal x Fin Dep + b Liberal x Tang + (3) ijkt 0 it 1 it k 2 it k
+ d Mkt Act + d Mkt Act x Fin Dep + d Mkt Act x Tang + 0 it 1 it k 2 it k
+ g Liberal x Mkt Act + g Liberal x Mkt Act x Fin Dep + g Liberal x Mkt Act x Tang + 0 it it 1 it it k 2 it it k
+ a + a GDP + a GDP + a DIST + h + h + h + h + e . 0 1 it 2 jt 3 ij i j k t ijkt

If liberalization (active stock markets) benefits disproportionately more external capital dependent
industries and sectors with low asset tangibility, we expect b > 0 and b < 0 (d > 0 and d < 0). 1 2 1 2
Moreover, if liberalization compensates for a poorly developed domestic financial system, the
effects of liberalization will diminish as market activity increases, and we should observe g < 0 1
and g > 0. Since opening equity markets to foreign flows may endogenously increase market 2
activity, I use a measure of the initial, pre-liberalization level of market activity and drop Mkt Act it
and Liberal x Mkt Act from (3). it it
While the gravity model is very successful empirically, trade theories predict that it should
only work at the level of total bilateral trade. The literature does not provide a theoretical
justification for why the gravity model explains bilateral exports by sector. Nevertheless, in most
specifications I observe a significant coefficient with a value close to 1 on the exporter’s and
importer’s logGDPs and –1 on the distance between them, as the gravity model would predict for
total bilateral flows. One way to understand this result is to think of these three variables as
proxying for total bilateral trade between the two countries by controlling for market-size effects
and transport costs associated with trade. In the absence of comparative advantage and with
differentiated goods in each industry, trade would be balanced across sectors and total bilateral
flows would predict my left-hand side variable (the volume of sectoral exports). We can then
attribute the deviations in the sectoral composition of trade from a balanced distribution to
comparative advantage arising from factor endowments and the level of financial development.

2.1 Measures of financial development
Financial development is measured by the relative size of the financial sector in the economy.
Conceptually, establishing a credit constraints channel due to moral hazard problems requires a
measure of the level of financial contractibility or, more generally, of the capacity of the