Credit constraints in Latin America

Published on November 2016 | Categories: Documents | Downloads: 115 | Comments: 0 | Views: 1561
of 318
Download PDF   Embed   Report

Comments

Content

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

This page intentionally left blank

Credit Constraints and Investment in Latin America
Arturo Galindo Fabio Schiantarelli
Editors

inter-American Development Bank Washington, D.C. 2003

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

©2003 Inter-American Development Bank 1300 New York Avenue, N.W. Washington, D.C. 20577 Produced by the IDB Public Information and Publications Section. To order this book, contact: IDB Bookstore Tel: 1-877-PUBS IDB/(202) 623-1753 Fax: (202) 623-1709 E-mail: [email protected] www.iadb.org/pub The views and opinions expressed in this publication are those of the authors and do not necessarily reflect the official position of the Inter-American Development Bank. Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Credit constraints and investment in Latin America / Arturo Galindo and Fabio Schiantarelli, editors. p. cm. Includes bibliographical references. 1. Commercial creditÑLati n America. 2. Financial institutionsÑLati n America. 3. InvestmentsÑLati n America. 4. CorporationsÑLati n America. I. Galindo, Arturo J. II. Schiantarelli, Fabio. III. InterAmerican Development Bank.

ISBN 1931003599

LCCN: 2003110757

332.3 C597--dc21

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Preface
This book contains new evidence on the nature, extent, evolution, and consequences of financing constraints in Latin America. Who has and who does not have access to credit markets and the impact of such access on firm performance are questions of paramount importance. Moreover, it is crucial to understand the dynamics of credit markets after significant events, such as sudden stops in international capital flows, or the adoption of policies to liberalize financial markets. Researchers inside and outside Latin America have devoted much attention to the macroeconomic effects of financial crises and financial policies, yet there has been little research on the microeconomic implications of such events. This book constitutes a serious, thoughtful, and important attempt to fill this gap. Written by a distinguished group of economists, the chapters provide empirical analysis of the factors that determine the access to credit and its composition, the role of credit information in easing financial constraints, the impact of organizational structures on financial constraints and investment, and the dynamic effects of crises and financial policies on access to credit and capital accumulation. The authors offer important new insights and bring new evidence to bear on key issues in the policy debate on the future of Latin American development. Guillermo Calvo, Chief Economist Inter-American Development Bank

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Latin American Research Network Inter-American Development Bank Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub
The Inter-American Development Bank created the Latin American Research Network in 1991 in order to strengthen policy formulation and contribute to the development policy agenda in Latin America. Through a competitive bidding process, the network provides grant funding to leading Latin American research centers to conduct studies on economic and social issues selected by the Bank in consultation with the region's development community. Most of the studies are comparative, which allows the Bank to build its knowledge base and draw on lessons from experiences in macroeconomic and financial policy, modernization of the state, regulation, poverty and income distribution, social services, and employment. Individual country studies are available as working papers and are also available in PDF format on the Internet at http://www.iadb.org/res.

Acknowledgments
The studies in this book were financed by the Latin American Research Network of the Inter-American Development Bank and would not have been possible without the collaboration of many friends and colleagues. The authors would like to thank the following for their comments and collaboration on the individual studies: Thorsten Beck, Ricardo Caballero, C. Calomiris, G. Caprio Jr., Mauricio Cardenas, S. Claessens, Jose Fanelli, Gaston Gelos, Carlos Ibarra, Maria Eugenia Ibarrar‡n, Tullio Jappelli, Leora Klapper, Asli DemirgŸ ς-Kunt, Ross Levine, Norman Loayza, Inessa Love, V. Maksimovic, Alejandro Micco, Daniel Oks, S. Ospina, Carlos A. Rodriguez, Mariano Rojas, Susana Sanchez, Sergio Schmukler, Eduardo Siandra, Kim Staking, and Fernando Tenjo. The authors are also indebted to Norelis Betancourt and Raquel Gomez for valuable administrative support and to John Dunn Smith for providing editorial expertise.

CHAPTER 1 Determinants and Consequences of Financial Constraints Facing Firms in Latin America: An Overview Arturo Galindo and Fabio Schiantarelli CHAPTER 2 The Effect of Bank Relationships on Credit FOR fIRMS IN aRGENTINA Jorge M. Streb, Javier Bolzico, Pablo Druck, Alejandro Henke, José Rutman, and Walter Sosa Escudero CHAPTER 3 Determinants and Consequences of Financial Constraints Facing Firms in Argentina José M. Fanelli, Ricardo N. Bebczuk, and Juan J. Pradelli CHAPTER 4 Credit, Financial Liberalization, and Manufacturing Investment in Colombia Maria Angelica Arbeláez and Juan José Echavarria CHAPTER 5 The Effects of Credit Constraints on Costa Rican Manufacturing Firms Alexander Monge-Naranjo and Luis J. Hall CHAPTER 6 Access to Long-Term Debt and Effects on Firm Performance: Lessons from Ecuador Fidel Jaramillo and Fabio Schiantarelli

1

23

71

117

151

199

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Contents

vi

CONTENTS

225

CHAPTER 8 Investment and Financial Restrictions at the Firm Level in Uruguay Julio de Brun, Nestor Gandelman, and Eduardo Barbieri BIBLIOGRAPHY

259

293

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 7 Internal Capital Markets and the Financing Choices of Mexican Firms, 1995-2000 Gonzalo Castañeda

Determinants and Consequences of Financial Constraints Facing Firms in Latin America: An Overview
Arturo Galindo and Fabio Schiantarelli
Bank credit plays a very important role for firms, especially in developing countries where equity markets are considerably underdeveloped. If access to bank loans is restricted, potentially profitable projects cannot be undertaken. Since technology is often embedded in new capital goods, the capacity of economies to absorb new methods of production and to grow is, therefore, adversely affected. Hence, the ability of the banking sector to pool resources and channel them efficiently to firms is an important determinant of the process of economic development and growth. The chapters in this volume study the determinants and consequences of credit supply restrictions at the firm level in Latin America using micro data.1 The book covers Argentina, Colombia, Costa Rica, Ecuador, Mexico, and Uruguay. The chapters provide quantitative evidence on firms' financing choices (access to bank loans, maturity structure, and currency denomination) and on how firm characteristics and past history affect these choices.2 They also analyze the effect of financing constraints on firms' investment
Arturo Galindo is a research economist at the Inter-American Development Bank; Fabio Schiantarelli is a professor of economics at Boston College. 1 All but one of the chapters were part of the project "Determinants and Consequences of Financial Constraints Facing Firms in Latin America and the Caribbean," financed by the IDB. The exception is Jaramillo and Schiantarelli (chapter 6), which had been prepared for the World Bank conference "Term Finance: Theory and Evidence," and appeared as World Bank Policy Research Working Paper 1725. 2 The role of information asymmetries in accessing credit (or not) has been amply discussed in the literature. See, for instance, the seminal contribution by Stiglitz and Weiss (1981).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 1

2

GALINDO AND SCHIANTARELLI

See Schiantarelli (1996) and Hubbard (1998) for a critical review. The chapters in this book complement previous research on the impact of the institutional framework surrounding credit systems on the supply of credit. See the contributions of La Porta, L—pez-de-Silanes , and Shleifer (1998); Levine (1998); Japelli and Pagano (2001); Padilla and Requejo (2001); and Claessens and Laeven (2002). For country-specific evidence, see the chapters in Pagano (2001).
4

3

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

choices and show that the severity of constraints depends on firm characteristics such as size, membership in a business group, and foreign ownership.3 The investigation of all these issues requires the availability of firm-level micro data. The use of such data is a common feature of all the chapters in this volume and one of its strengths.4 The results suggest that access to credit depends not only on favorable balance sheet characteristics, but also on the closeness of the relationship between firms and banks and credit history. Access to long-term loans and to loans denominated in foreign currency is positively related to the size and tangibility of firms' assets and negatively related to measures of country risk. Moreover, firms that have foreign participation appear to be less financially constrained in their investment decisions. The same is true for firms that are associated with business groups. Another issue that is investigated in some of the chapters is the evolution over time of financing constraints. In particular, the authors present evidence on the effect of financial reform on access to external finance, and on how this affects firms' real choices. The consequences of financial and banking crises on financing constraints are also addressed. One of the interesting issues studied is whether crisis episodes and financial reform have a differential effect on different types of firms. On the whole, it appears that financial liberalization tends to relax financial constraints for firms that were previously constrained, while financial crises tighten them. However, firms that have more access to external sources of finance, for instance, via exports or ownership links, appear to suffer less in post-crisis periods. This introductory chapter reviews the main issues concerning firms' financing choices and investment decisions in the presence of capital market imperfections. The focus then turns to the methodology and data sources used in the chapters. After presenting the main results, the chapter concludes with a discussion of the policy implications that can be drawn from this project.

FINANCIAL CONSTRAINTS IN LATIN AMERICA

3

The main issues in the discussion of financial constraints facing firms in Latin America are access to credit and financing constraints and investment. Access to Credit Stiglitz and Weiss (1981) formalize the effect of asymmetric information in the loan market and offer a rationale for the existence of limited access to credit. In essence, they assume that banks can only classify the creditworthiness of firms at a broad level; that is, they have a global perception of the distribution of returns across a certain variety of projects, but lack knowledge about the creditworthiness of specific firms that wish to undertake particular projects. In this setting, the interest rate charged on loans not only influences the amount of loans granted, but also the riskiness of the creditor's own portfolio of loans, either by sorting potential borrowers according to their risk (the adverse selection problem) or by affecting the behavior of borrowers (the moral hazard problem). The combined result is a credit supply curve that might not be monotonically increasing in the interest rate. Banks' profit maximization might then lead to an equilibrium where the market is not cleared and demand for credit exceeds supply.5 Although Stiglitz and Weiss's conclusion on the possibility of credit rationing is derived in a model that assumes debt is exogenous in the form of a contract, it also holds in costly state verification models where debt arises endogenously as the optimal contract (Williamson 1986, 1987). Moreover, the possibility of credit rationing is robust and survives the introduction of mechanisms that are designed to address the adverse selection or moral hazard problem, such as the use of collateral (Bester 1985). Although it has been shown that such mechanisms mitigate the problems derived by information asymmetries, they are not completely eliminated, especially if potential borrowers exhibit decreasing average risk aversion (Stiglitz and Weiss 1986). In such a case, wealthier agents are the only ones who would be granted credit, but they would also be the worst risks. Moreover, even if all agents have similar risk aversion, if asset markets are not developed, banks will still face the
5

Surveys and discussion of the literature on credit rationing can be found in Blanchard and Fischer (1989), Freixas and Rochet (1998), and Mazzoli (1998).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

The Main Issues

4

GALINDO AND SCHIANTARELLI

On the role of banks and stock markets in growth, see Levine (2002), Beck and Levine (2002), and the contributions in Demirgu9-Kunt and Levine (2001). For recent contributions on the more general issue of financial development and growth, see Demirguc-Kunt and Maksimovic (1998); Beck, Levine, and Loayza (2000); Levine, Loayza, and Beck (2000); Wurgler (2000); and Galindo, Schiantarelli, and Weiss (2002). See Levine (1997) for a review of earlier contributions. On maturity, see Demirguc-Kunt and Maksimovic (1999). On financial structure in developing countries, see Booth and others (2001).

6

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

risk selection problem and the credit rationing problem may resurfaceÑ even if all debts are completely collateralizedÑgive n the difficulty of assessing assets pledged as collateral. In addition to the use of collateral, there are several other mechanisms that can be used to screen good and bad risks, such as the use of credit bureaus and the development of credit scoring models. In many Latin American countries, however, credit bureaus are underdeveloped, the use of sophisticated credit scoring technologies is not a common practice, and banks rely on self-gathered information to sort out risks. Among the characteristics that might influence a firm's access to credit are its age and size, its property ownership structure (such as foreign versus domestic or individual versus group), and its ongoing business relationships with banks (Petersen and Rajan 1994). The studies in this book present empirical evidence on these factors. It is important to identify and discuss the issues related to project selection, since the efficiency of banks in analyzing the creditworthiness of firms determines how resources are allocated and which firms will eventually have the chance to test their projects in the market. From this perspective, the ability of banks to distinguish the firms with the greatest chance of success from the rest can determine a country's pattern of growth.6 Not only is the issue of access to bank credit important, but the maturity structure of loans also deserves further discussion. In particular, there has been a widespread perception both by domestic and international policymakers that asymmetric information and contract enforcement problems may lead to a shortage of long-term finance. This shortage is thought to have a cost in terms of productivity growth and capital accumulation, and it may justify some form of government intervention because firms are prevented from choosing projects with higher returns that may be illiquid and have delayed returns. The setting up in most developing countries of long-term credit institutions (development banks) and/or programs to foster the provision of long-term credit was indeed the policy response to this problem.

FINANCIAL CONSTRAINTS IN LATIN AMERICA

5

Financing Constraints and Investment In general, even if information asymmetries and contract enforcement problems do not lead to outright credit rationing, they make external funds imperfect substitutes for internal funds and invalidate the separation between financing and investment choices implied by the Modigliani-Miller theorem (Modigliani and Miller 1958). Many papers have explored the consequences of these information and incentive problems for investment.7 Although the models differ in their details, two main results emerge from this literature. First, unless the loans are fully collateralized, external finance is more costly than internal finance. Second, everything else equal, the premium on external finance is an inverse function of a borrower's net worth (liquid assets plus the collateral value of illiquid assets). It follows that any negative shock to net worth (due to technological reasons, shifts in investors' preferences, or changes in monetary policy) leads to an increase in the premium and, therefore, to a reduction in investment and production. For this reason, the initial impact of the shock may be amplified (the so-called financial accelerator effect). The problems associated with asymmetric information and contract enforcement affect firms differently, and several criteria have been used in
7

See, for instance, Bernanke and Gertler (1989, 1990); Gertler and Hubbard (1988); Calomiris and Hubbard (1990); Gertler (1992); Bernanke, Gertler, and Gilchrist (1996, 1999); Kiyotaki and Moore (1997); and Greenwald and Stiglitz (1988a, 1993).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

The emphasis on long-term finance and the potentially adverse consequences when it is in short supply is somewhat at odds with recent theoretical contributions that emphasize the fact that the use of short-term debt may be associated with higher-quality firms and may have better incentive properties (Diamond 199la). In particular, the possibility of premature liquidation may act as a disciplinary device that improves firms' performance. A rethinking of the role of long-term debt, particularly when heavily subsidized, has also been prompted by the problems development banks have encountered in many countries in terms of nonperforming loans, and by doubts about the selection criteria used in allocating funds. In any case, the issue of the determinants of the maturity structure of debt and its consequences for investment and productivity are important topics that deserve investigation.

6

GALINDO AND SCHIANTARELLI

8

See Schiantarelli (1996) and Hubbard (1998) for a review.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

the literature to partition firms into groups according to their likelihood of being financially constrained.8 The main cross-sectional criteria that have been used in this volume in order to identify firms for which information and agency problems are more or less severe are affiliation with industrial groups and banks, foreign ownership, and size. Business groups are a pervasive form of organization found in a variety of countries, both developed (such as Japan, Germany, and Italy) and developing (such as Indonesia, Korea, and several Latin American countries). Business groups can be seen as an organizational form that helps to cope with information and contract enforcement problems in the capital markets. The knowledge by financial intermediaries or individual investors that in case of financial distress individual firms may also rely on the financial resources of the group is likely to improve their access to external financial resources. The diversification of the group's activities is an added bonus in this respect. Moreover, even in the absence of financial distress, business groups allow the formation of an internal capital market that supplements the capital allocation function of the external market. In some countries, groups are organically linked with banks. Strong ties between banks and firms represent a possible way to reduce information costs. In this sense, firms affiliated with a business group would be expected to be less sensitive to cash flow both because of the mitigation of information problems in accessing external finance (especially if there are bank links) and because of the creation of an internal capital market. Direct foreign control or foreign participation in ownership can obviously alleviate financing constraints for similar reasons. In this case, financing constraints are alleviated because it is likely that firms with a degree of foreign ownership will find it easier to access international capital markets. Size is another criterion that some chapters use to identify firms that are more likely to be financially constrained. This is based on the presumption that size is highly correlated with the fundamental factors that determine the probability of being constrained. Smaller firms are more likely to suffer from idiosyncratic risk and, insofar as size is positively correlated with age, are less likely to have developed a track record that helps investors to distinguish good from bad firms. Moreover, small firms may

FINANCIAL CONSTRAINTS IN LATIN AMERICA

7

9

See Schiantarelli (1996) for a discussion of this issue. See Bernanke (1983), Bernanke and Blinder (1988), Kashyap and Stein (1994), and Hubbard (1994) for a fuller discussion of the consequences of shocks to credit supply and the implications for the transmission mechanism of monetary policy of imperfect substitutability between bank loans and other forms of credit.
10

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

have lower collateral relative to their liabilities and unit bankruptcy costs are likely to decrease with size. And it is likely that transaction costs for issuing securities decrease with size. In any case, these and other criteria used in sorting firms are to a varying degree potentially endogenous. Hence, care should be taken in addressing endogeneity issues in estimation.9 As described above, one of the implications of the information-based models of investment is that the severity of financial constraints is likely to vary with overall macroeconomic conditions and with the stance of monetary policy because they influence the value of firms' net worth. It would therefore be expected that during recessions or after a monetary tightening, the cost of external finance would increase and/or access to it would decrease. Similarly, negative shocks to balance sheets associated with depreciation, when part of the borrowing is in foreign currency, can be associated with tightening of financial constraints. The occurrence of banking crises, often associated with currency crises, can disrupt and destroy information capital that had been accumulated and leads to a restriction in the supply of loans. This may lead to severe financial constraints for those firms that derive their external financing mostly from banks, with negative consequences for their investment decisions.10 The tightness of financial constraints over time may vary not only following changes in business cycle conditions and monetary policy, but also because of structural changes in financial markets. In the 1980s and early 1990s, several developing countries introduced financial reforms to facilitate capital accumulation and growth. These reforms consisted mainly of the removal of administrative controls on the interest rate and in the elimination or scaling down of directed credit programs. The reforms lowered barriers to entry in the banking sector and stimulated the development of securities markets. The main objective of banking deregulation was to provide higher returns to depositors and increase the supply of funds for investment, although whether this is happening at the economy-wide level is a matter of controversy. It is likely, however, that the amount of saving intermediated by

8

GALINDO AND SCHIANTARELLI

Financial Constraints in Latin America: Methodology and Data The chapters in this volume provide novel and intriguing evidence on the nature and consequences of capital market imperfections in Latin America, using data from Argentina, Colombia, Costa Rica, Ecuador, Mexico, and Uruguay. All the chapters share the characteristic of being based on micro data, mostly from firm-level balance sheets. In addition to firm-level data, for Argentina the researchers had access to information on debt contracts and borrower characteristics collected by the Central Bank's Public Credit Bureau. For Costa Rica, information was collected by means of a specially designed survey administered to manufacturing enterprises. Some of the chapters investigate the determinants of firms' financing choices using firm-level panel data containing balance sheet information. In particular, they investigate econometrically how firm characteristics, macroeconomic conditions, and financial reform affect the overall degree of leverage and/or the maturity structure of firms' debt. Fanelli, Bebczuk, and Pradelli (chapter 3) take this approach for Argentina, and Jaramillo and Schiantarelli (chapter 6) for Ecuador. Fanelli, Bebczuk, and Pradelli also present results on the currency denomination of debt, while laramillo and Schiantarelli analyze the effect of the maturity structure of debt on productivity and investment. Monge-Naranjo and Hall (2002) investigate how the characteristics of firms and owners at a given point in time affect access to bank finance, and how such access affects several measures of firm performance, such as investment, employment, and profitability. Adopting and extending the approach by Petersen and Rajan (1994), Streb and others (chapter 2) provide evidence for

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

the banking system will increase. To the extent that there are economies of scale in information gathering and monitoring, it is possible that banking intermediaries may have an advantage over the curb (informal) market in allocating investment funds, and this may lead to a reduction in the premium of external finance over internal finance. However, the elimination of subsidized credit programs will increase the financing constraints on those firms that previously benefited from the system of administrative allocation of credit. This means that financial liberalization programs have distributional consequences, and whether they relax financing constraints for different categories of firms is ultimately an empirical question.

FINANCIAL CONSTRAINTS IN LATIN AMERICA99

In one case (Uruguay), in addition to the investment equations, the Euler equation for the capital stock is estimated, allowing for the presence of a ceiling on leverage and an interest rate premium related to leverage. 12 Note that cash flow captures both balance sheet conditions and expectations of future profitability.

11

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Argentina using data from the Central de Deudores on factors that affect the access to and cost of bank credit, including the closeness of bank relationships and past credit history. Other chapters focus on assessing the presence and severity of financing constraints by focusing on firms' investment choices. Arbelaez and Echavarria (chapter 4) take this approach for Colombia, as does Castaneda (chapter 7) for Mexico. De Brun, Gandelman, and Barbieri (chapter 8) focus on investment choices in Uruguay, and Fanelli, Bebczuk, and Pradelli (chapter 3) do the same with regard to Argentina. All of these chapters share a common methodological approach, in that they are based on panel estimation of an investment equation containing, in addition to a proxy for fundamentals, financial variables that capture the availability of internal sources of finance and the net worth position of the firm. The basic strategy, following the spirit of the seminal contribution by Fazzari, Hubbard, and Petersen (1988), is to test whether these variables are significant for the firms that a priori are thought more likely to face information and incentive problems. The measurement of fundamentals is based on either Tobin's average Q or proxies for the present value of the marginal product of capital based on the sales-tocapital ratio. Error correction models for investment or accelerator models are also estimated, in which case sales and sales growth capture profit opportunities.11 The measurement of net worth is a difficult problem in an intertemporal context. Some of the chapters use cash flow as a proxy for internal net worth; other chapters use the stock of liquid assets.12 Some chapters include additional balance sheet variables, such as leverage, in the investment equation. Whatever the choice, it is expected that firms that suffer more from asymmetric information problems are more sensitive to variation in their net worth or in the availability of internal funds. The estimation of both the financing and investment equations needs to address endogeneity issues. The availability of panel data is especially important because it allows the analyst to deal with the presence of (relatively) firm-specific and time-invariant unobserved characteristics that appear as

10

GALINDO AND SCHIANTARELLI

Overview of the Results What does the evidence suggest about firms' access to bank credit and the maturity structure and currency composition of debt? In order to discuss the main results and put them in the more general context of the literature, table 1.1 summarizes the data sources used and tables 1.2 and 1.3 summarize the models that have been estimated, the sample separation criteria used, and the econometric methods. Starting with the composition of debt, Fanelli, Bebczuk, and Pradelli (chapter 3) present evidence that size (proxied by the fixed capital stock) has a significant positive effect on the percentage of total debt that is of long duration (1 year or more) in Argentina.14 The maturity structure is also significantly related to the tangibility/duration of assets (measured by the ratio of fixed to total assets) and there is evidence

See Bond (2002) for a review of the econometric issues that arise in the estimation of dynamic panels. 14 The results are based on a smaller panel of 36 companies quoted on the Buenos Aires Stock Exchange and a larger one provided by INDEC of approximately 300 firms. The former has a quarterly frequency and covers most of the 1990s, while the latter is annual and is of shorter duration.

13

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

components of the error term in the equations. In addition, some variables, even after removing such components by an appropriate transformation, are correlated with the contemporaneous or lagged values of the idiosyncratic component of the error term. In the case of short panels, this calls for the use of instrumental variables or generalized method of moments techniques.13 Whereas there are well-developed techniques for addressing these problems in the context of dynamic panel data models with continuous data, the same is less true in dealing with models that have a discrete choice component. This affects, for instance, the estimation of equations dealing with access to finance, and it is therefore more difficult to give a structural/ causal interpretation of the results. The same caution must be exercised when results are based on only one cross section. However, even in that case, the correlations captured in estimation provide useful information on the financing problems firms face and on the factors that may be associated with different outcomes.

Table 1.1. Data Used in the Analyses, by Chapter Authors
Data Data source Superintendencia de Valores and 1978-99 1990-2000 1995-2000 Superintendencia de Sociedades Mexican Securities Market (BMV) Superintendencia del Mercado de Valores at the Central Bank of Uruguay and Liga de Defensa Comercial Buenos Aires Stock Exchange for quarterly data and Encuesta Nacional de Grandes Empresas (ENGE) by INDEC (Institute Nacional de Estadisticas y Censos) for annual data Superintendencia de Companies Own survey Central de Deudores del Sistema Financiero at the Central Bank of Argentina 1984-88 and 1984-92
2001

Chapter authors

Country

Sample period

Arbelaez and Echavarrfa

Colombia

Annual balance sheet data for

(chapter 4)

1,488 listed and unlisted firms

Castaheda (chapter 7)

Mexico

Annual balance sheet data for

176 listed firms

De Brun, Gandelman,

Uruguay

Annual balance sheet data for

and Barbieri (chapter 8)

56 listed and unlisted firms

Fanelli, Bebczuk, and

Argentina

Quarterly balance sheet data for

1986-2000 for quarterly data, 1994-98 for annual data

Pradelli (chapter 3)

45 listed firms and annual data

for 308 firms

Jaramillo and Schiantarelli

Ecuador

Balance sheet data on 731 or

(chapter 6)

850 firms

Monge-Naranjo and Hall

Costa Rica

Survey data for 150 manufacturing

(chapter 5)

firms October 2000

Streb, Bolzico, Druck,

Argentina

Balance sheet and debt information

Henke, Rutman, and

for 15,796 firms

Sosa Escudero

(chapter 2)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 1.2. Investment Model, by Chapter Authors
Estimation method Generalized method of moments and ordinary least squares Cash flow, cash stock Group membership, Financial crisis bank ties, export orientation Cash flow (contribution margin) Cash flow, cash stock American Depositary Receipts and bond issues, recently privatized Cash flow Size Financial liberalization Generalized method of moments Group membership, size Financial crisis Foreign ownership, Financial crisis Generalized method of moments and ordinary least squares Generalized method of moments Generalized method of moments and fixed effects within

Chapter authors Cash flow Group membership, foreign ownership, size liberalization financial Financial crisis,

Dependent variable Proxy for net worth3 Cross-sectional sample separation Macroeconomic events

Proxy for fundamentals

Arbelaez and

Investment/capital

Sales/capital

Echavarria

stock

(chapter 4)

Castaneda

Investment/capital

Production/capital

(chapter 7)

stock

De Brun, Gandelman,

Investment/capital

First differences of

and Barbieri

stock

log (sales)

(chapter 8)

Fanelli, Bebczuk,

Investment/capital

Tobin's q, sales/capital

and Pradelli

stock

(chapter 3)

Jaramillo and

Investment/capital

Growth in real sales

Schiantarelli

stock

(chapter 6)

Relative to the capital stock.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 1 .3. Models Used for Access, Debt Composition, and Performance, by Chapter Authors
Macroeconomic Firm characteristics
Yes

Credit history sample separation
Yes

events used for Estimation method Generalized method of moments and fixed effects within

Chapter authors

Dependent variable

Fanelli, Bebczuk,

Debt/equity, long-term

No

and Pradelli (chapter 3)
Yes

debt/total debt, dollar

denominated debt/total debt

Jaramillo and Schiantarelli
Yes

Prob(access to long-term debt),

No

Probit, logit, Heckman selection model, generalized method of moments (for performance)

(chapter 6)
Yes Yes

maturity, performance

(productivity, investment)

Monge-Naranjo and Hall

Prob(access to bank debt),

No

Probit, Tobit, Heckman selection model, and semi-parametric methods (for performance)

(chapter 5)

amount of bank debt.

performance (employment.
Yes Yes

investment, profitability)

Streb, Bolzico, Druck, Henke,

Overdraft interest rate, debt

No

OLS, TOBIT, Heckman selection model

Rutman, and Sosa Escudero

ratio, unused credit lines

(chapter 2)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

14

GALINDO AND SCHIANTARELLI

See Hart and Moore (1994) for a theoretical model. Moreover, they find that in general, debt ratios in developing countries are affected in a similar way by the same types of variables that appear significant in studies for developed countries. However, they note that the way country-specific factors tend to affect debt varies substantially across countries. 17 Their data source is the Superintendencia de Companias and consists of balance sheets for several hundred companies in 1984-92, therefore excluding the most recent crisis period.
16

15

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

that firms match the maturity structure of assets and liabilities.15 Size and tangibility are also positively related to the proportion of debt denominated in foreign currency. Finally, country risk, measured as the emerging market bond index spread, alters the maturity structure of debt in favor of shortterm debt denominated in domestic currency, while the opposite is true for financial development, which is captured by the ratio of private debt to gross domestic product. These results are in line with previous findings in the literature. For example, Booth and others (2000) find that for a sample of 10 developing countries (not including Argentina), size and tangibility are important determinants of debt ratios.16 Schmukler and Vesperoni (2001) analyze a sample of seven developing countries (including Argentina) and find similar results. Gallego and Loayza (2000) come to similar conclusions using a sample of Chilean firms. Notably, Fanelli, Bebczuk, and Pradelli (chapter 3) do not find significant evidence of an increase in the proportion of longterm debt for firms with access to foreign sources of funding (captured by American Depositary Rights or the ability to issue international bonds). Jaramillo and Schiantarelli (1997) find that size and tangibility are crucial in determining access to and amount of long-term debt for Ecuadorian firms.17 These results are consistent with several explanations. One is simply that collateral is a prerequisite for obtaining long-term credit. Moreover, larger firms tend to be more profitable, so this result may reflect a positive association between firm quality and long-term debt. Larger firms are likely to have more bargaining power and greater political influence in obtaining long-term financial resources, particularly through government-subsidized programs. Jaramillo and Schiantarelli find that estimation of an augmented production function suggests that the availability of long-term finance may have a positive effect on productivity. Perhaps the availability of long-term finance facilitates access to more productive technologies, and this effect

FINANCIAL CONSTRAINTS IN LATIN AMERICA

15

One disturbing result for Ecuador is that, conditional on size, greater profits do not increase the probability of receiving a long-term loan. Moreover, conditional on access, profitability is negatively correlated with the length of the maturity structure of debt. This raises some questions on the mechanism used in allocating long-term financial resources in Ecuador during the period covered by the study. Interestingly, the negative effect of profits is greater before financial liberalization, while afterward the profit coefficient increases, but not enough to make it positive. 19 For other work using information from the Central de Deudores, see Berger, Klapper, and Udell (2000). However, they do not use information on the interest rates and balance sheets of firms.

18

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

dominates the positive incentive effects generated by more intense monitoring and by the fear of liquidation associated with short-term debt.18 Monge-Naranjo and Hall (chapter 5) present interesting evidence on the source of credit for Costa Rican firms. They find that while banks are the most important source of credit for larger firms, nonbank credit (trade credit and informal credit) is the leading source of funds for smaller firms. Moreover, own funds and informal credit are very important for newly created firms. The probability of having access to bank credit (or its share of total credit) is positively related to firm characteristics such as size, having formal accounting statements, and the existence of a long-term relationship with a bank. Surprisingly, it is not significantly related to personal characteristics of the owners of the firm, such as education and age. Both parametric and semi-parametric methods fail to deliver statistically conclusive results on the effect of access to bank credit on firm performance. The results suggest that bank credit can have large positive effects on firm performance, but such effects are not precisely estimated. Streb and others (chapter 2) focus on the financing side of firms in Argentina. However, unlike Jaramillo and Schiantarelli (chapter 6) and Monge-Naranjo and Hall (chapter 5), Streb and others do not address the issue of access to bank credit. Instead, they investigate the determinants of the availability and cost of bank credit, conditional on access, for firms that have a relationship with the banking sector. The analysis uses the information contained in the Central de Deudores records collected from financial institutions by the Banco Central de la Republica Argentina.19 The data set is rich and the empirical work is based on approximately 4,000 observations. To measure the marginal cost of credit, Streb and others use overdrafts, the most expensive line of credit. The measure of the availability of

16

GALINDO AND SCHIANTARELLI

Petersen and Rajan (1994) measure credit constraints by the degree to which firms resort to trade credit, which is generally more expensive than bank credit. 21 The sample includes 54 firms and covers 1997-2000. 22 The sample for Colombia includes 140 quoted and 1,348 unquoted firms for 1970-99. The sample for Mexico includes 176 quoted companies for 1990-2000.

20

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

credit is unused credit lines as a proportion of total liabilities with the main bank.20 The availability of credit depends positively on the closeness of the relationship between firms and banks. Closeness is measured by debt concentration at the marginal bank and by the number of accounts with it. Favorable balance sheet characteristics (such as large assets, a high sales-toassets ratio, and low leverage) and a good credit history (a normal credit situation with no arrears and no bounced checks) lead to improved credit availability and lower cost. In addition, a good record in the credit register is associated with higher credit availability, suggesting that the information contained in the Central de Deudores eases credit constraints for healthy firms. This evidence supports the importance of credit registries as one of the institutions that can help relax financing constraints (Pagano and Japelli 1993; Japelli and Pagano 2001). Another interesting result is that, as the credit situation deteriorates, the interest rate does not increase monotonically. This is consistent with a credit rationing story in which increases in interest rates beyond a certain limit may lead to a decrease in bank profits by increasing the probability of bankruptcy. What can be learned from the estimation of the investment equations about the differences across firms and over time in the severity of financing constraints? The evidence presented by de Brun, Gandelman, and Barbieri (chapter 8) for publicly traded firms in Uruguay suggests that, even within this group of relatively large firms, size matters in the sense that smaller firms display greater sensitivity to cash flow.21 By contrast, the findings by Fanelli, Bebczuk, and Pradelli (chapter 3) for Argentina do not support the presence of significant differences related to size in their sample of quoted companies. Arbelaez and Echavarria's (chapter 4) study of Colombia and Castaneda's (chapter 7) study of Mexico, both based on large samples of several hundred firms,22 present evidence of greater sensitivity to financial variables such as cash flow or the stock of liquid assets for independent firms not affiliated with business groups, confirming the role of groups in mitigating

FINANCIAL CONSTRAINTS IN LATIN AMERICA

17

23

Harris, Schiantarelli, and Siregar (1994) obtain similar results for Indonesian establishments, as does Cho (1995) for Korean firms. 24 Castaneda suggests that group structure may have become tighter in the second half of the 1990s as a response to the problems of the financial sector. However, he also notes that one piece of evidence is not consistent with this story, namely, the fact that the coefficient of total group liquidity is significant only in the pre-1995 period. 25 Corroborating evidence is the fact that the coefficient for the stock of cash is not significant for exporting or nonexporting firms in the second period. It is significant only for exporting firms in the first period, which is somewhat puzzling.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

financing constraints.23 They also provide evidence that companies with foreign ownership (in Colombia) or those with affiliation with a bank (in Mexico in the first half of the 1990s) are less financially constrained. Some of the chapters in this volume also provide evidence on the timevarying nature of liquidity constraints. As predicted by many theoretical models of investment based on asymmetric information, there is evidence that episodes of financial and currency crises, such as those that occurred in the middle and at the end of the 1990s, are associated with a tightening of financial constraints. This is true in both Colombia and Uruguay. In the latter case, the worsening of financing constraints has affected mainly smaller firms. Note that this is the first hard econometric evidence, based on the estimation of an investment function, on the effect of financial crisis on the severity of financing constraints. It complements and extends the evidence in Domac and Ferri (1999) for Korea and Malaysia, based on estimation of a vector autoregression model containing various measures of the interest spreads and production for the aggregate of small and large firms. In chapter 7 by Castaneda, the results for Mexico are puzzling. In particular, they suggest that independent firms were less sensitive to cash flow after the 1995 crisis (coinciding with the North American Free Trade Agreement). Group members did not display excess sensitivity pre or post1995. The fact that group members did not display excess sensitivity during the second half of the 1990s, despite the problems affecting the banking sector, is consistent with the idea that groups lessen financing constraints by creating an internal capital market.24 It is more difficult to explain the result for independent firms, unless the analysis assumes that firms that have internal liquidity or access to capital markets, such as export-oriented firms, recycle funds to independent firms, for instance through trade credit, as suggested by Castaneda.25 An interesting result is that firms that were affiliated with a bank experienced greater financing constraints in the second half of

18

GALINDO AND SCHIANTARELLI

Policy Consequences and Conclusions The results contained in this volume help to explain how the tightness of financial constraints varies across different types of firms and over time.
26

Jaramillo, Schiantarelli, and Weiss (1996) find, instead, that financial liberalization did not significantly relax financing constraints for small firms in Ecuador. 27 However, Harrison and McMillan (2002) find that borrowing by foreign firms exacerbates the credit constraints of domestic firms in Cote d'lvoire.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

the 1990s, which is not surprising given the continued weakness of the financial sector after the crisis. The Mexican experience is a source of both useful lessons and unresolved puzzles that require further investigation. There is evidence that financial liberalization has relaxed financing constraints for investment in Colombia, where firms that are not members of a group have benefited more from the liberalization of the financial sector. This result complements and extends the findings by Harris, Schiantarelli, and Siregar (1994) for Indonesia. They find that smaller or independent firms experienced a relaxation in constraints, while larger firms or members of industrial groups were not constrained before or after liberalization.26 More recently, using data on quoted companies for several developing countries from Worldscope and a time-varying index of financial liberalization, Laeven (2000) finds that financial liberalization relaxed financing constraints for smaller firms. Love (2000), using a larger panel from Worldscope, including developed countries, provides evidence that time-in variant measures of financial development are associated with a relaxation of constraints for smaller firms in the context of Euler equations. More important for the present purpose, Harrison, Love, and McMillan (2001), using the same data set, find that measures of foreign direct investment in a country relax financing constraints for firms that are not members of multinationals in developing countries. This evidence shows that foreign direct investment, by bringing in scarce capital, may ease domestic firms' credit constraints. However, if foreign firms borrow heavily from domestic banks, they may crowd local firms out of domestic capital markets. The empirical results suggest that the first effect dominates.27

FINANCIAL CONSTRAINTS IN LATIN AMERICA

19

Galindo, Schiantarelli, and Weiss (2002) find that financial liberalization in fact increases the efficiency of investment. 29 For a discussion from a macroeconomic perspective, see Calvo, Izquierdo, and Talvi (2002).

28

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Firms that have access to foreign funds, for instance via ownership links, appear to be less constrained, as do firms that have access to internal credit markets of business conglomerates or that can use group membership as a way to improve access to external funds. Financial liberalization can have positive effects on real activity by relaxing financial constraints. A direct implication that is derived from this study is that policies that promote liberalization of financial markets (in dimensions such as removing interest rate controls, reducing the role of directed credit, and allowing for foreign participation in domestic markets) can increase firms' access to credit. On the one hand, eliminating restrictions on how financial institutions need to allocate credit or manage their risks allows them to increase their efficiency in allocating resources toward firms with higher returns to investment.28 On the other hand, liberalization is usually accompanied by capital account liberalization policies that allow firms to improve their access to foreign funding sources. In this respect, the studies undertaken for this volume find that these policies can help to ease constraints by allowing firms in a host country to access the financial markets of the home countries of their parent companies. Currency and financial crises increase the tightness of financial constraints and can have severe real costs. This emphasizes the importance of prudent monetary and budget policies that minimize the risk of a financial crisis. Moreover, it also puts in sharp relief the important role of a system of prudential regulation and supervision that reduces the probability of episodes of excessive credit expansion and risk-taking by banks (World Bank 2001). Sound macro policies and effective prudential regulations are crucial in avoiding the risk that financial liberalization may exacerbate the probability of a financial crisis, as suggested by Demirgu9-Kunt and Detragiache (1999). The results also suggest that the impact of a crisis is not equal across firms. Firms that have ties to external sources of funds, via exports or ownership links, appear to be less constrained in the post-crisis period. This result shows that policies that support openness are fundamental to alleviate the vulnerability of the real and financial sectors to international shocks.29

20

GALINDO AND SCHIANTARELLI

See also Levine (1998); Claessens and Laeven (2002); and Beck, Demirgiic-Kunt, and Levine (2002) on the effects of institutions on financial development and growth. 31 As shown by Lora, Cortes, and Herrera (2001), the size of firms all over the world tends to be positively associated with the quality of institutions, namely institutions that protect property rights. Where property rights tend to be protected, entrepreneurs are in less risk of expropriation and hence tend to increase their investments in their firms. This research project also suggests that those types of policies that allow firm building also alleviate credit constraints. 32 Note that accurate credit information can have greater predictive power for the performance of firms than the data contained in financial statements (Japelli and Pagano 2001).

30

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Moreover, policies that support foreign participation in domestic markets can reduce the vulnerability of firms, at least from external shocks of a moderate size. The debt structure of firms is strongly determined by their size and the tangibility of their assets. This reflects, among other things, the importance of the collateral that firms are able to pledge in accessing credit: firms with greater collateral have access to longer-term debt. From a policy perspective, the importance of collateral should attract attention to putting in place institutions, rules, and regulations that facilitate the effective use of various assets as collateral in Latin American countries. At a general level, the concern should focus on policies and institutions that enforce creditor rights, which are extremely unprotected in Latin America (La Porta and others 1997; La Porta, L6pez-de-Silanes, and Shleifer 1998).30 Specifically, there is the need to develop instruments and institutions that facilitate the process by which firms as well as individuals can register their property as assets that can then be used as collateral.31 Information sharing, documentation of credit history, and the adequate functioning of credit registries are important tools for reducing the impact of information asymmetries and, hence, financing constraints. The availability of credit history information has been shown to be crucial for sound lending decisions.32 Greater availability of information reduces default rates and increases access to credit, and better-informed lenders are able to provide better financial services to borrowers. In order to exploit the benefits of credit registries, an adequate legal framework that encourages information sharing among lenders must be in place. In this regard, bank secrecy laws, which can restrict information flows, have to be reviewed. Similarly, laws that impose limits on credit reporting can hinder the usefulness of credit reporting agencies. However, rules that

FINANCIAL CONSTRAINTS IN LATIN AMERICA

21

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

impede the improper use of credit information must exist in order to guarantee an adequate balance between the benefits derived from the protection of individual privacy and those from information sharing. Moreover, it is important to minimize the risk that information sharing may harm the security and well-being of the people who appear in the registry. Although there is still much to be learned, the studies collected in this volume represent significant contributions in understanding firms' financing and investment decisions and the constraints they face in Latin America. The evidence shows how firms' characteristics and the evolving nature of capital markets shape those choices and affect the severity of constraints. These results bear significant policy implications, and it is hoped that further empirical work based on micro data will make it possible to sharpen the present conclusions and provide answers to the many important questions that still need to be addressed in this area of research.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

This page intentionally left blank

The Effect of Bank Relationships on Credit for Firms in Argentina
Jorge M. Streb, Javier Bolzico, Pablo Druck, Alejandro Henke, Jose Rutman, and Walter Sosa Escudero
This chapter seeks to evaluate the determinants of the variation in the cost and availability of bank credit across firms in Argentina. This is a high profile issue, as many Argentine firms loudly protest the high cost of bank credit and the difficulty of obtaining it. While the same grievances are voiced in many countries, in Argentina these complaints arise against a backdrop of financial markets that are particularly underdeveloped, not only in comparison with those of OECD countries, but also with those of other emerging market countries, such as Chile (Caballero 2000). Stock market capitalization is low, as is financial intermediation in terms of the ratio of M3 money supply to gross domestic product (GDP) or of loans to the private sector to GDP. To complete this picture, high country risk and crowding out by the public sector have made credit constraints particularly acute for firms in Argentina since 2000. In light of the limited options presented by weak capital markets, bank credit is particularly critical for firms in Argentina.1 There are fewer than 100 firms listed in the stock exchange, while the financial system grants credit to more than 100,000 legally incorporated firms. Thus, as defined by Mayer (1994), Argentina can clearly be classified as a banking economy, with a small proportion of listed firms, as opposed to a market economy, which would have a high percentage of listed firms.
Jorge M. Streb and Pablo Druck are professors of economics at the Universidad del CEMA in Buenos Aires; Javier Bolzico, Alejandro Henke, and Jose Rutman are affiliated with the Banco Central de la Repiiblica Argentina; and Walter Sosa Escudero is affiliated with the Universidad Nacional de la Plata. 1 All financial institutions supervised by the central bank will hereafter be referred to as "banks" for short.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 2

24

STREB AND OTHERS

In Argentina, changes in the cost of and access to credit over the past decade or so have been substantial. There was practically no credit market in 1990. Banks had imploded and the few assets they had were concentrated in government securities. After the Convertibility Plan was launched in 1991, with the aim of achieving price stability, lending to firms had to start almost from scratch. Since it takes time to generate a track record for evaluating the risk of lending to firms, it is no surprise that firms should initially have been subject to large individual credit constraints. Over the course of the 1990s, banks compiled internal information on their client firms as their lending relationships developed, and private credit bureaus and public credit registers developed to pool this kind of privileged information. The present goal is to see how these different bits and pieces of information affect the credit constraints faced by firms. In the empirical literature, however, there is no straightforward procedure to detect the presence of credit constraints. Usually, the presence of credit constraints is detected indirectly, for example, by excess sensitivity of investment to liquidity (Schiantarelli 1996). This chapter proposes a different approach to analyzing credit constraintsÑon e that is closest in spirit to Petersen and Rajan (1994), who measure credit constraints by the degree to which firms resort to trade credit to finance their operations. Rather than using trade credit as a yardstick, however, we look at the lack of unused credit lines as an indicator of credit-constrained firms, using a cross section for October 2000 built with data collected from financial institutions by the Banco Central de la Republica Argentina (BCRA). The highest observed interest rate paid on the most expensive type of loan, overdrafts, is used to measure the marginal cost of bank credit faced by a firm. This credit option should be the last resort of the firm, which would rationally attempt to exhaust cheaper sources of credit first. The chapter therefore focuses on the intensive credit margin, which refers to how much credit is available to a firm that already operates within the financial system. Since data are not available on firms that have not received any credit from the financial system, the chapter cannot explain extensive credit margins that determine whether a firm is cut off from bank loans. The analysis finds that the availability of credit depends positively on a close relationship with a bank. Favorable characteristics of a firm (large assets, a high return to assets, a high sales/assets ratio, and a low debt/assets ratio), good credit history (normal credit situation and no bounced checks),

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

25

Conceptual Framework Theoretical Literature Information asymmetries are a central issue in credit markets. Freixas and Rochet (1998) distinguish between three sets of asymmetric information: ex ante (adverse selection), interim (moral hazard), and ex post (the costly state verification model, which can be related to ex post moral hazard). In the theoretical literature, asymmetric information has been identified as the source of equilibrium credit rationing because asymmetric information pushes the market away from a perfectly competitive equilibrium where the intersection of demand and supply clears the market. The classic contribution by Stiglitz and Weiss (1981) predicts that a lender's expected return on a risky loan is a nonmonotonic function of the interest rate. This can be due either to adverse selection, because as interest rates rise the best borrowers drop out, or to moral hazard, because at higher interest rates borrowers adopt riskier strategies. Hence, lenders may not be willing to supply more credit beyond a certain interest rate. This may lead to a credit rationing

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

and collateral also lead to higher credit availability. Significantly, a good median credit situation in the public credit register leads to higher credit availability, suggesting that the Central de Deudores eases credit constraints for healthy firms. With regard to the cost of credit, the findings mostly parallel those of credit availability. The cost is lower for a firm with a close relationship with a bank. Firms with large assets, a high sales/assets ratio, and a low debt/assets ratio pay a lower interest rate (returns on assets, however, turn out not to be significant). A good credit history and collateral also reduce the interest rate. An important difference between the effects on the price and quantity of credit is that as the credit situation deteriorates, the interest rate does not increase monotonically, although credit availability does decrease monotonically. Similarly, the median credit situation in the public credit register does not affect the cost. These findings are consistent with credit rationing, which predicts that beyond a certain point it makes no sense for the lender to raise interest rates because that would only increase the probability of default. Rather, as risk rises, lenders will cut back on the supply of credit.

26

STREB AND OTHERS

2

Banks can also be important in monitoring borrowers, thus solving corporate governance problems related to moral hazard (Allen 2001).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

equilibrium, where borrowers are willing to demand more than they are supplied at the equilibrium interest rate. This type of credit rationing is an intensive limit, or type 1 credit rationing. If lenders refuse outright to supply any loans at all to some of the prospective customers at the prevailing interest rate, the borrower faces instead an extensive limit, or type 2 credit rationing. For example, this is the case in the costly state verification version of credit rationing in Williamson (1987). Banks are precisely the organizations that specialize in collecting information on potential borrowers to price credit risks appropriately.2 To alleviate the degree of asymmetric information, banks resort to different mechanisms. A common practice by banks is to implement screening mechanisms (in which case good firms have an incentive to signal their type in order to differentiate themselves from bad firms and get better terms on their loans). Another response to the problem of asymmetric information is the development of private information through banking relationships. In this regard, Petersen and Rajan (1994) distinguish between public and private information in lending activities. Private information is the information that lenders acquire in the course of relationships with borrowersÑ information that is not easily transferable to others. Along the same lines, Berger, Klapper, and Udell (2000) contrast relationship lending with pure transactions lending. Relationship lending is based on information gathered by the lender through contact over time with the firm, its owner, and the local business community. Pure transactions lending is based on information from financial statements, credit scoring, and other similar quantitative techniques. This information is relatively public and transparent and only requires the analysis of currently available data. The idea of relationship lending is illustrated in Akerlofs (1970) example of local moneylenders in credit markets in India who can lend profitably because of their knowledge of local borrowers' creditworthiness. Outside middlemen trying to arbitrage in that market, lending at the same high rates, would lose money due to the risk of attracting borrowers with poor repayment prospects. The emergence of financial intermediaries such as banks can thus be explained as a way to solve the problem of asymmetric information in financial markets.

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

27

Relationship information is especially important for small firms because they produce less public information than large firms, which can instead resort directly to capital markets to place debt (Diamond 1991b). Relationships affect the bank-firm interaction in at least two dimensions. Concerning the amount of credit, Diamond (1984) shows that relationships allow a firm to have more access to credit. With respect to the cost of credit, there is no clear-cut association between interest rates and relationships. On the one hand, Diamond shows that a relationship with a single bank may reduce risk, leading to a lower interest rate charged. The reduction in interest rates is due both to the monitoring role of the bank, which reduces the incidence of moral hazard, and to an improved knowledge of the firm, which helps to overcome the problem of adverse selection. On the other hand, relationships provide a bank with inside information about the firm's financial health and prospects. The bank that lends to a firm learns more about it than other banks do. This information advantage gives the bank market power over the firm, allowing the bank to extract rents attributable to knowing that the borrower is less risky than average (Sharpe 1990). In summary, there are theoretical justifications for both increasing and decreasing a firm's interest rate with closer relationships. Whether other information is private or public depends on the regulations in place, such as laws of habeas data that guarantee privacy. Public credit registers and private credit bureaus that pool lender information reduce the degree of asymmetric information. Since information is costly to produce, if it becomes public other lenders might have an incentive to free ride and lure away the banks' best clients through competitive rates. This might reduce the incentives of banks to invest in relationship lending in the first place. While this may be the case for positive information on borrowers, negative information helps to discipline borrowers, reducing problems of moral hazard. Asymmetric information and agency problems are not the only source of credit market distortions. For example, an inefficient legal system may diminish the value of collateral, leading to a credit-constrained equilibrium where firms cannot take advantage of all their worthwhile investment opportunities despite the guarantees they can provide. Even if there is no credit rationing, firms suffer from credit constraints due to these inefficiencies because interest rates rise to clear the market. These problems can contribute to weak international financial links and an underdeveloped domestic financial market (Caballero and Krishnamurthy 1999).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

28

STREB AND OTHERS

Much of the literature on credit constraints faced by firms focuses on the effects on investment, which can be detected through the excess sensitivity of investment to liquidity.3 There is also a specialized literature that looks at how bank relationships can help ease liquidity constraints faced by firms. Hoshi, Kashyap, and Scharfstein (1990a), for instance, show that the benefit of bank association for Japanese firms is that it reduces the excess sensitivity of investment to cash flow in times of financial distress. This chapter takes a different tack. The present approach resembles Petersen and Rajan's (1994) study in attempting to look more closely at the issue of credit availability. The key insight in Petersen and Rajan is that the use of trade credit can be an indication of credit constraints. In their study of small business firms in the United States, the average rate on bank loans is 1 percent a month, or 12 percent a year. They calculate that firms that do not take discounts for early payment forsake a 2 percent discount to stretch the payment period by 20 days, equivalent to a 3 percent monthly interest rate, or a 44 percent annual interest rate. Accordingly, the authors use the percentage of discounts taken for early payment as an indication of firms that do not face credit constraints because they have access to cheaper credit from the financial system. Conversely, the percentage of trade credit paid late is an indication of credit-constrained firms. If trade credit discounts taken is an indirect indicator of credit availability in banks, an alternative is to try to assess directly the availability of credit in banks. This is meaningful if line-of-credit contracts or loan commitment contracts are important sources of loans. Melnik and Plaut (1986) state that loan commitment contracts are behind more than 70 percent of commercial and industrial loans in the United States. They study the ex ante trade-offs between the size of loan commitments and other variables, such as the interest rate premium charged and collateral offered. Since they concentrate on the ex ante determinants of line-of-credit contracts, Melnik and Plaut provide no evidence on credit constraints per se. They do mention that credit constraints can be interpreted as the ex post usage of credit beyond these loan commitment contracts when firms have to turn to spot loans at higher rates.
Schiantarelli (1996) provides a survey of this literature.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Empirical Literature

BANK RELATIONSHIPS AND FIRMS IN ARGENTIN229

Regarding the effect of banking relationships on access to credit, Petersen and Rajan (1994) show how their measures of lending relationships are related to reduced reliance on trade credit. They interpret this in the sense that firms with closer relationships face fewer credit constraints, having more access to cheaper bank credit. Later studies, using different measures of credit availability, also find that relationships lead to more access to credit. Cole (1998) finds that a potential lender is more likely to extend credit to a firm with which it has a preexisting relationship. Machauer and Weber (2000) find that a firm obtains a higher proportion of financing from a bank with which it has a closer relationship, using as dependent variable the ratio of the total credit line at each bank relative to the total assets of the firm. Using the interest rate paid on the last loan of each firm, Petersen and Rajan find that lending relationships affect the quantity of credit more than they affect the cost of credit. A high number of banks signals weaker relationships and is related to a higher cost of credit, but the length of the relationship does not affect interest rates at all. However, Berger and Udell (1995) also use the interest rate paid on the last loan to find that borrowers with longer-term banking relationships pay lower interest rates. They restrict their analysis to lending under lines of credit. A recent study by D'Auria, Foglia, and Marullo-Reedtz (1999) establishes that close relationships reduce the interest rate a firm is charged on uncollateralized overdraft facilities, using as a proxy of close relationships the share of each bank over the total credit lines granted by the banking system to individual borrowers. They also control for the number of banks as a measure of competition, which of course interacts with the concentration of credit in a bank. Their interpretation of the results implies that having closer relationships leads to lower rates, but there is a holdup problem if a firm operates exclusively with one bank. A small degree of competition among banks reinforces the reduction in the firm's interest rate, while further increasing the number of banks leads to higher interest rates. Combining the approaches in Petersen and Rajan (1994) and Melnik and Plaut (1986), this chapter utilizes the unused portion of precommitted credit lines as a measure of firms that are not credit-constrained. Unused credit lines can be interpreted as an intensive, not extensive, measure of credit constraints. An extensive credit constraint is the case when no bank is willing ex ante to sign a line-of-credit contract with a firm. All the firms in the present sample have bank credit, so they do not face this type of constraint

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

30

STREB AND OTHERS

Data A database was assembled, drawing information mainly from the Central de Deudores del Sistema Financiero (Debtor Center of the Financial System) of the BCRA. The information on firms includes incorporated firms that are large debtors or principal debtors of the financial system, which does not necessarily mean that they are large firms. The data set consists of a cross section of interest rates and debt of firms with each individual bank for October 2000. Previous researchers working with Central de Deudores, such as Berger, Klapper, and Udell (2000), did not use firms' interest rate and balance sheet information because they considered the information to be too unreliable, if not altogether useless. To overcome the drawbacks in the underlying information, extensive preparatory groundwork was carried out before the econometric estimates were run. Assets of Firms Balance sheet information was collected on 17,809 firms, of which 17,394 reported positive assets. However, not all the information was equally reliable. Two alternative approaches were followed to screen out noise in the information. First, filters were used in an attempt to eliminate problematic observations. An alternative validation process was then formulated that distinguished different degrees of reliability of the information.4
4

This process is discussed in the working paper on which this chapter is based, Streb and others (2002).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

(except a few firms in the sample that only have write-offs). However, even those firms that are given a credit line face a maximum credit limit or commitment. This limit is potentially an intensive credit constraint. Firms that ex post have unused credit lines are not credit-constrained, but the lack of unused credit lines may provide direct evidence on firms that face intensive credit constraints. It also appears that unused credit lines may be a less ambiguous measure of credit availability than the total debt figures that are often used in the literature.

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

31

(Thousands of pesos) Sector Unclassified Agriculture, fishing. or livestock Mining Industry Electricity, water, or gas Construction Wholesale or retail distribution Services
111 0.1 0.1

Cases

Minimum
0.1 0.1

Mean

Median

Maximum 780,710 24,400,000 4,363,000 126,000,000 9,719,920 5,915,785 57,000,000,000 426,000,000

1,776
3,076

6,108
15,106 96,351 68,018 369,145 23,406 19,400,000 207,216

1,574 2,130 4,411
3,409 63,253

3,548
125

11.9
0.1 0.1 0.1

1,052
2,947

2,519
2,039 2,820

3,161

Note: The complete sample consists of 15,796 firms in Argentina in 2000. Source: Banco Central de la Republica Argentina.

In the figures on the domestic financial system in October 2000, only 16,095 of the 17,809 firms appear, in part because the balance sheet information is mostly from 1997, 1998, and 1999. Of these 16,095 firms, 15,796 had positive assets in October 2000. These 15,796 firms form what we refer to as the complete sample. We also use three subsamples of information, which we regard as more reliable than the complete sample. Table 2.1 shows the assets of the firms in the complete sample classified according to economic sector (note that 1 peso = 1 dollar). Table 2.2 shows the assets of firms in the most reduced sample. This subsample satisfies a journalistic requirement, that is, that two or more different sources agree on the original information. In the most restrictive subsample, most of the outliers of the complete sample are eliminated.5 However, this reduced sample is made up of only a little more than a quarter of the firms, and the median size of assets is around twice as large as in the complete sample. An intermediate subsample adds to this reduced group the firms whose information has been reported on at least two different dates to the BCRA, in which case the most recent date was chosen. For this intermediate sample of 5,849 firms, table 2.3 shows the relation between their liabilities in the financial system and the book value of their assets in October 2000. Although there are still outliers, the information on assets is better
One of which exceeded the size of Argentina's GDP by a factor of 200.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.1. Assets of Firms in the Complete Sample

32

STREB AND OTHERS

(Thousands of pesos) Sector Unclassified Agriculture, fishing. or livestock Mining Industry Electricity, water, or gas Construction Wholesale or retail distribution Services
908 0.1

Cases
210 528

Minimum

Mean 10,242 11,209 84,632 145,819 280,561 24,532 16,978 44,842

Median

Maximum 332,296 482,676 1,371,330 126,000,000 1,993,481 773,847 1,720,693 6,281,000

12.0
1.0

4,201
4,333 6,282

37

226.6
0.1

1,352
66
369 947

8,271
132,417 5,325

12.0 58.4
0.1

4,314
5,696

Note: The most reduced sample consists of 4,417 firms in Argentina in 2000 that satisfy validation criteria 1/2/3/5/7 in appendix A in Streb and others (2002). Source: Banco Central de la Republica Argentina.

than in the complete sample (not shown). This subsample of firms is the preferred information set for the econometric estimations. Bank Liabilities The information on each firm's liabilities with the financial system covers four items: loans, unused credit lines, shares not quoted on the stock exchange, and write-offs. Although unused credit lines are a potential liability, we use bank liabilities to denote the aggregate figure for purposes of simplicity. This information on quantities is validated by the BCRA: the total for each item, added up over all the clients of each bank, has to agree with the amount reported by the bank to the BCRA on its balance sheet. If these totals do not coincide, the information is rejected. Table 2.4 shows the liabilities for the complete sample of firms (including those that only have write-offs) for October 2000. The average number of operations refers to the average number of banks with which firms have accounts for each line. According to Informacion de Entidades Financieras published by the BCRA, liabilities in the whole financial system in September 2000 totaled 102 billion pesos, of which 36 billion corresponded to partnerships and sole proprietorships (5,162,305 cases) and 66 billion to corporations (116,960 cases). The complete database of 16,095 firms adds up to 48 billion pesos in

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.2. Assets of Firms in the Most Reduced Sample

Table 2.3. Bank Liabilities and Firm Assets for the Intermediate Sample

(Thousands of pesos) Firm assets Median
38
1.0 1.0 3.0 1.0 1.0 170 263 379 542 807

Bank liabilities Mean 13,653 2,299 Median Minimum

Decile
42
166 264 382 550 813

Minimum

Maximum

Mean

Maximum 662,844 102,235 664,844 164,811 5,915,785

Cases
585 586 585 586 583

1
3,184
4,235

0.1

104.9 1,617 1,573
2,079

2

104.9 3,781
23,326 6,557 9,456 12,757 32,244 414,761

214.9

3

214.9 2,591 3,421 4,914 7,199
13,907 65,177

314.9

4

314.9

455.7

5

455.7

659.0

6

659.0

987.5

14.0 201.0
0.1 1.0 0.1

206,641 656,147 254,887 1,259,205 126,000,000

586 584 585 585 584

7

987.5

1,587.1

1,264 2,124
2,062 4,304 14,097

1,249 4,160

8

1,587.1

2,854.1

9

2,854.1

6,422.0

10

6,422.0

294,433.2

30,537

Note: Values are for firms that satisfy validation criteria 1/2/3/5/7/13 in appendix A in Streb and others (2002). The sample consists of 5,849 firms in Argentina in 2000. Source: Banco Central de la Republica Argentina.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

34

STREB AND OTHERS

Total (thousands Item Loans Unused credit lines Unquoted shares Write-offs Total 720,395 48,283,210 1,209,843 of pesos) 41,460,414 4,892,558

Total/ number of firms 2,576
304

Firms with line

Total/ firms

Average number of

Total/ number of operations
850 454

(percent) with line operations

96.2 41.9
1.1 9.3

2,677
726

3.1 1.6 2.1 1.9 3.3

75 45

7,117
483

3,467
251 905

3,000

100.0

3,000

Note: The complete sample consists of 16,095 firms with bank liabilities reported in October 2000 in Argentina. Source: Central de Deudores, Banco Central de la Republica Argentina.

October 2000. These firms represent nearly 15 percent of the corporations, but more than 75 percent of the debt of incorporated firms. Starting in October 2000, a detailed breakdown of loans is available. The loans are classified into 17 different types (table 2.5). Table 2.6 gives a breakdown of unused credit lines into four categories. Banks have to provision unused credit lines for losses in the same manner as with actual loans. If banks include a clause that allows them to revoke the credit line at any moment, they are not obliged to inform the BCRA of the unused portion. Hence, the unused credit lines that are reported represent unused portions of loan commitment contracts. Information on collateral for each type of loan and unused line-ofcredit line is also available. Collateral is divided into guarantees of type A (the best), type B, and type C (no collateral). Interest Rates on Loans Interest rates on each type of loan, as well as the average duration of these loans, are available for October 2000. The BCRA has not yet implemented a validation procedure similar to that applied to quantities, so interest rate figures are not very reliable.6 The validated interest rate information is presented in table 2.7.
6

A description of how this information was processed appears in appendix B in Streb and

others (2002).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.4. Bank Liabilities for the Complete Sample of Firms

Table 2.5. Breakdown of Loans Average Total/number Firms with Total/firms with line operations
2.3 1.0 0.0 2.1 1.0 1.1 1.0 1.3 1.1 1.2 584

Total number of of firms line (percent)
381

Total/ number of operations
259

(thousands

Type of loan

of pesos)

Overdrafts
0
0.1 0.0

6,138,594
62
0 0
869

65.3

Discounted bills - guarantee A1

1,310 58.2 1,494
195 590 0.6

62
0
717 191 519

Discounted bills - guarantee A2
1
148

0

Bills and promissory notes

13,990,700

House mortgages

19,663

Other mortgages
2
4.0

2,380,451
51
267

25.1 24.9
6.8

Car loans
67
3

32,772
41 73

50
203

Other pledges
12
116 0.3 0.0

1,072,312

Personal

44,741

37 60
3.9

Credit card
22
830

195,885

16.8

Interfinancial

1,866,267

33,326 59,273

8,640
1.2

To guaranteed public bank

355,638

50,805

Other loans
76 44
3 0
7.2 0.4 0.0

13,364,295

42.7 35.7

1,946
214 610 745 149

1.4 1.3 1.1 1.0 1.0

1,376
168 531 733 149

Other financial

1,228,170

Leases

702,075

Miscellaneous

43,224

Small personal

746

Source: Central de Deudores, Banco Central de la Republica Argentina.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

36

STREB AND OTHERS

Type of credit line
Overdraft facilities Eventual liabilities Guarantees granted Endorsements and external credit lines

Total Total/ Firms Total/ Average Total/ (thousands number with line firms number of number of of pesos) of firms (percent) with line operations operations
381,926 1,155,193 2,790,482 564,957 24 72 173 35
29.0

82
726

1.3 1.6 1.3

64
462 838

9.9
15.8

1,094

0.5

6,494

1 .0

6,348

Source: Central de Deudores, Banco Central de la Republica Argentina.

Bank overdrafts are, on average, the most expensive type of loans. Information on inter-financial loans is disregarded because these rates appear to be greatly underreported in the transformed sample, accounting for less than 2 percent a month in 2000. Furthermore, this type of debt does not qualify for the sample that targets nonfinancial firms. The only other rate that is close to overdrafts is the rate charged on credit cards. Marginal Cost of Credit The measure of the marginal cost of credit is the highest observed interest rate on overdrafts paid by a firm in the financial system. Applying this definition, the marginal rate turns out to be slightly above 3 percent a month (with small fluctuations according to the specific subsample of firms used). The highest observed interest rate charged on any type of loan could have been used instead, but in 85 percent of the cases of firms that use overdrafts, the highest rate is indeed the rate charged on overdrafts. As to the relevance of the interest rate for overdrafts, table 2.5 shows that 65 percent of the firms in the data set used overdrafts in October 2000. An advantage of only using overdrafts is that they represent a homogeneous type of loan that reduces the influence of unobserved loan characteristics on the cost of credit. The rate charged on overdrafts can be reviewed monthly. Although overdrafts can be rolled over and do not have

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.6. Breakdown of Unused Credit Lines

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

37

(Percent) Validated sample Standard Type of loan Overdrafts Discounted bills - guarantee A Bills and promissory notes House mortgages Other mortgages Car loans Other pledges Personal Credit card Interfinancial To guaranteed public bank Other loans Other financial Leases Miscellaneous Small personal
a b

Total Cases3 15,529
20

Mean

Median

deviation
1.6 0.4 0.5 0.2 0.6 0.3 0.9 0.7 0.9

casesb 23,720
21

2.54 2.22 1.28 1.04 1.14 1.41 1.27 1.82 2.43 4.97 0.91 1.39 1.07 1.25 1.09 1.12

2.04 2.17 1.19 1.02 1.08 1.46 1.20 1.75 2.59 0.75 0.90 1.28 1.00 1.28 0.99 0.99

13,465
77

19,517
103

2,724
509

4,589
660

2,530
530

5,279
1,198

1,727
132

3,283
216

19.9
0.1 0.8 0.5 0.3 0.8 0.2

7

7
9,712

5,488

1,282
481

7,327
1,323

47
5

59
5

Restricted to rates with monthly values greater than 0.1. Total cases in the complete sample. Source: Central de Deudores, Banco Central de la Republica Argentina.

a specified termination date, their duration would be expected to be quite short. Table 2.8 presents the data on maturity. The data have not been validated, so medians seem more reliable than averages. The median of one month for overdrafts confirms that their duration is short. Since overdrafts are short-term liabilities, they are overwhelmingly denominated in pesos. As the maturity lengthens, debt tends to be denominated in dollars. This is most clearly the case for the debts with longest maturity, house mortgages. Credit card lines are similar to overdrafts in several aspects. Credit cards are subject to both pre-established rates and credit limits characteristic of loan commitment contracts. However, this approach was rejected because they are not completely homogeneous lines. According to table 2.5, the average size of operations with credit cards is substantially smaller than that of overdrafts, and their use by firms is much less widespread.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.7. Interest Rates in Validated Sample

38

STREB AND OTHERS

(Months)
Type of loan Overdrafts Discounted bills - guarantee A Bills and promissory notes House mortgages Other mortgages Car loans Other pledges Personal Credit card Interfinancial To guaranteed public bank Other loans Other financial Leases Miscellaneous Small personal

Average
4

Median

Minimum
0
0 0 0 0 0 0 0 0 0 9 0 0 0 0 2

Maximum
590 119 770 833 963 237 450 200 590

Number of cases
23,720
21

11
7
131

1 1 1
93 10 16
1 0 1 1

19,517
103

42 20 12
5 1 2

4,589
660

5,279
1,198

3,283
216

64 82
925 999 139 455

31
7 9

28
1 1 1

7
9,712

7,327
1,323

12 72
7

21
6

59
5

15

Note: Values are for the complete sample of 16,095 firms. Source: Central de Deudores, Banco Central de la Republica Argentina.

Access to Additional Credit Two possible alternatives as measures of credit constraints are overdrafts over total bank credit and the percentage of authorized overdrafts effectively drawn on. With respect to the first alternative, the conjecture was that overdrafts, being the most expensive, would be the last type of loan a firm would resort to in the financial system. Moreover, their cost at the margin resembles the trade credit paid late variable in Petersen and Rajan (1994), which amounts to a rate of 3 percent a month. If trade credit paid late indicated that the firm faced credit constraints, it initially seemed possible that a high percentage of overdrafts in relation to total bank loans would indicate the same problem. As table 2.9 shows, however, the findings do not support this conjecture. Although the percentage of credit drawn through overdrafts is roughly constant across categories, the percentage of firms that actually use overdrafts tends to fall as the credit situation deteriorates. This seems to indicate

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.8. Maturity by Loan Type

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

39

Median credit situation3
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
a

Total cases
11,788
306 581 214 497 149 909 220

Firms using overdrafts (percent)
71.5 67.6 51.6 58.9 46.3 57.0 63.5 48.6 31.4 28.1 29.8

Overdrafts/ bank loans'3 (percent)
31.7 21.5 28.1 24.8 29.5 25.3 33.0

37.0
33.0 23.5 35.3

1,352
32 47

This is the median of a firm's credit situation in banks. It ranges from 1 (the best credit standing) to 5 (the worst); 6 is a technical category for delinquent loans with failed banks. b This refers only to firms actually using overdrafts. Note: Values are for the complete sample of 16,095 firms. Source: Central de Deudores, Banco Central de la Republica Argentina.

that firms are cut off from overdrafts when they run into financial trouble. Unlike trade credit, the supply of which might be fairly elastic, overdrafts are part of loan commitment contracts that are subject to a prior approval process. This illustrates why the use of overdrafts is an ambiguous indicator of credit rationing. Even if it is the most expensive source of credit in the financial system, the firms in the worst shape have less access to it. In this sense, using overdrafts poses problems similar to using leverage indicators. The second alternative measure of credit constraints, the percentage of authorized overdrafts effectively drawn on, takes into account that overdrafts are a line-of-credit contract subject to predefined quotas. The conjecture in this instance was that the amount of unused overdrafts could provide valuable information to identify firms that were not credit-constrained. Firms in good condition would tend to use overdrafts infrequently, while troubled firms would tend to exhaust their available credit. Since overdrafts are limited and could be expected to be the credit source of last resort, once that limit is reached it would be expected that the firm would be creditconstrained in the financial system.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.9. Overdrafts Drawn by Credit Situation

40

STREB AND OTHERS

Table 2.10. Unused Overdrafts and Credit Lines by Credit Situation
Median credit situation3
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0
a

Unused overdrafts Total cases
11,788
306 581 214 497 149 909 220

Unused credit lines Percentage of firms
52.6 33.0 22.2 16.4 18.1 10.7 10.1
7.3 4.0 6.3 6.4

Percentage of firms
37.2 23.5 15.8
9.3 9.7 6.7 2.9 1.8 0.7 0.0 4.3

Percentage of liabilities13
10.8
6.2 7.1

Percentage of liabilities'
21.3 11.2

9.8
13.1 15.3 10.3

11.8
8.2

14.7 13.2 17.3 10.9
0.2

8.3
11.4

1,352
32 47

9.3 0.2
25.3

24.4

This is the median of a firm's credit situation in banks. It ranges from 1 (the best credit standing) to 5 (the worst); 6 is a technical category for delinquent loans with failed banks. b Refers only to firms that actually have unused overdrafts. c Refers only to firms that actually have unused credit lines. Source: Central de Deudores, Banco Central de la Republica Argentina.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Both of these measurements depend on information for which records may not be readily available. In contrast to these limited sources of information, banks provide detailed information to the central bank on all unused credit lines that are contained in loan commitment contracts. Consequently, bearing in mind Petersen and Rajan's (1994) use of trade credit discounts as a shadow indicator of credit availability across banks, a similar measure based on unused credit lines was developed for this study. This perspective also relates to the view in Melnik and Plaut (1986) that credit constraints are related to ex post use of credit in excess of the amounts agreed on in loan commitment contracts. The BCRA requires banks to provision unused credit lines for credit risk unless the credit line is revocable at any time, so the figures of unused credit lines basically reflect loans backed by loan commitment contracts that have been granted but not drawn. Table 2.10 shows the behavior of unused overdrafts and total unused credit lines, including unused overdrafts. For example, 53 percent of firms with a median credit standing of 1 have unused credit lines. Thus, a large percentage of firms with the best credit records have unused credit lines, representing a large share of their

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

41

Bank Relationships Argentina has an underdeveloped capital market. Few firms are quoted on the stock exchange, there is a negligible domestic corporate bond market, and most firms basically have to rely on bank debt. The country resembles what Mayer (1994) calls a banking economy, with a small proportion of quoted companies, a high concentration of ownership, and long-term relations between banks and industry, as opposed to a market economy where the opposite would hold. This fact makes the impact of banking relationships particularly relevant. More generally, banking relationships are important for small and medium enterprises (SMEs) because they are typically the firms that are most likely to resort to bank credit. Most of the Argentine firms in the data set are SMEs. Despite this high lending activity, SMEs have aptly been described as information opaque firms because they provide limited public information (Berger, Klapper, and Udell 2000). Although the private information generated in relationships is not available for analysis, the literature uses several variables as signals of close or weak relationships. The analysis uses proxies to represent bank relationships. Table 2.11 shows the behavior of three measuresÑnumbe r of banks, number of credit lines, and liabilities in the main bank as a percentage of liabilities in the financial system)Ña s the size of financial liabilities increases. These measures of bank relationships refer to the number of banks with which a firm has accounts, the extent of the business it conducts with each bank, and the importance that the main bank has in covering the firm's financial needs. The number of financial institutions (no. banks) is inversely related to the strength of bank relationships. The number of banks rises with the size of liabilities, so larger firms rely less on any given financial institution (as a rule, they also rely less on bank credit to finance their needs). The number of credit lines (loans, unused credit lines, and unquoted shares, no. credit lines) is directly related to the strength of relationships. The average number of operations with each financial institution shows no

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

bank liabilities. Unused overdrafts follow approximately the same pattern, but they are less representative of available sources of funds in banks. For that reason, we use unused credit lines as a percentage of liabilities within the financial system as the final measure of credit availability.

Table 2.11. Bank Relationship by size of Liablilities

Liabilities Number of banks Average Median Median
1.0 1.0 2.0 2.0 2.0 1.7 1.7 1.8 1.8 2.0 2.0 2.0 1.8

Number of credit lines3 Average
1.2 1.5 1.6 1.6 1.6

Liabilities bank/system3 Average (percent)
92 83 81 81 78 74 71 67 64

Decile
13
9
1.5 2.1 2.3 2.4 2.7 3.0 3.3

Cases
1
2 2 2 2 3 3 4 4 6

Minimum (thousands of pesos)
84 83
168 235 312 425 606 982 4.0 4.8 7.0 168 237 314 427 613

Maximum (thousands of pesos) Mean (thousands of pesos) Median (thousands of pesos)

Median (percent)
100

1

1,610

0.1

41

2

1,611

41

127

94 89 90 81 74 71 64 61

3

1,610

127

205

4

1,609

205

272

5

1,609

272

363

6

1 ,608

363

502

7

1,610

502

757

8

1,609

757

1,342 1,007 2,127 2,001
7,856 25,017

9

1,610

1,342

3,413

10

1,609

3,413

1,702,432

1.7

1.7

56

51

Firms that only register write-offs in the financial system are assigned a zero. Source: Central de Deudores, Banco Central de la Republica Argentina.

a

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

43

Econometric Evidence This section uses the cross-sectional variation among firms to analyze the high cost of credit and the limited access of firms to credit in Argentina. The analysis is based on measuring the cost and availability of bank credit at the margin. Although the methodology is inspired by Petersen and Rajan (1994), we use several different right-hand variables to extend the study in order to examine the impact of credit history. We also use different left-hand variables that relate to the idea of loan commitment contracts in Melnik and Plaut (1986). The marginal interest rate that a firm has to pay on overdrafts captures the cost of credit. With respect to access to credit, the regressions are based on leverage, similar to what is commonly used in the literature. However, high leverage is inherently ambiguous because it may indicate either that a firm enjoys high credit availability or that it is in financial distress. An alternative approach is to detect credit-constrained firms based on the proportion of outstanding credit lines.
7

See appendix A in Streb and others (2002).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

marked upward trend with size. Note that there are up to 17 types of loans, four types of unused credit lines, and unquoted shares count as 1, so this variable can reach a maximum value of 22. The concentration of operations with the largest creditor institution (liabilities bank/system), measured by the ratio of the liabilities in the main bank to liabilities in the financial system, exclusive of write-offs, is directly related to the strength of the relationship with the main bank. The concentration decreases with the size of liabilities. However, even in the largest decile, the median firm conducts 50 percent of its business with a single institution. Note that only in the last three deciles in table 2.11 are there firms with more than 1 million pesos in liabilities in the domestic financial system. Taking as a working definition of SMEs those firms that receive less than 2.5 million pesos in loans from the financial system (Burdisso and others 2001), more than 85 percent of the firms in the complete sample are SMEs. In the more restrictive set of firms, table 2.3 shows that, applying this same criterion of bank debt, more than 75 percent are SMEs.7

44

STREB AND OTHERS

Overdraft Rate at Marginal Bank In the interest rate regression, the dependent variable is the interest rate paid by firms on their overdraft account, which is considered informative. In case a firm uses more than one overdraft account, the overdraft account that pays the highest interest rate is used. Using this definition, the marginal overdraft interest rate is on average slightly more than 3 percent a month. Because a firm can switch overnight from one account to another or from one bank to another, thereby canceling its outstanding overdrafts, it seems rational to use the most expensive overdraft account used as the marginal account. The interest rate on overdrafts is a relevant interest rate for

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

In each case, we run regressions on subsets of information to see if more restrictive validation criteria lead to more precise estimates. Naturally, a more restrictive validation criterion leaves fewer observations, hopefully of better quality for the problem analyzed. There is thus an implicit tradeoff when restricting the sample, in which degrees of freedom are sacrificed to obtain more relevant observations. Of particular interest is how information affects credit conditions. If banks can screen between good and bad credit risks more easily, this should reduce the problem of "lemons" and improve good firms' access to credit. Do close relationships matter? It seems they do, reducing the cost of credit and increasing the access to credit. It appears that firms with closer ties to financial institutions not only share a considerable amount of information with banks, but also show they are good credit risks. Otherwise, if a firm were bad, the bank would lose interest in developing or maintaining the relationship once it got to know the firm better. Does credit history matter? It seems to matter, but whereas the negative effect on access to credit seems to increase monotonically as the credit history deteriorates, the effect on interest rates is more of a step function. This makes sense from the perspective of credit rationing because raising interest rates for firms in very bad shape only increases the probability of default. Public information also seems to matter because the credit history in other financial institutions impacts bank clients' access to credit. The analysis controls for the characteristics of a firm from financial statements as well as characteristics of banks and loans and, when possible, the industrial sector to which the firm belongs.

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

45

Highest overdraft interest rate = |3

industry dummies
Credit history

measures of relationships

loan and lender characteristics

firm-specific information

The equation introduces controls for firm-specific information, industry dummies, and loan and lender characteristics. The analysis follows Petersen and Rajan (1994) in using relationship characteristics to capture the influence of private information, and introduces variables on credit history that are not in Petersen and Rajan (1994). As Greene (1992) emphasizes, credit history plays an important role in interest rate formation and access to credit. We estimate parameters by ordinary least squares using the Eicker-White heteroskedasticity-consistent estimator for standard errors. Table 2.12 displays results for four sets of firms. Regression 1 corresponds to the complete sample, while regression 2 applies a filter to eliminate problematic observations (those with no return on assets and cases where either assets, liabilities, or net worth are between 0 and 1,000 pesos). Regression 3 corresponds to registers where at least two sources agree on the information. Regression 4 also includes firms whose balance sheet information has been regularly updated by a bank. Although it cannot be cross-checked with other reports, this information seems to behave much like the information that can be checked. Given the large heterogeneity of the sample considered, the possible presence of outliers that can seriously affect least squares estimates cannot be dismissed in advance. In addition, all models are estimated using robust methods, for instance, Huber's biweight function based on Cook's distance (Wilcox 1997). The robust estimates

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

most of the firms with relationships with the financial sector, and it is used more often than any other single credit line. In those firms actually using overdrafts, it is almost always the highest recorded rate. From an econometric perspective, the advantage of concentrating on the rate of a specific credit line lies in not having to model the influence of line-specific factors that affect the underlying rate. Those controls would be necessary if the analysis were to be made on the highest rate charged on any line. In addition, the monthly pricing of this account facilitates the estimation procedure because there is no need to introduce extra variables to measure the underlying cost. The constant of the regression should capture it. The following regression function is estimated:

46

STREB AND OTHERS

Complete sample Variable
roa

(D
0.000 (-1.12) -0.100 (-9.18)*** -0.246 (-6.39)***
0.011

Filtered complete sample3 (2)
0.000 (-1.12) -0.125 (-8.72)*** -0.245 (-5.32)***
0.154

Reduced sample6

(3)
0.076 (1.06) -0.150 (-6.85)*** -0.189 (-2.36)** 0.624 (2.83)*** -0.160 (-1.15) -0.004 (-0.04) -0.434 (-1.31) -0.306 (-1.12)
0.163

Reduced sample with extra firms' (4)
0.075 (1.07) -0.129 (-6.98)*** -0.180 (-2.68)*** 0.452 (2.45)** 0.020 (0.19)
0.110

Inassets Insales/assets Indebt/assets sector 0 sector 1 sector 3 sector 5 sector 6 sector 7 sector 8 situation 2 situation 3 situation 4 situation 5 situation 6 Inbounced/liabilities unwarranted/overdrafts

(0.12)
0.215

(1.41)
0.110

(3.25)***
0.129

(1.40)
0.130

(2.24)** -0.033 (-0.16) -0.059 (-0.29) 0.272 (3.78)***
0.110

(2.03)** -0.075 (-0.35)
0.010

(1.22) -0.237 (-0.87) -0.244 (-0.96) 0.220 (2.09)** -0.003 (-0.04)
0.198

(0.05) 0.223 (2.84)*** 0.066 (1-19) 0.266 (4.55)*** 0.539 (4.71)*** 0.258 (2.44)**
0.501

(1.38) -0.086 (-1.05)
0.158

(2.16)** 0.302 (5.65)*** 0.652 (6.32)*** 0.252 (2.62)*** 0.422 (5.87)*** 0.228 (2.22)**
0.811

(1.76)*
0.511

(2.54)** 0.577 (3.72)*** 0.084 (0.56) 0.475 (4.45)*** 0.205 (1.23) 0.767 (1.90)* 0.066 (4.67)*** 0.963 (6.90)***

(2.98)*** 0.002 (0.01) 0.538 (4.66)***
0.170

(6.23)***
0.121

(0.97) 0.994 (2.41)** 0.049 (4.88)*** 0.903 (9.09)***

(0.95) 0.698 (1.69)* 0.072 (4.62)*** 0.893 (5.49)***

(2.55)** 0.048 (5.42)*** 0.896 (10.19)***

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.12. Interest Rate Regressions: Subsamples

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

47

Complete sample Variable
foreign bank=1 public bank=l wholesale bank=1 bank's market share liabilities bank/system no. banks no. credit lines _cons

(D
0.198
(5.09)***

Filtered complete sample3 (2)
0.153
(3.54)***

Reduced sample6 (3)
0.165
(2.51)** 0.295 (1.70)* -0.989 (-7.99)*** -9.484 (-9.84)*** -0.653 (-5.39)*** 0.092 (8.45)*** -0.133 (-4.38)***

Reduced sample with extra firms' (4)
0.146

(2.55)**
0.261

0.110
(1.27) -1.000 (-11.33)*** -8.779 (-15.54)*** -0.690 (-11.21)*** 0.069 (9.06)*** -0.129 (-7.23)*** 3.525 (24.65)***

0.213
(1.99)* -1.005 (-10.62)*** -9.118 (-14.56)*** -0.647 (-9.27)*** 0.084 (10.11)*** -0.136 (-6.84)*** 3.646 (20.68)***

(1.63) -1.011 (-8.90)*** -8.724 (-10.36)*** -0.646 (-6.45)*** 0.090 (8.89)*** -0.142 (-5.37)*** 3.396 (14.33)*** 3,963 33.52
0

3.617
(12.73)***

Number of observations

8,548 63.44
0

7,010
55.34
0

3,112 25.91
0

F
Prob > F

R2
Adjusted R2 Root MSB
a b

0.1569 0.1544

0.1653 0.1623 1.5859

0.1735 0.1668 1.6279

0.1755 0.1703 1.5897

1.58

Excludes roa=0 and balance sheets with problems. Validation criteria 1/2/3/5/7 in appendix A in Streb and others (2002). c Validation criteria 1/2/3/5/7/13 in appendix A in Streb and others (2002). * Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: The dependent variable is the interest rate. Equations are estimated using ordinary least squares; t-statistics are in parentheses. Source: Authors' calculations.

(not shown) are fairly similar to those obtained by ordinary least squares, so outliers do not affect the estimation significantly. Although the samples of regressions 3 and 4 appear more reliable, the estimates in table 2.12 are not very sensitive to the specific sample used. In the initial estimates, many coefficients in the first two regressions had signs

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.12. (continued)

48

STREB AND OTHERS

8

All the variables defined as Inx are calculated adding 1,000 pesos to x, so ln(l+x) is the actual explanatory variable. This transformation is due to the many zeros that appear in the regression with the complete sample. Otherwise, a large number of observations would be lost. 9 The dummies for type of bank might in part capture characteristics of firms. For example, prime firms tend to operate with large banks. It is not clear why foreign banks charge higher rates. Although Claessens, Demirgiic-Kunt, and Huizinga (2001) find that foreign banks charge higher spreads than domestic banks in developing countries, they do not specifically relate it to higher loan rates.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

significantly different from the last two regressions. Putting more controls in place, the most noteworthy differences are that in the first two regressions, the firm's debt/assets ratio (Indebt/assets) is not statistically significant, and the return on assets (roa) has a negative sign (although it is never statistically significant).8 The main results involving the cost of credit (skipping the controls for the industry sector dummy) are as follows. Looking at a firm's characteristics, the cost of credit is negatively and significantly associated with the amount of assets (Inassets) and the sales/assets ratio (Insoles/assets), while it is positively and significantly related to the firm's debt/assets ratio (Indebt/assets). With respect to credit history, the cost of credit is positively related to an unfavorable credit situation (situation*} and to bounced checks at the marginal bank (Inbounced/liabilities). However, in the case of credit situation, the coefficients are not always significant and they do not increase with the deterioration of the credit situation. Relationship variables show that the cost of credit is negatively and significantly associated with debt concentration at the marginal bank (liabilities bank/system) and with the number of accounts at the marginal bank (no. credit lines], but positively and significantly related to the number of banks (no. banks}. Although table 2.11 shows that the number of banks increases with the size of liabilities in the financial system, the regressions control for the size of firms (Inassets}, so a size effect can be ruled out. Other significant controls show that less collateral (unwarranted/over drafts) leads to higher interest rates (table 2.12). Marginal banks that are wholesale (wholesale bank} or large (bank's market share} charge lower interest rates, and marginal banks that are foreign (foreign bank) charge higher interest rates.9 The effects of credit history merit further comment. The variable for bounced checks is measured as the ratio in the marginal bank between the amount in bounced checks from January to September 2000 and liabilities. There are five dummy variables for the credit situation of each firm.

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

49

10 Note that the number of banks (no. banks) is not completely independent of the concentration of business at the marginal bank (liabilities bank/system). This is emphatically so when no. banks=l, which implies liabilities bank/system=l. However, the mean value of no. banks in regression 4 is 5.1 banks. Furthermore, the two variables do not have a high positive correlation.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Situation 1 is omitted, so it acts as the benchmark. Although not all parameters are significantly different from zero in the four estimates, all of the coefficients are positive. This indicates that firms with poor credit histories tend to pay higher interest rates. Furthermore, if a dummy for credit situation 1 is included, it is negative and statistically significant (we do not show the regression due to space considerations). Interest rates do not increase as the credit situation worsens to situation 4, which corresponds to firms under debt rescheduling proceedings, or situation 5, which corresponds to firms filing for bankruptcy. Since situation 6 is a technical category that represents arrears with financial institutions that went bankrupt, it cannot be ranked as worse than either situation 4 or 5. These results maybe interpreted as an indication of credit rationing. As the risk of default increases, beyond a certain point it makes no sense to increase the interest rate further. Rather, banks will simply cut back the amount of credit. As for the relationship variables, the variable liabilities bank/system is an indication of how deep the marginal bank's involvement in the firm is. An increase in this variable means that the bank has a higher stake in the firm. This higher exposure assumed by a bank also generates the incentive to monitor the firm more closely. However, it could possibly indicate that the firm does not have access to other banks. In this case, the parameter of this variable should have a positive sign, signaling the firm's risk. According to table 2.12, the effect of debt concentration at the marginal bank is explained by the first hypothesis. A second variable to control for relationships, no. credit lines, is the number of accounts the firm has at the marginal bank. Each type of bank account generates information. Although not all accounts are informative, we consider the total number of accounts as a proxy for informative accounts. The results of the regressions show that firms with more accounts are charged lower interest rates. A third variable for relationships, no. banks, does not have a straightforward sign. The issues are similar to those for liabilities bank/system.10 On the one hand, this variable could signal that a firm wants to diversify its

50

STREB AND OTHERS

Leverage Measuring access to credit is not an easy task. Leverage is commonly used in the applied literature. For the purposes of this chapter, the analysis uses

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

financing sources so as not to be a captive client of a single bank. If this is the case, an increase in the number of banks should reduce the interest rate. On the other hand, this variable could signal that the relationship is weaker, in which case an increase in the number of banks should go hand in hand with an increase in the interest rate paid. The interest rate paid could also rise if more banks signaled that the marginal bank could not monitor the firm as effectively, or if the marginal firm signaled that the main bank cut the firm's access to credit, forcing that firm to seek other lenders. According to the results in table 2.12, firms that deal with more banks are charged higher interest rates, which signals either weaker relationships or moral hazard and quality problems. Taking into account the quantity effects, the additional effects of the variable no. banks may indeed be an indication that firms need to seek financing from another bank because they are cut off by their main lender. In addition to the basic regressions in table 2.12, we ran alternative regressions using the preferred set of firms. The only significant change in table 2.13 is that when a quadratic term for number of banks is introduced (no. banks squared), the linear term, although still positive, loses its statistical significance. The quadratic term is positive and significant, so the effect on the number of banks is convex, leading to increasingly higher marginal rates. If anything, this confirms the interpretation of the number of banks as a distress signal, where a firm that is more in need of funds faces steeply increasing financing costs at the margin. A dummy variable for bounced checks at other banks (bounced other bank) turns out to be significant, although this was not public information released at the time by the BCRA. Although this information only became public in mid-2001, bounced check records may previously have been shared among banks or revealed by clearinghouses. There is a dummy for differences between the rating of the firm by the marginal bank and the median of the financial system, system situation, which takes the value 1 if the median situation in the system is worse and -1 if it is better. This variable turned out to be not statistically significant.

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

51

Variable
roa

(D
0.075 (1.07) -0.129 (-6.98)*** -0.180 (-2.68)*** 0.452 (2.45)** 0.020 (0.19) 0.110 (1.22) -0.237 (-0.87) -0.244 (-0.96) 0.220 (2.09)** -0.003 (-0.04) 0.198 (2.54)** 0.577 (3.72)*** 0.084 (0.56) 0.475 (4.45)*** 0.205 (1.23) 0.767 (1.91)* 0.066 (4.67)*** 0.963 (6.90)*** 0.146 (2.55)** 0.261 (1.63)

(2)
0.079 (1.13) -0.120 (-6.39)*** -0.176 (-2.64)*** 0.437 (2.37)** 0.033 (0.32) 0.111 (1.23) -0.249 (-0.92) -0.234 (-0.92) 0.210 (2.00)** -0.004 (-0.06) 0.195 (2.51)** 0.525 (3.38)*** 0.014 (0.09) 0.379 (3.46)*** 0.178 (1.07) 0.695 (1.73)* 0.049 (3.31)*** 0.971 (6.96)*** 0.157 (2.75)*** 0.262 (1.64)

(3)
0.080 (1.14) -0.120 (-6.41)*** -0.178 (-2.66)*** 0.441 (2.39)** 0.034 (0.32) 0.111 (1.23) -0.249 (-0.91) -0.234 (-0.92) 0.211 (2.01)** -0.004 (-0.06) 0.195 (2.51)** 0.504 (3.16)*** 0.000 (0.00) 0.376 (3.43)*** 0.160 (0.95) 0.645 (1.57) 0.049 (3.34)*** 0.973 (6.98)*** 0.157 (2.74)*** 0.263 (1.65)*

(4)

0.080 (1.15) -0.119 (-6.36)*** -0.173 (-2.59)*** 0.445 (2.41)** 0.027 (0.26) 0.100 (1.11) -0.249 (-0.91) -0.237 (-0.93) 0.209 (1.99)** -0.011 (-0.15) 0.182 (2.39)** 0.505 (3.17)*** 0.004 (0.02) 0.394 (3.59)*** 0.174 (1.03) 0.630 (1.53) 0.048 (3.24)*** 0.973 (6.98)*** 0.162 (2.83)*** 0.247 (1.54) (continued)

Inassets Insales/assets Indebt/assets sector 0 sector 1 sector 3 sector 5 sector 6 sector 7 sector 8 situation 2 situation 3 situation 4 situation 5 situation 6 Inbounced/liabilities unwarranted/overdrafts foreign bank= 1 public bank=l

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.13. Interest Rate Regressions with Additional Variables

52

STREB AND OTHERS

Variable
wholesale bank= 1 bank's market share liabilities bank/system no. banks no. credit lines _cons bounced other bank= 1 system situation (-1,0, 1) no. banks squared

(D

(2)
-1.004 (-8.85)*** -8.749 (-10.41)*** -0.624 (-6.23)*** 0.086 (8.46)*** -0.139 (-5.27)*** 3.282 (13.76)*** 0.295 (3.79)***
-

(3)
-1.005 (-8.86)*** -8.771 (-10.42)*** -0.623 (-6.22)*** 0.086 (8.47)*** -0.140 (-5.28)*** 3.285 (13.76)*** 0.297 (3.81)*** -0.049 (-0.58)
-

(4)

-1.011 (-8.90)*** -8.724 (-10.36)*** -0.646 (-6.45)*** 0.090 (8.89)*** -0.142 (-5.37)*** 3.396 (14.33)***
-

-1.007 (-8.88)*** -8.864 (-10.52)*** -0.720 (-6.63)*** 0.033 (1.32) -0.133 (-5.00)*** 3.444 (13.87)*** 0.308 (3.95)*** -0.049 (-0.58) 0.003 (2.31)** 3,963 30.77
0

Number of observations
F

3,963 33.52
0

3,963 32.89
0

3,963
31.68

Prob > F
R2

0

0.1755 0.1703 1.5897

0.1785 0.1731 1.5871

0.1786 0.1729 1.5872

0.1797 0.1738 1.5863

Adjusted R2 Root MSE

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: Equations are estimated using ordinary least squares; t-statistics are in parentheses. The validation criteria 1/2/3/5/7/13 in appendix A in Streb and others (2002). Source: Authors' calculations.

the ratio of several different measures of debt to assets, from broader to more restricted measures. The dependent variable is one of four measures of a firm's debt ratio: ¥ Indebt/assets: natural log [l+(debt/assets}} • lnliabilities/a: log [l+(liabilities in main bank/assets}] • lnloans/a: log [l+(loans in main bank/assets}] • Inoverdraft/a: log [^(overdrafts in main bank/assets)}.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.13. (continued)

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

53

Except for the first variable, where liabilities and assets are taken from balance sheets at the same point in time, in the other cases there is some noise. A figure of debt with the financial system in October 2000 was divided by the assets reported in the last available balance sheet statement. The balance sheet information is mostly from 1998 and 1999. But since the economy on average has not grown since 1998 due to a prolonged slump (followed by a strong downward slide in 2001), it is not clear that there is any bias. Table 2.14 shows the estimates with one of the subsamples. The results are similar with the complete sample of firms. Using total bank debt (column (1)) instead of debt from the main bank (columns (2), (3), and (4)) as the dependent variable does not substantially change the results. Petersen and Rajan (1994) point out that estimates where the dependent variable is some measure of a firm's debt ratio suffer from identification problems. Changes in the debt ratio can be due to changes in the demand for credit (the supply curve is observed) or to changes in the supply of credit (the demand curve is observed). For example, Petersen and Rajan expect credit availability to be greater for higher-quality firms, and in this sense they find that larger firms tend to have a high debt-to-asset ratio. However, older firms and more profitable firms (which they also expect to be of higher quality) have lower, not higher, debt ratios. The problem is that it is difficult to tell, for instance, whether older firms are rationed by creditors (a supply effect) or have a lower demand for external credit. Therefore, Petersen and Rajan discard indebtedness as a measure of credit availability, turning instead to the percentage of discounts on trade credit taken by the firm. In this sense, the signs in table 2.14 are sensitive to the definition. Firms in bad credit situations tend to be highly indebted, which seems to indicate a demand factor related to financial distress. However, bank credit tends to be smaller when the credit situation deteriorates (except for situation 6, a special technical category). This is particularly clear for overdrafts and seems to indicate a supply effect, that is, that firms are cut off from credit by banks. Although good firms may need to resort less to overdrafts, a supply factor such as willingness to lend can reduce the amount of overdrafts available to bad firms. Thus, overdrafts have the same problems of identification as indicators of indebtedness. Another example is the sales/assets ratio, which has a positive sign in column (1) in table 2.14. This can be interpreted as meaning that firms with good investment opportunities have greater access to credit. The negative

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

54

STREB AND OTHERS

Dependent variable Indebt/ assets Variable
roa
Inassets Insales/assets sector 0 sector 7 sector 3 sector 5 sector 6 sector 7 sector 8 situation 2 situation 3 situation 4 situation 5 situation 6 Inbounced/liabilities guarantees-! foreign bank= 1 public bank= 1

^liabilities/ assets

Inloans/ assets

Inoverdrafts/ assets

(D
-0.009 (-1.52) 0.009 (6.81)***

(2)
0.055 (3.51)*** -0.133 (-37.40)*** -0.142 (-11.42)*** -0.032 (-1.51) -0.040 (-2.27)** -0.022 (-0.41) 0.230 (5.00)*** 0.045 (2.18)** -0.008 (-0.57) 0.037 (2.49)**

(3)
0.055 (3.44)*** -0.135 (-36.44)*** -0.124 (-9.68)*** -0.035 (-1.66)* -0.034

(4)
0.003 (0.17) -0.069 (-19.02)*** -0.030 (-2.47)** -0.010 (-0.51) -0.008 (-0.43)

0.112
(23.14)*** -0.002 (-0.26) -0.016 (-2.31)** -0.043 (-1.98)** -0.021 (-1.15)

(-1 .84)*
-0.040 (-0.69) 0.237 (4.96)*** 0.049 (2.29)** -0.013 (-0.87) 0.022 (1.43) 0.026 (1.01) -0.020 (-0.67) -0.032 (-1.51) -0.113 (-4.31)*** 0.444 (6.48)*** -0.010 (-2.49)** 0.006 (0.50) 0.054 (4.46)*** 0.069 (3.71)***

0.016 (0.3)
0.044 (0.89) 0.023 (1.12) -0.019 (-1.35) -0.004 (-0.27) -0.072 (-2.68)*** -0.048 (-1.56) -0.034 (-1.59) -0.130 (-4.52)*** 0.484 (7.78)*** 0.007 (1.92)* -0.071 (-6.58)*** 0.096 (8.58)*** -0.188 (-8.78)***

0.010
(1.26)

0.019
(3.39)*** 0.020 (3.49)***

0.061
(6.16)***

0.010
(0.38) -0.045 (-1.57) -0.052 (-2.53)** 0.009 (0.36) 0.496 (7.53)*** -0.013 (-3.14)*** -0.011 (-0.98) 0.039 (3.37)*** 0.044 (2.43)**

0.071
(6.33)***

0.051
(6.31)*** 0.047 (5.02)*** 0.065 (2.57)**

0.001
(0.68) 0.008 (1.86)*

0.013
(2.95)*** -0.003 (-0.38)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.14. Debt Ratio Regressions

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

55

Dependent variable
Indebt/ assets

Variable
wholesale bank= 1 bank's market share liabilities bank/system no. banks no. credit lines jcons

(1)

^liabilities/ assets (2)
0.107
(5.02)*** 0.000 (0.13) 0.266 (9.96)*** 0.035 (14.72)***

Inloans/ assets (3)
0.091
(4.10)*** 0.000 (-0.04) 0.347 (12.20)***

Inoverdrafts/ assets (4)
-0.007 (-0.34) -0.017 (-10.56)*** 0.084 (3.04)*** 0.020 (8.43)*** 0.084 (18.32)*** 0.230 (5.48)*** 5,860
1,263
0.00 0.24

0.019
(2.24)** 0.000 (0.12) 0.038 (3.62)*** 0.004 (4.26)***

0.041
(16.52)*** 0.024 (5.03)*** 0.876 (20.27)*** 5,860

0.001
(0.47)

0.010
(2.25)**

0.181
(11.22)*** 5,860
851

1.026
(24.77)*** 5,860

Number of observations LR chi2 Prob > chi2 Pseudo R2 Log likelihood

1,528 0.00 0.22
-2,660

1,514 0.00 0.21
-2,834

0.00 -0.17 2,810

-1,996

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: Equations are estimated using a Tobit specification; t-statistics are in parentheses. Source: Authors' calculations.

sign for the other columns is not necessarily an inconsistency because it could mean that the firm does not need to resort to more credit from the domestic banking system, substituting it for other forms of credit. However, if that is so, what is being measured is a demand effect, not availability of funds. There are several other examples along these lines. Additional complications arise with the use of these ratios as measures of access to credit. In accordance with the results of other studies, the debt/ asset ratio rises with the size of the firm (Inassets). However, reliance on bank creditÑb y any of its definitionsÑwoul d be expected to fall if SMEs were the firms most reliant on banking relationships for finance. Hence, there are composition effects. Although larger firms rely less on bank debt, they more than compensate for this with nonbank sources of finance. Thus, the ratio of bank liabilities to assets, which might appear promising as a

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.14.

(continued)

56

STREB AND OTHERS

Proportion of Unused Credit Lines in the Main Bank To avoid the problems that leverage presents as a measure of availability of credit, the analysis uses unused credit lines in the financial system. In principle, this measure indicates not how much credit a firm has used, but rather how much it has available through loan commitment contracts. Although the chapter began with unused credit lines at the system level, it is preferable to focus here on an alternative formulation in which the dependent variable is the portion of unused credit lines in the main bank,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

measure of access to credit because it includes not only loans drawn but also the total of loans granted, loses its appeal due to this composition effect. Nonetheless, if a figure were available not only of actual debt, but also of overall potential credit, the picture might be different. It is difficult to sort out demand and supply effects in table 2.14. The procedure of using debt-to-assets ratios to estimate access to credit is the one typically used in the applied literature. A series of instruments and econometric procedures are applied to get around the simultaneity problems. In this regard, Gallego and Loayza (2000) represent current best practice. We propose a different procedure to get around these problems, singling out unused credit lines, which can be related to Petersen and Rajan's (1994) idea of why discounts taken is informative in relation to credit constraints. Their firms had the possibility of using trade credit to finance their purchases, paying 20 days later and forsaking a 2 percent discount. If a firm decides to pay earlier to get a discount, it has a less expensive credit line available in the financial system. Alternatively, it could have lots of cash on hand so that its demand for external funds would be low. Petersen and Rajan's measure of the percentage of discounts taken on trade credits does not isolate demand and supply factors. What it achieves, instead, is an indicator that is not affected in the same direction by both factors. Better quality firms are offered more credit in the financial system, so for supply reasons these firms have to resort less to trade credit. But if a firm demanded more funds, it would eventually have to resort more to trade credit, so this cannot be confused with an outward supply shift, as could happen with the figures on indebtedness discussed above. Therefore, a dependent variable that displayed the same behavior as discounts taken could work. The measure of unused credit lines provides such a variable.

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

57

Unused credit line in main bank = (3 firm-specific information industry dummies loan and lender characteristics
credit history

measure of relationships

We used a standard two-part selection model and estimated it by both full information maximum likelihood and the Heckman two-step procedure.11 On the one hand, the likelihood function for the normal selection model is well known to be numerically difficult to handle, calling for two-stage methods. On the other hand, as in the present case, when the selection equation shares many explanatory variables with the regression equation, the identification of the second stage of Heckman's method may suffer from high multicollinearity between the explanatory variables in the regression equation and the additional regressor included to correct for the selection bias. We used both methods to handle this issue and compare estimation results. Although only the results of the Heckman
Initially, since the dependent variable is limited to the [0,1] closed interval, parameters were estimated as in Petersen and Rajan (1994) using a two-limit Tobit specification. This specification implicitly assumes, in terms of the general selectivity model, that the regression equation is also the selection equation. In this final version, by allowing these two equations to differ it is possible to distinguish between two effects. First, what determines whether a firm has an unused credit line or not (a 0/1 decision)? Second, of those firms with unused credit lines, what determines the amount of unused credit lines? The determinants turn out to be similar in both steps. Hence, at least qualitatively, results of the selectivity model replicate those of the initial two-limit Tobit specification.
11

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

that is, the bank where the largest share of liabilities is concentrated. The characteristics of the lender correspond to the main bank, except for the few cases in which two banks were tied in first place and both were included. This allows for a better estimation of the effect of the relationship variables because variables taken from the main bank are weaker in explaining available credit lines on a system-wide basis, and it makes no sense to define some of the relationship variables in terms of the whole system. The hypothesis is that unused credit lines indicate a firm that is not credit-constrained, while the opposite represents a firm that is. Therefore, the estimation should consider both zeros and the nonzero unused credit lines because both are informative. The following formulation captures the access to credit as the amount of unused credit lines granted by financial institutions to firms:

58

STREB AND OTHERS

two-step procedure are shown, the estimates from the two methods do not differ substantially (see Nawata and Nagase 1996 for more details on estimating selection models). Consistent standard errors for the selection model are computed as in Greene (1997, p. 981). Table 2.15 presents the results of the regression equation. With regard to firm characteristics, the availability of credit is positively and significantly associated with the amount of assets (Inassets), the sales/assets ratio (Insales/assets), andÑi n the last regression equationÑ the return on assets (roa). It is negatively and significantly associated with the debt-to-assets ratio (Indebt/assets). At the main bank, a worse credit situation (situation*} is negatively and significantly associated with less credit availability, and the same happens with bounced checks (Inbounced/liabilities). The relationship variables show that credit availability is positively and significantly associated with more debt concentration at the main bank (liabilities bank/system) and with the use of more accounts at the main bank (no. credit lines), and it is negatively and significantly associated with the number of banks (no. banks). Finally, more collateral (guarantees/liabilities), a wholesale main bank (wholesale bank), and a large main bank (bank's market share) are significantly related to greater credit availability. Some issues related to the interpretation of these results require further clarification. Petersen and Rajan (1994) divide firm characteristics into investment opportunities, which should reduce credit availability, and measures of cash flow, which should increase it. They admit that the dividing line is not always clear and the effect may be a priori ambiguous (for example, return over assets and sales over assets can be considered either as measures of internal cash flow or as measures of investment opportunities). However, these distinctions may not matter at all: if banks see that the firm has greater investment opportunities, they may be willing to lend more. Indeed, as Galindo, Schiantarelli, and Weiss (2002) show, an efficient financial system will try to channel funds to firms with more profitable investment opportunities. The proxies they use for investment opportunities are return over assets and sales over assets, precisely the variables that Petersen and Rajan consider might reduce credit availability. As far as these investment opportunities are common knowledge to both parts, because they can be inferred from the accounting statements of firms, this should increase the supply of funds coming from profit-maximizing financial institutions. Hence, it should be expected that return over assets and sales

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

59

Complete sample Variable
roa

(1)
0.003 (0.88) 0.063 (5.16)***
0.117

Filtered complete sample3 (2)
0.002 (0.48) 0.083 (11.05)***
0.147

Reduced sampleb

(3)
0.119

Reduced sample with extra firms' (4)
0.140

(1.14) 0.096 (7.78)***
0.100

(1.78)* 0.090 (8.81)***
0.105

Inassets Insales/assets Indebt/assets sector 0 sector 1 sector 3 sector 5 sector 6 sector 7 sector 8 situation 2 situation 3 situation 4 situation 5 situation 6 Inbounced/liabilities

(4.50)*** -0.096 (-1.80)* -0.045 (-0.94) -0.269 (-4.25)***
0.104

(5.50)*** -0.130 (-2.32)** 0.000 (-0.01) -0.222 (-6.44)***
0.145 (1.37)

(2.10)** -0.279 (-2.25)** -0.022 (-0.29) -0.222 (-3.71)*** 0.286 (1.71)* -0.042 (-0.29) 0.002 (0.03) -0.070
(-1 .42)

(2.73)*** -0.253 (-2.54)** -0.047 (-0.86) -0.219 (-4.47)*** 0.204 (1.43) -0.065 (-0.48) -0.036 (-0.62) -0.092 (-2.20)** -0.081 (-1.90)* -0.257 (-2.81)*** -0.617 (-4.61)*** -0.802 (-7.17)*** -0.778 (-5.39)*** -0.395 (-1.56) -0.038 (-2.26)** 0.475 (6.10)*** -0.003 (-0.08) (continued)

(1.03) -0.055 (-0.54) -0.026 (-0.59) -0.083 (-1.67)* -0.068
(-1 .42) -0.299

-0.085 (-0.80) -0.007 (-0.16) -0.092 (-2.89)*** -0.069 (-2.14)** -0.259 (-4.58)*** -0.461 (-5.61)*** -0.790 (-10.20)*** -0.730 (-7.05)*** -0.720 (-2.89)*** -0.047

(-5.17)*** -0.457 (-5.58)*** -0.774 (-7.20)*** -0.773 (-5.73)*** -0.606 (-2.85)*** -0.040 (-3.61)*** (8.25)***

-0.063 (-1.25) -0.286 (-2.78)*** -0.663 (-4.44)*** -0.845 (-6.97)*** -0.804 (-5.01)*** -0.388 (-1.51) -0.047 (-2.24)** (5.30)*** 0.007 (0.16)

(-4.29)*** uation (situation*} is negatively and significantly associated 0.427 0.429 with less credit 0.475 (8.27)*** -0.022 (-0.86) foreign bank= 1 -0.012 (-0.40)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.15. Basic Unused Credit Ratio Regressions

60

STREB AND OTHERS

Complete sample Variable
public bank= 1 wholesale bank= 1 bank's market share liabilities bank/system no. banks no. credit lines _cons

Filtered complete sample3

Reduced sampleb

Reduced sample with extra firrnsc

(D
0.069 (2.15)** 0.247 (5.38)*** 0.029 (5.84)***

(2)
0.061
(1.74)* 0.259 (5.18)*** 0.032 (9.09)***

(3)
0.081
(1.32) 0.293 (4.24)*** 0.032 (5.86)*** 0.238 (2.43)** -0.030 (-3.92)***

(4)
0.057 (1.11) 0.303 (4.75)*** 0.035 (7.31)***
0.187

0.214

0.197
(3.20)*** -0.031 (-5.45)***

-0.027 (-7.28)***

(2.27)** -0.033 (-4.70)***
0.115

0.132
(6.90)*** -1.394

0.129
(12.91)*** -1.582 (-16.50)*** 11,772 8,242 3,530

0.124
(8.00)*** -1.646 (-10.74)*** 4,426 3,078

(8.73)*** -1.523 (-11.82)*** 5,860 3,998
1,862

Number of observations Censored observations Uncensored observations Wald chi2 Prob > chi2 Log likelihood
a b

15,822 11,679

4,143 4,186 0.00
-8,313

1,348
380

876

481
0.00

0.00
-6,880

0.00
-2,629

-3,514

Excludes roa=0 and balance sheets with problems. Validation criteria 1/2/3/5/7 in appendix A in Streb and others (2002). c Validation criteria 1/2/3/5/7/13 in appendix A in Streb and others (2002). * Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: Equations are estimated using the Heckman two-step procedure; z-statistics are in parentheses. Source: Authors' calculations.

over assets unambiguously increase the availability of credit, which is exactly what happens in our estimates. A possible shortcoming of our line of reasoning is that good firms with a large amount of cash might be excluded. They simply might not need, and therefore would not apply for, line-of-credit contracts in the first place. Thus, they would not have unused credit lines. In this regard, large firms immediately come to mind because they are identified in the literature with high-quality firms. Incidentally, it has been shown that larger firms are

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.15.

(continued)

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

61

12

See appendk E.2 in Streb and others (2002).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

less likely to be credit-constrained (Schiantarelli 1996). However, table 2.15 shows that large firms have more unused credit lines available than small firms (large firms are also more likely to have unused credit lines). This happens despite the fact that large firms in Argentina might have fewer incentives to sign loan commitment contracts with domestic banks because access to cheaper credit abroad decreases firms' dependence on the domestic financial system. This seems to support the conjecture that firms that are not credit-constrained have unused bank credit available. It remains to be seen if this result is robust in other data sets. For instance, if the measure of unused credit lines is introduced into the interest rate regressions, it is significantly and negatively related to marginal interest rates. This is sensible if firms that are cut off from credit by their main bank have to "shop around" and other banks may be more expensive. However, using data from the Italian Centrale dei Rischi, which distinguishes between bank loans granted and bank loans drawn, D'Auria, Foglia, and MarulloReedtz (1999) find the opposite. The share of loans drawn is negatively related to the rate of interest, which means that firms with fewer unused credit lines pay lower interest rates. A difference between their study and this chapter is that we consider the interest rate in the marginal bank, whereas their study explains the interest rate at each bank and examines the effect of unused credit lines available in that same bank. However, a priori a firm might also have been expected to exhaust its cheaper sources of funds at a bank first, so it is hard to make sense of their results on the cost of credit. In summary, although the proportion of unused credit lines in total bank liabilities might be easier to replicate for a wider number of countries than the indirect measures drawn from trade credit by Petersen and Raj an (1994), it remains to be seen if unused bank credit turns out to be a sensible indicator of credit availability, as seems to be the case for the firms in our sample from Argentina. As in the case of the cost of credit, we re-estimated the regression equations with some additional explanatory variables.12 To control for possible nonlinearity in the number of banks, we introduced a quadratic term. This term is positive and significant (table 2.16). Simulating the composite effect of changing the number of banks gives the results shown in table 2.17. The

62

STREB AND OTHERS

Complete sample Variable
roa
Inassets Insales/assets Indebt/assets sector 0 sector 1 sector 3 sector 5 sector 6 sector 7 sector 8 situation 2 situation 3 situation 4 situation 5 situation 6 Inbounced/liabilities guarantees/liabilities foreign bank=l

(1)
0.140 (1.78)* 0.090 (8.81)*** 0.105 (2.73)*** -0.253 (-2.54)** -0.047 (-0.86) -0.219 (-4.46)*** 0.204 (1.43) -0.065 (-0.48) -0.036 (-0.62) -0.092 (-2.20)** -0.081
(-1 .90)*

Filtered complete sample3 (2)
0.134 (1.72)* 0.086 (8.45)*** 0.104 (2.73)*** -0.249 (-2.52)** -0.049 (-0.91) -0.220 (-4.50)*** 0.207 (1.46) -0.058 (-0.43) -0.022 (-0.38) -0.089 (-2.12)** -0.076 (-1.78)* -0.235 (-2.57)*** -0.578 (-4.32)*** -0.730 (-6.41)*** -0.743 (-5.16)*** -0.353 (-1.41) -0.024 (-1.45) 0.474 (6.12)*** -0.005 (-0.13)

Reduced sample13
(3)

Reduced sample with extra firms' (4)
0.141 (1.85)* 0.086 (8.64)*** 0.100 (2.67)*** -0.240 (-2.49)** -0.056 (-1.06) -0.223 (-4.66)*** 0.210 (1.51) -0.069 (-0.52) -0.017 (-0.3) -0.091 (-2.22)** -0.080 (-1.91)* -0.300 (-2.99)*** -0.599 (-4.32)*** -0.716 (-6.38)*** -0.772 (-5.39)*** -0.438 (-1.72)* -0.022 (-1.31) 0.477 (6.30)*** 0.003 (0.08)

0.142 (1.82)* 0.086 (8.43)*** 0.103 (2.68)*** -0.248 (-2.51)** -0.051 (-0.95) -0.221 (-4.51)*** 0.210 (1.48) -0.059 (-0.44) -0.021 (-0.36) -0.089 (-2.14)** -0.077 (-1.81)* -0.305 (-2.98)*** -0.618 (-4.36)*** -0.745 (-6.49)*** -0.788 (-5.37)*** -0.453
(-1 .74)*

-0.257 (-2.81)*** -0.617 (-4.61)*** -0.802 (-7.17)*** -0.778 (-5.39)*** -0.395 (-1.56) -0.038 (-2.26)** 0.475 (6.10)*** -0.003 (-0.08)

-0.022 (-1.29) 0.473 (6.09)*** -0.004 (-0.12)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.16. Unused Credit Ratio Regressions with Additional Variables

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

63

Complete sample Variable
public bank= 1 wholesale bank= 1 bank's market share liabilities bank/system no. banks no. credit lines __cons bounced other bank= 1 system situation (-1,0, 1) no. banks squared Number of observations Censored observations Uncensored observations Wald chi2 Prob > chi2 Log likelihood
a b

Filtered complete sample3
(2)
0.060 (1.16) 0.307 (4.83)*** 0.034 (7.17)*** 0.175 (2.13)** -0.031 (-4.42)*** 0.114 (8.66)*** -1.470 (-11.43)*** -0.192 (-3.45)***

Reduced sampleb
(3)
0.061 (1.19) 0.307 (4.81)*** 0.034 (7.14)*** 0.177 (2.15)** -0.031 (_441)*** 0.114 (8.67)*** -1.471 (-11.41)*** -0.193 (-3.44)*** -0.127 (-1.72)*

Reduced sample with extra firms'
(4)
0.065 (1.29) 0.301 (4.84)*** 0.032 (6.97)*** 0.055 (0.61) -0.071 (-4.47)*** 0.112 (8.64)*** -1.258 (-9.07)*** -0.183 (-3.34)*** -0.130 (-1.80)* 0.002 (2.85)*** 5,860 3,998 1,862
496

(1)
0.057 (1.11) 0.303 (4.75)*** 0.035 (7.31)*** 0.187 (2.27)** -0.033 (-4.70)*** 0.115 (8.73)*** -1.523 (-11.82)***

— — — — — —
5,860 3,998 1,862
481

— — — —
5,860 3,998 1,862
490

— —
5,860 3,998 1,862
488

0.00 -3,514

0.00 -3,497

0.00 -3,495

0.00 -3,457

Excludes roa=0 and balance sheets with problems. Validation criteria 1/2/3/5/7 in appendix A in Streb and others (2002). c Validation criteria 1/2/3/5/7/13 in appendix A in Streb and others (2002). * Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: Equations are estimated using the Heckman two-step procedure; z-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.16. (continued)

64

STREB AND OTHERS

Percentile

Number of banks

Composite effect -0.069 -0.069 -0.069 -0.134 -0.195 -0.354 -0.440 -0.510 -0.615

1
5

10 25 50 75 90 95 99

1 1 1
2 3 6 8

10 15

Source: Authors' calculations.

negative effect of banks on available credit increases as the number of banks rises, reaching a minimum when no. banks =18. After that, the value starts rising. However, 99 percent of the firms do business with 15 or fewer banks. The dummy for bounced checks from other banks is significant and negative, but it makes the bounced checks at the main bank lose significance (table 2.16). The variable on system situation is significant, indicating that if the median rating in the system is worse, then available credit at the main bank is smaller, and if it is rated better, available credit is larger. This suggests that the credit situations reported by the Central de Deudores have an impact on credit constraints and the public credit register helps to ease problems of asymmetric information. In summary, the two-step procedure is interpreted as providing evidence that firms that are not credit-constrained have unused credit lines available in the financial system, and that a larger proportion of unused credit available is an indication of easier access to credit. More precisely, when the credit limits in loan commitment contracts are exhausted, the firm faces an intensive credit constraint. Economic Significance In this section, we estimate the economic significance of the parameters that are statistically significant. We use regression 3 in table 2.16 to estimate the impact on the access to credit, and regression 3 in table 2.13 to estimate the impact on the cost of credit.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.17. Effect of the Number of Banks on Credit Availability

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

65

Table 2.18. Economic Significance of Variables Affecting Credit Availability
(Mean plus two Mean times Variable parameter 0.0059 0.7276 0.0736 -0.1028 -0.3050 -0.6180 -0.7450 -0.7880 -0.4530 0.0063 0.3070 0.1483 0.1158 -0.1310 0.2556 -0.1927 0.0318 -0.0529 0.3969 0.2114 -0.3275 0.5492 0.0914 standard deviations) times parameter 0.0966 1.0149 0.1651 -0.1823 Effect of treatment (percent)
9

roa
Inassets Insales/assets Indebt/assets situation 2 situation 3 situation 4 situation 5 situation 6 guaran tees/liabilities wholesale bank- 1 bank's market share liabilities bank/system no. banks no. credit lines bounced other bank= 1 system situation (-1,0, 1)

29
9

-8
-31 -62 -75 -79 -45

9

31 25 10
-20

29
-19

-8

Note: Values are calculated using regression equation 3 in table 2.16. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 2.18 presents the economic significance of the parameters estimated using the quantity regression, where the first column is the estimated parameter times the mean of the variable, and the second column is derived by multiplying the estimated variable times the mean plus two standard deviations. For dummy variables, the table measures a switch from 0 to 1. The dependent variable varies between zero and one. The table shows that the main factor affecting the amount of credit available is the credit situation of the firm. A firm's access to credit decreases sharply when the credit situation worsens, except for situation 6 (a technical category that is not worse than situation 4 or 5). Close relationships are quite important for access to credit: there is a positive effect from doing more business with the main bank and having a larger number of credit lines there, and a negative one from working with many banks, which signals a more distant relationship. The no. banks variable indeed reinforces the effect of the liabilities bank/system variable. It is also significant that large firms and firms that

66

STREB AND OTHERS

Table 2.19. Economic Significance of Variables Affecting the Cost of Credit
(Mean plus two Mean times Variable parameter
0

standard deviations) times parameter
6

Effect of treatment (basis points)
6
-39 -15

roa
Inassets Insales/assets Indebt/assets situation 2 situation 3 situation 4 situation 5 situation 6 Inbounced/liabilities unwarranted/overdrafts foreign bank=1 wholesale bank=1 bank's market share liabilities bank/system no. banks no. credit lines bounced other bank= 1 system situation (-1,0, 1)

-103
-13

-142
-29

19 50
0

32

13 50
0

38 16 65
3

38 16 65 21
130

19 36 16
-100

93 16

-100
-37 -19

-100
-60 101 -66

-63 -41

44
-32

57
-34

30
0

30 -3 -3

Note: Values are calculated using regression 3 in table 2.13. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

operate with wholesale banks and large banks (which may also provide an indication that these are prime firms) have more credit available. Table 2.19 shows the economic significance of the interest rate regression parameters, measuring the price effects in basis points. Although lender characteristics have not been emphasized in this discussion, the main explanatory factor is whether a firm works with a wholesale bank, which is related to a reduction of 100 basis points in the interest rate. Many wholesale banks tend to work with affiliates of multinational corporations, so this may capture characteristics of the clients they choose. Working with a large bank is also important in obtaining lower rates. Hence, becoming a client of the right type of bank seems relevant, and there appears to be some matching between types of firms and types of banks. In relation to a firm's credit situa-

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

67

Implications We constructed a cross section of data for October 2000 in order to study the variation of both the cost of credit and the access of firms to credit. Of particular interest is how banks use different pieces of private and public information to screen firms and overcome information asymmetries in the credit market. Some private information is transferable, such as balance sheet data. Private information generated in relationships is not. We also considered public information available from the Central de Deudores. The marginal cost of credit is measured by a very specific indicator, the interest rate charged on marginal (most expensive) bank overdrafts. To measure firms' access to credit, the first intention was to use the percentage of authorized overdrafts effectively drawn. The overdraft authorizations are a loan commitment contract or credit line. However, we came up with a more comprehensive measure: unused credit lines over total financial system liabilities. This is the other side of trade credit if, as Petersen and Rajan (1994) conjecture, firms with credit in the financial system do not resort to trade credit and pay early in order to get a discount. It is also related to the loan commitment contracts in Melnik and Plaut (1986) in the sense that firms that have exhausted their bank commitments have to look for more expensive credit elsewhere.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

tion, after passing situation 2, there is no strong punishment in terms of interest rates (situation 6 is a special technical category for firms in arrears with failed banks). Bounced checks lead to appreciably more expensive credit. Relationship variables are again important, despite controlling for credit history (Greene [1992] expects credit history to wipe out the significance of other types of information in credit scoring). The information in tables 2.17 and 2.18 suggests the existence of equilibrium credit constraints in Argentina's formal credit market. The parameters for a firm's situation variable suggest that the firm faces a quantity effect that is stronger than the price effect as the credit situation deteriorates. The relationship variables appreciably affect both quantities and prices. A close relationship is related to more access to credit and lower marginal interest rates.

68

STREB AND OTHERS

The findings indicate that the cost of credit is smaller for a firm with a close relationship to the marginal bank, where a close relationship is linked to a large share of credit and a large number of accounts in the marginal bank and to a small number of total banks. Firms with large assets, a high sales/assets ratio, and a low debt/assets ratio pay a lower interest rate at the margin. A good credit history (no debt arrears and no bounced checks) and collateral also reduce the marginal interest rate. Although credit history matters, as Greene (1992) emphasizes, it is not the only factor. It does not wipe out the significance of the relationship variables, which are extremely important. Using the proportion of unused credit lines at the main bank as a measure of firms that are not credit-constrained, the analysis shows that a good credit situation increases the availability of credit, and that relationships are also very important in increasing the access to credit. The proxies for relationships might signal that banks have private information, indicating the firm to be a good credit risk rather than a lemon. This chapter's measure of unused credit lines is less ambiguous than traditional measures such as leverage, which may indicate financial distress rather than availability of credit. Large assets, a high return to assets, a high sales/assets ratio, a low debt/assets ratio, a good credit history, and collateral lead to higher credit availability. The chapter does not consider the macroeconomic determinants of credit constraints. Economy-wide credit constraints that reflect doubts about the government's prospects are reflected in the baseline interest rate charged to firms. The baseline interest rate has a premium related to the interest rate spread between Argentine government bonds and U.S. government bonds. This spread, or country risk, was on average above 750 annual basis points in October 2000. An observation from credit constraint theory is that a high probability of default gives financial institutions an incentive to eliminate credit lines to the borrower. Since November 2000, the international capital market has been closed due to international investors' belief that Argentina would default on its debt. Since then, the spread has fluctuated widely, reaching 3,500 basis points in November 2001 and even higher levels since then. This impacts firms, which are mostly cut off from new credit. Hence, it is worthwhile to explore the economy-wide determinants of credit constraints faced by firms in addition to the individual determinants stressed in this study. The regressions do not capture this result because the estimation

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

BANK RELATIONSHIPS AND FIRMS IN ARGENTINA

69

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

process is a cross section for October 2000. Further study on economy-wide constraints might compare the October data with data for January 2001 after international markets were closed. This chapter may be useful for studying the implications of the law of habeas data, which went into effect recently in Argentina. This law puts substantial costs on the acquisition of information by lenders, limiting credit history to five years and limiting access to the database. In particular, the study shows that a good median credit situation in the Central de Deudores eases credit restrictions. A policy that restricts disclosure may end up hurting good firms.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

This page intentionally left blank

Determinants and Consequences of Financial Constraints Facing Firms in Argentina
Jose M. Fanelli, Ricardo N. Bebczuk, and Juan J. Pradelli
In the 1990s, Argentina implemented an ambitious structural reform program that brought about profound changes in the economy, including the monetary and exchange rate regimes and the banking sector. Throughout the decade, the country displayed a unique combination of characteristics: ¥ The exchange rate/monetary regime was a currency board. ¥ There were no obstacles to capital flows and tighter prudential regulations were introduced. ¥ Private portfolios and bank balance sheets were highly dollarized. The effects of the reforms on the financial side of the economy were encouraging in the first half of the 1990s. In 1991-95, the economy grew quickly and there was a marked increase in the level of financial deepening. But in spite of this, the macroeconomic environment remained volatile and there were important credit crunch episodes when the so-called Tequila effect and Russian and Brazilian crises hit the economy in 1995 and 1998, respectively. In the second half of 1998, the economy entered a lengthy period of persistent recession, which ultimately resulted in a financial crisis at the end of 2001. In lanuary 2002, the currency board was formally abandoned and the process of financial reform was reversed. Capital controls were reintroduced and prudential regulations were softened.

Jose M. Fanelli and Juan J. Pradelli are researchers at the Centre de Estudios de Estado y Sociedad, and Ricardo N. Bebczuk is the director of the Department of Economics at the Universidad Nacional de la Plata.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 3

72

FANELLI, BEBCZUK, AND PRADELLI

Evidence suggests that in spite of market-friendly financial reforms and a higher level of financial development, the financial constraints facing Argentine firms remained tight throughout the 1990s. Market segmentation was important, that is, different kinds of firms enjoyed different access to capital. The causes and consequences of the financial constraints that firms face are better understood today than in the past (for a survey, see Hubbard [1998] and Schiantarelli [1996]). In the case of Argentina, the few existing studies point to the relevance of financial constraints (Fanelli and Damill 1988; Fanelli and Keifman 2002; Bebczuk 2000; Schmuckler and Vesperoni 2001). This research suggests that: (i) credit markets are segmented; (ii) firms are dependent on own funds; (iii) credit Granger causes the activity level, and country risk matters for real decisions; and (iv) the volatility of the environment and external shocks affect the capital structure of firms. However, there is limited knowledge of the precise way in which financial constraints affect the financial structure and investment decisions of firms in Argentina. Taking into account the well-known influence financial factors have on both business cycles and long-run growth, the little effort devoted to research in this area represents a restriction on policy design and policymaking. This chapter explores the determinants and consequences of financial constraints on firms in Argentina, focusing on the years that preceded the most recent crisis. While most of the literature in this field relies on listed firms, this chapter analyzes an additional set of 500 large firms (although the number of usable observations for econometric purposes is lower than that figure). Since the number of listed firms in Argentina is small and this group may exhibit different behavior than other firms, studying a new set of firms is bound to shed further light on the subject. The caveat is that, as expected, the set of available variables and the time frame differ across the databases, which precludes running the same econometric exercises and strictly comparing the results from each sample. Although this imposes a cost in terms of the desirable structure of the analysis, the additional insights derived from the new sample outweigh the disadvantages. Similarly, by covering a wide range of issuesÑstylize d facts, investment, and capital structure decisionsÑthi s chapter attempts to fill some gaps in the empirical literature on Argentina.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCIAL CONSTRAINTS IN ARGENTINA

73

¥ Present the stylized facts characterizing the capital structure of Argentine firms based on the analysis of balance sheet items ¥ Assess the relevance of financial constraints on investment at the firm level ¥ Obtain quantitative evidence of the financial structure and choices of firms ¥ Investigate the effects of financial imperfections on different types of firms according to different sample splits ¥ Evaluate the effects of macroeconomic volatility and shocks on the financial structure of firms. The Convertibility Regime and Firms' Financial Decisions: Stylized Facts The convertibility law instituted a currency board regime in 1991. The peso was pegged to the U.S. dollar at a one-to-one parity, and it was established that the central bank would hold an amount of international reserves at least equal to the amount of currency in circulation. The most remarkable result of the convertibility regime was the reduction in the inflation rate. Under convertibility, Argentina ceased to be a high-inflation country and the rate of inflation settled to a level below international standards. Another important fact is that in the 1990s the economy recorded a substantial average annual growth rate of 4.1 percent. However, this average growth rate reflects two periods separated by the Tequila effect in 1995. In the first years of the reform, the increase in gross domestic product (GDP) was strong. But after 1995, the evolution of the economy showed several disappointing features: the activity level followed a stop-and-go pattern, the average increase in GDP was low, and the unemployment rate soared. Likewise, the fiscal deficit and the stock of external debt experienced an upward trend. In such a context, Argentina faced increasing difficulties in meeting its external obligations. Finally, at the end of 2000, the country was forced to resort to the International Monetary Fund (IMF). The financial agreement (the blindaje) was reached in December 2000. But the effects have not been what were expected. The situation continued to worsen in 2001 and culminated in a financial crisis that obliged the government to abandon the currency board.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Specifically, the chapter focuses on micro data to do the following:

74

FANELLI, BEBCZUK, AND PRADELLI

High inflation, frequent maxi-devaluation, and uncertainty were the rule rather than the exception during the so-called lost decade following the debt crisis in 1982. In this context, the domestic demand for financial assets fell systematically in the 1980s. As a result, in 1991 the degree of financial deepening of the economy was low and total deposits amounted to around 5 percent of GDP. The changes induced by the convertibility plan in this financial scenario, a legacy of the lost decade, were as significant as those in price dynamics. The stabilization of the exchange rate and disinflation favored recovery in the demand for domestic assets. This recovery also benefited from substantial improvement in capital market conditions for emerging countries in the early 1990s. Figure 3.1 shows the continuous improvement in financial deepening in 1991-94 as measured by the increase in the demand for deposits and total credit. These developments not only loosened the tight credit rationing of the 1980s, but also opened up new opportunities for firms to innovate in the form of financing capital projects. The process of increasing financial deepening under convertibility, however, has certain features that are important for firms' financial decisions. In the first place, there has been an increasing dollarization of portfolios. Figure 3.2 shows the evolution of the stock of dollar-denominated credit and deposits in the domestic financial system as a proportion of the
Figure 3.1. Evolution of Deposits and Credit in Argentina, 1991-2000 (Millions of pesos)

Source: Banco Central de la Republica Argentina.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Convertibility, Dollarization, and Country Risk Premium

FINANCIAL CONSTRAINTS IN ARGENTINA

75

Source: Banco Central de la Republica Argentina.

total stock of credit and deposits. The proportion of dollar-denominated instruments grew continuously. At the end of 2000, more than 60 percent of credit and deposits was denominated in dollars and this tendency increased in 2001. However, the proportion of dollarized credit is greater than the proportion of deposits. This implies that, in fact, banks are hedged against a devaluation of the currency. A second feature is that the evolution of the demand for domestic assets proved to be highly dependent on external conditions. Figure 3.1 shows that external shocks rapidly impacted the demand for domestic assets and the credit supply. The Mexican crisis interrupted the upward trend in deposits and credit. After the recovery in 1996-97, the Russian crisis had the same effect. Note that the speed of the recovery in deposits and credit differed after the Tequila and Russian shocks. Although the recovery was very rapid in the former case, credit and deposits were more sluggish in the latter. In fact, the stock of deposits began to fall in 2001. External shocks, both positive and negative, also influenced the cost of domestic credit. In this regard, the main link between external and domestic credit markets is the country risk premium. Changes in the conditions in capital markets in emerging countries and/or in the domestic macroeconomic scenario are reflected immediately in changes in the country risk premium. The volatility of both domestic and external conditions was echoed in the evolution of the country risk. Through its influence on the cost of credit, this volatility increased the variance of aggregate demand. Figure 3.3 compares

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.2. Dollarization of Deposits and Credit in Argentina, 1991-2000 (Percentage of total)

76

FANELLI, BEBCZUK, AND PRADELLI

(Percent)

Source: Banco Central de la Republica Argentina.

the evolution of the country risk premiumÑa s measured by the Emerging Market Bond Index (EMBI) spreadÑwit h the economy's quarterly rate of growth. Both variables show high volatility and there is a marked and negative association between changes in the country risk premium and changes in the quarterly growth rate of GDP. The third feature of increased financial deepening is the close association between the supply of credit and the activity level. Indeed, given that Argentina's capital markets are far from perfect, it seems plausible that changes in the availability of credit do matter for the level of activity. Using an error correction model, Fanelli and Keifman (2002) find results that are consistent with the hypotheses of a relevant positive association between credit and output in the short run and of a negative correlation between the country risk premium and the evolution of the macro economy. The features analyzed suggest that, under convertibility and free movement of capital, there is a close association among capital flows, the generation of credit, and the activity level. This is an important potential source of macroeconomic and financial uncertainty, as international capital flows into emerging countries are far from stable. It must be taken into account, nonetheless, that the economic authorities' degrees of freedom under convertibility are not equal to zero. In fact, it seems that the depth of the recession since 1998 has not been independent of

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.3.

Country Risk Premium and Growth Rate in Argentina, 1991-2000

FINANCIAL CONSTRAINTS IN ARGENTINA

77

Capital Structure of Argentine Firms Figure 3.4 shows the evolution of the capital structure of Argentine firms listed on the stock exchange in the pre and post-convertibility period. In the years preceding convertibility, it is difficult to identify a definite pattern in the relationship between net worth, total assets, and liabilities. In the 1990s, by contrast, there is a clear tendency for the level of leverage to increase. Between 1992:1 and 2000:3, net worth increased by 22 percent in real terms and total debt grew by 221 percent. It would be difficult to explain such growth in indebtedness after the implementation of convertibility without referring to macroeconomic factors. In the early 1990s, following two hyperinflationary episodes in 1989 and 1990, foreign capital markets were closed to most Argentine firms and the

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

some policy actions on the financial and fiscal sides. There has been a persistent tendency for the rate of growth of credit to lag behind the rate of growth of deposits since 1995. In 1999, the line representing deposits crosses the credit line in figure 3.1. The tightening in the central bank's prudential and liquidity regulations in the second part of the 1990s is closely associated with this result. However, the credit squeeze in the private sector since 1998 has been stronger than figure 3.1 suggests. The figure shows that the aggregate stock of credit has stagnated since 1998, but the aggregate conceals the fall in the stock of private credit that was offset by an increase in the amount of public sector credit demand. The increase in the fiscal deficit (which was associated with the political cycle) increased the public sector's borrowing needs and, as a consequence, the government crowded out the private sector. The private/ public credit ratio fell from 7.7 when the Russian crisis hit the economy in 1998 to 4.4 at the end of 2000. The tightening of credit conditions for the private sector was undoubtedly a major factor that deepened the recession. The funds available for financing the private sector suffered simultaneously from the pressure exerted by the fall in capital inflows, the tightening in prudential regulations, and mounting public demand for credit. In such a context, it is not surprising that demand plummeted for investment and consumer durablesÑ a major factor in the stagnation of aggregate demand at the end of the 1990s.

78

FANELLI, BEBCZUK, AND PRADELLI

(Index, 1992.1 = 100) a. Smaller sample, 1986-2000

b. Larger sample, 1992-2000

Note: In order to cover as wide a time span as possible and preserve the homogeneity of the panels, the number of firms in panel a is lower than in panel b. Panel a refers to the SE sample of 45 listed firms; panel b refers to the ENGE sample of 308 large firms. Source: Authors' estimates based on data from the Buenos Aires Stock Exchange.

domestic credit/GDP ratio was extremely low. Under such circumstances, it seems logical to assume that firms were in disequilibrium and the observed leverage ratio did not reflect long-run equilibrium values. As stability consolidated in the 1990s and capital inflows recovered, firms sought to reduce the gap between existing and preferred levels of leverage. In order to highlight the relevance of macroeconomic factors for microeconomic decisions,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.4. Assets, Liabilities, and Net Worth for the Smaller and Larger Samples, Argentina, 1986-2000

FINANCIAL CONSTRAINTS IN ARGENTINA

79

Source: Authors' estimates based on data from the Buenos Aires Stock Exchange.

figure 3.5 plots the observed values of the leverage ratio (debt over total assets) against the medium-run trend (using the Hodrick-Prescott filter).1 Substantial deviations from the trend are associated with major macroeconomic shocks. The two largest downward deviations coincide with the hyperinflationary period and the Tequila effect. In both cases, however, the leverage ratio recovered rapidly as the effects of the shock faded, particularly in the years of booming capital inflows (1992-94). These fluctuations in leverage suggest that negative shocks tend to worsen credit conditions rapidly, driving firms' leverage to suboptimal levels. They also suggest that firms may resort to liquidating assets in order to smooth the effects of shortrun credit crunches. By contrast, firms take advantage of tranquil periods in credit markets to correct deviations from long-run equilibrium. The dynamics of short and long-term debt held by firms throughout the cycle also suggest that credit conditions can quickly react to changes in investor sentiment. Figure 3.6 shows the evolution of long and short-term debt. Comparison of figures 3.5 and 3.6 shows that the behavior of the longterm liabilities/asset ratio tends to mimic the behavior of the gearing ratio, but fluctuations in the latter ratio are smoother. This implies that agents

The reference to the trend is only illustrative. It follows from the arguments in the text that, after hyperinflation and stabilization, the Argentine economy is in a period of adjustment.

1

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.5. Leverage Ratio, Argentina, 1986-2000

80

FANELLI, BEBCZUK, AND PRADELLI

(Percent)

Source: Authors' estimates based on data from the Buenos Aires Stock Exchange.

tend to resort to short-term debt when they face increasing costs in the markets for long-term debt or rationing. Under convertibility, the stock of short-term debt held by firms is higher than that of long-term debt during almost all of the sample period, and the proportion is comparable to that in other developing countries (see Booth and others 2000). The telling presence of instruments with short maturity implies that Argentina's debtor/creditor relationships are characterized by staged finance (Stulz 2000). The behavior of the long-term debt/total assets ratio reproduces the general shape of the leverage ratio, but is more volatile, reflecting the fact that the short-term debt ratio is more stable. The coefficient of variation of the long-term ratio is more than twice that of short-term debt. Figure 3.7 shows the co-movement of the proportions of short and long-term debt in total debt. In figure 3.7, the ratio between long and short-term liabilities tends to move procyclically, so that negative shocks not only reduce leverage, but also tend to shorten the duration of debt. If average maturity falls, to maintain their liquidity position (that is, to keep the ratio of liquid assets to short-term debt constant), firms should increase their demand for liquid assets during downturns. However, the available evidence casts doubt on this hypothesis. It seems that in the case of Argentina, firms' liquidity position tends to worsen in periods of macroeconomic instability. To illustrate this point, figure 3.8 compares the evolution of liquidity (liquid assets/short-term liabilities) with the evolution of country risk, which is interpreted as a proxy

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.6.

Short and Long-Term Debt Ratios, Argentina, 1992-2000

FINANCIAL CONSTRAINTS IN ARGENTINA

81

Source: Authors' estimates based on data from the Buenos Aires Stock Exchange.

for macroeconomic disequilibrium. The behavior of these two series is compatible with the conjecture that, under convertibility, liquidity constraints move countercyclically.2 In bad times, firms are forced to rely on internal sources of liquidity. The behavior of the debt components throughout the cycle raises interesting analytical questions. An important one is why this increase in the proportion of short-term liabilities is observed after a shock. The present hypothesis is that negative shocks reduce firms' net worth, increasing the probability of financial distress. Under such circumstances, creditors react by shifting their demand toward assets with short-term maturity to better monitor the behavior of debtors because the liquidity premium rises in uncertain environments. But if it is assumed that the duration of assets is somewhat constant throughout the cycle, when the shortening in the term to maturity of debt occurs, the financial position of firms further deteriorates and default becomes more probable. Creditors perceive this increase as an upward movement in the costs of financial distress (if these costs are calculated as the probability of default multiplied by its cost). Under these circumstances, a
Liquidity conditions seem to respond quickly to changes in foreign capital markets. In figure 3.8, for example, the worsening in liquidity conditions occurs well before the Tequila effect hits the economy. In fact, there is a clear worsening in the liquidity indicator after the tightening of monetary policy in the United States in the first quarter of 1994. This suggests that the Tequila effect and the change of orientation in monetary policy in the United States were not independent phenomena.
2

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.7. Proportion of Short and Long-Term Debt, Argentina, 1992-2000 (Percentage of total debt)

82

FANELLI, BEBCZUK, AND PRADELLI

(Percent)

(Percentage points)

Source: Authors' estimates based on data from the Buenos Aires Stock Exchange and the Banco Central de la Republics Argentina.

logical result is that creditors will try to shorten maturity to better monitor and discipline debtors. In sum, if this reasoning holds, there are endogenous factors that tend to reduce maturity and increase financial duress during recessionary periods. This hypothesis of maturity shortening as a disciplinary device is fully consistent with the hypothesis of staged finance as an antimoral hazard mechanism in contexts where institutional underdevelopment impedes the precise definition of property rights. The increase in the proportion of dollar-denominated liabilities under convertibility that is observed at the aggregate level is also clear in the firm data. The proportion of dollar-denominated debt in total debt rose from 52 percent in 1992 to 77 percent in 2000 (figure 3.9). The growth rate of dollar-denominated debt was even higher than the rate corresponding to total liabilities; between 1992:1 and 2000:3, the total amount of dollardenominated debt almost quadrupled. There is a close link in the evolution of dollar-denominated and longterm liabilities, which suggests that for Argentine firms, domestic dollardenominated credit and external capital markets are critical sources of long-run funds. Under the assumption that firms prefer to match the duration of their assets and liabilities, dollarization and capital inflows must have had a positive influence on capital formation. However, it is also true that as the proportion of dollar-denominated liabilities increases, so does exposure to unanticipated changes in the real exchange rate. Hence, there is a

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.8.

Liquidity and Country Risk, Argentina, 1992-2000

FINANCIAL CONSTRAINTS IN ARGENTINA

83

Source: Authors' estimates based on data from the Buenos Aires Stock Exchange.

trade-off between the benefits of matching the duration of the two sides of the balance sheet and the increased currency risk taken because of higher mismatching in the currency denomination of assets and liabilities. The existence of currency risk implies an inverse relationship between the expectations of a change in the real exchange rate and the degree of currency mismatch between assets and liabilities. If this argument is valid, every time the firms seek to hedge against devaluation, their balance sheets will indicate a fall in participation of long-run liabilities. Figure 3.10 presents evidence that this conjecture merits investigation in the Argentine case. The figure plots the ratio between short and long-term debt against the ratio between dollardenominated liabilities and assets. Figure 3.10 shows a significant decline in dollar-denominated liabilities and a rise in the importance of short-term debt every time an important shock hits the economy. The gap widens between the two variables during hyperinflation (1989/90) and the Tequila effect (1995/96) and after the Russian and Brazilian crises (1998/99). This means that when the macroeconomic setting worsens, there is a concurrent increase in the demand for foreign exchange and higher pressure on markets for short-term financing. It also means that economic downturns create pressure on both foreign exchange and domestic financial markets. When the exogenous macroeconomic shock is strong enough and the regulatory framework is weak, this combination of events can trigger the so-called twin crises that have in fact occurred in Argentina. Domestic markets for short-term credit are unable to

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.9. Dollarized Debt, Argentina, 1992-2000 (Percentage of total debt)

84

FANELLI, BEBCZUK, AND PRADELLI

Argentina, 1986-2000 (Percent)

(Percent)

Source: Authors' estimates based on data from the Buenos Aires Stock Exchange.

make up for the fall in dollar-denominated and long-run loans and firms face increasing difficulties in meeting their short-run obligations. Exogenous macroeconomic shocks play a significant role in Argentina because they are both sizable and frequent. Agents must make financing decisions in a highly uncertain environment in which substantial wealth losses can result from errors in expectations. When a fiscal or external shock leads to an unexpected currency devaluation (and, eventually, to a change of exchange rate regime), those agents facing severe losses typically feel as if the authorities had violated their property rights. This fear of losses and the need to protect property rights from moral hazard underlie the tendency to dollarize that is observed under increasing macroeconomic instability.

Capital Structure, Investment, and Dollarization The stylized facts suggest the following points: 1. Prima facie financial constraints matter in Argentina. 2. Dollarization is a structural feature of financial intermediation. 3. Macroeconomic factors have a bearing on the tightness of financial constraints.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 3.10.

Short-Term and Dollar-Denominated Debt,

FINANCIAL CONSTRAINTS IN ARGENTINA

85

¥ Affiliation with a group ¥ Ability to issue obligadones negociables (equity) or Eurobonds, that is, to participate in bond markets ¥ Foreign ownership ¥ Quotes in foreign markets, that is, issuance of American Depositary Receipts (ADRs) ¥ Privatized enterprise. Dollarization of financial instruments (point 2) characterizes many emerging economies. But the literature on capital structure emphasizes the decision on the level of leverage and the proportion of long-term debt. Not much research has been done on the factors that determine the proportion of dollar-denominated debt a firm decides to hold. In this regard, attention has focused on whether financial reform affects capital structure and investment decisions. But dollarization is a structural change whose consequences can be even stronger than those induced by financial reform. Consequently, we consider whether the factors that the literature identifies as relevant in determining the capital structure and the mix of short and long-term debt are also relevant to the choice of the proportion of dollar-denominated liabilities

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Point 1 means that circumstances do not correspond with a Modigliani-Miller (1958) world of perfect capital markets and that it is necessary to approach the Argentine case within a theoretical framework where information asymmetries, contract enforceability, and other frictions matter. In such a framework, lenders may be willing to provide additional financing for investment only at an increasing interest premium (Bernanke and Gertler 1990), credit rationing may be observed (Stiglitz and Weiss 1981), and changes in investor sentiment may trigger sudden flight-to-quality episodes with significant consequences for the level of activity and the country's macroeconomic stability (Bernanke, Gertler, and Gilchrist 1994). A key characteristic of imperfect capital markets is that they are segmented. This means that different firms face financial constraints of varying intensity and, therefore, the partition of the sample of firms may uncover important differences regarding the relevance of financial constraints. The sample is split according to the likelihood that the firm will suffer from incentive and information problems. The dummy variables used to partition the sample are based on the following characteristics of the firm:

86

FANELLI, BEBCZUK, AND PRADELLI

Investment and Financial Constraints Financial constraints are likely to influence the investment process. Hence, one way to test the hypothesis that market imperfections are relevant is to go beyond the traditional approach based on Tobin's Q and include, in addition to variables related to profit maximization, variables representing financial constraints in the investment equation (Fazzari, Hubbard, and Petersen 1988; Hubbard 1998; Schiantarelli 1996). If markets are imperfect, not only does profitability matter, but so do liquidity and the cost of external financing. Accordingly, a first step is to estimate the following regression model using a panel of firms that are quoted on the stock market:

where I is investment as a ratio of the capital stock, q is Tobin's Q (market value of the firm over its book value), c is cash flow as a ratio of the capital

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

in the capital structure. We also examine whether debt dollarization has a bearing on investment. Point 3 is clearly stated in the literature. It is well known that the severity of financial constraints is likely to vary with overall macroeconomic conditions and the stance of economic policy because these factors influence net worth. In this way, analysts stress that monetary policy works not only through the traditional cost-of-capital channel, but also through collateralizable net worth. However, little is known about the way in which changes in macroeconomic conditions affect financial constraints in the context of a currency board where, by definition, there is no room for monetary policy. When expectations change, as reflected in a variation of the country risk premium, for example, there will be variations in the rate at which returns are discounted and more pessimistic forecasts may result. Although this is true of all economies, volatility is higher in emerging markets. For example, shocks induce sizable changes in both the level of leverage and the composition of debt. These questions remain underresearched in the literature because of the lack of data on emerging markets. To take these effects into account, we introduce country risk, an index of financial deepening (the aggregate private banking credit/GDP ratio), and the proportion of dollar debt in total debt as explanatory variables in the regressions.

FINANCIAL CONSTRAINTS IN ARGENTINA

87

3 Following a reviewer's suggestions, dynamic equations are specified with lagged dependent variables, and estimated using the program DPD OX version 3 2001. The GMM results present the one-step estimations and the Sargan statistic from the two-step estimations. GMM estimates are based on instrumenting the differenced equation with the lagged level value of the endogenous and predetermined variables and with the first-differenced values of the exogenous variables. The GMM technique is not appropriate when the number of periods is large relative to the number of firms. In the stock exchange panel, there are 34 quarterly observations for 45 firms. Given the significant number of instruments derived from the long time dimension, the Sargan tests do not reject the null, probably reflecting the overfitting problem. FEW results are preferable for the analysis.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

stock, and d is the debt/capital ratio (leverage). The subscript i is an index for firms and t for time. In the absence of financial restrictions and agency problems, investment depends exclusively on q, provided it adequately captures the fundamentals. Hence, if coefficients y2 and y3 are significant, this will indicate the existence of market imperfections. Coefficient J2 is assumed to reflect the influence that variations in internal funds, liquidity, and net worth have on investment. Coefficient y3 should be significant if it is true that an increasing debt-to-capital ratio accentuates incentive problems because the growth rate of debt is higher than growth in the value of collateral. However, since cash flow may also contain information about future profitability, imperfectly represented by q, it is important to check whether there are differences in the coefficients corresponding to different sample groups and periods. We add some macro financial indicators to test whether they have an independent effect in addition to microeconomic channels. Therefore, Hjis introduced, which stands for the set of variables that will be used to test different splits of the sample and macroeconomic effects. Results are presented for two techniques: the fixed-effects within estimator (FEW) and the generalized method of moments estimator (GMM). In the present case, Tobin's Q, cash flow, and leverage are uncorrelated with present and future errors after taking deviations from the firm's mean. It should be noted that a large number of time periods (35 quarters) are available. We also present results based on GMM estimation of the model in differences.3 This model is standard in the literature (Arellano and Bond 1991; Mairesse, Bronwyn, and Mulkay 1999; Harris, Schiantarelli, and Siregar 1994). The Argentine case can thus be compared with results for other countries, showing the sensitivity of the results to the estimation technique. The estimations are summarized in tables 3.1 and 3.2. One solid result of the research is that qs coefficient is highly significant, independent of the

Table 3.1. Basic Regression Results on Firm Investment, Stock Exchange Firms, Argentina, 1990s

Fixed-effects within estimation

Generalized method of moments estimation
(4) (7)

Instrumental variables estimation
(8) (9)

Variable

(D
-0.029324 0.003336 -0.000561 (-0.58) 0.027654 (5.01) 0.09753.09 (1.33) 0.012560 (0.55) -0.012633 (-0.46) (-1.43) 0.000059 (0.41) -0.004303 (-0.59) -0.008697 (-0.95) (0.23) (-0.95) 0.026719 (4.87) 0.094757 (1.30) 0.010453 (0.45) -0.013983 (-0.51) -0.179242 0.034499 (10.13) 0.134176 (2.84) -0.011375 (-0.85) 0.031796 (1.28) -0.000487 (-2.94) 0.032225 (9.34) 0.146273 (3.10) -0.000785 (-0.06) 0.037669 (1.51) 0.000026 (0.05) 0.027061 (4.93) 0.097680 (1.34) 0.014603 (0.63) -0.012976 (-0.48)

(2)

(3)

(5) (6)

Constant

-0.016563

-0.057012

-0.001089 (-2.03) 0.025275 (4.66) 0.075336 (1.04) 0.009008 (0.40) -0.020493 (-0.76)

-0.021928 (-1.10) 0.032899 (3.99)

(-1.19)

(-2.53)

q Tobin

0.037107

0.034177

(9.78)

(9.60)

Lagged cash flow/capital stock

0.145320

0.146500

(3.08)

(3.10)

Initial debt/capital stock

-0.004440

-0.006626

0.018153 (0.69) 0.005009 (0.17)

(-0.33)

(-0.49)

Lagged investment/capital stock

0.037107

0.036989

(1.49)

(1 .48)

Country risk (EMBI spread)

-0.156007

(-1.42)

Private credit by banks/GDP

0.000070

(1.29)

Crisis effect

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Average of past values of -0.342719 (-3.15) (-2.85) (-2.03) 0.397132
1.98

-0.43727 -0.315867

country risk (EMBI spread)

Current cash flow/capital stock

Firms

45 45 45
1,440 1,440

45 45
1,530 1,440
8.43 9.02

45
1,530 1,440

45

45

45
1,530
6.52

Observations

1,530

1,530

Fixed-effects R 2 (percent)

8.54

5.52

GMM specification tests 39.81 39.91 39.76
484 484 484

Sargan text 39.81
484

Chi-statistic 1.000 1.000 -26.04 0.0000 0.4274 0.6690 -26.01 0.0000 0.4163 0.6770 1.000 -25.99 0.0000 0.3950 0.6930

Degrees of freedom

P-value

1.000 -26.01 0.0000 0.2500 0.8059

Serial correlation, first order

z-statistic

P-value

Serial correlation, second order

z-statistic

P-value

Note: The dependent variable is investment/capital stock; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.2. Basic Regression Results with Cross-Products, Firm Investment, Stock Exchange Firms, Argentina, 1990s Fixedeffects within Instrumental variables
(4)
-0.050010 (-3.60) 0.040320 (4.40) 0.034420 (10.09) 0.194133 (1.40) 0.009429 (0.35) 0.010857 (0.38) -0.010393 (-0.76) 0.031455 (1.26) -0.613638 (-0.46) 0.004787 (0.21) 0.014008 (0.47) (0.07) -0.001142

Fixedeffects within Instrumental variables
(6)
-0.020924 (-0.82) 0.034725 (3.34)

Instrumental variables
(2)
-0.034836 (-3.36) 0.033449 (9.74) 0.104490 (1.73) 0.003016 (0.22) 0.037529 (1.51) (0.24) (0.63) (-3.68) (3.22)

Fixedeffects within
(5)

Fixedeffects within
(7)
-0.033464 (-3.19) 0.032919 (9.42) 0.096882 (1.59) 0.005600 (0.39) 0.037119

Instrumental § variables 2
(8) g
-0.046409 ~ (-3.11) 0.039458 (4.40)

Variable

(1)

(3)

Constant

-0.032031

-0.040926

(-3.29)

q Tobin

0.032720

0.035258

(9.72)

Lagged cash flow/capital stock

0.328371

(2.51)

Initial debt/capital stock

-0.000514

0.005513

0.011037 (0.42) 0.010590

(-0.04)

Lagged investment/capital stock

0.035903

0.018180

(1 -44)

(1 .49)

(0.37)

Average of past values of

-1.876381

country risk (EMBI spread)

(-1.48)

lagged cash flow/capital stock

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Crisis effect * lagged cash flow/ 0.098105 (1.14) 0.200881 (0.82) -4.367133 (-0.45) 0.124514 0.136683 (1.55) -0.325010 -0.230759 (-0.88) -0.006233 (-0.84) -0.007162 (-0.89)
45 45 45

0.112960 (1.28) 0.628992 (0.68) 0.181789 (0.73)

capital stock (0.81) (-0.59)

Current cash flow/capital stock

0.712828

Average of past values of country

-5.289394

risk (EMBI spread) * current cash

flow/capital stock

Crisis effect * current cash flow/

capital stock

(1 .46)
(-2.82)

Average of past values of country

risk (EMBI spread)

Crisis effect

Firms 1,530 6.53 8.49 7.45 9.04 1,530 1,530 1,530

45

45

45

45

45

Observations

1,530

1,530 6.89

1,530 8.54

1,530 7.54

Fixed-effects R2 (percent)

8.55

Note: The dependent variable is investment/capital stock; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

92

FANELLI, BEBCZUK, AND PRADELLI

method used. This contrasts with Gallego and Loayza's (2000) results, which indicate that investment does not significantly depend on the firm's q value. This suggests that the firms in the sample are less constrained in capital markets than Chilean firms are. This is an interesting finding because Chile is generally characterized by greater financial depth than Argentina. However, it must be taken into account that the firms in the present sample are larger than their Chilean counterparts and that the Argentine firms have relatively good access to international capital markets. In light of the noticeable delisting process in the Argentine stock market in the 1990s, the higher sensitivity of investment to q in Argentina may reflect some selection bias, in the sense that only the larger and stronger firms in the country continued to be quoted. In any case, it would be interesting to compare Argentine and Chilean firms, controlling for size and other characteristics to isolate the effects of financial deepening. The coefficient on cash flow, y2, is highly significant when fixed effects are used. Using GMM, the coefficient is smaller and less significant. However, it is usually significant at the 10 percent level in table 3.3. The coefficient on debt, y3, is not significant. It must be taken into account that most of the firms in the panel are large enterprises. It is possible that firm size is highly correlated with the fundamental factors that determine the probability of being constrained. Smaller firms are likely to suffer from idiosyncratic risk and have lower collateral, and are less likely to have developed a track record. Unit bankruptcy costs are likely to decrease with size and smaller firms face higher transaction costs in capital markets. In this regard, the results should be interpreted as a signal that financial constraints due to incentive problems have a lower impact on large firms. The analysis includes three aggregate measures to test for an independent influence of macroeconomic factors: the private banking credit/GDP ratio, a dummy for crisis effect, and country risk. Since these variables do not change across firms, they are similar to time-specific effects. It was not possible to detect any significant influence of the crisis effect and private banking credit/GDP variables. Nor were these variables relevant on their own or as multiplicative variables affecting the q, cash flow, or leverage variables (table 3.2 reports only the results corresponding to the cash flow variable). Note that the point estimates using the interactions with the crisis dummy suggest that the size of the coefficient on cash flow increases substantially during the crisis years compared with the noncrisis years (between 75 per-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCIAL CONSTRAINTS IN ARGENTINA

93

cent and 117 percent, depending on the specification). However, the difference is not statistically significant at conventional levels (the asymptotic t-statistic on the difference equals 1.55 at best). This result is similar to the findings of Gallego and Loayza (2000) for Chile and would indicate that macroeconomic factors work through microeconomic variables. However, the coefficient corresponding to country risk (defined as an annual simple average) is significant (table 3.1). This supports the hypothesis on the relationship between overall macroeconomic conditions, changes in the country risk premium, and the value of collateralizable net worth. The coefficient of the interactions of the country risk premium with q, cash flow, and leverage is not significant (table 3.2 shows the results for cash flow). To test whether debt composition affects investment, we included the proportions of long-term and dollar-denominated debt in total debt on the right-hand side of the equation. Table 3.3 summarizes the estimations. The ratio between long-term and total debt is relevant in both FEW and GMM exercises. This result suggests that the availability of long-term external funds, given the desire of matching long mature assets with long mature financial obligations, is associated with an increase in investment. The ratio between dollar-denominated and total debt is not significant in any case. All firms in the sample are compared with subgroups that are expected to have differential access to financial markets. Multiplicative dummies are used to compare the coefficients for different groups. Specifically, the sample is split according to whether the firm is affiliated with a business group, whether the firm quotes its shares in foreign markets, whether the firm participates in the market for obligaciones negodables and Eurobonds, and whether the firm is a privatized enterprise (privatized enterprises had contractual clauses that obliged them to invest heavily in the years following privatization). As a general rule, the results reported in table 3.4 are sensitive to the method of estimation used. When FEW results are considered, whether a firm is privatized does not alter the influence of q or cash flow on investment. Affiliation with a business group improves the positive influence of cash flow on investment. This is also true using GMM. This result is puzzling because it would be expected that group membership would reduce financing constraints. However, two points must be taken into account. First, affiliation with a business group may be a signal that the firm is facing some kind of imperfection

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.3. Debt Composition and Firm Investment, Stock Exchange Firms, Argentina, 1990s
Generalized method of moments estimation
(3) (4) (5)
-0.000283 (-0.63) 0.028459 (5.35) 0.125899 (1.85) 0.023615 (0.94) -0.016546 (-0.61) 0.000633 (1-30) 0.045549 (2.15) -0.017139 (-0.94)
45 45 45 45

Fixed-effects within estimation
(2)

Variable
-0.043452 -0.000155 (-0.33) 0.029685 (5-63) 0.105116 (1.65) 0.018038 (-3.83) (-1.53) 0.032036 (9.33) 0.150802 (3.18) -0.000060 (-0.01) 0.037608 (1.51) 0.033070 (9.82) 0.138961 (2.94) -0.003113 (-0.24) 0.034059 (1.37) -0.021332

(D

(6)
-0.000469 (-1.07) 0.026158 (4.84) 0.105463 (1.54) 0.013094

Constant

-0.030869

(-3.18)

q Tobin

0.032704

(9.71)

Lagged cash flow/capital stock

0.147843

(3.13)

Initial debt/capital stock

-0.002253

(-0.17)

(1.10) -0.024165 (-0.90)

(0.63) -0.021612 (-0.81)

Lagged investment/capital stock

0.038076

(1.53)

Initial total debt/equity

0.000173

(0.38)

Initial long debt/total debt

0.128322 (3.51) -0.011157 (-0.37)
45

Initial dollar debt/total debt

Firms

45

Observations

1,530

1,530

1,530

1,440

1,440

1,440

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Fixed-effects R2 (percent)
8.70 8.47

8.42

GMM specification tests

Sargan test

Chi-statistic

40.91
605

39.22

Degrees of freedom

35.92 605
J .000
1.000
-25.7300 0.0000 0.2621 0.7930 -26.0200 0.0000

605

P-value

1.000
-26.1200 0.0000 0.3076 0.7580

Serial correlation, first order

z-statistic

P-value

Serial correlation, second order

z-statistic

P-value

0.3907 0.6960

Note: The dependent variable is investment/capital stock; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.4. Effects of Type of Firm on Investment, Stock Exchange Firms, Argentina, 1990s
American Economic group Depository Receipts Generalized method of Generalized method of Bonds Generalized method of Generalized method of moments

Privatized firms

Fixed-

effects

within

FEW
moments

FEW
moments

FEW (7)
-0.033283 (-3.09)

moments

Variable

(1)
0.000138 (0.24) (-3.45) 0.030195 (7.16) 0.003835 (0.54) 0.092190 (1.81) 0.270707 (2.39) 0.030254 (1.38) -0.055546
(-1 .99)

(2) (3)
-0.033900 (-0.55) 0.037079 (4.70) -0.020328 (-1.34) -0.271015 (-2.26) 0.548140 (3.57) 0.073297 (2.05) -0.062213 (-1.27) (9.03) 0.032509 (1.98) 0.164230 (3.42) -0.259902
(-1 .05) 0.034366

(4)
-0.000267 (-2.82) 0.031045 -0.028629

(5)

(6)
-0.000109 (-0.23) 0.028996 (5.08) -0.030319 (-0.83) 0.108204 (1.48) -0.538519 (-0.61) 0.016186

(8)
-0.000301 (-0.62) 0.031032 (8.94) 0.023323 (1.65) 0.152374 (2.99) 0.278523 (1.21) -0.006473 0.026033 (4.43) 0.030983 (1.09) 0.035152 (0.38) 0.268239 (1.57) 0.007020

Constant 0.026661 (4.66) -0.038264 (-0.23) 0.099202 (1.36) 2.423900 (0.22) 0.015819 (0.68) -1.058010
(-1 .02)

-0.028546

(-2.12)

q Tobin

0.032583

(9.65)

q Tobin * dummy

-0.014480

(-0.28)

Lagged cash flow/capital stock

0.148454

(3.14)

Lagged cash flow/capital stock

0.564472

* dummy

(0.22)

Initial debt/capital stock

-0.000780

(-0.06)

(0.26) -0.186742 (-2.03)

(0.70) -0.210713 (-1.07)

(-0.37) -0.013872 (-0.48)

(0.23) 0.022154 (0.39)

Initial debt/capital stock

-0.158306

* dummy

(-1.02)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Lagged investment/Capital stock

0037261

-0.152960 (-0.56) (1.41) (-0.37) (1.25) (-0.38) (1.38) (-0.53)

0.035270

-0.009931

0.031322

-0.010304

0.034405

-0.014426

(1.49)

Firms 1,440 1,530 8.89 8.75 1,440 9.05 1,530 1,440 1,530 1,440

45

45 45 45

45

45 45 45

Observations

1,530

Fixed-effects R2 (percent)

8.50

GMM specification tests 37.94 37.71
481 481

Sargan test 39.47 1.000 -25.92 0.0000 0.4249 0.6710 39.57
481 481

Chi-statistic 1.000 1.000 -24.94 0.0000 -0.1262 0.9000 -25.89 0.0000 0.3814 0.7031

Degrees of freedom

P-value

1.000 -25.88 0.0000 0.3774 0.7060

Serial correlation, first order

z-statistic

P-value

Serial correlation, second order

z-statistic

P-value

Note: The dependent variable is investment/capital stock; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

98

FANELLI, BEBCZUK, AND PRADELLI

4

In fact, this argument deserves a qualification. Like other developing countries, Argentina's stock market is considerably volatile and thin, which in turn poses some doubts about market efficiency. Under these circumstances, the observed q may lose informative power at the time of making investment decisions. Consequently, a low sensitivity of investment to q may reflect this problem in addition to the financial constraints approach. 5 Due to the contemporaneous relationship between investment and sales, the FEW estimator is inconsistent. Therefore, the instrumental variables estimator (ASIV) for the SE panel is also presented. The ENGE panel has a shorter time dimension (four years) and includes more firms (308), so the GMM estimations are appropriate in this case. The so-called GMM system estimator is also used for the ENGE panel. GMM system estimates are based on instrumenting both the differenced and level equations. For the differenced equation, instruments are the lagged level value of the endogenous and predetermined variables and the first-differenced values of the exogenous variables. For the level equation, instruments are the lagged firstdifferenced value of the endogenous and predetermined variables. Again, estimations of equations are one-step estimations and the Sargan statistic comes from the two-step estimations. Given the shorter time dimension, Sargan tests do not exhibit an over-fitting problem. 6 Note that contemporaneous sales are included as a proxy for fundamentals here. See Hsiao (1986).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

in financial markets. The organization of an internal capital market within the group will be profitable only to the extent that it reduces the cost of funding for the firms in the group. Second, the firms in the sample are large; small firms are implicitly underrepresented. Consequently, the group variable basically reflects the difference between large firms and firms affiliated with a group. If the latter are more financially constrained than the former, it is reasonable to expect a positive sign for the coefficient on the interaction variable. An interesting result is that market segmentation (reflecting the incidence of information and incentive problems) is relevant in relation to the market/book ratio but not so regarding cash flow or leverage. Actually, those firms that do not have access to more sophisticated markets (ADRs and bonds) are less sensitive to q. This is consistent with the predictions of the approach stating that financial constraints matter.4 Finally, some of the regressions use the current sales/capital ratio instead of Tobin's Q as a proxy for returns. In this case, the available data are richer because another panel of firms is used in addition to the panel of firms listed on the stock exchange. To distinguish between the two, the stock exchange panel is referred to as SE, and the new panel as ENGE (Encuesta Nacional de Grandes Empresas). Table 3.5 presents the estimations and the instrumental variable estimation of the model in differences as suggested by Anderson and Hsiao.5'6

FINANCIAL CONSTRAINTS IN ARGENTINA

99

Capital Structure and Dollarization The evidence in Fanelli and Keifman (2002) and Schmukler and Vesperoni (2001) and the stylized facts suggest two points about the interaction of leverage, maturity, and currency denomination in Argentina. First, the variables that appear in the literature on capital structure should play a relevant role in the Argentine case. In this sense, Argentina would be consistent with the findings of Booth and others (2000). Second, the analysis of balance sheets shows specific features that are probably associated with the volatility of the Argentine context and the particular characteristics of the currency board regime. To study these issues, we use a regression model like the one Gallego and Loayza (2000) use in the Chilean case and Booth and others (2000) use for a set of emerging countries. We add an equation to analyze the firm's decision regarding the proportion of dollar-denominated debt in total debt. Specifically, the equations are the following:

The variable Z is the leverage ratio (defined here as debt to equity), and I and F are the proportions of long-term debt and dollar-denominated debt in total debt. S denotes the log capital stock, which proxies for size; B represents the operational profits/assets ratio, which proxies for the firm's profitability. K is the ratio of fixed-to-total assets that measures the tangibility of total assets. Variable D is introduced to distinguish firms with different degrees of market access, and M to test for macroeconomic effects. The strategy
7

The investment model is also estimated using current cash flow instead of lagged cash flow in the SE panel. The results do not change significantly. See column (9) in table 3.1.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

For the SE panel, the instrumental variable (ASIV) estimations support the previous results on cash flow and leverage. In both GMM and ASIV results, the coefficient on sales is not significant, possibly because of poor instruments. The same arguments apply to the ENGE panel, in which none of the other variables seems to be relevant.7

Table 3.5. Stock exchange firms Generalized Instrumental method of moments variables Generalized method of moments

Effects of Sales/Capital on Firm Investment, Stock Exchange and ENGE Firms, Argentina, 1990s ENGE firms3 Generalized method of moments system

Fixed-

effects

within

Variable

(1)
-0.022648 (-1.45) (-1.85) 0.003203 (0.14) 0.114331 (1.51) 0.078544 (3.29) -0.007404 (-0.27)
45

(2) (3)
-0.000914

(4)
0.016029 (1.02) 0.029727 (1.37) 0.234703 (1.73) 0.116407 (1.82) 0.117839 (1.97)
308

(5)
-3710.39 (-0.08) -0.026723 (-0.59) 0.130602 (1.16) 0.093127 (0.99) 0.096992 (1.36)
308

Constant 0.032790 (0.39) 0.431368 (2.44) 0.048502 (1.71) 0.015654 (0.52)
45

-0.008702

(-0.90)

Sales/capital stock

0.055061

(3.53)

Lagged cash flow/capital stock

0.198669

(3.95)

Initial debt/capital stock

0.024545

Lagged investment/capital stock

(1.81) 0.051670

(2.02)
1,215

Firms

45

Observations

1,530

1,440

616

924

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Fixed-effects R2 (percent)

3.40

2.01

GMM specification tests 43.88
484

Sargan test

Chi-statistic

8.54
8

13.55

Degrees of freedom 1.000 -7.16 0.0000

P-value

0.382

16 0.633

Serial correlation, first order

z-statistic

P-value 0.3751 0.7080

-26.02 0.0000

-1.768 0.0770

Serial correlation, second order

z-statistic

P-value

' Encuesta Nacional de Grandes Empresas.

Note: The dependent variable is investment/capital stock; t-statistics are in parentheses. Data for ENGE firms are from the Encuesta Nacional de Grandes Empresas.

Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

102

FANELLI, BEBCZUK, AND PRADELLI

for estimating these equations is similar to that followed in the case of investment, although the GMM system method is used instead of GMM, following Gallego and Loayza (2000). Results are presented in tables 3.6 to 3.11. The logarithm of total assets is used as a proxy for size. This variable is included in the equations because it alleviates information asymmetries. But since larger firms are likely to be more diversified, it also reduces repayment risk. It must be taken into account that diversification may have a premium in economies where markets for allocating risk are incomplete. In the Argentine case, for example, the stock exchange can only be partially used to diversify risk. Owing to the reduced number of firms listed, many sectors in the economy are not represented. In the leverage equation, the coefficient corresponding to size is not significant in either panel (see tables 3.6 and 3.9). The results are very different regarding the other two equations. For the SE panel, the coefficient is positive and significant. For the ENGE panel, the results are the same when using the FEW method, but this is not the case when controlling for joint endogeneity (tables 3.7 and 3.10). It seems that larger firms have a higher preference and/or face softer constraints in accessing long-term and external markets. A greater degree of tangibility (fixed assets over total assets) in the composition of the asset mitigates asymmetric information problems and favors the use of long-run debt owing to the desire of firms to match the duration of assets and liabilities. Consequently, a positive sign is expected in the leverage and long-term debt equations. In the case of dollar-denominated debt, a negative sign is expected. For a foreign creditor, it is much more difficult to liquidate tangible assets to recover a nonpaid dollar-denominated loan. For both foreign and domestic creditors, the existence of currency risk may imply that the liquidation value of some tangible assets is low due to irreversibility. On the demand side, if those firms holding tangible assets belonged to the nontradable sector, limiting currency risk would result in a preference for domestic currency-denominated debt. The overall results on the role of tangibility in the leverage equation are weaker than in the case of the size variable. It was not possible to obtain definite conclusions about the sign of the variable or its significance because those factors change according to the estimation method and the panel used (tables 3.6 and 3.9). These results are consistent with the evidence presented in Booth and others (2000); they also found an unstable coefficient.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCIAL CONSTRAINTS IN ARGENTINA

103

In the case of the equation explaining the behavior of debt with longer maturity, the evidence suggests that tangibility exerts a positive influence (tables 3.7 and 3.10). This favors the matching argument. It seems that firms with more tangible assets try to lengthen the maturity of their liabilities. In the case of the dollar equation, tangibility is included to test the hypothesis that, ceteris paribus, external creditors find ex ante distress costs higher as the proportion of tangible assets in the capital structure increases. Likewise, firms holding such assets may prefer to reduce currency risk exposure. It is interesting here to recall the previous argument. Given the de facto association between dollar debt and long-term debt, firms seeking to finance longrun assets face a troublesome trade-off between the benefits of extending the duration of liabilities and the higher currency risk that long-term debt entails. However, the empirical findings are mixed. In both panels, the variable is significant when the FEW method is used but it is not when the GMM system method is employed. It is difficult to tell ex ante the sign of the coefficient of the return on assets in each of the equations. The higher the return on assets, the lower the repayment risk is. However, higher profits reduce the need for the more expensive external funds and firms with better growth prospects may want to avoid the possibility that (bank) creditors will extract rents from them. In order to control for firms' growth prospects, the market-to-book ratio is included. As a proxy for growth, it is expected that a higher q will reduce perceived repayment risk. The results in tables 3.6 and 3.9 suggest that profits do not play a relevant role in explaining leverage either in the SE or the larger ENGE panels. In the case of the market-to-book value, the results only refer to the SE panel because many of the firms in the ENGE panel are not listed. It was not possible to detect a significant influence on leverage (table 3.6), but it appears that growth prospects have a bearing on decisions regarding the proportion of long-term and dollar-denominated debt. The results indicate that q has a significant and positive effect on the proportion of debt of longer maturity held by firms as a proportion of the total stock of debt (table 3.7) and a negative effect on dollar debt (table 3.8). This suggests that while growth prospects ease access to long-term credit markets, currency exposure makes lenders and/or borrowers reluctant to write dollar-denominated debt contracts. Business groups may help to cope with information and contract enforcement and, in the case of financial distress, individual firms may also rely

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.6. Generalized method of moments system estimation
(4) (5)
(6) (7) (8)

Effects of Type of Firm on Firm Debt/Equity Ratio, Stock Exchange Firms, Argentina, 1990s

Fixed-effects within estimation
(2) (3)

Variable
5.095653 3.077778 -2.595810 4.009580 17.394300 (1.68) -0.844897 (-1.50) -1 .469340 (-0.78) 10.244300 (1.48) -0.701646 (-1.70) (0.83) -0.029849 (-0.14 -1.988320 (-1.23 8.245820 (1.30 -0.519335
(-1 .48

(D

Constant (0.58) (0.35) -0.978872 (-0.20) -0.024937 (-0.01) 0.327134 9.500530 (1.33) -0.189372 (-1.09) -1.350360 (-0.25) 0.009891 0.008867 (1.19) 0.798516 (0.06) 0.022390 (0.23) 0.026048 (0.26) -0.059188 (-0.06) 0.018683 (0.19) 0.024955 (2.99) (0.08) -0.101551 (-0.52) (-0.07) -0.151146 (0.29) 0.056828 -0.477916 (-0.92) 2.132631 (0.80) 0.802845 (0.20) -0.009352 (-0.05) (-0.61)

2.212353

-0.425791 (-0.10) 0.212060 (1.09) -2.016530
(-1 .28)

13.060700 (1.30) -0.570914
(-1 .07)

(0.25)

Size, ln(capital stock)

0.010634

(0.02)

Fixed assets/total assets

-0.486420

-0.927312 (-0.59) 8.831300 (1.40) -0.443845 (-1.37) 8.935910 (1.20) -0.633295 (-1.52)

(-0.19)

Profits/total assets

-0.577885

(-0.14)

q Tobin

-0.157474

(-0.82)

Country risk (EMBI spread)

-5.943777

(-0.94)

Private credit by banks/GDP

Crisis effect

Lagged ratio of debt to equity

0.021597 (0.25) (0.22)

Privatized firms

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Economic group (1.72) -1.306900 (-2.07) 2.219380 (1.22) 45 1,530 1,485 1,485 0.66 0.30 1,530 1,485 1,485 45

2.782010

American Depository Receipts

Bonds

Firms

45

45

45

45 45

45
1,485

Observations

1,530

Fixed-effects R2 (percent)

0.12

GMM specification tests 41.29
765 765

Sargan Test 38.81 1.000 -1.5960 0.1100 -0.2312 0.8170 37.51
765

Chi-statistic 1.000 -1.6040 0.1090 -0.2580 0.7960

38.47
765

38.77
765

Degrees of freedom

P-value

1.000 -1.6040 0.1090 -0.3197 0.7490

1.000 -1.5960 0.1100 -0.2500 0.8030

1.000 -1.6040 0.1080 -0.2696 0.7870

Serial correlation, first order

z-statistic

P-value

Serial correlation, second order

z-statistic

P-value

Note: The dependent variable is debt/equity; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.7. Effects of Type of Firm on Firm Long-Term Debt/Total Debt Ratio, Stock Exchange Firms, Argentina, 1990s Genenilized metho(j of moment;; system estiination
(4) (6) (7)
-0.327914 (-3.12) 0.016826 (2.77) 0.111525 (2.88) 0.060426 (0.46) 0.006930 (1.75) (0.70) 0.006521 (1.61) -0.431440 (-3.85) 0.023899 (3.62) 0.116034 (3.00) 0.037260 -0.091273 -0.048535 (-1.78) 0.020150 (2.99) 0.116759 (2.87) 0.063163 (0.75) 0.007882 (1.75) -0.391883 (-3.28) (-1.46) 0.006671 (1.89) 0.112365 (3.15) 0.064867 (0.78) 0.002212 (0.711) -0.294276 (-3.47) 0.000241 (3.51) -0.013485 (-1.53) 0.807581 (31.3) 0.784102 (24.8) -0.048535 (-1.78) 0.777166 (26.1) 0.789537 (26.8) 0.789918 (27.3) -0.000061 (-1.24)

Fixed-eff ects within e<stimation

Variable

(D
-0.720068 -0.862975 (-4.76) 0.051543 (5.02) 0.313484 (6.01) 0.313963 (3.67) 0.006946 (1.71) (-3.97) 0.035197 (3.27) 0.401877 (7.28) 0.362872 (4.25) 0.011711 (2.84)

(2) (5)

(3)

(8)
-0.256474 (-2.56) 0.012607 (2.14) 0.117764 (3.12) 0.063793 (0.73) 0.005622 (1.47)

Constant

-0.673078

(-3.72)

Size, ln(capital stock)

0.041823

(4.13)

Fixed assets/total assets

0.372005

(7.20)

Profits/total assets

0.329224

(3.90)

q Tobin

0.008531

(2.15)

Country risk (EMBI spread)

-0.639780

(-4.89)

Private credit by banks/GDP

Crisis effect

Lagged ratio of long debt to total debt

Privatized rirms

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Economic group (-3.24) -0.007065 (-0.38) 0.014974 (1.10) 45 1,530 8.42 7.80 1,530 1,485 1,485 1,485 45

-0.046139

American Depository Receipts

Bonds

Firms

45

45

45

45

45
1,485

45
1,485

Observations

1,530

Fixed-effects R2 (percent)

9.13

GMM specification tests 34.94
765 765

Sargan test 35.38 1.000 -3.6870 0.0000 0.3463 0.7290 35.98
765

Chi-statistic 1.000 -3.6330 0.0000 0.3358 0.7370

36.76
765

36.05
765

Degrees of freedom

P-value

1.000 -3.6550 0.0000 0.3394 0.7340

1.000 -3.6770 0.0000 0.3374 0.7360

1.000 -3.6970 0.0000 0.3295 0.7420

Serial correlation, first order

z-statistic

P-value

Serial correlation, second order

z-statistic

P-value

Note: The dependent variable is long-term debt/total debt; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.8. Effects of Type of Firm on Firm Dollar Debt/Total Debt Ratio, Stock Exchange Firms, Argentina, 1990s Generalized method of moments system estimation
(4)
(5) (6) (7)

Fixed-effects within estimation
(2) (3)

Variable
-1.192748 -1.277908 -0.244544 -0.809955 -1.026260 (-4.80) 0.072981 (5.07) -0.056488 (-0.95) -0.234048 (-1.42) -0.002359 (-0.44) (-3.71) 0.055794 (3.91) -0.034342 (-0.65) -0.135650 (-0.90) -0.000780 (-0.14) (-2.39) 0.022160 (2.91) 0.002020 (0.05) -0.084960 (-0.73) -0.009054 (-2.1) -0.018131 (-0.16) 0.000293 (3.61) 0.014747 (1.42) 0.702655 (10.8) 0.665567 (9.35) -0.138803 (-2.68) 0.639652 (9.01) 0.670517 (10.3) 0.000058 (1.09) (-5.96) 0.107443 (8.84) -0.195748 (-3.18) 0.077374 (0.76) -0.015490 (-3.23) (-5.56) 0.094265 (7.40) -0.122122 (-1.87) 0.101568 (1.01) -0.012077 (-2.48) -0.747306 (-4.07) 0.052606 (4.26) -0.038932 (-0.74) -0.127667 (-0.88) -0.002459 (-0.44)

(D

(8)
-0.499053 (-3.25) 0.038533 (3.69) -0.033523 (-0.70) -0.137225 (-0.99) -0.006103 (-1.24)

Constant

-1.244826

(-5.78)

Size, ln(capital stock)

0.107240

(8.90)

Fixed assets/total assets

-0.190071

(-3.09)

Profits/total assets

0.060653

(0.60

q Tobin

-0.016343

(-3.46)

Country risk (EMBI spread)

-0.316954

(-2.04)

Private credit by banks/GDP

Crisis effect

Lagged ratio of dollar debt to total debt

0.685329

(10)

Privatized firms

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Economic group (-3.94) -0.062895 (-1.55) -0.003267 (-0.11) 45 1,530 1,485 1,485 8.48 7.80 1,530 1,485 45 45

-0.127107

American Depository Receipts

Bonds

Firms

45 45 45

45
1,485

45
1,485

Observations

1,530

Fixed-effects R2 (percent)

7.93

GMM specification tests 35.31 32.31
765 765

Sargan test 37.41
765

Chi-statistic 1 .0000 -2.7130 0.0000 0.8380 0.4020 -2.7560 0.0060 0.7377 0.4610 1.000

36.41
765

38.55
765

Degrees of freedom

P-value

1.000 -2.7590 0.0060 0.6769 0.4980

1.000 -2.4190 0.0160 0.8779 0.3800

1.000 -2.7530 0.0060 0.8142 0.4160

Serial correlation, first order

z-statistic

P-value

Serial correlation, second order

z-statistic

P-value

Note: The dependent variable is dollar debt/total debt; t-statistics are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

110

FANELLI, BEBCZUK, AND PRADELLI

Fixedeffects within
Variable Constant Size, ln(capital stock) Fixed assets/total assets Profits/total assets Lagged ratio of debt to equity Country risk (EMBI spread) Private credit by banks/GDP Economic group Access to international capital markets Percentage of capital owned by foreigners Firms Observations Fixed-effects R2 (percent) GMM specification tests Sargan test Chi-statistic Degrees of freedom P-value Serial correlation, first order z-statistic P-value 0.075052 (0.02) -0.340793 (-0.11) 0.016385 (0.46)
308

Generalized method of moments system
(2)
2283580.0 (1.74) -1.284730 (-1.39) 670466.0 (1.24) -4.790850 (-1.25) 0.514380 (5.12) 828726.0 (0.21) -848902000 (-0.15)

(D
-0.640412 (-0.03) 0.321177 (0.24) -10.268620 (-2.08) 0.844767 (0.21)

(3)
2252530.0 (1.74) -1.255680 (-1.37) 650650.0 (1.21) -4.687860 (-1.24) 0.514071 (5.13) 858803.0 (0.22) -922469000 (-0.16) -6607.0 (-0.06)

308 924

308 924

1,232 0.10

18.74

18.9

16

16

0.282 -3.824 0.0000

0.274 -3.824 0.0000

Note: The dependent variable is total debt/equity; t-statistics are in parentheses. Data for ENGE firms are from the Encuesta Nacional de Grandes Empresas. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.9. Effects of Type of Firm on Firm Total Debt/Equity Ratio, ENGE Firms, Argentina, 1990s

FINANCIAL CONSTRAINTS IN ARGENTINA

111

Fixedeffects within Variable
Constant Size, ln(capital stock) Fixed assets/total assets Profits/total assets Lagged ratio of long debt to total debt Country risk (EMBI spread) Private credit by banks/GDP Economic group Access to international capital markets Percentage of capital owned by foreigners Firms Observations Fixed-effects R2 (percent) GMM specification tests Sargan test Chi-statistic Degrees of freedom P-value Serial correlation, first order z-statistic P-value 0.022147 (0.60) 0.166206 (4.75) 0.000656 (1.57)
308

Generalized method of moments system
(2)
0.101539 (0.88) 0.000000 (0.44) 0.131848 (2.96) 0.000000 (0.47) 0.806792 (23.5) 0.432910 (1.12) -1513.57 (-2.53)

(1)
-1.585029 (-5.74) 0.101684 (6.46) -0.106377
H.82)

(3)
0.096602 (0.84) 0.000000 (0.45) 0.134983 (3.03) 0.000000 (0.43) 0.805608 (23.5) 0.432961 (1.12) -1516.82 (-2.54) 0.011930 (1.11)

-0.064518 (-1.39)

308 924

308 924

1,232

27.90

11.06

11.06

16

16

0.806 -4.593 0.0000

0.806 -4.619 0.0000

Note: The dependent variable is total debt/equity; t-statistics are in parentheses. Data for ENGE firms are from the Encuesta Nacional de Grandes Empresas. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.10. Effects of Type of Firm on Firm Long-Term Debt/Total Debt Ratio, ENGE Firms, Argentina, 1990s

112

FANELLI, BEBCZUK, AND PRADELLI

ENGE Firms, Argentina, 1990s Fixedeffects within Variable Constant Size, ln(capital stock) Fixed assets/total assets Profits/total assets Lagged ratio of dollar debt to total debt Country risk (EMBI spread) Economic group Access to international capital markets Percentage of capital owned by foreigners Firms Observations Fixed-effects R2 (percent) GMM specification tests Sargan test Chi-statistic Degrees of freedom P-value Serial correlation, first order z-statistic P-value -2.451 0.0140 -2.461 0.0140
13.09 12.93

Generalized method of moments system

(1)
-1 .040294 (-3.45) 0.080088 (4.66) -0.234532 (-3.68) -0.043680 (-0.86)

(2)
151517000 (1.59) -133.769 (-2.09) 108706000 (1.92) -43.197200 (-0.34) 1.056850 (8.22) -117884000 (-0.47)

(3)
1 6987800 (1.29) -133.393 (-2.09) 1 1 3454000 (1.95) -54.726300 (-0.43) 1.051870 (8.09) -116999000 (-0.46) 1.698780 (1.29)

-0.072955 (-1.81) 0.171761 (4.49) 0.000342 (0.75)
308 308 924

308 924

1,232 14.38

16

16

0.666

0.678

Note: The dependent variable is dollar debt/total debt; t-statistics are in parentheses. Data for ENGE firms are from the Encuesta Nacional de Grandes Empresas. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 3.11.

Effects of Type of Firm on Firm Dollar Debt/Total Debt Ratio,

FINANCIAL CONSTRAINTS IN ARGENTINA

113

1

Note that affiliation with a group is a time-varying variable in the case of the ENGE panel.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

on the financial resources of a group. It is reasonable to assume that multinationals operating in Argentina can also count on internal capital markets. It is likely that both groups and multinationals have better access to foreign credit markets. However, independent of whether the firm is affiliated with a group, it is reasonable to assume that firms that gained access to bond markets or placed an ADR face fewer constraints when deciding on leverage and debt composition. Several dummy variables were designed to split the sample and test these hypotheses. Although the significance and size of the effect may vary with the method of estimation and the panel utilized, the results show that firms in different categories face financial constraints of diverse intensities. Affiliation with a group tends to reduce the proportion of dollar-denominated debt (tables 3.8 and 3.11 ).8 Having access to the market for ADRs and bonds may be relevant for market access. The results of some of the exercises indicate that firms that participate either in the ADR or bond markets tend to reduce leverage and increase long-term and dollar-denominated debt (tables 3.6, 3.10, and 3.11). This means that less constrained firms prefer to rely on internal funds or equity to meet their financial needs and tend to use credit markets to increase the duration of their liabilities. They also have better access to international markets and can take advantage of the better conditions offered by dollar-denominated credit markets in terms of both maturity and price. The analysis tested whether the financial structure for all types of firms was dependent on the macroeconomic situation and the evolution of financial deepening. To examine the importance of macroeconomic disequilibria, we introduced the following variables: the country risk premium, private credit/GDP, and a dummy for crisis periods among the right-hand variables in the three equations. Regarding financial development, the hypothesis is that increasing financial deepening and capital inflows increased credit supply in the 1990s, thus allowing firms to elevate their leverage after a long period of tight rationing. The higher increment in the stock of total liabilities was noted above. Tables 3.6, 3.7, and 3.8 show that both the macro economy and financial deepening matter for debt composition (in terms of maturity and currency denomination), but not for total leverage. Specifically, the country

114

FANELLI, BEBCZUK, AND PRADELLI

Concluding Remarks The 1990s was a singular period in Argentina, characterized by structural reforms, the existence of a currency board, and increased capital flows. This chapter has examined the financing and investment decisions that Argentine firms made in that period. The econometric analysis focused on the investment and financial structure equations. The findings suggest that firms faced important financial constraints in a context of imperfect capital markets, high dollarization in financial contracts, and significant interactions between macroeconomic factors and financial constraints. In a context of financial imperfections, some firms may forego profitable opportunities because they do not have fluent access to credit markets. In this way, financial market failures became a source of inefficiency and a deterrent to growth. Capital market imperfections imply, among other things, that information and agency problems affect investment and financing decisions. The present findings suggest that firm size and tangibility of assets, two variables associated with these problems, have a significant effect in both the long-term debt and dollar debt ratio equations. As it turns out, the long-term financing opportunities appear to promote investments. Cash flow also has a positive influence on the investment equations. This is an indication of financial market imperfections to the extent that Tobin's Q does an efficient job of controlling for expectations of future returns. This conclusion should be softened if observed cash flow contains information on future profitability. It has also been possible to detect some differences in the severity of financing constraints across firms using dummy variables for firm characteristics. When financial imperfections are pervasive, macroeconomic fluctuations tend to affect the financial environment and it is difficult for the firms to manage risk and the consequences of cyclical downturns. This affects the

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

risk coefficient is significant and negative in some of the exercises (implying a negative association between the proportion of long-term debt or dollardenominated debt and country risk), while the influence of the credit/GDP ratio is significant and positive. The crisis dummy does not seem to be relevant in any case. It seems, therefore, that there is a direct link between aggregate variables and decisions at the micro level.

FINANCIAL CONSTRAINTS IN ARGENTINA

115

financial position of the firms throughout the cycle. In the Argentine case in the 1990s, shocks and cyclical events were associated with changes in firms' liquidity, access to long-run finance, and net worth value. Thus, it seems reasonable to hypothesize that firms' investment and financing opportunities are influenced by changes in the financial environment. The econometric exercises detected that fluctuations in the country risk premium are relevant to the firms' decisions on investment and debt ratios. In the 1990s, long-term and dollar-denominated debt contracts in domestic and international markets allowed firms to finance investment and growth with better gearing ratios than in the 1980s. The present findings suggest that these decisions implied facing a trade-off between maturity and currency balance sheet mismatches. The risks associated with this trade-off and financial dollarization in general have become evident in the years that followed the period analyzed here. After the abandonment of convertibility (in January 2002), which was accompanied by a strong real depreciation, the debt burden of agents with dollar-denominated liabilities and peso income-generating assets was so huge that there was an across-the-board breakdown of financial contracts. To soften such a burden, dollar debts were converted into pesos and partially indexed. But, as a consequence, the banks' capitalization heavily eroded and is now practically nil. At the same time, some important firms operating in the nontradable sector of the economy have defaulted on their external debts. The present financial crisis in Argentina shows that even bank assets of reasonable quality may deteriorate heavily when the economy experiences a resilient recession and sudden changes in relative prices in a context of dollarization. These facts appear to be closely related to the kinds of financial constraints that firms face, which were detected in this study. These facts further suggest that the improvement of financial policies may contribute to fostering growth in Argentina. Are there any economic policy lessons that can be drawn from our research? The following considerations appear to be the most relevant. First, in the case of Argentina, fluctuations in capital flows and conditions for accessing international capital markets are closely associated with macroeconomic fluctuations. Hence, the development of mechanisms to stabilize capital flows may contribute to avoiding (or at least smoothing) abrupt changes in flows that can induce quick changes in the macroeconomic environment and financial constraints faced by firms. Options

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

116

FANELLI, BEBCZUK, AND PRADELLI

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

for such mechanisms include contracting contingency credit lines with foreign banks and building regional funds (in the Latin American or Mercosur context) oriented toward stabilizing financial conditions when a shock occurs. Nonetheless, the IMF's role will continue to be critical for Argentina. Second, private sector access to foreign markets and long-run finance deteriorates quickly during cyclical downturns. Policymakers must take this into account. Fiscal policy and public sector deficit financing should avoid replicating the pattern of private sector finance during the cycle. The Argentine authorities did not take this into account in the late 1990s. In the period following the Asian crisis, the government crowded out the private sector when a significant worsening in the financial conditions of the private sector was taking place. Finally, the Argentine case provides lessons on the process of structural reforms that could be useful for other developing countries. In spite of Argentina's efforts to reform and liberalize its financial markets, the findings suggest that firms continue to face financial constraints. To a large extent, some financial markets are weak and others are missing because of weak institutions, underdevelopment of the legal and regulatory frameworks, corruption, deficient skilled human resources, and insufficient experience in screening, monitoring, and enforcing contracts in a free market environment. To soften the financial constraints facing firms, it is necessary to create the financial markets that are lacking and to improve the functioning of the existing ones. For this, institutional strengthening is a key policy goal.

Credit, Financial Liberalization, and Manufacturing Investment in Colombia
Maria Angelica Arbeldez and Juan Jose Echavarria
Taking into account information asymmetries, costly monitoring, contract enforcement, and incentive problems modifies Modigliani and Miller's view on the determinants of investment. In this alternative scenario, the capital structure of the firm, average taxes, current profits, and wealth matter; interest rates affect both the use of capital and the availability of funds; and the evolution and efficiency of the financial sector produce a financial accelerator that can affect cycles and growth (Hubbard 1998). Schumpeter (1934) recognized the potential effect of the financial sector in promoting economic growth, one of the robust conclusions in King and Levine's (1993, 2001) analysis. He also argued that credit affects productivity more than capital accumulation does, a conclusion recently verified by Levine, Loayza, and Beck (2000). A strong financial sector reduces global risk, allows progress in the mobilization of savings and the allocation of capital funds, and increases the monitoring of managers. In addition, the financial sector seems to play a large role in the determination of cycles. Kindleberger (1978), for example, illustrates the importance of irrational financial markets in economic crises and Bernanke (1983) shows that debt crises and the collapse of the banking system were central factors in the explanation of the Great Depression of the 1930s. A strong financial sector is associated with macroeconomic stability and poverty reduction. Fazzari, Hubbard, and Petersen (1988) provide a pioneer analysis of the relation between financial constraints, investment, and firm growth. More recent studies include Gallego and Loayza (2000) on Chile; Harris,
Maria Angelica Arbelaez and Juan Jose Echavarria are researchers at Fedesarrollo in Bogota.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 4

118

ARBELAEZ AND ECHAVARRlA

Financial Liberalization and the Relative Development of the Colombian Financial System The 1970s and 1980s During the 1970s, the Colombian financial sector operated under restrictive conditions. The regulatory framework was rigid, institutions were overregulated, and the government tightly controlled the sector. Interest rates and credit allocation were subject to strict administrative controls. Directed credit to specific sectors at subsidized interest rates was an important proportion of total credit, and forced investment played a negative role in credit and credit institutions. Most forced investment had a below-market return and intermediation margins were high (and tended to increase), negatively affecting the
1

See also Tybout (1983) and Echavarria and Tenjo (1993).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Schiantarelli, and Siregar (1994) on Indonesia; Jaramillo, Schiantarelli, and Weiss (1996) on Ecuador; Gelos and Werner (1999) on Mexico; Laeven (2001) on a group of 13 developing countries; Love (2001) on 40 developed and developing countries; and Demirg{i9-Kunt and Levine (1996, 1999) on a large sample of developed and developing countries. Common results in the literature show that financial constraints have decreased after most liberalization episodes, and have been especially large for those firms where information and monitoring are more costly: small and recently created companies, firms not belonging to conglomerates, and domestic as opposed to foreign firms. This chapter uses the Colombian experience of the past two decades to analyze the impact of credit and financial liberalization on manufacturing investment. Domestic investment grew much more in Colombia than in any other Latin American country during the liberalization period of 1990-95, when capital flows and domestic credit expanded drastically, before the sharp reduction in 1998-99. Were the liberalization episodes of the 1990s important in explaining the investment boom? Did financial restrictions increase during the crisis of 1998-99?l What are the implications of the Colombian experience?

FINANCIAL LIBERALIZATION IN COLOMBIA

119

2

The dominant presence of government in the financial sector was evident. The government owned 57 percent of the capital in commercial banks, 81 percent in financial corporations, 27 percent in mortgage corporations, and 19 percent in commercial financing corporations. 3 Financial capitalization (Resolution 42, 1983 and Resolution 60, 1984 of the Board). Board Resolutions 16 and 116 of 1983 (of the Board) created a capitalization fund for firms and gave subsidized credit to textiles, steel, and the construction sector. 4 Board Resolution 33 of 1984 of the Board.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

ability of financial institutions to make profits. Reserve requirements were the main tool for monetary policy, with high levels and large volatility. Four commercial banks accounted for 43 percent of deposits and 45 percent of total assets.2 The Mexican debt crisis of 1982 and its aftermath hit Colombia much less than it hit other Latin American countries, but there were important effects nonetheless, particularly during the first part of the decade (Fischer 1988; Edwards 1995). The first casualties were observed in 1982, when deteriorating macroeconomic conditions negatively affected the performance of financial entities. A precarious regulatory framework, which combined repressive features with weak supervision, aggravated the distress of the underdeveloped financial sector. The financial crisis of 1982-85 was challenging. It was essentially a solvency crisis, with a portfolio deterioration aggravated by a simultaneous weakening of equity bases. Nonperforming loans increased significantly in relation to total assets. The profitability of the institutions was strongly affected by the high provisions they had to undertake in 1985. Of the 111 functioning entities operating in 1980, only 99 remained by 1986. Private commercial banks, financial corporations, and some commercial financing corporations were the most adversely affected, and the profitability of the financial sector fell abruptly in 1985. The measures adopted to overcome the crisis were oriented toward reducing the solvency risk of the financial institutions, giving the government power to nationalize without compensation any financial entity undergoing severe crisis, lowering reserve requirements, and increasing interest rates on forced investment. The central bank played a primary role by providing liquidity and solvency support, giving credit to the shareholders of financial institutions and to firms,3 and relieving firms that had acquired foreign debt.4 All these policies helped to overcome the crisis, but the state ended up as the main owner of the financial system as a consequence of the whole package of measures. By

120

ARBELAEZ AND ECHAVARRIA

Financial Liberalization The country implemented a broad-based package of reforms in the early 1990s, aimed at enhancing competition, allowing the operation of foreign banks in the country, increasing reliance on market instruments, and reducing government and monetary authorities' intervention in the financial system.5 The cornerstone of the liberalization was the financial reform introduced by Law 45 in 1990, followed by Law 35 in 1993. The financial reforms covered four main fronts: interest rate policy, credit policy, forced investment, and monetary policy. The reforms liberalized interest rates on savings deposits, mortgage loans, and a large part of other loans, and limited the central bank's capacity to intervene in interest rates. Liberalization phased out credit subsidies. It converted most interest rates from fixed to variable terms at the beginning of 1987, and gradually increased rates to market levels. Ceilings remained and are still in place today. Concerning monetary policy, the Constitutional Reform of 1991 increased the autonomy of the central bank (Alesina, Carrasquilla, and Steiner 2000). The authorities also made efforts to strengthen the role of open market operations and to reduce reserve requirements, which currently average approximately 5 percent compared with 16.7 percent in January 1994. Finally, the authorities took several measures in the areas of supervision and prudential regulation to adjust the balance sheets of financial intermediaries to correctly reflect price changes and new investment in the sector. The reforms made stricter norms on provisions and nonperforming loans, and required financial entities to maintain minimum solvency ratios. The overall result was a more liberalized and better-supervised financial sector. Figure 4.1 presents the evolution of the liberalization process, using a compound index of domestic liberalization, which is based on the relation between reserves and deposits, the evolution of interest rate controls, and the imposition of Basel-type supervisory practices (Lora and Barrera 1997; Lora 2001). The domestic liberalization index appears almost like a dummy
5

The package included tax reform (Law 75 of 1986), foreign investment reform (Law 9 of 1991), labor reform (Law 50 of 1990), and social security reform (Law 100 of 1993).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

the end of the decade, the government owned approximately 66 percent of the banking sector's assets and close to half of the total financial system's assets.

FINANCIAL LIBERALIZATION IN COLOMBIA

121

Colombia, 1978-2000

a

Average of Lora and Barrera's financial liberalization index and the Morley, Machado, and Pettinato capital account liberalization index. Source: Lora and Barrera (1997); Morley, Machado, and Pettinato (1999); and authors' calculations.

variable, with a value of 0 during the 1980s and 1 in the 1990s. By contrast, Morley, Machado, and Pettinato's (1999) capital account liberalization index shows a smoother trend, with the highest slope in 1990-95. The compound index is an arithmetic average of the domestic liberalization and capital account liberalization indexes. Figure 4.2 presents the Laeven (2001) index of liberalization. It assigns a value of 0 or 1 to each of six variables: interest rates, entry barriers, reserve requirements, credit controls, privatizations, and prudential regulation. The index is the sum of the six variables; it takes a minimum value of 0 in one extreme case and a maximum value of 6 when all six variables are liberalized and prudential regulations are adopted. The pattern obtained is relatively similar to that in figure 4.1, although it suggests that the liberalization process started earlier (1988) and lasted longer.6

6

The principal laws considered for this exercise were interest rates (Decree 2994 of December 14, 1990), entry barriers (Law 45 of December 18, 1990), reserve requirements (Central Bank Resolution 7, 1993), credit controls (a series of decrees that partially eliminated forced investment and directed credit; FINAGRO's investment was regulated in 1990), and privatization. The most important prudential regulation measures were adopted in 1989.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.1. Domestic and Current Account Liberalization Indexes,

122

ARBELAEZ AND ECHAVARRlA

Note: The index is an average of Lora's financial liberalization index and the Morley, Machado, and Pettinato capital account liberalization index. Source: Lora and Barrera (1997); Morley, Machado, and Pettinato (1999); and authors' calculations.

Despite a great deal of liberalization on some fronts, a number of policies have gone in the opposite direction. As shown in figure 4.3, for example, forced investment decreased until 1992 but increased in the following years.7 The importance of directed credit decreased until 1997 but increased thereafter. FINAGRO (agriculture) and BANCOLDEX (exports) increased their directed credit in absolute terms. FOGAFIN (the financial sector guarantee fund) also exerted additional influence, particularly after 1997, when a large amount of resources were used to alleviate the pressure on the financial sector.

The only forced investment that still remains is the type A and B bonds for the agricultural sector (FINAGRO). In 2000, the Board of the Central Bank established that this forced investment would be realized as a function of the liabilities subject to reserve requirements, deducting from these the amount corresponding to required reserves. According to Hernandez and Tolosa (2001), this captures the spirit of Law 16 of 1990, whereby FINAGRO was created and investment was made compulsory inversely to the cost of the liabilities.

7

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.2. Laeven's Index of Financial Liberalization, Colombia, 1980-2000

FINANCIAL LIBERALIZATION IN COLOMBIA

123

(Percentaae of total assets)

(Percentaae of GDP)

Source: Banco de la Republics, Superintendenda Bancaria; and authors' calculations.

Size and Activity Levine, Loayza, and Beck (2000) consider activity, the stock of credit from the financial to the private sector, the best index of financial development. Gallego and Loayza (2000) use activity and size—the relation between assets of the financial sector and gross domestic product (GDP)—in their analysis of the impact of the financial sector in Chile. Figure 4.4 shows the evolution of both variables in Colombia. The financial system was relatively small during the 1980s, with size accounting for 35-40 percent of GDP. The same trend is observed in activity, with a weak expansion of credit during the 1980s and a large expansion after the financial liberalization of the 1990s. The Colombian credit boom of 1991-97 was followed by a deep contraction in the following years, when a credit crunch could have taken place (Fischer 1988; Echeverry and Salazar 1999). Financial reforms not only increased credit directly, but also indirectly through their impact on capital flows. The promotion of a larger role for

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.3. Directed Credit and Forced Investment, Colombia, 1980-2000

124

ARBELAEZ AND ECHAVARRf A

(Percent)

Source: Banco de la Republica, Superintendenda Bancaria, Ministerio de Hacienda; and authors' calculations.

foreign capital in the national financial system (in 1990 and Law 9 of 1991) could be considered among the pull factors that increased capital flows and the presence of foreign banks in Colombia.8 Other important pull factors were the package of structural reforms undertaken during the first part of the 1990s and the high interest rates of mid-1991 (39 percent for deposit rates). Calvo, Leiderman, and Reinhart (1995) and Corbo and Hernandez (2000) consider the relative importance of pull and push factors in Latin America during the 1990s. Figure 4.5 shows the close association between capital flows and domestic credit in Colombia, a relevant issue because capital flows to the country increased much faster than to other Latin American countries between 1990 and 1997, and also fell faster in 1997-2000. Calvo (2000) stresses the close association between capital flows, credit, and growth. The Crisis of 1998-2000 The financial crisis of 1998-2000 was much worse than any other crisis recorded in Colombia in recent decades, even worse than the crisis of
8

Before the liberalization of the 1990s, the entrance of foreign capital was restricted to 49 percent of the capital of the financial entity, a limit later eliminated. The external indebtedness norms for local agents were made more flexible, allowing for limit-free acquisition of debt with foreign banks.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.4. Size and Activity of the Financial Sector, Colombia, 1980-2000

FINANCIAL LIBERALIZATION IN COLOMBIA

125

Source: International Monetary Fund; Superintendencia Bancaria; and authors' calculations.

1982-85 when the financial sector was also badly bruised. The deterioration of some indicators that began in 1996, especially in the mortgage sector, was aggravated in 1998.9 The financial crisis was mainly induced by the economic recession of 1998 and 1999, the significant drop in national income, a spectacular increase in interest rates, and the crisis in the construction sector (Carrasquilla and Arbelaez 2000). Bad loans and nonproductive assets grew significantly (nonperforming loans over total loans reached 12 percent in 1999). The increase in credit risk and liquidity risk caused by deposit reduction led to a fall in credit, which reached a negative real growth rate of-7.0 percent in December 1998 and averaged -12.5 percent in 1999. Solvency deteriorated and losses increased. The recent crisis has been deeper in public banks and mortgage corporations than in domestic and foreign private banks, with bailout measures mainly oriented toward those sectors. In any case, it seems the government's package has helped to ease financial distress. Solvency has recovered
9

Mortgage corporations were hardly affected by the financial reform, which included the elimination of the liquidity fund FAVI and the exclusive role of mortgage corporations in having remunerated deposit accounts.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.5. Capital Flows and Real Credit, Colombia, 1980-2000 Annual growth rate (percent) (Millions of U.S. dollars)

126

ARBELAEZ AND ECHAVARRf A

Note: Profitability is profits/total assets. Source: Superintendencia Bancaria and authors' calculations.

and credit has stopped decreasing. As shown in figure 4.6, profitability in the financial sector dropped abruptly after 1997, with negative figures in 1998,1999, and 2000. It is no lower today than in 1985, the worst year of the 1980s, but it is low because of changes in provision policies.10 The Colombian Financial Sector and the Stock Market Today: Still Far Behind Despite the liberalization process described above, international comparisons suggest that the Colombian financial sector remains small and inefficient. Thus, Colombia is behind the Latin American average in seven of the eight financial development indicators considered in Demirgiic-Kunt and Levine (1999). Figure 4.7 presents comparisons for five of the indicators. Colombia is behind Chile and the world mean for all eight variables, and behind Brazil for five of them. These quantitative results agree with those of a recent qualitative survey the Inter-American Development Bank and Fedesarrollo conducted
10 101

Provisions were increased only later in 1999, as a preventive and prudential measure.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.6. Profitability of the Financial Sector, Colombia, 1980-2000

FINANCIAL LIBERALIZATION IN COLOMBIA

127

SOutce

among 50-100 representative business people in 10 countries.11 The results are reported in Arbelaez and Echavarria (2001). Colombia lags in most areas, from the influence of government controls to the possibility of obtaining a loan with only a good business plan and no collateral (minimum value for Paraguay). The importance of retained earnings among the sources of funds is also greater in Colombia. The available literature on the region suggests that the equity market is still more underdeveloped than the banking-financial sector (Demirgu9-Kunt and Levine 1999). Evolution and Impact of Financial Constraints in Colombia Investment Function and Main Results Following Laeven (2000), we estimate an investment equation derived from the first-order conditions of a firm's value-maximizing problem in a finanColombia, the Dominican Republic, Guatemala, Honduras, Nicaragua, Panama, Paraguay, Uruguay, Jamaica, and Trinidad and Tobago.
11

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.7. Financial Development Indicators, Colombia, 1990s (Percent)

128

ARBELAEZ AND ECHAVARRIA

where i denotes the firm; f, the year; I, investment (gross investment, absolute change in K); K, capital stock (machinery, plant, and equipment); MPK, marginal productivity of capital; FIN, a proxy for liquidity; LEV, leverage; /, a firm-specific effect; and d, a time dummy. This specification requires a measure of MPK. As in Gilchrist and Himmelberg (1998), we assume that the underlying production function is Cobb-Douglas. Under this setup, the ratio of net sales to capital can be used as a proxy for MPK. A distinctive feature of this model is that lagged investment appears as a determinant of current investment. In a framework with perfect capital markets, current investment should not depend on lagged investment, but investment ratios can show high persistence when firms make arrangements that are costly to cancel (Laeven 2001). The interest rate does not appear here because it cancels out when solving the model. This framework can test for departures from the basic Modigliani and Miller (1958) framework, where a firm's capital structure is independent of its value, internal and external funds are perfect substitutes, and investment decisions rely exclusively on expected profitability rather than financing choices. With imperfect capital markets, financing constraints may arise and internal and external funds are not usually substitutes. The nature of these could come from various sources, such as information asymmetries, costly monitoring, and contract enforcement and incentive problems. These factors can lead to explanations of why investment decisions are in practice linked to firms' value and finance, especially in firms with higher information costs. Hence, in a financially constrained environment, the signs on sales/k (proxying MPK) and liquidity in equation (4.1) should be positive. There is no a priori expected sign for leverage. A positive relation could be obtained in the estimation if firms obtained new loans (and became more indebted)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

daily constrained environment. Based on Gilchrist and Himmelberg's (1998) basic setup, we assume quadratic and persistent adjustment costs as in Love (2000), and linearize the underlying functions of the marginal productivity of capital and liquid assets to obtain an investment equation of the following form:

FINANCIAL LIBERALIZATION IN COLOMBIA

129

Gallego and Loayza (2000), Devereux and Schiantarelli (1989), and Jaramillo, Schiantarelli, and Weiss (1996) find a negative relationship between investment and debt. Harris, Schiantarelli, and Siregar (1994) find a negative relationship for small firms. Laeven (2001) does not find evidence that small firms suffer from leverage costs. 13 Their definition of cash flow (income after tax and interest plus depreciation) and liquidity (cash flow minus dividends) differs from that presented here. 14 Their definition of the variables differs from the definitions used here. They define cash flow as income after tax plus depreciation minus dividend payments and the stock of liquidity as short-term securities.

12

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

before they invested, or if debt in the past acted as a signal of creditworthiness to the financial intermediaries. Harris, Schiantarelli, and Siregar (1994) use this reasoning for the positive sign they find for large firms and conglomerates in Indonesia. Most authors find a negative sign for this relation, however, indicating that very indebted firms do not get credit easily.12 Two alternative proxies are used for FIN: the stock of liquidity (current assets minus current liabilities) and cash flow (operational profits). The best results are obtained for the stock of liquidity, suggesting that firms expecting high investment in the future will accumulate cash stock to use when opportunities arrive. Since holding cash is costly to the firm (it offers a low return), firms will accumulate cash stock only if they expect to be financially constrained in the future. This position agrees with the arguments in Greenwald and Stiglitz (1988b), the concept of financial slack in Myers and Majluf (1984), and Love (2001). However, there is no consensus in this area. Hsiao and Tahmiscioglu (1997), for example, argue that both variables are important: investment is determined by profitability considerations in the long run and liquidity is an important determinant in the short run.13 Hoshi, Kashyap, and Scharfstein (1990b) find a positive influence for both variables,14 and the results in Devereux and Schiantarelli (1989) and Harris, Schiantarelli, and Siregar (1994) are the opposite of those presented here. It is interesting to note the cross products of MPK, FIN, and LEV with the macro indexes on liberalization, size, and activity. The cross products show that the favorable development of the financial sector during the 1990s indeed decreased the financial constraints faced by the firms, or affected conglomerates and multinationals in different ways. The results are partially consistent with other studies. In particular, the liberalization process reduced financial constraints in Chile for all firms, and in Mexico and Indonesia for small firms (Gallego and Loayza 2000; Gelos and Werner 1998; and Harris,

130

ARBELAEZ AND ECHAVARRf A

The Data The information used in this chapter is provided by the Superintendencia de Sociedades and the Superintendencia de Valores (after 1995) in Colombia, with characteristics similar to the information that Jaramillo, Schiantarelli, and Weiss (1996) use for Ecuador. The Superintendencia database contains balance sheets and income statements for 8,000 to 10,000 firms reporting each year, with close to 25 percent of firms and 40 percent of sales in manufacturing. The Supervalores database contains balance sheets and income statements for close to 140 very large firms,15 42 percent of them in manuAverage sales in 1999 were 22 times larger for the firms listed (Supervalores) than for unlisted firms (Supersociedades).
15

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Schiantarelli, and Siregar 1994). Laeven (2001) finds similar results for small firms in his sample of 12 countries. This was not the case in Ecuador, however, because subsidized credit for small firms disappeared after the liberalization process (Jaramillo, Schiantarelli, and Weiss 1996). Results using the liberalization indexes are more significant than those using size or activity of the financial sector. This result in part confirms that a successful financial reform goes beyond increased credit availability but also has an impact on credit allocation. A successful financial reform removes controls on market allocation and leads to greater access to credit, reducing the premium paid on external finance. It reduces information asymmetries between borrowers and lenders and facilitates the reallocation of funds between firms. In more concrete terms, the reforms allow banks to set interest rates, abolish directed credits from official banks to preferential sectors, eliminate credit ceilings and forced lending, reduce reserve requirements, improve creditors' rights, and stimulate securities markets. Domestic financial reforms were also accompanied by capital account liberalization in Latin America. A final area of interest includes the cross products between conglomerates and multinationals with the indexes of liberalization and financial development. Those groups are less financially constrained and benefited less from liberalization. Hoshi, Kashyap, and Scharfstein (1991); Schiantarelli and Sembenelli (1995); and Cho (1995) also find important interactions between conglomerates and financial liberalization.

FINANCIAL LIBERALIZATION IN COLOMBIA

131

16 During 1978-99,1,972 firms reported information in at least some years; 484 of them were removed because of lack of sufficient information in the regression analysis. 17 Further research could shed light on the importance of different groups through time, but it is not easy to reconcile the information in the few studies available. See Superintendencia de Sociedades (1978). Slightly more is known in the area of foreign investment. Thus, the information derived from the Superintendencia de Sociedades reveals that 175 firms with (some) foreign investment in 1999, but not in 1995, represent 13.5 percent of sales. Of course, some firms disappeared: the 77 firms with FDI in 1995 but not in 1999 represent 6.7 percent of total sales. 18 The information derived from Superintendencia de Sociedades on the relative weight of firms with (some) FDI is relatively consistent with figures provided by the central bank. In 1999, the percentage of firms with FDI was 47.1 percent in manufacturing; 62.6 percent in energy, water, and gas; 60.9 percent in fishing; 44.8 percent in transport; 33.2 percent in banks and financial activities; 28 percent in services; 26.3 percent in commerce; 15 percent in agriculture, 12.6 percent in hotels and tourism; 11.9 percent in health; and 7 percent in construction.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

facturing (47 percent in sales, figures for 1999). In total, the unbalanced panel includes 1,488 firms.16 The information on conglomerates comes from a special study undertaken by the Superintendencia de Sociedades (2000). The study presents the list of 887 parents and 1,983 subsidiaries in the country in 2000; it was possible to identify balance sheets and income statements for most of the firms and work with those firms in manufacturing. It was also possible to classify groups according to their ownership (or not) of banks (here defined as firms in the financial sector in general). The information on foreign investment comes from the census available at the central bank in 1998. Unfortunately, it was not possible to track the history of each firm through time, and a firm not belonging to a conglomerate or with foreign direct investment (FDI) in previous years would be erroneously considered in those two special groups.17 Table 4.1 shows the relative weight of conglomerates and firms with FDI (in 1999, although results do not change much when other years are considered). Those firms belonging to a conglomerate represent 10.6 percent of the number of firms, and 36.9 percent of sales. Firms with FDI represent 16.4 percent of firms and 45 percent of sales,18 and the weight of firms belonging to conglomerates with and without banks is similar. Firms created before 1970 represent more than half of sales, and those created during the 1990s represent only 9.2 percent. The comparison between number of firms and sales indicates that firms belonging to conglomerates, firms with some FDI, and old firms are relatively large.

132

ARBELAEZ AND ECHAVARRf A

(Percent) Indicator Type of firm Conglomerate with bank Conglomerate without bank Nonconglomerate Total Firms with foreign direct investment Year firm was created Before 1 970 1970s 1980s 1990s Total
2.3 8.3
18.6 18.2 63.1

Number of firms

Sales

89.4 100.0 16.4 30.2 25.8 31.0 13.0 Credit history

100.0
45.0

60.5 14.0 16.3

9.2

100.0

Source: Superintendencia de Sociedades and Supervalores, firms with information in 1999.

Evolution of Key Variables Gross investment in Colombia reached its highest level in decades in 1995 and its lowest level in 1999 (figure 4.8). These were extreme levels, even compared with other Latin American countries. Contrary to the 1980s, the volatility of investment was also higher than in most Latin American counFigure 4.8. Gross Investment, Colombia and Latin America and the Caribbean, 1978-99
(Percentage of GDP)

Source: World Bank (various years).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.1. Conglomerates and Firms with Foreign Direct Investment, Colombia, 1999

FINANCIAL LIBERALIZATION IN COLOMBIA

133

(Percentage of gross domestic fixed investment)

Source: World Bank (various years).

tries.19 Figure 4.9 shows that the participation of the private sector in total investment has been increasing for decades in the average Latin American country, but not in Colombia. Figure 4.10 shows the evolution of the median values of 7/Kand //A, where A denotes total assets. The pattern does not differ much from that in figure 4.8, except that the peak occurs in 1992 instead of 1995. The value reached in 1999 is lower than in any other year in both figures. Figure 4.11 (panels a-d) shows the evolution of sales, debt, liquidity (stock of current assets minus current liabilities), and cash flow (operational profits) in manufacturing for those firms reporting each year.20 The variable S t/Kt^ decreases between 1979 and 1983, increases during the rest of the 1980s, increases between 1990 and 1992, and decreases after 1992. The pattern for S t/At_i is similar, although more stable during the 1980s. Liquidity has been decreasing over the long run, with some expansions in particular subperiods, such as 1981-85 and 1990-92. Cash flow minus profits decreased in 1983, 1986, and during the 1990s. It seems that firms' level of debt (denoted D) is much lower today than in the past;

19 The standard deviation of investment/GDP in Colombia was 0.93 and 4.09 in the 1980s and 1990s, respectively. The value for the 1980s is less than half, and the value for the 1990s is more than double that of Argentina, Brazil, Chile, Costa Rica, Ecuador, or Mexico. 20 The number of firms increased from 196 in 1979 to 474 in 1990 and 1,672 in 1999.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.9. Private Fixed Investment, Colombia and Latin America and the Caribbean, 1978-98

134

ARBELAEZ AND ECHAVARRlA

Note: I denotes investment in machinery and equipment; K, plant and equipment at the end of the year; and A, total assets at the end of the year. Values are medians for firms with information in every year. Source: Authors' calculations.

A/-Kf-i has been decreasing since 1985, especially in 1990-94. The decrease in D f /A f _j has been much less marked. There is the idea in Colombia that the real sector massively contracted debt during the 1990s, and that the recent recession of 1998-2000 was partially the result of the bubble bursting. The results do not confirm this idea; on the contrary, they show that levels of debt decreased during the 1990s.

Figure 4.11. Median Values for Manufacturing Firms, Colombia, 1979-99 a. Sales

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.10. Investment in Manufacturing, Colombia, 1979-99

FINANCIAL LIBERALIZATION IN COLOMBIA

135

b. Debt

c. Liquidity

d. Operational profits

Note: K denotes plant and equipment at the end of the year; A, total assets at the end of the year; S, sales; D, debt; Liq, liquidity, and CF, cash flow. Values are medians for firms with information in every year. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 4.11.

(continued)

136

ARBELAEZ AND ECHAVARRIA

Table 4.2 presents the median values for the relevant variables (outliers removed) for close to 900 and 1,900 firms reporting in 1992 and 1999, respectively. It shows that sales (S/K) are larger for multinationals, firms belonging to conglomerates, and firms created before 1970. Multinationals are more liquid (Liq/K) than others, and firms created during the 1990s are less liquid than others. Firms created during the 1980s have higher debt ratios (D/K) in 1999. The other differences are not statistically significant, although all variables considered in the table are larger for multinationals in 1999, and most are larger for multinationals in 1992 (except debt). Firms belonging to conglomerates have higher liquidity (L/K) and lower debt (D/K). All variables except S/K are lower for the firms created during the 1990s. All figures are lower in 1999 than in 1992, especially for I/K, a result consistent with the information provided in figure 4.11. Thus, IIKm 1999 is only 6 percent and 13 percent of the value in 1992 for conglomerates and nonconglomerates, respectively; it is 9 percent and 6 percent for firms with and without FDI, respectively; and only 5 percent for firms created during the 1990s. Nonconglomerates have the largest decrease among all groups in CF/K and Liq/K (CF denotes cash flow). The decrease in D/K was largest for firms created before 1970 and during the 1970s. Correlation Matrix Table 4.3 shows the Spearman rank correlation among the variables considered in the regression analysis for close to 17,400 observations included in the unbalanced panel data. Correlations are relatively high (greater than 70 percent) for S/K and D/K. The correlation with sales (S/K) is similar for the stock of liquidity (Liq/K) and cash flow minus operational profits. Other criteria will be needed to choose between those two variables. Estimation Techniques The dynamic investment models considered above are likely to suffer from endogeneity because investment and cash flow or liquidity could be simultaneously determined and investment may feed back into sales. In fact, most

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Firm Medians by Group

Table 4.2 Key Variables: Median Values for 1992 and 1999

1992
1999 CF/K

Indicator Liq/K
Dt/K
I/K S/K

I/K

S/K

CF/K

Liq/K
0.501
0.759 0.705** 0.849

Dt/K
2.076
1.911 1.894

Type of firm 0.493 0.388 2.665 0.026 4.424 4.274* 4.867 4.914*** 0.220 0.208 0.027 0.035 0.030 2.700 2.585 0.363 0.962 0.578 0.543 0.384 0.286 0.575 0.025 0.526 2.295 0.882 2.764 0.052 3.772 0.099

Conglomerate

0.378

15.095**

Nonconglomerate

0.389

16.157

1.072 0.186

Multinationals

Without FDI

0.384

16.018

With FDI

0.392

15.998

1.280 1.266 2.913
2.695

2.029

Year firm was created

Before 1970

0.417 1.061 2.316 0.018
4.074 3.853

17.714

0.215 0.211 0.188
0.136***

0.910
0.783 0.677 0.485***

1.930 1.841

1970s

0.341 0.031 4.179 0.218

16.044

1980s

0.341

10.746

2.000** 1.904*

1990s

0.327

33.693

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: I denotes investment; K, stock of plant, machinery, and equipment at the beginning of the year; 5, sales; CF, cash flow (operational profits); Liq, liquidity (current assets minus current liabilities); and D, total liabilities. Significance levels were obtained using a regression between the variable and dummies for each category after removing outliers. Values are for 900 firms in 1992 and 1,900 firms in 1999. Source: Superintendencia de Sociedades, Superintendencia de Valores, and authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

138

ARBELAEZ AND ECHAVARRlA

Variable
UK S/K CFIK Liq/K D/K

I/K
1

S,/JCW

CFIK

Liq/K

D/K

0.26 0.09 -0.07 -0.04

1
0.58 0.59 0.71
1

0.52 0.37

1
0.52
1

Note: Calculations are based on close to 17,400 observations. / denotes investment; K, stock of capital at the end of the period; 5, sales; CF, cash flow (operational profits); Liq, liquidity (current assets minus current liabilities); and D, total liabilities. Source: Authors' calculations based on data from Superintendencia Bancaria and Supervalores.

variables pertaining to the firm, such as output and cash flow, are potentially endogenous because they depend on the technology stock (Hayashi andlnoue 1991). Arellano and Bond's (1988) generalized method of moments technique (GMM) allows the use of lagged dependent variables and controls for unobserved individual effects and endogeneity of explanatory variables. Their methodology considers the possibility of simultaneous determination and reverse causality. The GMM estimator in differences, the technique we use in this section, uses (y,>2> yi>3> • • • yn) and (x,-jt_2, x J i t _ 3 ... XH) as instruments.21 Arellano and Bond (1988) suggest a first and second-order correlation test to assess the validity of those instruments and the Sargan test for overidentifying restrictions. First-order serial correlation is expected by construction when first differences of the variables are used, and only second-order serial correlation will be a sign of misspecification. The empirical evidence suggests that the GMM estimator in differences provides the most sensible results, but the annex system (Arellano and Bover 1995) and ordinary least squares (OLS) estimators are also reported, given the potential shortcomings of each methodology (Mairesse, Hall, and Mulkay 1999).
Only instruments lagged two, three, and four periods are used. There is not much additional information in going back further, and the tests explained below are more stringent when fewer instruments are included in the regression. Those instruments will be valid when the error term is serially uncorrelated (or at least follow a moving average process of finite order); and future innovations of the dependent variable do not affect current values of the explanatory variables, although they can be affected by the current and past realizations of the dependent variable (jointly endogenous). See Gallego and Loayza (2000).
21

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.3.

Spearman Rank Correlations

FINANCIAL LIBERALIZATION IN COLOMBIA

139

Basic Model Estimation of the basic model (equation (4.1)) uses the unbalanced panel data for 1,488 of the 1,972 listed and nonlisted firms that provided some information in 1981-99. The estimation results are from GMM in first differences with It/Kt as the dependent variable with two lags, and sales, liquidityt/Kt as the dependent variable with two lags, and sales, liquidity (or cash flow), and debt as the independent variables. We report the Wald and Sargan tests (assuming homoskedastic errors) and the p values for the first and second-order serial correlation tests. Instruments are lagged two, three, and four periods. Results are compared for two alternative definitions of liquidity: current assets minus current liabilities and operational profits. Liquidity is sometimes preferred to CF because sales and cash flow are highly correlated, but table 4.3 shows that there are no important differences between the two variables in this respect. There is no consensus in the theoretical literature in this area. The relation between sales and capital stock at the beginning of the year (S t/Kt-i) is used and, after considering different lag structures, oneperiod lagged values (beginning of period) are used for liquidity, cash flow, and debt (Liqt^/Kt_ly CFt_JKt_ly and Dt_JKt-i}. Second-order serial correlation does not seem to be a problem in the regressions, and the Sargan test suggests that the restrictions are valid. Formally, the Sargan test does not reject the null hypothesis of valid overidentifying restrictions.22 The results indicate that investment depends on lagged investment (+, net for the coefficients of/ f _i and Jf_2), sales (+), liquidity (+), and debt (+), the three variables being significant at the 1 percent level (table 4.4). The results for sales and liquidity are robust to the inclusion/exclusion of debt. The stock of liquidity (current assets minus current liabilities) gives better results than CF (operational profits). Both variables are used in models (2), (3), and (4) in table 4.4. Liquidity is highly significant and has the
Unfortunately, the values reported for the Sargan test using first differences are too good. That is, they equal 1 in all the tables, a problem that could be related to over-fitting bias. The results did not change when using two and three alternative maximum lags for the instruments. The problem disappears, however, when system estimators are used, as in the annex in Arbelaez and Echavarria (2001).
22

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Empirical Results: Financial Constraints and Liberalization

140

ARBELAEZ AND ECHAVARRlA

Variable
/f,/Kf,
H2/Kt2

(D
-6.010E-05 -(2.46) 1.127E-04 (5.69)*** 0.030 (9.34)***
0.051

(2)
-6.500E-05 -(2.69)*** 1.169E-04 (6.44)*** 0.033 (12.92)*** 0.054 (6.38)***

(3)
-5.180E-05 -(2.23)*** 1.131E-04 (5.99)*** 0.034 (11.24)***

(4)

-5.540E-05 -(2.20)*** 1.192E-04 (6.07)*** 0.030 (9.36)*** 0.055 (5.66)***

St/KM (Liq/K)t_i

(6.22)***
(CTWOr-i

-0.005 -(0.23) 0.022 (2.61)***
X X

-0.025 -(1.17)
0.021

(D/K), D_year Number of observations Number of firms Wald test of joint significance Specification tests (p values) Sargan test First-order serial correlation Second-order serial correlation

0.025 (3.00)***
X

(2.62)***
X

5,377
1,488

5,377
1,488

5,377
1,488

5,377
1,488

733.9

733.76

786.7

798.51

1.0

1.0

1.0

1.0

0.000 0.285

0.0001 0.2409

0.000 0.345

0.0001 0.336

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: The model is estimated following Arellano and Bond (1988) in first differences with two lags. The dependent variable is I/K,. I denotes investment; K, total assets in machinery and equipment at the end of the year; 5, sales; Liq. liquidity (current assets minus current liabilities); CF, cash flow (operating profits); and D, debt. Outliers were excluded. Heteroskedasticity-consistent t values are in parentheses; the constant is not reported. The Sargan test was calculated for homoskedastic errors. The maximum number of lags allowed for the predetermined variables used as instruments was 4. Source: Authors' calculations based on data from Superintendencia de Sociedades and Supervalores.

expected positive sign, but CF is not significant and has the wrong sign. The following sections use the specification in model (1) as the best model. Loglog regressions are run for this model with elasticity 1.3 for sales, 0.16 for the stock of liquidity, and 0.36 for debt. The standardized beta coefficients are 9.3 for sales, 6.2 for liquidity, and 2.6 for debt. The system and OLS estimators in the appendices in Arbelaez and Echavarria (2001) are similar to those in table 4.4, except for the negative (and significant) sign on D/K. However, the sign becomes positive when

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.4. The Basic Model

FINANCIAL LIBERALIZATION IN COLOMBIA

141

The Impact of Financial Development and Reforms Table 4.5 adds the cross products with the macroeconomic indexes of financial development. The results for sales and liquidity are consistent with those of the previous section, but the coefficient for D/K is no longer significant. The cross products indicate that financial and debt constraints decreased with financial liberalization.23 The results are similar for the four financial variables used, but liberalization indexes produce better results when additional variables are included, as is done below. Regressions with size and activity for the stock market (not shown) have the expected poor results, given the precarious development of the stock market in Colombia. The Wald, Sargan, and second-order serial correlation tests are satisfactory.24 System and OLS estimators are consistent for most coefficients and D/K is (again, as in table 4.4) significant (Arbelaez and Echavarria 2001). The cross products between debt and the macro financial variables are significant and have the correct signs for the system estimators and also for Lora and Laeven (but not for size or activity) when OLS is used. However, the size of the coefficients in table 4.5 indicates that financial liberalization had an important impact, larger for the liberalization indexes (Lora and Laeven) than for size or activity. Thus, the coefficients indicate that financial restrictions decreased 70 percent (Lora, column (3)) and 52 percent (Laeven, column (4)) during the liberalization episode in 1990-97.25 The

23

It would be more rigorous to refer to a liquidity constraint and a debt premium. The coefficient of debt is positive, suggesting that indebted firms get easy credit because of their good history with the bank. 24 However, for the Sargan test see footnote 23. 25 Thus, the results in column (4) in table 4.5 (Laeven) indicate that the liquidity constraint changed from 0.0511 to 0.02432, a reduction of 52.44 percent; 0.0511 + 4*(-0.0067) = 0.02432. The Laeven index changes 4 units between 1990 (2) and 1997 (6). For the other calculations, we use the fact that size changed 0.22 points (from 0.384 in 1990 to 0.606 in 1997), activity changed 0.112 points (0.264 in 1990,0.376 in 1997), and Lora's index changed 0.744 points (0.203 in 1990, 0.947 in 1997). See figures 4.1 and 4.2.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

additional variables are included. OLS estimators also shift coefficients for D/Ky with the additional difference that cash flow produces better results than liquidity.

142

ARBELAEZ AND ECHAVARRf A

Variable

(D
-4.27E-06 -(0.20) 1.13E-04 (7.19)*** 0.043 (12.70)*** 0.048 (6.05)*** -0.068 -(5.24)***

(2)
-5.87E-06 -(0.27) 1.19E-04 (7.48)*** 0.043 (12.69)*** 0.048 (6.07)***

(3)
2.96E-05 (1.28) 1 .34E-04 (7.37)***
0.041

(4)
1.63E-05 (0.70) 1.27E-04 (7.98)*** 0.039 (11.68)*** 0.051 (6.29)***

/V/cr,
lt-2/Kt-2

St/KM (Liq/K)^ Liqt_i/Kt-i x Fin_size /./qt-i/Kf-1 x Fin_activity Liqt-t/Kt^ x Finjiberalization (Lora) Liqt_i/Kt--\ x Finjiberalization (Laeven) (D/K), D/K,xFin_size D/K,xFin_activity D/K,x Finjiberalization (Lora) D//Cfx Finjiberalization (Laeven) D_year Number of observations Number of firms Wald test of joint significance Specification tests (p values) Sargan test First-order serial correlation Second-order serial correlation

(12.12)*** 0.050 (6.07)***

-0.098 -(4.96)
-0.46 -(5.47)***

-0.007 -(5.25)*** 0.005 (0.65) -2.04 -(9.85)*** -0.301 -(10.13)*** -0.137 -(8.32)*** -0.020 -(7.78)
X 5,377
X

0.005 (0.62)

0.009 (1.07)

0.012

(1.49)

5,377

X 5,377
1,488

X 5,377
1,488

1,488
1,233.6
1.0 0.0000

1,488
1,221.1
1.0

662.22
1.0

699.41
1.0

0.000 0.7261

0.0001 0.5825

0.7369

0.0001 0.3254

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: The model is estimated following Arellano and Bond (1988) in first differences with two lags. The dependent variable is I/K,. I denotes investment; K, total assets in machinery and equipment at the end of the year; 5, sales; Liq, liquidity (current assets minus current liabilities); D, total debt; fin_size, total assets of the financial sector/GDP; and fin_activity, stock of credit from the financial sector to the private sector/GDP. For finjiberalization, see Lora and Barrera (1997) or Laeven (2001). Outliers were exluded. Heteroskedasticity-consistent t values are in parentheses; the constant is not reported. The Sargan test was calculated for homoskedastic errors. The maximum number of lags allowed for the predetermined variables used as instruments was 4. Source: Authors' calculations based on data from Superintendencia de Sociedades and Supervalores.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.5. The Effect of Liberalization and Financial Market Development on Financial Constraints

FINANCIAL LIBERALIZATION IN COLOMBIA

143

The Crisis of 1998-99 Liberalization and financial development reduced the constraints faced by firms, and it is interesting to see whether those constraints increased again during the deep financial crisis of 1998-99. Table 4.6 adds to the basic model the cross products Liqt/Kt • D98_99 and Dt/Kt • D98_99, where D98.99 takest/Kt • D98_99 and Dt/Kt • D98_99, where D98.99 takes the value 1 during 1998 and 1999. As expected, the coefficients are significant and positive, suggesting that financial constraints did indeed increase during the financial crisis of 1998-99. The results for liberalization (column (1) in table 4.6) seem to be similar to those for activity (column (2)). The results for lagged investment, sales, and liquidity are consistent with those of the previous sections, as are the cross products between liquidity and debt with liberalization. Dt/Kt is not significant. However, the size of the coefficients indicates that financial constraints increased 54 percent in 1998-99, with almost identical results for the Laeven index and for activity.27 Empirical Results: Conglomerates and Multinationals Table 4.7 adds the effect of conglomerates to the variables included in the basic model of table 4.4. The analysis uses a dummy variable Dcongiom (which takes the value 1 when the firm belongs to a conglomerate in 2000 and 0 otherwise) and cross products between the previous variables and the dummy. The Laeven liberalization index and activity are considered in the

26

The system estimators of table A.4 in Arbelaez and Echavarria (2001) indicate that financial restrictions decreased 56 percent (Lora), 41 percent (Laeven), 16.3 percent (size), and 26 percent (activity). 27 In column (2) in table 4.6, the new coefficient is calculated as 0.0426 + 0.0232, and in column (1) as 0.0454 + 0.0249. As noted in Arbelaez and Echavarria (2001), the results of system estimators in table A.6 in that work are consistent for most coefficients and D/K is (again, as in table 4.4) significant, but the OLS estimators in table A.7 are less satisfactory for some variables and cross products related to liquidity.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

reduction is smaller but still significant for size (30.9 percent) and activity (22.8 percent).26

144

ARBELAEZ AND ECHAVARRIA

Variable
/t-i/Kf-1
lt-2/Kt-2

(D
-1.56E-05 -(0.56) 1 .34E-04 (2.99)*** 0.042 (11.91)*** 0.045 (5.77)***

(2)
-3.16E-05 -(1.12) 1.26E-04 (2.42)** 0.045 (12.77)*** 0.043 (5.61)*** -0.083 -(4.41)***

SI/*,., /./qrM/Kf-i Liqt--\/Kt-i x Fin_activity /JO/M/KM x Finjiberalization (Laeven) i/gM/Kt-i x D_9S99 (D//C)f D,//(t x Fin_activity D/K, x Finjiberalization (Laeven) DJK, x D_9S99 D_year Number of observations Number of firms Wald test of joint significance Specification tests (p values) Sargan test First-order serial correlation Second-order serial correlation

-0.006 -(4.90)*** 0.025 (2.86)*** 0.006 (0.81) 0.023 (2.67)***
0.001

(0.09) -0.299 -(10.15)***

-0.020 -(7.83)*** 0.030 (3.43)***
X

0.022 (2.65)***
X

5,377
1,488

5,377
1,488

827.85

903.51

1.0

1.0

0.0001 0.1838

0.0000 0.5349

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: The model is estimated following Arellano and Bond (1988) in first differences with two lags. The dependent variable is IJKt. I denotes investment; K, total assets in machinery and equipment at the end of the year; 5, sales; Liq, liquidity (current assets minus current liabilities); D, total debt; fin_size, total assets of the financial sector/GDP; and fin_activity, stock of credit from the financial sector to the private sector/GDP. D_9899 is a dummy variable that takes the value 1 for years 1998 and 1999. For finjiberalization, see Laeven (2001). Outliers were excluded. Heteroskedasticity-consistent t values are in parentheses; the constant is not reported. The Sargan test was calculated for homoskedastic errors. The maximum number of lags allowed for the predetermined variables used as instruments was 4. Source: Authors' calculations based on data from Superintendencia de Sociedades and Supervalores.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.6. The Crisis of 1998-99

FINANCIAL LIBERALIZATION IN COLOMBIA

145

Variable
UKt-,
lt-2/Kt-2

(D
6.3200E-06 (0.28) 1.2580E-04 (8.00)*** 0.042 (12.02)*** 0.051 (6.40)***

(2)
-7.21E-06 -(0.34) 1.19E-04 (7.02)*** 0.043 (12.57)*** 0.047 (5.96)*** -0.069 -(1.36)

SI/KM
(Liq/K)^ Liqt-\/Kt-\
x

Fin_activity -0.011 -(3.42)*** -0.027 -(2.62)*** 0.009 (2.44)***

I/C/M/KM x Finjiberalization (Laeven) /./C/M/KM x D_conglom /./qrt_i//C,_i x Finjiberalization x Djconglom /JO/M/KM x Fin_acti x D_conglom (D/K) t D t/Kt x Finjiberalization D,/Kt x Fin^acti Dt/K,D_conglom Dt/Kt x Finjiberalization x D_conglom Dt/Kt x Fin_acti x D_conglom D_year Number of observations Number of firms Wald test of joint significance Specification tests (p values) Sargan test First-order serial correlation Second-order serial correlation

-0.009 -(0.21)

0.009 (1.06) -0.033 -(3.58)***

-0.006 -(0.04) 0.005 (0.67)

-0.071 -(9.60)*** 0.026 2.71***

-0.363 -(3.90)*** -.006 -(0.15)

X 5,377 1,488
810.67

0.088 (0.61) X 5,377 1,488 1,302.17

1.0 0.0001 0.8564

1.0 0.0000 0.6323

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: The model is estimated following Arellano and Bond (1988) in first differences with two lags. The dependent variable is It/Kt. I denotes investment; K, total assets in machinery and equipment at the end of the year; 5, sales; Liq, liquidity (current assets minus current liabilities); D, total debt; fin_size, total assets of the financial sector/GDP; and fin_activity, stock of credit from the financial sector to the private sector/GDP. Djconglom is a dummy variable that takes the value 1 when the firm belongs to a conglomerate in 2000. For finjiberalization, see Laeven (2001). Outliers were excluded. Heteroskedastidtyconsistent t values are in parentheses; the constant is not reported. The Sargan test was calculated for homoskedastic errors. The maximum number of lags allowed for the predetermined variables used as instruments was 4. Source: Authors' calculations based on data from Superintendencia Bancaria and Supervalores.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.7. The Basic Model with Conglomerates

146

ARBELAEZ AND ECHAVARRf A

Conglomerate Finances and Firm Investment Do the finances of the conglomerate affect investment at the firm level? Do firms invest more when their conglomerate is more liquid or has a good history of debt? Liquidity (current assets minus current liabilities) and debt (total liabilities) were calculated for all firms, in manufacturing and in other sectors, belonging to each of 14 conglomerates reported in Dinerv in 1998 (table 4.8).28 Two conglomerates, Sindicato Antioqueno and Bavaria, account for almost half of the number of firms and close to 60 percent of sales (in 1998) of the 14 groups. Table 4.9 presents the basic model from table 4.4, adding (Liqt^/ Kt-Jcongiom and (AV^-iWom> liquidity, and debt for the whole conglomerate. The results indicate that the firm invests more when the liquidity of the whole conglomerate increases, but we did not obtain significant values
Those firms were eliminated that did not have information on liquidity and debt for all years in 1978-99; six of the 20 groups considered by Dinero were therefore excluded.
28

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

regressions. Liberalization seems to produce better results for capturing the impact of financial development than does the amount of credit activity. The coefficients and levels of significance of lagged investment, sales, and liquidity are similar to those discussed above, as are the cross products between liquidity and financial liberalization. D/K has the correct positive sign but is not significant. The evidence indicates that conglomerates are less financially constrained and less debt constrained than other firms since Liq t/Kt • Dcong}om and Dt/Kt • Dcmgiom are negative and significant. It also suggests that conglomerates benefited less from financial liberalization since the triple cross products are positive and significant. The size of the coefficients indicates that a conglomerate is 53 percent (column (1) in table 4.7) or 19 percent (column (2)) less financially constrained than a nonconglomerate firm. The Wald and Sargan tests are satisfactory (high): the group of variables explains the behavior of investment and the identifying restrictions are valid. The null hypothesis of no second-order serial correlation cannot be rejected. Results for system estimators and OLS for the cross products between liquidity and debt with Dcongiom are not encouraging. The signs are as expected, but the coefficients are not significant.

FINANCIAL LIBERALIZATION IN COLOMBIA

147

Conglomerate Sindicato Antioqueno Bavaria Ardila Sanford Chaid Neme Hermanos S.A. Mundial Corona El Tiempo Haime Aval Olimpica Coca Cola Cafetero Lloreda Total

Number of firms
67 36 26 19 11
9 9 8 7 7 5 3 3 2
212

Sales (percentage of total)
31.2 27.3 13.1

5.8 0.7 3.8 2.1 1.5 1.6 0.9 5.3 3.8 1.7 1.1
100.0

Source: Dinero, Superintendence de Sociedades, Superintendencia de Valores, and authors' calculations.

for debt. The results are consistent with the previous findings related to the lower constraints faced by firms belonging to conglomerates. Foreign Direct Investment Little work has been carried out in this area, although it is traditionally assumed that the domestic affiliates of multinationals can partially use the resources of their parent company. The results in table 4.10 corroborate such a hypothesis, with coefficients relatively similar to those obtained for conglomerates. A dummy variable takes the value 1 when the firm has some foreign investment in 1998 and 0 otherwise. Again, the discussion refers mainly to the results using a Laeven index of liberalization (column (1)), which are more solid than those of activity (column (2)). The coefficient of Lt/Kt • Ddfi is negative and significant at the 5 percent level in table 4.10, and the coefficients of the triple products Liqt/Kt • Finiiberataation * Ddfi and Dt/Kt • Finiiherdization - Ddfi are positive and highly significant. This suggests that domestic affiliates face lower financial constraints than the average firm, but that they also benefited less from financial liberalization. However, the size of the coefficients indicates that a firm with

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.8.

Conglomerates: Sales and Number of Firms, 1998

148

ARBELAEZ AND ECHAVARRlA

Variable
(/M/Kr-i)/

(D
-.003
-(0.07) -0.012 -(0.37) 0.036 (5.05)*** 0.054 (3.11)***

(2)
-0.010 -(0.20) -0.014 -(0.43) 0.035 (4.76)*** 0.046 (2.90)*** 0.000 (2.10)***

(UKt-2)i (S«/Kt-i)/ (Liqt.i/K^)iI (/./q;M//CM)/ conglom (Dt/Kt)i

0.054 (3.47)***

0.052 (3.45)*** -8.85E-07 -0.02)

(D//Ct)/ conglom
D-year Number of observations Number of firms Wald test of joint significance Specification tests (p values) Sargan test First-order serial correlation Second-order serial correlation * Significant at 10 percent. ** Significant at 5 percent.
0.91

X
588

X
588

94
206.82

94
260.31

0.92

-4.11
0.15

-4.11
0.35

*** Significant at 1 percent. Note: The model is estimated following Arellano and Bond (1988) in first differences with two lags. The dependent variable is l,/Kt. I denotes investment; K, total assets in machinery and equipment at the end of the year; 5, sales; Liq, liquidity (current assets minus current liabilities); D, total debt; /, a single firm; and conglom, all the firms in a particular conglomerate. Information on liquidity and debt for all the firms in each conglomerate was added to the basic model. Outliers were excluded. Heteroskedasticity-consistent t values are in parentheses; the constant is not reported. The Sargan test was calculated for homoskedastic errors. Source: Authors' calculations based on data from Dinero on the 20 largest conglomerates in Colombia.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 4.9. The Effect of Increased Conglomerate Liquidity or Debt on Investment

FINANCIAL LIBERALIZATION IN COLOMBIA

149

Table 4.10. Foreign Direct Investment

/t^ATf,
lt-2/Kt-2

(S/KM) (Liq/lQt-, Liqt-i/Kt-i x Fin_activity /./O/M/KM x Finjiberalization /./ovVKf-i x D_DFI Liqt-i/Kt-i x Fin_activity x D_DFI Liqt--\/Kt-i x Finjiberalization x D_DFI (D/K)< D//Ct x Fin_activity Dt/Kt x Finjiberalization
D//C, x D_DF/

3.14E-05 (1.23) 1 .34E-04 (7.66)*** 0.041 (12.52)*** 0.052 (6.30)***

4.24E-06 (0.19) 1 .20E-04 (7.29)*** 0.043 12.91)*** 0.047 (5.97)*** -0.109 -(4.71)***

-0.008 -(5.52) -0.024 -(1.85)**

0.005 (0.11) 0.032 (0.25)

0.009 (2.97)*** 0.012 (1.46)

0.006 (0.71) 0.337 -(8.82)***

-0.023 -(7.55)*** -(0.07) -(7.14)***

D/Kf x Fin_activity x D_Df/ D/Kt x Finjiberalization x Of/ D_year Number of observations Number of firms Wald test of joint significance Specification tests (p values) Sargan test First-order serial correlation Second-order serial correlation 0.024 (5.36)*** X 5,377
1,488

-0.001 -(0.04) 0.135 (1.15)

X

5,377
1,488

748.38
1.0 0.0001 0.3504

1,341.16
1.0 0.0000 0.6383

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: The model is estimated following Arellano and Bond (1988) in first differences with two lags. The dependent variable is l,/Kt. I denotes investment; K, total assets in machinery and equipment at the end of the year; 5, sales; Liq, liquidity (current assets minus current liabilities); D, total debt; fin_size, total assets of the financial sector/GDP; and fin_activity, stock of credit from the financial sector to the private sector/GDP. D_DFI is a dummy variable that takes the value 1 if the firm has some foreign investment (registered in the central bank) in 1998. For finjiberalization, see Laeven (2001). Outliers were excluded. Heteroskedasticity-consistent standard errors are in parentheses; the constant is not reported. The Sargan test was calculated for homoskedastic errors. The maximum number of lags allowed for the predetermined variables used as instruments was 4. Source: Authors' calculations based on data from Superintendencia Bancaria, Supervalores, and the Central Bank.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Variable

(D

(2)

150

ARBELAEZ AND ECHAVARRlA

Conclusions Financial constraints play an important role in investment in the developed economies where the financial sector is strong and diversified, and should play a larger role in some less developed countries, where the financial sector remains behind. Firms are indeed financially constrained in Colombia, where they are limited by the availability of external funds despite the important reforms undertaken during the 1990s. The liberalization process of the 1990s decreased liquidity and debt requirements for investment, but constraints increased again during the recent crisis of 1998-99. There is not an important relationship between operational profits minus cash flow and investment; instead, firms build a stock of liquidity before investment takes place. The amount of credit irrigating the economy is an important variable, but the relevant story is much more complex. In particular, the liberalization indexes seem to capture the whole picture. A successful financial reform removes controls on market allocation and leads to greater access to credit, reducing the premium paid on external finance, reducing information asymmetries between borrowers and lenders, and facilitating the reallocation of funds between firms. As expected, there is strong evidence that firms belonging to conglomerates and multinational firms are less financially constrained. The conglomerate is an organization partially designed to cope with information and contract enforcement problems, and firms belonging to a conglomerate are less likely to be financially constrained; they can rely on the financial resources of the group. Multinational firms can use resources of the parent company and should be less constrained when they want to invest in new machinery and equipment. As expected, these two groups of firms benefited less than the average firm from financial liberalization.

As noted in Arbelaez and Echavarria (2001), however, system estimators (table A.10) and OLS (table A.ll) are not as encouraging. The coefficients for Liq^/K^ x D_DFI and Liq,_i/ £,_! x Fin_Liberalization x D_DFI have the wrong signs, and are significant in some cases.

29

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

some foreign direct investment faces liquidity restrictions 46 percent lower than other firms.29

The Effects of Credit Constraints on Costa Rican Manufacturing Firms
Alexander Monge-Naranjo and Luis J. Hall
Economists are familiar with the notion that credit constraints and other credit market imperfections may severely limit the investment and operations of firms. Credit constraints limit the size of firms, as well as their growth, profits, start-up, and liquidation; their scope of operations may also be limited. Understanding the implications of credit constraints is of first-order importance for the performance of aggregate economies, especially for developing economies because capital market imperfections can impair the aggregate accumulation of capital, the rate of return on investments, and innovation. This chapter investigates the existence, determinants, and consequences of credit constraints for firms operating in Costa Rica. Although the existence of credit market imperfections maybe self-evident, the chapter aims to empirically examine their nature and relevance. The analysis uses data from a survey of a relatively large sample of manufacturing firms operating in metropolitan areas in Costa Rica. The survey questions covered firms' current finances as well as their sources of funds at the time they were established. On the basis of these data, the chapter explores the relationship between a firm's finances and its characteristics and performance. Monge, Cascante, and Hall (2001) explore institutional arrangements and banking practices in Costa Rica for enforcing financial contracts. That study documents a rather sophisticated information network among lenders, finding that banks seem to actively screen and keep track of the projects they finance. Interestingly, banks use the value and liquidity of the collateral posted by the entrepreneur as a key criterion for granting credit. In fact,
Alexander Monge-Naranjo is a professor of economics at Northwestern University, and Luis J. Hall is a professor of economics at the Universidad de Costa Rica and New York University.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 5

152

MONGE-NARANJO AND HALL

1 Bank credit includes credit from savings and loan cooperatives and other quasi-bank financial intermediaries.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

collateral plays a crucial role in the interaction between the creditor and the borrower, including cases of default and their resolution via civil courts. Monge, Cascante, and Hall also find that previous experience with borrowers represents another major determinant of banks' decisions to grant credit. Their results suggest some of the main reasons why some entrepreneurs may not receive credit. This chapter concentrates on firms and examines differences in the sources of funds for firms with different characteristics. Among other things, this information can further an understanding of the importance of internal versus external finance as well as the different sources of external finance, in particular, the importance of formal (bank) versus trade credit and informal credit for firms with different characteristics. Especially relevant is the question of which factors determine whether a firm has access to formal financial markets. It is well known that, typically, credit from formal institutions is less expensive than credit from informal creditors or commercial partners. We apply simple and standard econometric methods (probits and Tobits) to the data, with the goal of determining whether any of the characteristics of the firm or the entrepreneur determine access to bank credit.1 This question is examined not only in relation to firms' current finances, but also for their reported finances at the time they started operating. The survey also asked firms to provide measures of their performance. That information makes it possible to adopt econometric methods from the literature on treatment effects in order to assess the effect of access to bank financing on the performance and behavior of firms. However, it is necessary to address a key econometric problem: The characteristics of firms that determine their performance may also determine their access to credit from banks. It would be misleading to simply run an ordinary least squares regression including firms with and without access and estimate the effect of bank credit from the difference in the average performance measure. We use two methods to correct for potential selection bias. The first is the widely used two-step estimator developed by Heckman (1974, 1979). The method consists of first estimating the probability of access to credit and then using the predicted value to correct a regression on the performance of the firm.

COSTA RICAN MANUFACTURING FIRMS

153

Credit Constraints and Firm Behavior Early theoretical models of entrepreneurship assume directly that credit contracts for business start-ups and ongoing financing are limited. For example, in Bernhardt and Lloyd-Ellis's (2000) model, there are no credit possibilities. In their economy, the operation and formation of firms have to be funded by entrepreneurs' accumulated savings and firms' past profitability. In other models, the maximum credit agents can obtain to fund their productive ventures is modeled as a direct function of wealth or available collateral. Examples of those models are Evans and Jovanovic (1989), Hart and Moore (1994), and Banerjee and Newman (1991). Some of these models allow trade credit, that is, funds that are backed by the goods supplied. More recent studies are much more explicit on the way credit markets work and on the role of private information, contract enforcement, and

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

The second method is less well known, but its application in economics and other social sciences is growing rapidly. This is a nonparametric method, which was developed by Manski (1995) and Manski and Horowitz (1995). The method consists of estimating the worst-case and best-case scenarios of the effect of access to bank credit on the performance of firms. The second estimator is more robust but typically less conclusive (less statistically efficient given correct functional form assumptions) than the parametric methods. With the data at hand, the results obtained seem to indicate that access to bank credit indeed affects firms' behavior, and it appears that the effect is stronger on young firms. It is important to highlight the limitations of the chapter. Unfortunately, availability of data on firms is the major limitation in Costa Rica. That is precisely why the main task for this study consisted of collecting the data. However, we were able to compile only a cross-sectional database, with some retrospective questions on previous dates. The results hinge on the cross-sectional variation of active firms to identify the effect of credit constraints. The lack of panel data makes it impossible to apply generalized method of moments estimations to test credit constraints on the investment of firms. Those methods have been discussed and applied with relative success by authors including Jaramillo, Schiantarelli, and Weiss (1996) and Schiantarelli (1996).

154

MONGE-NARANJO AND HALL

2 See, for example, Banerjee and Newman (1991) and Lehnert (1998).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

renegotiation in shaping the form of contracts and access to lending. Moral hazard is the incentive problem that has received the most attention. If the bank takes too much of the project's returns, it might not be in the best interest of the borrower/entrepreneur to exert much effort or care. However, rational lenders would foresee the borrowers' poor incentives and consequently restrict lending. In general, incentive problems can affect the operation of active firms and not only the establishment of firms. For example, some models predict that because of the incentive problems, firms with different net worth will choose different technologies. Agents who do manage to borrow, as compared with those that rely exclusively on savings, may choose technologies or activities with lower variance but lower mean returns. For example, Monge (2001), Morduch (1995), Stiglitz and Weiss (1981), and Lehnert, Ligon, and Townsend (1999) all present variations on this argument. Another branch of the literature focuses on limited contract enforceability as the origin of credit constraints. Dynamic general equilibrium models with limited contract enforcement have been successfully applied for asset prices by Kehoe and Levine (1993) and Alvarez and Jermann (2001), for consumption by Krueger and Perri (2001), for international capital flows by Kehoe and Perri (2000), for human capital accumulation by Lochner and Monge (2002), and for firm and job creation and destruction by Monge (2001). In the context of firm financing, in the models by Hart and Moore (1994), Albuquerque and Hopenhayn (2001), and Ligon, Thomas, and Worrall (2001), the temptation to renege imposes limits on credit. Because the temptation to repudiate and default is a direct function of the net worth of the firm, those models provide explicit predictions on the links between firm age and size and growth, survival, and profits, as well as the dividends distributed to owners. These kinds of obstacles to the smooth operation of credit markets can make a difference in occupational choice, and therefore to small firms' levels of activity, success, and growth. The level of inequality, the overall rate of growth, and the level of employment are all functions of the nature of credit markets. Thus, improvements in credit markets could have beneficial implications for growth, employment, and the distribution of income.2

COSTA RICAN MANUFACTURING FIRMS

155

Bank Behavior and the Allocation of Credit in Costa Rica Before reviewing the information on firm behavior, it is convenient to review the results in Monge, Cascante, and Hall (2001) on bank practices in Costa Rica. That work studies the interaction of banks in all the stages of the lending relationship: analysis and approval criteria for loan applications, contractual terms, control, follow-up and enforcement, as well as renegotiation in cases of default. These findings are from the point of view of the banking institutions; we obtained information from a detailed questionnaire submitted in 1998 to a sample of intermediaries. Monge, Cascante, and Hall find significant differences in the default rates of financial intermediaries. On the one hand, production activities in Costa Rica are heterogeneous and financial alternatives are diverse, as there is a large variety of financial intermediaries.3 Traditionally, public institutions dominated the allocation of credit as part of the political-economic model. But the waves of liberalization of the 1980s and 1990s and structural changes in the economy have given more room to private banks and intermediaries. Indeed, manufactures and services, the fastest-growing sectors, have relied more on private financing, while public banks remain more specialized in agricultural sectors. Equilibrium in the credit market determines which types of borrowers obtain credit from which type of lender. Such matching can be vitiated by adverse selection, explaining part of the differences in the performance of banks. Yet, the ultimate determinant of the differences must be found in the credit policies of intermediaries; the evidence on Costa Rica compiled
3

Formal institutions are composed of three commercial public banks, 23 private banks, 35 savings and loan cooperatives, 17 nonbank private financial companies, nine housing mutual funds, and three other intermediaries created under special laws.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Furthermore, in a world where small firms are innovators, limitations in the allocation of credit could severely impair the ability of the whole economy to adopt new technologies and economic activities. All of the many different incentive problems emphasized by theory may be of relevance in practice, and this agnostic attitude will guide the interpretation of the findings in this chapter.

156

MONGE-NARANJO AND HALL

The Manufacturing Sector in Costa Rica We selected the manufacturing sector for several reasons. First, the importance of this sector has remained stable and indeed has increased lately. Second, the available data are better for this sector than for agriculture, services, or commerce. Third, the sector is largely located in metropolitan areas,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

by Monge, Cascante, and Hall (2001) contradicts the commonly held view of banks as dormant lenders. In fact, the banks in the sample are active, and measures such as visiting firms, analyzing projects, and increasing customer incentives and capacity for repayment are not uncommon. The high frequency of these actions may be due to the regulation of the superintendency of banks and the reserves that banks must hold to meet risk qualification criteria for loans. Banks pay particular attention to entrepreneurs' collateral. The overwhelming majority of banks make some assessment of the existence, type, market value, and liquidity of collateral, and unsecured lending is almost nonexistent. Moreover, the reputation of the borrower is a key element in evaluating an application. Monge, Cascante, and Hall (2001) asked about the importance of a variety of criteria in deciding whether to grant a loan, and all elements related to the warranty put on the project and the solvency and references of the borrower play a critical role. Monge, Cascante, and Hall also find that banks look for information on the entrepreneur from alternative sources. Indeed, the use of credit bureaus is widespread, although banks in metropolitan areas have a greater need for those references, while banks in rural areas have more first-hand information on creditors. The operational characteristics of credit bureaus indicate the existence of a sophisticated information network. Monge, Cascante, and Hall's findings support the relevance of a variety of incentive problems at different stages of the bank-entrepreneur relationship. Banks devote resources to scrutinizing applications, controlling the development of ongoing projects, and enforcing contracts, even at the level of the courts. This chapter attempts to complete the picture by using survey data on a large set of firms and information on their behavior and financing characteristics. We begin with some background on the manufacturing sector in Costa Rica.

COSTA RICAN MANUFACTURING FIRMS

157

Table 5.1. Economic Participation of the Manufacturing Sector, Costa Rica, 1985-99 (Percent) Indicator Share Share Share Share in GDP in total employment in total exports in exports out of zonas francas
1985
21.9 15.6 28.1
— —

1990
21.5 18.0 36.5


1994
21.8 17.9 45.8 29.4

1997
21.3 15.6 47.1 33.8

1999
27.2 15.7 70.7 36.8

(export processing zones) Number of formal enterprises 4,463 4,629 5,069 4,884 — Not available. Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

which made it possible to obtain a representative sample with the limited resources available. Fourth, private institutions dominate bank credit for manufacturing firms. This suggests that, if there are constraints on credit, they are less likely to be affected by political considerations. Working with this sector is also convenient because the results are likely to be relevant for the future. Manufactures and private intermediaries are bound to increase their relative importance in Costa Rica. The manufacturing sector has a strong presence in the Costa Rican economy, as shown in table 5.1. During the 1990s, the sector averaged approximately 21 percent of GDP and 16 percent of the labor force (table 5.1). The importance of the manufacturing sector in exports is not only significant but has increased and is expected to continue to increase over time. This is true even excluding the firms with special tax treatments (zonasfrancaSy export processing zones). The chapter analyzes information from the Registry of the Costa Rican Social Security Fund (Caja Costarricense de Seguro Social, or CCSS) for May 2000. We use a stratified random sampling method to extract a representative group of firms with different sizes and levels of economic activity. As shown in figure 5.1, small firms are a highly relevant segment in the manufacturing sector: of all active firms in 2000, more than 70 percent have at most five employees. There are few large firms. Moreover, as figure 5.2 indicates, small firms are important in terms of employment. Roughly speaking, firms with fewer than 10 employees account for 10 percent of total manufacturing employment; those with at most 50 employees account for more

158

MONGE-NARANJO AND HALL

(Number of firms)

Number of employees in the firm

Note: The total number of manufacturing firms was 4,884 in 2001. Source: Authors' calculations based on survey data.

Figure 5.2 Employment in Active Manufactrung Firms in Costa Rica, 2000

Number of employees in the firm

Note: Total employment in manufacturing firms was 138,311 in 2001. Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

figure

COSTA RICAN MANUFACTURING FIRMS

159

than 30 percent. These figures are significantly higher than those for developed economies (Ulate 2000; Bolanos and Gutierrez 1999; Yong 1988). Available information on firm financing is fragmented and outdated, and most is for large firms registered with the National Stock Exchange or the Electronic Exchange. To address these deficiencies, the original survey used for this chapter includes both large and small firms and focuses on different financing decisions. The survey provides information on a cross section of firms for 2000. The chapter does not make extensive use of the data provided by Coyuntura Industrial, a section of the Institute de Investigaciones Economicas at Universidad de Costa Rica, which has periodically surveyed the manufacturing sector since 1980, producing a quarterly index of manufacturing activity. Although these surveys are rich in terms of production, employment, and other indicators, they are inadequate for present purposes for two reasons. First, they do not look into financing conditions. Second, they include mostly larger firms. The most comparable study dates back to 1994, when the U.S. Agency for International Development and the Academia de Centroamerica conducted a survey of the small business sector (commerce, industry, and services). The survey centered on financial aspects of small firms, including start-up financing, and sampled 808 firms with fewer than 20 employees and monthly sales less than US$13,000. Villalobos (1996) reports the results, which suggest a financing profile for small firms as well as the determinants of their access to formal credit. We compare the results of the survey conducted for this chapter with those of Villalobos below. Small businesses have limited access to bank credit. As indicated in table 5.2, formal credit has a small role in the start-up of small firms; formal credit provides less than 14 percent of the funds required to establish such a firm. The lion's share of the funds originates from entrepreneurs' personal savings. Villalobos reports that as many as one-third of entrepreneurs do not have access to formal credit. For them, the main source of credit is a supplier and/or advance payments by customers. Even firms with bank credit use banks infrequently and on a small scale. Moreover, the firms tend to use only one source of funds, and 70 percent use only one provider of credit. Table 5.3 shows that formal credit is much lower than commercial credit.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

160

MONGE-NARANJO AND HALL

Businesses, Costa Rica Source of financing Start-up costs Personal funds Friends or relatives Gifts Trade credit Formal credit Ongoing financing Personal funds All sectors Manufacture Service Commercial Trade credit All sectors Manufacture Service Commercial Advance payments All sectors Manufacture Service Commercial Source: Authors' calculations based on survey data. Percent

59.5 11.1 10.3
5.2

13.9

38.4 40.0 33.0 40.2 29.0 12.8 23.0 42.4 19.1 30.6 33.0
4.3

Table 5.3. (Percent)

Sources of Financing for Small Businesses with

Multiple Providers of Credit, Costa Rica

Type of financing Formal Informal Trade credit Informal lender Friends and relatives Source: Villalobos (1996).

1995-2000 10.6 5.8 58.7 2.8 10.9

2000
5.8 3.8
57.4

2.5 5.9

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.2.

Start-up Costs and Ongoing Sources of Financing for Small

COSTA RICAN MANUFACTURING FIRMS

161

The objective in designing the sample was to depict the financing profile of various types of firms. For firms with 20 or more employees, we constructed a representative sample of 150 firms from the universe of firms that Coyuntura uses for its index of industrial activity. The sample of these "large" firms was stratified according to the different sectors of the index of industrial activity and the number of employees in the firm; firms from each sector were randomly chosen. Two substitute firms were selected for each firm. A similar sampling scheme was applied to smaller firms (fewer than 20 employees). The universe consisted of the set of manufacturing firms registered with CCSS as of January 2001. A representative sample of 500 firms was extracted from a universe of approximately 5,000 firms, and each selected firm was assigned two substitutes, a step that proved useful later. The survey was limited to metropolitan areas, which include the country's main cities (San Jose, Alajuela, Heredia, and Cartago) and most of its industrial production. The survey contained different questions for large, medium, and small firms. Conducting the survey involved approximately 2,900 calls, 500 faxes, and more than 600 visits to firms. As some of the originally selected firms were closed by the time of the interview, or their phone number was incorrect, the sample of substitutes was used extensively. In the end, we collected 355 questionnaires from large and small firms. There was a low rate of response for access to accounting statements because many firms considered such information confidential. This was the most stringent limitation on the information collected. The questionnaire was based on the questionnaires that Hall and Lopez (2000) and Hall and Madrigal (2000) employed for the borrowers of two commercial banks in Costa Rica and the questionnaires that Bond and Townsend (1996) and Huck and others (1999) used for the financing options of minority groups in Chicago and Thailand.4 The rest of this section classifies the information collected in seven categories.

4

We are grateful to Rob Townsend for providing us with the survey of La Villita (Little Village) in Chicago.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Firm Survey and Sample Selection

162

MONGE-NARANJO AND HALL

The general description of the business includes the most basic information on the firms, such as type of activity, location, size (number of employees and total assets), age, type of ownership, employees, type of hiring, tenure of current owner, and tenure of current management. Business Performance and Financial Conditions Information was gathered on production, sales, profits, investments, debt, net worth, and total assets. Human Capital and Related Issues Because the characteristics of the manager and/or owner can affect both access to credit and the efficiency of the firm, we constructed indicators of education, previous experience in related activities, ownership of other businesses, family composition, and other businesses and occupations. Previous Performance Previous performance (such as a good or bad record on loans in the past) can determine whether agents would have access to credit. Entrepreneurs were asked about previous relationships with creditors. Ongoing Financing To identify the main forms of financing by different types of firms, information was gathered on production, sales, size of investments, inventory holdings and other working capital, and payroll. We also collected data on sources of finance (internal funds or external finance, including banks and other formal intermediaries), trade credit, type of relationship with lender (frequency and types of services), suppliers, and informal credit, including personal and family sources as well as other social networks. One way to learn about whether credit constraints may be binding is to ask a battery of questions, such as those in Bond and Townsend (1996) and Huck and others (1999). The survey included questions such as the following:

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

General Description of the Business

COSTA RICAN MANUFACTURING FIRMS

163

The idea of including these qualitative indicators is to extract information that would not be available even if ideal information from firms' financial statements were available. For example, in the first question, investing a windfall in the business would indicate that there are productive investments available to the entrepreneur. If the entrepreneur would invest, it would mean that at the very least the expected return on the investment would be above the market interest rate. (However, the answer to this question may not be as telling in an environment such as Costa Rica, where there is a high spread between deposit and lending rates.) The second question serves a similar purpose. However, risk aversion and not necessarily credit constraints could be a factor. Yet, with a complete-markets (Arrow-Debreu) economy as a benchmark, risk aversion would not be an issue because agents could fully insure; investment and consumption decisions would be separated in those cases. In any case, the answers to these questions could be invalidated by issues of risk aversion due to the lack of insurance. The second and third questions attempt to investigate whether the composition of liabilities is directly affected by the lack of some markets or lack of access to them. A similar objective drives the fifth question, which is specifically geared to bank credit. Start-up Financing It is worthwhile to investigate credit rationing in the entry (extensive) margin. Questions like those stated above can be asked with respect to the date when the firm was activated or purchased. Information is collected on the firm's financing at the time of its establishment. As in Bond and Townsend (1996) and Huck and others (1999), the questions used will distinguish new firms from those acquired by the entrepreneur.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

1. Would you invest a windfall in your business? 2. Would you be willing to risk all or some of your assets on a new business? 3. Would you be willing to swap part of your firm in exchange for a reduction in debt? 4. Would you like to change the maturity of your debt? 5. Would you like to exchange some of your trade credit for bank credit? 6. Do you maintain a long-term relationship with a bank?

164

MONGE-NARANJO AND HALL

Finally, the survey includes a battery of questions regarding the possibility of using credit to shield the firm in case of sudden need due to liquidity, business opportunities, or other shocks.

Financing Profiles Banks do not provide the majority of credit (tables 5.4 to 5.7). Indeed, their participation in financing the start-up of firms is scant. Capital for the industrial sector is mostly obtained from entrepreneurs' own funds. Firms with larger value or larger employment exhibit larger co-participation of partners. Both sources of funds add up to roughly 75 percent of total initial needs, while only 14 percent of start-up capital is provided by banks. However, banks provide 48 percent of total debt for ongoing firms. Notice that while banks remain the single most important source, trade and informal credit jointly outdo banks. These two sources account for 42 percent and 10 percent, respectively, of the average credit of ongoing firms. With some variation, the pattern holds for other firms. Older firms, as well as firms with higher value and larger employment, finance their activities with a larger portion of private bank credit and a lower share of trade credit and informal credit than their counterparts. The tables report simple averages of debt composition over the total number of firms. Moreover, firms were grouped according to three characteristics: age, number of employees, and the reported value of the firm. According to age, firms are classified as young, mature, or old, at 0 to 10 years, 11 to 25 years, and more than 25 years, respectively. For employment, firms were grouped into those with fewer than 10 employees, 10 to 20 employees, and more than 20 employees. Finally, 50 million colones (approximately $1.67 million at the time of the survey) was the dividing line for firms with high and low value. Slight variations in the cut-off points did not affect the numbers in a significant way. Tables 5.5 to 5.7 present the percentage of firms that have used alternative sources of funding at least once. The percentages were calculated by dividing the number of firms that used this source of funding at least once by the number of firms that reported having employed funding. In grouping

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Shocks and Insurance

Table 5.4. Composition of Start-up Debt for Industrial Firms, Costa Rica
Number of employees

(Percent)

Firm value

Firm age (years)

Source

Average

Young
Old
1.0 1.0 1.0 1.0 1.0 4.0 2.0 6.0 3.0 6.0 0.0 3.0 0.0 3.0 1.0 2.0 1.0 1.0 4.0 1.0 4.0

Mature
14.0 12.0
2.0 1.0 3.0

Less than 10
13.0 16.0

Between 10 and 20

More than 20

Less than 50 million colones
15.0
2.0 1.0 3.0 3.0 7.0

More than 50 million colones
14.0

Banking credit

14.0

12.0

Private lender

1.0

2.0

1.0 1.0 4.0 2.0 3.0

Government program
.0
8.0

1.0

2.0

Relatives

3.0

1.0

Suppliers and clients

2.0

1.0

Cooperatives

6.0

5.0

Total loans
2.0 7.0 5.0

26.0 47.0 21.0 38.0 19.0 33.0 51.0

23.0

31.0 26.0

23.0

31.0
5.0

23.0
3.0

30.0
5.0

24.0

Gifts

4.0

6.0

2.0

Personal resources

44.0

50.0

51.0 14.0

23.0 51.0

49.0 17.0

31.0 42.0

Partners

25.0

22.0

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.5. Firms Using Alternative Sources of Start-up Financing, Costa Rica Firm value
Number of employees

(Percent)

Firm age Mature
Old
17.0 16.0
3.0 0.0 4.0 2.0 6.0 0.9 3.5 0.9 3.5 1.7

Source

Average

Young
17.4 10.4
2.1 0.9 0.0 1.3 1.9 3.8 7.5 1.9 0.9 1.9 1.3 0.7 4.9 3.5 9.0

Less than 10
17.0

Between 10 and 20

More than 20

Less than 50 million colones
21.1
3.0 0.6 3.6 4.2 9.6

More than 50 million colones
14.9

Bank credit

19.9

15.2

Private lender

2.4

2.9

0.7 1.5 4.5 2.2 3.0

Government program

1.0

1.9

Relatives

4.0

2.9

Suppliers and clients

3.4

3.8

Cooperatives

7.7

7.6

Total loans
2.8 5.7 5.7

35.0 50.0 22.2 30.2 20.1 28.3 57.2

28.6 29.6

35.4

21.7

31.0
5.0

20.0
1.7

39.2
6.6

23.9

Gifts

5.4

5.7

2.2

Personal resources

53.2

53.3

51.0 14.0

20.0 37.4

57.8 19.3

28.4 37.3

Partners

30.0

23.8

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.6. Composition of Ongoing Firm Finance, Costa Rica Firm value Number of employees Between More than 20
67.0 20.0 33.0 14.0
3.0 1.0 3.0 2.0

Firm age

Less than 50 million colones
34.0 19.0
3.0

More than 50 million colones
57.0 19.0 22.0

Source Less than 10
34.0 17.0
4.0 3.0

Average
53.0 53.0 43.0 25.0 15.0 10.0
7.0

Young
28.0 16.0 25.0 11.0
3.0 2.0 1.0 6.0

Mature

Old

10 and 20

Banking sector
11.0 14.0 13.0 17.0 10.0 49.0 12.0
8.0 3.0

48.0

35.0

Public banks

20.0

12.0

Private banks

14.0

7.0

Other formal

14.0

16.0

13.0 14.0
8.0 6.0

17.0 10.0

Informal sector

10.0

16.0

Relatives
35.0 44.0

6.0

8.0

6.0 4.0

Other informal

4.0

8.0

Suppliers

42.0

49.0

47.0

30.0

52.0

33.0

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.7. Firms Using Alternative Sources of Ongoing Financing, Costa Rica (Percent) Firm value Number of employees
Old
43.0 28.0 14.5
3.8 6.2

Firm age Less than 10
43.0 61.0 23.5 36.5 18.3
7.0 3.5 6.2 3.5

Source
45.0 23.6 17.9 24.7 17.3 17.3 12.3 54.3 25.5 13.2 11.9 15.7
8.8 6.9 4.7 3.8 0.9

Average

Young

Mature

Between 10 and 20 More than 20

Less than 50 million colones
32.0 17.5
3.0

More than 50 million colones
59.0 23.9 29.1 13.3 15.7
8.4 7.2

Banking sector 11.1 14.6 15.3 10.4
4.9

60.0

37.0

Public banks

27.8

16.2

Private banks

21.0

9.5

Other formal

21.4

18.1

21.6 15.7 10.4
6.0

Informal sector

18.7

19.0

Relatives 41.0 47.2 42.1

11.1

8.6

Other informal

7.9

11.4

Suppliers

64.7

51.4

45.2

47.6

50.0

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

COSTA RICAN MANUFACTURING FIRMS

169

Start-up Capital In general, older firms, firms with high value, and firms with high employment finance their start-up capital by combining not only the entrepreneur's own resources (33 percent), but also partners' resources (38 percent). In contrast, younger firms, firms with low value, and firms with low employment base their funding more on entrepreneurs' own resources (50 percent) and partners' funds are used in only 22 percent of the cases. Bank credit in the initial development of the firm represents only 14 percent of total startup capital, and this percentage does not vary greatly when firms are grouped by age, employment, or value. A minor difference is observed when cooperatives (banking firms specializing in individual customers) are taken into consideration. Cooperatives account for 5 and 8 percent of the start-up capital for small and medium-size firms, respectively, but only 2 percent of the start-up capital for larger firms. Ongoing Financing Firms finance their ongoing activities primarily through banks and suppliers. The banking sector represents on average 48 percent of firms' total ongoing financing, while suppliers provide 42 percent (table 5.6). The informal sector finances 10 percent of total resources. When the sample is decomposed by age of the firm, number of employees, and value of the firm, some important points deserve to be mentioned. In particular, the role of private bank credit increases considerably as the sample moves to older firms, firms with a large number of employees, and firms with larger value. Moreover, firms with larger value and a large number of employees use less trade credit than their counterparts, as well as less informal credit. As firms grow older, they increase their use of the banking sector and reduce informal financing. Their use of private banks increases from 7 to 25 percent when firms move from young to old, and the informal sector's share declines from 16 percent to 3 percent (table 5.6). In the case of trade

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

the firms by age, number of employees, and value of the firm, these figures were divided by the number of firms within each group. The tables report only on the extensive margin of financing. A similar picture arises based on the percentage of debt of firms by source of funding.

170

MONGE-NARANJO AND HALL

Determinants of Access to Bank Credit Access to formal bank credit is far from widespread. While the evidence collected is only for the manufacturing sector, it is likely that the results apply in general. Indeed, the results indicate that things have not changed dramatically since the Villalobos (1996) study. The lack of formal bank credit does not involve only intensity of use; many firms do not use bank credit at all. This section treats the intensive and extensive margins explicitly, with simple econometric tools. If the objective were only to examine whether a firm receives credit at all, dichotomous probit models would be sufficient. In such models, firms would be classified in two groups: those with some bank credit and those with no bank credit. We would then estimate the probability that a firm belongs to either group as a function of the firm's observable characteristics. However, a probit model does not make use of all the information available; it neglects the intensity of use of credit by firms with bank credit. To include that information, we use a Tobit model. We measure the intensity of use by the share of bank credit in total credit. Let y* indicate the fraction of debt of firm debt i that is owed to banks; let y, be an indicator variable of whether the firm has formal bank credit at all (that is, y{ = 1 if y* > 0 and /,- = 0 if y * = 0); and let *,- be a vec-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

credit, a U-shaped curve is observed over time; young and old firms have a larger participation of trade credit in total financing than mature firms do. As the number of employees increases, financing by means of bank credit tends to be higher. This result is especially strong for private bank credit, which moves from 4 percent to 33 percent (table 5.6). In contrast, firms use less informal and trade credit as the number of employees increases. For instance, the informal sector averages 17 percent in small firms and only 3 percent in large ones. Trade credit declines from 49 percent to 30 percent as the number of employees increases. As the value of firms increases, the percentage of firms using private bank credit increases and the percentage using trade credit decreases. The private bank credit sector finances 3 percent of the activities of firms with low value and 22 percent for firms with higher value. The values are 52 percent and 33 percent, respectively, for trade credit.

COSTA RICAN MANUFACTURING FIRMS

171

Here (3 is a vector of parameters and e, is the unobserved and random heterogeneity of the firm. A firm will have no bank credit at all if e,< -P'x,. Letting/and F denote the probability density function (pdf) and cumulative density function (cdf), respectively, of 8, expressions can be written for the probabilities. The probability that firm i receives no credit at all is JF(-p'xj). Thus, if the parameters of P are known, the probability of observing in the sample a firm with given characteristics (y,,x,) is F(-p/x1-)1"yi'[l-F(-p'Xj)]>''. The probability of observing a firm with characteristics (y*,y;,x,) is given by F(-p'x,)^/(-p'*,X'. To estimate the determinants of access to formal financial markets, we use maximum likelihood estimation, that is, the functions: LProbit =

are maximized by parameter estimates of P. The variables x, contain information on the firm (age, assets, employment, type of ownership, and total debt), including its industrial sector, as well as characteristics of the owner or manager (age, gender, education, ownership of a house, and previous experience) and the owner's response to the question about changing the financing profile. Table 5.8 defines the variables used in the analysis. Results for Bank Credit in Ongoing Financing We experimented with various combinations of the variables obtained in the survey. This section reports only the most interesting results.5 The purpose is to find out whether any of the firm and entrepreneur characteristics can explain the use of bank credit. Firm characteristics that are included are the firm's age, size (log of the number of employees) and leverage (total debt/assets). Entrepreneur
' The database is available from the authors on request.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

tor of observable characteristics of the firm. (All of these variables are obtained from the survey.) It is assumed that the relationship between x, and y* is given by the simple form

172

MONGE-NARANJO AND HALL

Variable
Bankrel Bactdeb Bancred Exastdbt Exsupdbt Finstat Ftypown Hgsch House Initdebt LnEmploy LnAge LnFage LnFvalue Onbkdbt Othinc Prexp Profrate Reinvest Secalim Secmad Secmetal Secpapel Secquim Sectex Sex Special Startcredt Startperc Unived Uselott a bank

Definition
Dummy equals 1 if the entrepreneur reports having a long-term relationship with Ratio of total banking credit to total debt 1 if the firm has ongoing banking credit 1 if the owner would exchange assets for lower debt 1 if the owner would exchange supplier debt for banking debt 1 if the firm has accounting statement 1 if the firm is stock company 1 if the owner has high school education 1 if the entrepreneur owns a house Total start-up investment Log of employment of the firm Log of age of the owner of the firm Log of age of the firm Log of value of the firm in colones Total amount of total ongoing debt in colones 1 if the owner reports other sources of income besides the firm Previous experience dummy variable is 1 if the owner had a business before Profit rate Rate of reinvestment out of total profits Dummy variable equals 1 if the firm is in the food sector Dummy variable equals 1 if the firm is in the wood sector Dummy variable equals 1 if the firm is in the metal sector Dummy variable equals 1 if the firm is in the paper sector Dummy variable equals 1 if the firm is in the chemical sector Dummy variable equals 1 if the firm is in the textile sector Gender of the owner of the firm: 1 if female 1 if the manager and the owner are different persons 1 if firm used start-up banking credit Ratio of start-up banking credit to total start-up investment 1 if the owner has university education 1 if the owner would invest a lottery prize in the firm

indicators are age, other sources of income, ownership of a house, previous experience as an entrepreneur, and the fraction of the firm initially financed by banks. Estimates are reported for specifications that focus exclusively on the characteristics of the firm, those that focus on the characteristics of the en-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.8. Variable Definitions

COSTA RICAN MANUFACTURING FIRMS

173

trepreneur, and those that include both. The following four models are estimated. Model 1 includes almost all information on the characteristics of the entrepreneur and the firm. Model 2 excludes from model 1 the size of the firm and its leverage; the rationale is that these variables could be better seen as the outcome of access to credit and not as a factor responsible for it. Model 3 focuses on the characteristics of the entrepreneur and thus excludes firm characteristics. Model 4 focuses on firm characteristics and includes the value of the firm as an explanatory variable. Although the (self-assessed) value of the firm may be affected by credit constraints, this variable is included in order to determine whether its inclusion in the regression affects the results for the other variables, including their values. Tables 5.9 and 5.10 show the results for the probit and the Tobit models, respectively. All the estimations include dummies for the industrial sector of the firm, but they are not included in the tables for three reasons. First, these dummies are not statistically significant. Second, intersector differences per se are not a matter of direct interest for this chapter. Third, omitting the dummies keeps the tables a manageable size. As can be seen in both tables and all four specifications, personal characteristics of entrepreneurs do not appear to have a significant effect on either the probability of having bank credit or the share of credit that comes from banks. In all cases, it seems that the age of the entrepreneur has a negative effect, but it is never significant. Consistent with the view that women may experience greater difficulty in obtaining credit, the estimates on the gender dummy frequently show a negative sign, but are never statistically significant. Ownership of a house shows a positive sign, but the result is not significant. With respect to other income, it is less obvious what sign to expect. On the one hand, individuals with other sources of income should have better access to banks. On the other hand, these individuals can more easily self-finance. The results, although not significant, tend to support the second hypothesis. With the exception of one regression, the percentage of initial capital from bank credit shows a negative relationship with banks' share in the current credit of firms. This may appear odd, as firms that were funded by banks in the first place would seem more likely to maintain an ongoing relationship with banks. We included in the regressions a dummy variable indicating whether the entrepreneur considers that he or she has an ongoing relationship with banks. However, the negative point estimates remain even

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

174

MONGE-NARANJO AND HALL

Variable
Constant STARTPER BANKREL FINSTAT LNAGE LNFAGE
SEX

Model 1
-1.3453 (-0.642) -0.2763 (-0.746) 0.4544 (1.898*) 0.4926 (1.739*) -0.3376 (-0.602) 0.2736 (1.482) -0.2196 (-0.611) 0.1698 (0.485) -0.1081 (-0.432) -0.1925 (-0.694) 1.2157 (2.571**) 0.3339 (2.754***)

Model 2
0.4056 (0.226) -0.0643 (-0.205) 0.4247 (2.055**) 0.5239 (2.127**) -0.5437

Model 3
0.4664 (0.267) 0.0030 (0.01) 0.4120 (2.025**)

Model 4
-7.6305 (-4.59***)

0.3535 (1.56) -0.1009 (-0.376)

-0.4706 (-1.021) 0.1011 (0.632) 0.1072 (0.339) 0.3527 (1.123) 0.0344 (0.16) 0.0935 (0.43) 1.6796 (3.617***) 0.0573 (0.432) 0.3558 (3.496***)

(-1 .097)
0.0418 (0.273) 0.0670 (0.207) 0.3599 (1.129) -0.0196 (-0.09) -0.0854 (-0.367)

HOUSE OTHINC FTYPOWN LEVER LNEMPLOY LNFVALUE Number of observations

164

189

189

180

* Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: t-statistics are in parentheses. Source: Authors' calculations.

if the dummy is not included. In any event, the results are never statistically significant. Firm characteristics are more significant. Tables 5.9 and 5.10 report only the results for the models discussed above. However, many different variations were estimated. The main problem in extracting conclusions

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.9. Determinants of Access to Credit for Manufacturing Firms in Costa Rica: Probit Model on Ongoing Financing

COSTA RICAN MANUFACTURING FIRMS

175

Variable Constant STARTPER FINSTAT LNAGE LNFAGE
SEX

Model 1 -0.4092 (-0.247) -0.4181

Model 2
0.1423 (0.085) -0.1720 (-0.577) 0.4767 (2.018**) -0.4590 (-0.987) 0.0838 (0.584) -0.0358 (-0.118) 0.3540 (1.163) 0.3987 (2.018**) -0.0227 (-0.111) -0.0353 (-0.162)

Model 3
0.1471 (0.088) -0.1021 (-0.338)

Model 4
-3.3700 (-3.025***)

(-1 .408)
0.4018 (1.754*) -0.3643 (-0.812) 0.1655 (1.211) -0.2662 (-0.92) HOUSE BANKREL OTHINC FTYPOWN LNEMPLOY LEVER LNFVALUE Number of observations * Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: t-statistics are in parentheses. Source: Authors' calculations.
164

0.1041 (0.474) -0.3597 (-0.809) 0.0585 (0.472) 0.0224 (0.074) 0.3598 (1.175) 0.3886 (1.953*) 0.0337 (0.163) 0.1268 (0.607) 0.1628 (1.575) 0.0325 (2.064**) 0.1228 (1.796*) 0.2179 (1.181)

0.2210 (0.783) 0.3791 (1.966**) -0.1119 (-0.561) -0.1058 (-0.483) 0.2675 (2.767***) 0.0246 (1.705*)

189

189

180

from these regressions is the high degree of collinearity among firm characteristics, including age, number of employees, and value. All these variables tend to move in the same direction. Estimating the probit and Tobit models with the probability of bank credit and its fraction over total credit using only size (number of employees),

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.10. Determinants of Access to Credit for Manufacturing Firms in Costa Rica: Tobit Model on Ongoing Financing

176

MONGE-NARANJO AND HALL

Start-up Financing It is widely believed that credit constraints are more stringent for younger firms than for older and better established ones. If that is the case, then

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

value, or age as the explanatory variable always renders positive and statistically significant estimates. With the exception of firm age, those results, which are not shown here, are robust to inclusion of the entrepreneur's characteristics. The collinearity problem arises when several of these characteristics are included at the same time. The value of the firm is the most robust predictor of the firm's access to bank credit. Regardless of which of the other variables are included, the estimated effect of the value of the firm remains positive and statistically significant. However, as long as the value of the firm is not included, both the size of the firm in terms of the (log of) number of employees and the dummy for formal accounting procedures have a positive and significant effect. Including the value of the firm eliminates both results, an indication that the value of the firm provides the same information. Model 4 includes leverage in order to control for the total debt of the firm and to analyze what determines the share that is financed by formal financial intermediaries. The point estimates are positive in both the probit and Tobit models, but they are more significant in the probit model. These results suggest that firms that need more credit would try harder to obtain it from cheaper sources (banks). Alternatively, causality may work in the opposite direction, so that firms with access to banks may make more intense use of credit. With the available data, it is impossible to distinguish which direction of causality is the most relevant. A word of warning is in order. As indicated above, characteristics such as number of employees, value, use of formal accounting procedures, and even type of ownership are all outcomes of the past, current, and expected future behavior and performance of the firm, and, obviously, these cannot be assumed to be independent of access to credit. Yet, at any point in time, those characteristics must determine access to bank credit in the period. If panel data were available, it would be possible to attempt different identification schemes to estimate the direct effect of firms' outcome characteristics on their access to bank credit. However, the lack of panel data on firms was precisely the main limitation encountered in this project.

COSTA RICAN MANUFACTURING FIRMS

177

credit limitations should be most stringent precisely at entry, that is, when an entrepreneur starts a new business. Indeed, the previous section found that bank credit is a significantly more important source of resources for established firms than for newly created ones. In this sense, the evidence is consistent with the view that new firms have greater difficulty obtaining bank credit. As in the previous subsection, the objective here is to investigate which characteristics of firms and entrepreneurs explain their access to banks. The inherent limitations of using retrospective data must be acknowledged from the beginning. The ideal would be to collect information on firms just entering at the time of the survey, but a good registry of new firms is not available. Moreover, the small size of the country would likely limit the applicability of statistical methods. As with ongoing financing, probit and Tobit models are used to estimate the share of bank credit as a function of observable characteristics; table 5.11 shows some of the results. We tried many specifications involving most of the recorded characteristics of entrepreneurs and firms. Contrary to expectations, the variables on the schooling attainment of the owner were never significant; moreover, they tended to change signs depending on the other regressors. Therefore, it did seem worthwhile to report any estimates on the owner's education. Moreover, those variables did not affect the significance of the other variables. Other characteristics of the owner, such as gender and home ownership, were not significant, but the sign of the estimate remained mostly unchanged with the different sets of regressors. Indeed, although not significant, having a house is positively associated with obtaining bank credit. Women appear to have greater difficulty in obtaining credit. The direction of these results is as expected, but again, the estimates are not statistically significant. Table 5.11 also shows that in the case of start-up financing, the sector of the firm can significantly affect access to bank credit. This is contrary to the case of ongoing financing, partly because in that instance more characteristics of the firm were included. The most robust finding is that there is a negative and significant relationship between the use of bank credit and whether the owner had previous experience as an entrepreneur at the time of starting the firm. To some extent this is surprising. It would be expected that entrepreneurs with previous experience might have accumulated useful skills and knowledge to successfully manage the new firm. Creditors would be expected to be willing to lend

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

178

MONGE-NARANJO AND HALL

in Costa Rica: Start-up Financing Variable Constant PREXP FINSTAT FTYPOWN HOUSE
SEX

Probit -1.5203 (-2.475***) -0.4715 (-1.977**) 0.2267 (0.996)

Tobit -1.414 (-2.316***) -0.5016 (-2.15**) 0.3172
(1 .445)

0.19
(0.833) 0.1972 (0.643) -0.105 (-0.328) 0.2758 (0.576) 0.6883

0.1298 (0.598) 0.1257 (0.431) -0.0077 (-0.025) 0.2265 (0.508) 0.5337 (1.194) 0.2762 (0.578) 0.8248 (1.801*) -0.3672 (-0.664) 0.357 (0.834)
222

SECALIM SECTEX SECMAD SECPAPEL SECQUIM SECMETAL Number of observations * Significant at 10 percent. ** Significant at 5 percent. *** Significant at 1 percent. Note: t-statistics are in parentheses. Source: Authors' calculations.

(1 .446)
0.3664 (0.72) 1 .0047 (2.064**) -0.0857 (-0.152) 0.4017 (0.874)
225

resources if they had a positive estimate of previous experience. A negative, significant effect could be explained by very different reasons, which cannot be determined on the basis of the data collected. One possibility is that previous experience indicates failures in the past. As such, a poor record as an entrepreneur could convey negative information (a stigma) from the point of view of bankers. An alternative explanation is that bank credit is difficult to obtain for young firms, but for reasons completely independent of the

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 5.11.

Determinants of Access to Credit for Manufacturing Firms

COSTA RICAN MANUFACTURING FIRMS

179

Credit Constraints and Firm Performance What is the effect of access to formal bank credit on the behavior and performance of firms? In general, bank credit is less expensive than other types of credit, such as trade and informal credit. Thus, having access to bank credit affects firms in a variety of margins, ranging from their profits (and hence their net value) as well as their size and investment decisions. Moreover, credit market frictions affect the creation, liquidation and growth of the population of firms in the economy.6 Consequently, better access to credit implies that more small firms will be created and fewer will be destroyed. Firms with better access to credit will grow faster, and active firms will be larger. Thus, in equilibrium, the extent of firms' access to bank credit would enhance the mass of active firms. Still, the implications for the shape of the cross-section distribution of active firms are not easy to determine. A controlled experiment would provide the ideal method for assessing the effect of access to formal credit markets. Such an experiment would include two large groups of individuals, identical in all respects except for access to formal credit markets. If the two groups could be followed over time, it would then be possible to record and compare firms' size, growth, profits, and investment, as well as their entry (firm activation decisions), exit (firm liquidation), and life span. In such an ideal scenario, the analyst could unambiguously assess the effect of having access to formal markets in all these dimensions of firm behavior and even make strong welfare conclusions.
' See, for example, Albuquerque and Hopenhayn (2001) and Monge (2001).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

previous experience (failure or success) of the firm. A negative sign could be explained if previous experience denoted previous successes that allowed the entrepreneur to accumulate the resources necessary for the new firm. Those entrepreneurs would self-finance their projects, eliminating the need for banks. Another alternative is that entrepreneurs with previous experience may find it easier to obtain credit from other firms in the sector. All these hypotheses remain possible because the available data do not allow distinctions to be made among them.

180

MONGE-NARANJO AND HALL

Of course, such experiments are not available. Yet, thinking of such hypothetical exercises helps the analyst to situate the limitations and visualize the potential biases of estimations using actual data. For example, data are available only on active firms. Indeed, credit-constrained firms are more likely than others to liquidate early, and the lack of good credit alternatives might prevent the establishment of such firms. Credit-constrained firms thus may not be observed at the time of collecting the sample. Therefore it is a tough problem to predict how a cross-section of firms would look if financial markets were different. A more limited objective would be to study the effect of credit constraints on the behavior of surviving firms if a set of firms with access to credit markets could be followed. In this case, omitting survival biases, the analyst would be able to contrast the behavior of investment and other measures of firm behavior. The main challenge in this case would be to identify variables that determine access to formal bank credit and do not affect the performance of the firm directly. Schiantarelli (1996) discusses panel generalized method of moments methods to address this problem specifically, but the aforementioned lack of available panel data on Costa Rica unfortunately makes it impossible to adopt these methods. Using the cross-section data collected in the survey, this section explores two econometric methods in an attempt to isolate the effects of credit constraints on the behavior and performance of firms. Of particular interest is the effect of having access to bank credit. The econometric problem that arises is that of sample selection: firms with and without access may be inherently different, and measures of their behavior and performance may determine the extent to which firms have credit. As made clear by the dynamic limited enforcement models of Albuquerque and Hopenhayn (2001), Hart and Moore (1994), and Monge (2001), the characteristics of firms at any point in time are the result of their previous behavior and access to credit. Those models also imply that the value and (observable) productivity and profits of a firm explicitly determine the credit that it can obtain. Thus, anyone interested in estimating the effect of credit constraints on firm behavior must necessarily face the identification problem of controlling for the effect of those observable characteristics on the credit received. The following subsection discusses a methodology that imposes functional and distributional form assumptions to explicitly handle the identification problem.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

COSTA RICAN MANUFACTURING FIRMS

181

The most natural way to assess the effects of having access to bank credit would be to run a simple regression of the form:

where p, denotes alternative measures of interest on the performance of firm z, y{ denotes whether firm i has bank credit, and X, denotes observable characteristics of the same firm. Here E, indicates random, unobserved heterogeneity. Although it is intuitive, such an approach will not necessarily render consistent estimates of y, the effect of access on the performance of the firm. At the very least, it must be recognized that we do not observe a purely random sample of (p^X,-). Imagine that there is a set of firms with the same characteristics X, and randomly some firms are allowed to have credit (y; = 1) and others are not (y, = 0). Under these circumstances, it would be possible to estimate consistently the effect of credit by the difference in the estimated means of the performances. If there were firms with different characteristics X,, as long as the sampling were random, ordinary least squares would consistently estimate the effect of credit. The problem is that having access to bank credit is indeed the result of market equilibrium and, as such, it is quite possible that a set of variables affects both firm performance and access to credit. Using the previous equation, the problem is that whether a firm has credit may depend on (X, e,). The key problem is that there is no way to observe the counterfactual performance that firms that received credit would have displayed if they had not had access. There is no way to observe the performance that firms with no credit would have displayed if they had enjoyed access to it. This problem could be solved if such a counterfactual could be estimated. This sample selectivity problem is well known in the economics literature. This subsection adopts a strategy originally developed by labor economists, most notably Heckman (1974, 1979). Consider having a sample of firms, a cross section (p^X^W,-) where, as before, p, indicates some measure of performance, y, indicates whether the firm has bank credit, and X;, W{ are vectors of observable characteristics of the firms.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Effect of Bank Credit I: Two-Step Parametric Method

182

MONGE-NARANJO AND HALL

lt is also assumed that the condition of whether a firm has access is given by an index model. There is a latent variable y* given by

where v, is a random component. Whether a firm has bank credit is given by

0

otherwise.

In order to parametrically estimate the model, Heckman assumes that (£j,Vi) are jointly normally distributed, with means zero and variancecovariances given by G%, <TV, poeav for some p e [-1,1]. Under these assumptions, the conditional expectations of the performance of the firms with and without access could not be computed. Indeed, the mean performance, conditional on having credit and the observable characteristics of the firms, is

where (() and O are, respectively, the pdf and cdf functions of a standard normal. The last line follows from the normality assumption. A similar expression can be obtained for £[p,|X,,y, = 0]. Thus, if the value of the parameter a were known, it would be possible to simply add the term X,-= [<|> (ocW,-)/O((xW,-)] on the right-hand side of the equation. In this way, consistent estimates of (Y>P) would be obtained. The problem is that the exact value of a is not known. This discussion suggests a method for estimating the model because a consistent estimate of oc can be obtained by estimating a probit (which,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

As before, assume that

COSTA RICAN MANUFACTURING FIRMS

183

Results The previous methods are applied to estimating the effect of bank credit on several measures of performance. Here it must be emphasized that separating exogenous characteristics and measures of behavior and performance is necessarily an arbitrary exercise. All observable characteristics recorded for each firm are derived from its history. In any event, the analysis attempts to determine the effect of bank credit on the following measures of behavior and performance: • • • • Log of employment Profit rate as a fraction of initial net assets Total investment8 Investment as a fraction of net earnings.

These exercises take the following characteristics as exogenous: • Firm indicators: age, accounting system, and type of ownership • Entrepreneur indicators: age and gender • Dummy variables to control for sector. In all cases, we included indicators of the characteristics of the owner and of whether the firm is managed by the owner or someone else. The results reported here do not include leverage, as it could be highly correlated with access to bank credit. The results did not change dramatically, however,
In this last step, the standard error has to be corrected to account for the fact that an estimate of a has been used instead of its actual value. 8 Total investment is used instead of log of investment in order to include firms with zero investment.
7

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

again, is warranted, given the normality assumption). This is precisely what the methods advocated by Heckman and others do. The first step is to estimate a via maximum likelihood on a probit. The second step is to obtain the values for A,;, that is, the inverse Mill's ratios ((|)/O) for each firm of each type, firms with and without formal bank credit. The third step is to use the observations on all firms to estimate the performance equation.7

184

MONGE-NARANJO AND HALL

Table 5.12. Bank Credit and Firm Performance, Costa Rica
Employment (log)
3.21
(3.07***)

Variable
Constant

Total investment
-7.81E+07 (-0.46) -6.63E+07 (-2.06***) 2.83E+07 (0.65) -1.73E+07 (-0.62) -6.12E+07 (-0.80) 3.31E+07 (1.54) 4.23E+08 (1.22) -2.24E+08

Profit rate
0.87
(0.51)

Reinvested earnings (percent)
94.16
(2.26***)

SEX
LNAGE LNFAGE FINSTAT FTYPOWN BANCRED

0.43
(2.07***)

-0.04
(-0.11)

-5.04
(-0.63)

-0.53
(-1.95**)

-0.26
(-0.59)

-8.37
(-0.77)

-0.08
(-0.48)

-0.08
(-0.29)

-6.48
(-0.97) -20.34 (-1.07)

0.06
(0.11)

-0.72
(-1.02)

0.20
(1.46)

-0.03
(-0.13)

11.74
(2.23***) 107.98 (1.27) -55.61 (-1.09)

2.28
(1.01)

3.57
(1.09)

A,
Number of observations ** Significant at 5 percent. *** Significant at 1 percent. Note: t-statistics are in parentheses. Source: Authors' calculations.

-1.08
(-0.81)

-1.97
(-1 .03) 109

(-1 .08) 173

185

185

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

when leverage was included. We did not include the value of the firm in the performance equation or the probability equation. Sector dummies are not statistically significant at conventional levels, and their exclusion from the equations did not change the main results. Table 5.12 reports the results of the regressions including sector dummies, but their estimates are not included in the table. Regardless of the specific measure of behavior/performance, carrying out these exercises requires imposing assumptions on which variables belong to both the access equation and the performance/behavior equation. Specifically, it is necessary to assume that some variables affect only the performance and not the probability of accessing credit. The results reported here are for the case in which the probability of having access to bank credit is specified as a function of the (log of) age of the firm and whether the firm

COSTA RICAN MANUFACTURING FIRMS

185

A previous version of the working paper on which this chapter is based reported the results of the exercises including qualitative indicators—such as willingness to exchange assets for debt, whether the entrepreneur would use a lottery windfall to invest in the firm, and whether the entrepreneur would want to exchange trade credit for bank credit—affecting the probability of receiving bank credit. Leverage was also excluded in the performance equation but not the probability equation because it has so much predictive power in the probits that it would cause singularities in the performance equation. The qualitative results in terms of sign, magnitude, and significance of the estimates on firms' ongoing access to bank credit and sample selection were the same as those discussed in this chapter.

9

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

has formal accounting practices.9 The previous section found that the use of financial statements and the value of the firm are the best predictors of access to credit. The current exercises use the age of the firm instead of the value of the firm because endogeneity problems are more severe with firm value than with firm age. We specify firm performance as a function of the age of the firm, whether it has financial statements, and other characteristics, such as type of ownership and the gender and age of the manager. As explained above, the effect of bank credit can be estimated by including the dummy variable for access to ongoing bank credit, and the estimates are consistent as long as the estimated Mill's ratio from the probability equation is included. Under these assumptions, the model is identified. While the identification assumptions are in principle ad hoc, similar results were obtained under a wide variety of alternative identification assumptions. Table 5.12 shows the results for all four measures of performance. The table shows that for the most part the characteristics of the firms do not affect their performance. The same applies for the sector dummies (not reported here). In the versions that included the total leverage of the firm, this variable had in general a significant positive effect on firm performance. In the versions in which the variable for whether the owner would invest a windfall in the firm was included in the performance equation, that variable was also a significant predictor of the reinvestment rate, but not of the other performance measures. The results suggest that access to bank credit has a positive effect on performance. In all cases, the estimates on access to ongoing bank credit have a positive sign and are large (although employment is logged). Unfortunately, the results are not statistically significant. For instance, the point estimates indicate that just having access to bank credit would increase the (natural) log of employment by 2.28, almost 10 more employees in each firm. The

186

MONGE-NARANJO AND HALL

Effect of Bank Credit II: Nonparametric Bounds The previous methods hinge heavily on functional and distributional form assumptions. The methods are parsimonious and commonly used, and there is no doubt that they are an essential exercise in investigating the effect of credit
10 As with the data set, the LIMDEP codes used for these regressions are available from the authors on request.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

implied size of the effects on total investment, profit rate, and reinvestment rate are so large that they cannot be taken seriously. But they signal that, if statistically significant, the effects are large. However, the t-statistics are low. In general, the values are around one but not significant at any relevant significance level. The present data do not permit strong conclusions, but the results are highly suggestive. In all cases, the correction for selection provides a negative estimate for the coefficient on A,. Directly including access to bank credit in the performance equation, the fact that the firm is likely or unlikely to have access to bank credit does not enhance or diminish firm performance. Indeed, the estimated effect goes in the opposite direction. However, the results are not significant. It turns out that the results reported in table 5.12 are robust to changes in the variables included in both regressions. This is not surprising because the right-hand-side variables are rarely significant. Experiments with eliminating some of the variables or including indicators of human capital of the owner, previous experience, or credit indicators at the time of the start-up of the business resulted in no substantial change: the signs for access to bank credit and A remained positive and negative, respectively. In a few cases, the coefficients turned out to be significantly different from zero for some of the performance measures, but those cases were easy to overturn by small changes in the set of regressors.10 To summarize, the results tentatively suggest that having access to credit can have large effects on the size (employment), investment, and profits of firms. However, even under the strong functional form assumptions inherent in the method, the data do not provide enough information to statistically reject the alternative hypothesis of no effect at all, at least at the significance levels traditionally used.

COSTA RICAN MANUFACTURING FIRMS

187

Results are reported based on three methods developed by Manski and Horowitz that are routinely used in this literature: worst-case bounds,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

constraints. It is important to emphasize that in the present application the equations are not derived from any economic structural model; therefore, the estimates are simple reduced-form effects, not directly interpretable parameters. As such, functional forms and distributional assumptions are not fundamentals of the problem, but rather additional assumptions imposed to solve it. The methods explored here dispense with functional and distributional form assumptions. Imposing less structure increases the robustness of the conclusions, but this comes at the price of necessarily reducing the sharpness of the possible conclusions. The general methods adopted here were developed mostly by Manski (1995) and Manski and Horowitz (1995) to analyze response to treatments. As before, access to bank credit is seen as a treatment. It is also explicitly recognized that there is a selection bias problem because the characteristics of firms endogenously determine whether they have access to credit from banks. Because these methods are not yet common tools, they are explained below in some detail. Specifically, consider a population of/ firms. Each firm ; e / has observable characteristics Xj and performance behavior y>j(t). That performance behavior can occur in two mutually exclusive cases: the firm has no access to bank credit (t - 0) or the firm has access to bank credit (t - 1). Firm j has a realized access to credit Zj e {0,1} and a realized outcome y;. As before, the selection problem arises because the latent outcomes y/f)> t ^ Zj, are not observable, that is, the analyst does not observe the (counterfactual) performance that firms that received credit would have displayed if they did not have access, as well as the performance that firms that did not receive credit would have displayed if they had enjoyed access. From a random sample of the population of firms, the analyst can learn the empirical distribution P(x, z, y) of covariates, realized performance behavior measures, and realized access to banks. The problem is to combine this empirical evidence with (identification) assumptions in order to learn about the distribution of response functions. Of particular interest is the average effect of access to bank credit,

188

MONGE-NARANJO AND HALL

Worst-Case Bounds This procedure estimates the worst-case bounds. The outcome variable is assumed bounded and normalized so that the lowest value is y - 0 and the highest is y - 1. Let y be a vector of performance data; z a vector of binary variables indicating whether firms have access to credit; and x data on the covariates, the observable characteristics of firms. This method computes for each treatment t e{0,l} the worst-case bounds on £[y(l)|x] and £[/(0)|x]:

and

Combining these equations, the resulting upper and lower bounds on the average treatment £[y(l)|x]-£[/(0)|x] are

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

exogenous selection, and instrumental variable bounds. The methods look to extract the most robust conclusions from the data in the sense that they look for the worst and best-case scenarios for the effects of the treatment (access to bank credit). Thus, if it can be established that access to bank credit has a positive effect in the worst-case scenario, then the data available will strongly indicate that it has a positive effect on firm performance. All the methods are based on nonparametric estimation of probability functions. Thus, they are free of functional form assumptions.

COSTA RICAN MANUFACTURING FIRMS

189

Exogenous Selection

A different set of bounds can be estimated if additional identification assumptions are imposed. One set of assumptions commonly used (implicitly or explicitly) is to assume that the selection of firms according to z = 1 or z = 0 is an exogenous process. In the present application, this assumption takes the form of the condition that the selection of firms in terms of their access to credit amounts to:

This assumption, which is not testable, is equivalent to assuming that the sample comes from a randomized experiment. Under this assumption, the effect of having access to credit is

As before, if the observable characteristics x contain a continuous variable, the method uses kernel estimation ofP('[) and £[•[]. In case all x are discrete, they are computed using cell averages. Instrumental Variable Bounds Sharper bounds may be obtained by making further identification assumptions. One possibility that is widely used in other econometric exercises is to assume that there is a set of (instrumental) variables that can be used to sharpen the estimation of the bounds. For nonparametric bounds, the procedure here was developed in Manski (1995) and Manski and Horowitz (1995).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

While the average or point estimate effect on the treatment cannot be estimated from the data, both lower and upper bounds can be estimated. If the observable characteristics x contain a continuous variable, the method uses kernel estimation of P(- •) and E['\-]. In case all x are discrete, they are computed using cell averages.

190

MONGE-NARANJO AND HALL

The IV bound on £[y(0)|w] is defined analogously. The lower (upper) bound on the treatment effect is the lower (upper) bound on £[)/(! )|w] minus the upper (lower) bound on E[y(0)|w]. And, as before, if the observable characteristics x contain a continuous variable, the method uses kernel estimation of P('l') and £[•]•]• In case all x are discrete, they are computed using cell averages.
Results: Bounds on the Effects of Bank Credit

In terms of firm behavior and performance, we apply these methods to the performance behavior measures used before: employment (number of employees), reinvestment rate, profit rate, and firm value. We did not consider total investment because it would have been necessary to control for firm size (assets or employment). The sample size is not large enough to allow a reliable use of nonparametric methods with the rich structure of covariation across independent variables. The method makes it possible to investigate the effect of access to bank credit on different classes of firms. A first dimension to explore is firm age. Thus, an attempt is made to capture the effect of credit constraints on firms in different stages of their life cycle. A second dimension involves distinguishing among firms with different types of organization and management. Given the available information, the most relevant variable for separating firms along these lines is whether the firm is a stock company. Unfortunately, the sample is not large enough to permit a reliable estimate of the joint distribution function of firm age, type of management, and access to bank credit. The key problem is that firm age is a continuous variable and kernel estimators, to be reliable, need a large number of firms for each age. To circumvent this problem, we classified firms in deciles of age, then estimated the probabilities for each cell of firms represented in

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Specifically, let v be instrumental variable (IV) data and w the covariates not used as instruments. The IV bound on £[y(l)| w] is

COSTA RICAN MANUFACTURING FIRMS

191

each of the deciles. This avoids uninteresting large sensitivities in the observations of the characteristics of one or two firms in ranges where few firms are observed in the sample. Using quintiles instead of deciles did not significantly change the results. Based on results obtained from Heckman's two-step estimators, whether the owner would invest a windfall in the firm is utilized as an instrument for estimating IV bounds. In general, this instrument sharpens the bounds just slightly. It is certainly not enough to yield point estimation and, moreover, it cannot rule out zero or negative values as a possibility. Similar results were obtained using other variables as instruments. Figures 5.3 to 5.6 show the results. The first panel of each figure shows the values for the upper and lower bounds according to firm age and whether worst or best-case bounds are estimated or whether the owner would invest a windfall in the firm is used to sharpen the bounds. The horizontal axis of the figures is the mean age of the deciles where the firms belong. The second panel of each figure shows the (point) estimate of the effect assuming that the sorting between firms with access and without access is exogenous. This estimate is shown for firms of all 10 age groups and for the two types of management. With worst and best-case bounds, there is always the possibility that the effect of having access to banks is zero, as the interval defined by the bounds includes zero in general, which is a problem with this methodology. When so few restrictions are imposed, most data sets fail to rule out the possibility that average performance under the two alternatives (bank credit or no bank credit) may be the same. However, an advantage of these methods is that they explicitly describe the entire set of possibilities allowed by the data for the effects. In this sense, it is interesting that the worst-case bound is closer to zero for younger firms than for older firms. Therefore, this method suggests, without clearly implying, that bank credit is more likely to benefit younger firms than older ones. Interestingly, using instruments such as whether the owner would invest a windfall in the firm does not greatly sharpen the bounds, but it does suffice to move their range to exclude zero. Indeed, with the IV bounds, zero is sometimes outside the admissible range for younger firms managed by the owner. These facts help to support the view that bank credit is more likely to have positive effects on the performance behavior of firms at younger ages. An important point is that, in general, the bounds allow the possibility of

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

192

MONGE-NARANJO AND HALL

(Effect on percent)

large effects of bank credit on the performance behavior of firms. Although in the worst-case scenario it cannot be ruled out that the effect of having access to bank credit is negligible or even negative, the bounds indicate that the effects can potentially be huge. These methods do not provide a direct way to obtain a point estimate. One possibility is to take the average between the worst and the best-case bounds, or between the upper and lower bounds of the IV estimator. But such a selection criterion does not have an explicit, conceptually sound basis. Accepting exogenous selection as an identification assumption, the analysis could directly compute a point estimate for the effect of bank credit. Those

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 5.3. Bounds on the Effect of Bank Credit on Firm Reinvestment Rate, Costa Rica

COSTA RICAN MANUFACTURING FIRMS

193

(Effect on percent)

Source: Authors' calculations based on survey data.

estimates are reported in the lower panels in figures 5.3 to 5.6 for each of the performance behavior measures. In general, the point estimates are positive and, indeed, large for all age groups and management types. They appear larger for older firms. However, as measured firm age increases, the sample includes smaller fractions of firms with own-management and nonbank credit. This fact reduces the reliability of the estimated effects not only for the point estimates with exogenous selection, but also for worst and best-case bounds and IV bounds. Moreover, the bounds seem more symmetrical for the last age decile,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 5.3.

(continued)

194

MONGE-NARANJO AND HALL

(Average effect on number of employees)

(Number of Employees)

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 5.4. Bounds on the Effect of Bank Credit on Firm Payroll Size, Costa Rica

COSTA RICAN MANUFACTURING FIRMS

195

(Value in 2001 colones)

(2001 colones)

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 5.5. Bounds on the Effect of Bank Credit on Firm Profit Rate, Costa Rica

196

MONGE-NARANJO AND HALL

(Effect on precent)

(Effect on precent)

Source: Authors' calculations based on survey data.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Figure 5.6. Bounds on the Effect of Bank Credit on Firm Value, Costa Rica

COSTA RICAN MANUFACTURING FIRMS

197

Conclusions This chapter has examined firm financing of firms in Costa Rica, investigated the variables that explain access to formal bank credit, and explored the effect of credit limitations on the behavior and performance of firms. Those objectives were served by directly collecting information on firm size in a variety of manufacturing sectors in metropolitan areas in Costa Rica. Given the lack of good data sources in Costa Rica, the first contribution of this chapter is precisely the data that were collected. Applying a variety of econometric techniques to the data has provided interesting results. Access to bank credit is far from widespread. Indeed, consistent with older studies, Costa Rican firms still depend to a large extent on informal credit and trade credit to finance their operations. Moreover, this dependence is only a matter of intensity, as many firms do not obtain bank credit at all. There is also strong evidence that smaller and younger firms have significantly less bank credit than older and larger firms. The small importance of bank credit is most vividly observed for entering firms; the data show that a large share of the funds of those firms comes from own savings, transfers from relatives, and informal credit. We explored the factors that determine access to bank credit. The use of simple econometric methods demonstrated that the probability of having bank credit and the fraction of bank credit with respect to total debt are mostly affected by firm characteristics, not entrepreneur characteristics. The main determinants seem to be firm value, size in terms of number of employees, age, and whether it uses formal accounting procedures. A serious limitation of this part of the study is that those firm characteristics are the outcome of previous, current, and future behavior and performance, which, in principle, are affected by the accessibility to credit. The longitudinal data needed for identifying the direct effect of those factors are not available. Consequently, research centers in Costa Rica should recognize the importance of setting up a longitudinal survey on the production and financing decisions of small and large firms.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

which means the data are rather uninformative for that group. With this limitation of the data set in mind, the results are more robust for younger firms.

198

MONGE-NARANJO AND HALL

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

It was originally expected that personal characteristics, such as the education and age of the entrepreneur, would be key factors behind access to credit. Surprisingly, the data generally do not single out characteristics of the entrepreneur that would explain access to credit for ongoing or new firms. Interestingly, previous experience as an entrepreneur has a negative and significant effect on the participation of banks in the financing of new firms. This finding is compatible with several hypotheses, all of them highlighting the importance of credit constraints for new firms. A similar result, but without statistical significance, holds for entrepreneurs with other sources. Again, it would be useful to have available longitudinal data in order to discern between these alternative hypotheses. Finally, adopting ideas from the econometric literature on treatment effects, results from two methods to correct for selection biases were reported: a parametric two-step point method and a nonparametric method that estimates upper and lower bounds for the effect of having access to bank credit. Both methods failed to render sharp or statistically conclusive results, but pointed in the direction that having access to bank credit could have large effects on the behavior of firms, increasing their size, investment, and profits. However, it appears that the results could have been much more conclusive had better data been available.

Access to Long-Term Debt and Effects on Firm Performance: Lessons from Ecuador
Fidel Jaramillo and Fabio Schiantarelli
The theoretical and empirical analysis of firm financing has mainly emphasized the choice of debt versus (internal or external) equity.1 Although the idea of debt as a homogeneous source of funds is a powerful theoretical construct and a useful first step, the analysis must go beyond the leverage decision and investigate other dimensions of the debt choice. In particular, the nature of debt and its incentive properties can vary according to maturity (long or short) and providers (banks or markets).2 This chapter addresses the issue of the maturity structure of firm debt and provides some empirical evidence for Ecuador.3 Although the maturity structure of debt is important for both developed and developing countries, there are some aspects of the problem that have been more often (although not exclusively) raised with respect to the latter. In particular, both domestic and international policymakers have the widespread perception that asymmetric information and contract enforcement problems may lead to a shortage of long-term finance. This shortage is thought to have a cost in terms of productivity growth and capital accumulation and it may justify some form of government intervention. Most developing countries have
Fidel Jaramillo is chief economist of the Corporation Andina de Fomento, and Fabio Schiantarelli is a professor of economics at Boston College and a research fellow of the Institute for the Study of Labor (IZA) in Bonn. 1 See Harris and Raviv (1990) for a comprehensive critical review. 2 On the maturity choice, see Myers (1977), Diamond (1991a, 1993), Kale and Noe (1990), Hart and Moore (1994), and Barclay and Smith (1995). On the role of intermediated debt, see Diamond (1984), Calomiris and Kahn (1991), and Rajan (1992). 3 See also Schiantarelli and Sembenelli (1996) for a parallel analysis for the United Kingdom and Italy, and Schiantarelli and Srivastava (1996) for India.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 6

200

JARAMILLO AND SCHIANTARELLI

4

See Jaramillo, Schiantarelli, and Weiss (1993, 1996) on the relationship between credit allocation and firm characteristics and the effects of financial constraints on investment pre and post financial liberalization. See Calomiris and Himmelberg (1995) on subsidized credit in Japan.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

responded by setting up long-term credit institutions (development banks) and/or programs to foster the provision of long-term credit. The emphasis on long-term finance and the potentially adverse consequences when it is in short supply is somewhat at odds with recent theoretical contributions that emphasize the fact that the use of short-term debt may be associated with higher-quality firms and may have better incentive properties. In particular, the possibility of premature liquidation may act as a disciplinary device, improving firm performance. A rethinking of the role of long-term debt, particularly when heavily subsidized, has also been prompted by the problems development banks have encountered in many countries in terms of nonperforming loans and doubts about the selection criteria used in allocating funds. Although the empirical analysis constitutes a useful preliminary step, it must be emphasized that the chapter does not answer the ultimate question of whether the provision of long-term finance should be subsidized (directly or indirectly) and, if so, which is the best way to provide subsidies. The issue is complex because programs of subsidized or directed credit generate distortions that must be compared with imperfections in the capital market due to information problems that would exist even in the absence of administrative controls.4 Moreover, government intervention in promoting the supply of long-term resources often has multiple objectives, such as redressing regional discrepancies or promoting greater equality in income distribution, which this chapter does not address. Finally, in spite of the chapter's narrow focus, the available data fall short of giving definitive answers concerning the effects of government-supported long-term credit. In particular, such analysis would require detailed information at the firm level on the amount of subsidized long and short-term credit, the terms and conditions of each loan, and repayment rates by type of program. However, empirical analysis of the determinants and consequences of the maturity structure of debt is a useful first step that highlights some interesting problems and issues in the allocation of long-term debt that are relevant for many developing countries.

LESSONS FROM ECUADOR

201

Until the first half of the 1980s, the Ecuadorian financial system was characterized by widespread regulation, including interest rate controls, directed credit programs, and other government interference in the allocation of finance. As a result, Ecuador exhibited poor measures of financial depth. For instance, the ratio of M2 money supply to gross domestic product (GDP) was low and declined from 20 percent in 1976 to 17 percent in 1983. One of the most important determinants of the weakness in mobilizing resources through the financial system was the interest rate policy followed in the 1970s and early 1980s. During this period, the government fixed interest rates at or below the inflation rate. Zero or negative real interest rates discouraged financial savings and limited the ability of banks to mobilize private funds. However, directed credit programs from public institutions, in particular the central bank, compensated the inability of the financial system to generate funds for investment. In 1984, these credit programs represented approximately 50 percent of the total credit in the economy. This explains why, despite the situation of financial repression, total credit in the economy increased during the 1970s and early 1980s and peaked in 1983 at 23 percent of GDP (see figure 6.1 and table 6.1).

Figure 6.1. Total Debt/Gross Domestic Product, Ecuador, 1980-95
(recent)

Source: Banco Central del Ecuador (various years).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Maturity Structure and the Role of Subsidized Credit

Table 6.1. Evolution of Aggregate Debt, Ecuador, 1980-95 (Percent) Short-term debt Over GDP Real growth rate Over GDP Real growth rate Long-term debt Long-term debt/ total debt

Total debt

Year

Over GDP

Real growth rate

1980

14.6

— — — — — — — — —
2.7 2.8 2.2 2.4 2.0 1.7 1.5 1.0 0.9 6.9

1981

15.5

10.7

1982 18.5 16.6 -6.3 -0.3 -7.0 -26.2 -22.5 -5.3 33.8
7.5

18.2

18.7

1983

23.1

23.4

— — — —
12.6 14.4 12.1

1984 16.1 15.8 10.5
8.2 7.6 9.7

21.2 -18.9
1.8

-4.4

— — — — — — — — — —

1985

19.4

-4.7

1986

18.3

-2.9

1987

18.2

-6.0

13.3

1988

12.5

-24.3

-11.2 -5.4 -8.5 -31.1 -10.8

16.0 17.2 16.8

1989

10.0

-20.0

1990 10.0
— — — — — —

9.1

-5.9

1991

10.7

22.9

9.4 7.9

1992

10.9

5.8

1993

14.4

35.0

— — —

— — —

— — —

1994

20.4

47.1

1995

25.0

25.9

— Not available. Note: For short-term debt, the maturity is one year or less; for long-term debt, it is more than one year. Source: Banco Central del Ecuador (1980-95); Government of Ecuador (1980-95).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

LESSONS FROM ECUADOR

203

p

Figure 6.2. Nominal and Real Lending Rates, Ecuador, 1980-95
(Percent)

Source: Banco Central del Ecuador (various years).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Beginning in 1984, Ecuador introduced a set of reforms that gradually liberalized the financial market. These reforms eliminated or scaled down directed credit programs and removed administrative controls on interest rates. The reforms led to an increase in real interest rates and improved the ability of the financial system to mobilize resources (see figure 6.2). As a result, the M2 to GDP ratio increased from 17 percent in 1983 to 23 percent in 1987, mainly due to the introduction of the polizas de acumulacion. However, the supply of credit was drastically reduced due to the contraction of government-provided loanable funds. As figure 6.1 shows, total credit in the economy decreased steadily during the second half of the 1980s and was as low as 9 percent of GDP in 1990. The main explanation for the decrease was the reduction in directed credit programs. Public sector institutions decreased their share in total credit from 52.7 percent in 1984 to 9 percent in 1992, as shown in table 6.2. In 1988 there was a similar setback in the process of financial deepening, reflected in a decline in the M2 to GDP ratio. This was followed, however, by a continuation of previous improvements in the early 1990s.

204

JARAMILLO AND SCHIANTARELLI

(Percent) Year Percentage of total credit
52.7

Real growth rate
-37.1

1984

1985
1986 1987 1988 1989 1990
1991 1992

35.2 29.0 21.4 27.7 17.9
16.1 12.5
9.2

-15.0
-26.3

22.6
-41.4

-15.4 -8.3 -26.8

Note: Values are based on aggregate figures. Source: Government of Ecuador (various years).

Table 6.1 shows that long-term debt is quantitatively less important than short-term debt. In the early 1980s, long-term credit (maturity greater than a year) accounted for 12 percent of total debt. Its share of total debt increased to 17 percent in 1989, but it dropped to 8 percent in 1992. It is difficult to assess exactly the role of directed credit programs in the availability of long-term credit. However, most programs of the Corporation Financiera Nacional and the Banco Nacional de Fomento supported sectors and activities such as exports, small industry, and agriculture, with long-term lines of credit for the purchase of machinery and fixed assets. The programs financed by rediscount lines that commercial banks could use with the central bank were instead typically short term, although it was a common practice to renew credit lines extended to firms. The programs financed directly by the central bank were important in the first half of the 1980s (89 percent of total directed credit) and decreased in importance throughout the 1980s and early 1990s (representing 32 percent of total directed credit in 1992), as shown in table 6.3. Table 6.3 reports the proportion of directed long-term credit relative to total long-term credit and the proportion of directed shortterm credit relative to total short-term credit. The data confirm that the percentage of directed credit is higher for longer maturities. This percentage decreased from 59.3 percent in 1985 to 35.9 percent in 1990. It then increased to 63.4 percent and 78.7 percent in the following two years, in spite of the real decline in directed long-term credit, since market-provided longterm credit declined even faster. The percentage of directed short-term credit

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.2. Directed Credit Programs, Ecuador, 1984-92

Table 6.3. Directed Credit by Type of Program, Ecuador, 1984-92 (Millions of current sucres)
1985
5,932 764 5,036 17,349 2,940 13,541
646 733

Program
1986 1987
30,772 5,735 24,610 37,089 3,960 32,720 56,212 10,423 43,028

1984

1988

1989 1990 1991

1992
81,031 9,529 66,590

— — — —
109 23

Corporaddn Finandera Nacional Agriculture Manufactures Construction Commerce Transportation Finance

— — —

11,223 1,218 9,102 20 546 195 18,236 2,424 14,907 24 642 85

4,317 1,727
52 125 124

Services Other
82 61
154 222

283 35 43 67 283 35 43 50



109 3 110 374

Banco Nacional de Fomento Agriculture Small industry Fishery Transportation Tourism Commerce 135,583 121,032 186,817
9.8

22,971 16,061 2,641 599 1,670 98 1,902 129,629 183,706
6.1

37,376 25,882 3,704 523 4,048 323 2,897 42,854 32,595 3,376 611 1,216 249 4,808 47,549 31,456 5,880 1,286 1,321 312 7,294 59,280 39,172 6,250 1,303 1,876 242 10,440 194,221 270,850
6.4

87,279 59,106 9,440 3,196 2,275 418 12,845 230,723 348,774
8.8

132,546 96,732 12,731 2,831 2,358 476 17,418 284,326 453,961
8.2

226,238 1 59,702 19,506 7,241 4,232 1,110 34,447 311,840 594,290
9.5

366,345 261,848 28,448 10,261 8,388 1,176 56,224 215,723 663,098

Central Bank 178,890
3.3

194,304

Total

217,276



10.6 89.4

Percentage share Corporaci6n Financiera Nacional (CFN) Banco Nacional de Fomento (BNF) Banco Central del Ecuador (BCE) (BNF+CFN)/long-term credit BCE/short-term credit

20.9 75.8 59.3 31.1

23.3 70.6 70.8 23.3

38.1 52.5 63.4 25.5 64.8 56.7 16.0 21.9 71.7 38.0 18.4
7.3

12.2 55.2 32.5 78.7 25.0 66.2 35.0 14.3 29.2 62.6 35.9 12.1
3.3

— Not available. Source: Government of Ecuador (various years).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

206

JARAMILLO AND SCHIANTARELLI

Firm-Level Evidence on Access to Long-Term Debt and Debt Maturity The analysis uses two samples based on accounting data collected by the Superintendencia de Companias. The first (unbalanced) sample (hereafter SCI) includes 731 Ecuadorian manufacturing companies from 1984-88 Table 6.4. Directed Credit Program Interest Rates and Market Rates, Ecuador, 1983-91
(Percent)
FOPEX and other CFN credit programs3 Year Market rates (2)
19 23

(D
12 18 18 18 23 23 32 39 47

Size of the subsidy (2)/(1)

1983
1984

1.58 1.28
1.42 1.71 1.69 1.94 1.54 1.36 1.19

1985
1986

25.6
30.7

1987
1988 1989 1990

38.79
44.57

49.16
53.09 55.82

1991
a

FOPEX is the Fondo de Promoci6n de Exportaciones no Tradicionales) and CFN is the Corporaci6n Financiera Nadonal. Source: Banco Central del Ecuador, Information Estadistica Mensual, various issues.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

decreased from 31.1 percent in 1985 to 3.3 percent in 1992, and in 1991 and 1992 real total short-term credit expanded, following the real credit crunch at the end of the 1980s. At the beginning of the 1980s, real (ex post) market lending rates were negative, even in the absence of subsidies. They became positive, on average, until 1987, and negative again in 1988 and 1989, following a March 1987 earthquake, which disrupted oil exports, and a period of fiscal laxity. A fall in the M2 to GDP ratio ensued. In the first half of the 1990s, real rates were mostly positive and increasing (see figure 6.2). Interest on directed credit programs was significantly lower than lending market rates, as table 6.4 shows. Market rates were 1.58 times subsidized rates in 1983, 1.94 in 1988, and only 1.19 in 1991. The spread between the two rates was 21.57 percentage points in 1988 and 8.82 percentage points in 1991.

LESSONS FROM ECUADOR

207

Descriptive Statistics for the Two Panels For the firms in the SCI sample, in 1984 long-term financial debt represented 11.5 percent of total debt (see table 6.5). This figure is of the same order of magnitude as the one obtained from aggregate financial data. The share of long-term debt in total debt increased until 1987, reflecting the faster decline in short-term debt during that period. In the SC2 sample, after 1987 there was a decrease in the length of maturity as a result of the real decline in longterm debt and the real expansion in short-term debt at the aggregate level. One striking fact regarding maturity structure is that a large number of firms appear to be cut off altogether from access to long-term debt. In the SCI sample, which we use to draw inferences on access because of its detailed figures on debt, 214 firms (29.3 percent of the total) never received long-term financial credit, 311 firms (42.5 percent) had long-term debt during some years, and 211 firms (28.2 percent) always had long-term debt (table 6.6). In the SC2 sample, which, because of the more aggregate nature of the debt variables is bound to present a rosier picture, 25 firms (2.9 percent of the total) never had long-term liabilities, 538 (63.3 percent) had

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

and contains more detailed information on firms' real and financial variables. Moreover, for the companies in this sample, it has been possible to identify whether they belong to an industrial group associated with a bank. This sample includes separate figures for short-term (nontrade) debt, long-term debt, and trade debt, so that the measure of length of maturity equals long-term debt divided by the sum of long-term, shortterm, and trade debt. The second sample (hereafter SC2), which unfortunately cannot be linked to the first, is also derived from the data of the Superintendencia de Companias. It includes 850 companies from 1984-92. The period covered by this sample is longer, making it possible to investigate more convincingly changes in the allocation mechanism of long-term credit, both before and after financial liberalization. However, this data set contains fewer variables overall and more aggregate variables. In this case, only data on total long-term liabilities are available, including shareholder debt, which is quite important in smaller companies, and other deferred liabilities unrelated to financial or trade debt. The measure of length of maturity is total long-term liabilities divided by total liabilities.

208

JARAMILLO AND SCHIANTARELLI

(Percent) Year Aggregate data SC1 sample3 SC2 sample13
27 27 28 28 27 25 25 25 20

1984 1985 1986 1987 1988 1989 1990 1991 1992

12.6 14.4 12.1 13.3 16.0 17.2 16.8
9.4 7.9

11.5 13.8 17.7 19.1 17.0
— — — —

— Not available. a Long-term credit/(long-term + short-term + trade debt). b Long-term liabilities/total liabilities. Source: Authors' calculations.

Table 6.6. Firms' Access to Long-Term Debt, Ecuador, 1984-92 SC1 sample Indicator Observations Percent SC2 sample Observations Percent

/Access of firms over the entire period Never Sometimes Always Total Firms with long-term debt
214 311 206 731 171 226 307 381 362

29.3 42.5 28.2
100

25
538 287 850 623 617 649 658 657 649 668 654 584

2.9

63.3 33.8
100

1984 1985 1986 1987 1988 1989 1990 1991 1992
Total

37.1 42.5 56.0 59.2 58.9
— — — —

73.3 72.6 76.4 77.4 77.3 76.4 78.6 76.9 68.7 74.8

— — — —

1,447

51.7

6,996

— Not available. Source: Jaramillo and Schiantarelli (2002, data appendix).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.5. Long-Term Debt/Total Debt, Ecuador, 1984-92

LESSONS FROM ECUADOR

209

5

We define bank association as when management or important shareholders of manufacturing firms were also members of the board of directors of a financial institution.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

long-term liabilities during some years, and 287 (33.8 percent) had longterm liabilities during the whole period. According to the SCI sample, access to long-term financing improved over time (table 6.6). In 1984, only 37 percent of firms had long-term debt. This number increased to 59.2 percent in 1987 and decreased slightly to 58.9 percent in 1988. Unfortunately, the question of access in more recent years cannot be tracked because the SC2 sample is less informative in this sense. If the SCI sample is split by firm size, it can be seen that access to longterm credit varies positively with size (table 6.7). Among the largest companies, 58 percent had long-term debt in every year of the period. Conversely, only 11 percent of the micro firms and 17 percent of the small firms had similar access to long-term financing. Half of the micro firms and 44 percent of the small firms never had long-term debt, while only 1.9 percent of large firms never had long-term debt. Sample SC2 yields similar conclusions concerning the correlation between access to long-term debt and firm size. Access to long-term credit also has a positive (simple) correlation with age: older firms have better access to this type of financing than younger firms, as shown in table 6.7. For instance, 35.7 percent of the youngest firms have never had long-term debt, compared with 22.7 percent of the oldest firms. There is also a relationship between access to long-term credit and bank association.5 Almost half of the firms associated with banks had longterm loans during the whole period. In contrast, only 25 percent of the firms with no bank association had regular access to long-term financing (table 6.7). It is important to mention that older firms and companies with bank association are usually the larger ones. Finally, table 6.8 investigates the association between the maturity structure of liabilities and other firm characteristics for the SCI and SC2 samples. The first three columns of the table give mean values of characteristics for firms that never, sometimes, and always had access to long-term debt. The last three columns give mean values of characteristics for firms with debt maturity below the median, between the median and the third quartile, and above the third quartile. Obviously, the share of long-term credit to total credit is

210

JARAMILLO AND SCHIANTARELLI

SC1 sample Indicator
Firm size Smallest Never Sometimes Always Small Never Sometimes Always Large Never Sometimes Always Largest Never Sometimes Always Firm age Youngest Never Sometimes Always Young Never Sometimes Always
171 100 282

SC2 sample Percent
100

Number of firms
28 14 11
3
216

Number of firms
212

Percent
100 7.5

50

16
161

39.3 10.7
100

75.9 16.5
100 2.3

35
213

94 86 36
355

43.5 39.8 16.7
100

5
157

73.7 23.9
100 1.9

51
213

94
165

26.5 46.5
27
100 9.1

4
133

62.4 35.7
100

96
132

71
212

12 49 71

0

0

37.1 53.8

87
125

41 59
100 3.2

61 66 44
260

35.7 38.6 25.7
100

9
194

68.8 28.0
100 3.1

79
456

82
112

31.5 43.1 25.4
100

14
285 157

62.5 34.4
100 2.2

66
150

Old
Never Sometimes Always Oldest Never Sometimes Always Bank association Not associated Never Sometimes Always

92
2

34 69 47
150

22.7 46.0 31.3
100

47 43 20
0

51.1 46.7
100

37 64 49
606 184 268 154

24.7 42.7 32.6
100

0

12
8 — — — —

60 40
— — — —

30.4 44.2 25.4

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.7. Access to Long-Term Debt by Firm Characteristic, Ecuador, 1984-92

LESSONS FROM ECUADOR

211

SC1 sample Indicator Associated Never Sometimes Always Unknown Never Sometimes Always Number of firms
109

SC2 sample

Percent
100

Number of firms
— — — — — — — —

Percent
— — — — — — — —

19 40 50 16 11
3 2

17.4 36.7 45.9
100

68.8 18.8 12.4

— Not available. Source: Jaramillo and Schiantarelli (2002, data appendix).

higher for firms that always had long-term financing compared with other firms. Firms that always had long-term debt are larger, judging by their mean or median real capital stock or real sales. Such firms are more leveraged, have a higher proportion of fixed assets to total assets, and are more profitable and dynamic, judging by the real sales growth rate. However, the investment rate (the ratio of investment to capital stock) is lower for this group of firms. The ratio of liquid assets to total capital is no different for firms that never had access to long-term credit and those that always had it, although it is higher for companies that have had long-term financing during some years. The results obtained when firms are divided according to quartiles are similar, with two exceptions. First, the (operating) profit rate is lowest for the group with a maturity length in the top quartile relative to the other firms. Second, the ratio of liquid assets to total assets seems to monotonically decrease with maturity. The results for the SCI and SC2 samples are similar. Econometric Evidence on Access and Maturity This subsection discusses some of the econometric evidence on access to long-term debt. Table 6.9 reports the results of probit and logit analysis of the probability of obtaining long-term credit for the SCI sample. In order to take into account the panel nature of the data, the table reports the results for the probit random effects model (Butler and Moffit 1982) and the fixedeffects logit model (Chamberlain 1980). In order to minimize endogeneity

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.7. (continued)

212

JARAMILLO AND SCHIANTARELLI

Firm debt maturity Firm access to long-term debt Characteristic
SC7 sample Maturity (long-term debt/total debt) Leverage (total debt/ total capital) Liquid assets/ total capital Clients/total capital Fixed assets/ total capital Sales (real growth rate) Operating surplus/ capital stock Investment/ capital stock Capital stock (millions of 1975sucres) Sales (millions of 1975sucres) SC2 sample Long-term liabilities/ total liabilities Total liabilities/ total assets Liquid assets/ total assets Profits/total assets Total assets (millions of 1975sucres) Source: Jaramillo and Schiantarelli (2002, data appendix).
00

Never Sometimes Always
0.000 0.505 0.069 0.236 0.534 0.044

Between Above Below median and third median third quartile quartile
0.023 0.563 0.080 0.254 0.504 0.053

0.165
0.603 0.083 0.227 0.540 0.066

0.324 0.644 0.067

0.179 0.616
0.077

0.430 0.605 0.062
0.173

0.213
0.598 0.067

0.221
0.559 0.059

0.650 0.078
0.120

0.120
0.241 1.386
5.486

0.151
0.226 3.027 10.706

0.152 0.214
7.184
18.145

0.150
0.238 2.224

0.152

0.212
4.479 14.190

0.223 3.824 8.074

9.571

0.22
0.54 0.44
22.135 22.135

0.35 0.59 0.43
62.255 62.255

0.04

0.29
0.64 0.61
0.5

0.65
0.84 0.85 0.14
120.437

0.31 0.47 2.212

0.35
0.14 -0.2
2.457

2.212

15.054

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.8. Mean Values of Firm Characteristics for Firms Grouped by Access to Long-Term Debt and Debt Maturity, Ecuador, 1984-92

LESSONS FROM ECUADOR

213

Probit Maximum likelihood Maximum likelihood Variable
Age + 2
a

Logit Logit maximum Maximum likelihood model likelihood fixed-effects model

random effects model

model

0.03
(0.38)

0.00
(0.00)

0.05
(0.37)

Age + 3 a Age + 4a

-0.04
(-0.49)

-0.13
(-0.58)

-0.07
(-0.44)

-0.13
(-1.32)

-0.34
(-1.49)

-0.21
(-1.31)

Bank association13 Liquid assets lagged Leverage lagged Operating profits laggedc Firm size, capital stock laggedd

-0.05
(-0.60)

-0.01
(-0.04)

-0.10
(-0.67)

0.51
(1.59)

1.32
(2.41)

0.78
(1.46)

2.72
(2.39)
0.13

0.44
(6.28)

0.53
(4.73)

0.72
(6.16)

(0.39)

-0.20
(-0.94)

-0.46
(-1.19)

-0.33
(-0.93)

-0.58
(-0.60)

0.29
(12.99)

0.50
(9.32)

0.48
(12.56)

0.58
(1.86)

P
Chi2 (p-value) Observations Firms
a b

0.68
(10.47) 314.417 (0.0000) 2,069 353.626 (0.0000) 2,069 314.045 (0.0000) 2,069
1,869

731

731

731

531

Dummy variable for the age of the firm in deviation from the youngest firms. Dummy variable for association with a business group with bank links. c Proportion of total assets (fixed capital plus inventories plus liquid assets). d Size proxied by the logarithm of the real capital stock at the beginning of the period. Note: Year and industry dummies are included. Source: Authors' calculations.

problems, at least in the logit models with fixed effects, we included all regressors as beginning-of-period values for stocks and last-period values for flows. All equations in this subsection include sector and year dummies (not reported for the sake of brevity), with the exception of the logit model with fixed effects, which includes only year dummies.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.9. Determinants of Firm Access to Long-Term Credit, SC1 Sample

214

JARAMILLO AND SCHIANTARELLI

6

See Schiantarelli and Sembenelli (1996) for a more detailed discussion of the theoretical models of maturity choice.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Although the significance of the coefficients varies across estimators, the direction of the effects of the various variables is in most cases consistent across estimators. One of the clearest results is the importance of size (proxied here by the logarithm of real capital stock at the beginning of the period) in determining access to long-term credit. More specifically, the probability of obtaining long-term credit is greater for larger firms. This result is consistent with one of the predictions in Diamond's (1991b) model, which shows that for firms with low credit ratings (presumably the small firms in this case), an increase in quality is associated with gaining access to long-term debt.6 The basic trade-off is between the benefits of short-term debt, because it allows firms to take advantage of favorable news, and the liquidation risk they have to bear, since opportunistic lenders may try to appropriate the surplus by forcing the firm into bankruptcy. However, caution is needed before linking the empirical results for Ecuador to the theoretical models that have been proposed in the literature. One problem is that the model assumes that lenders are profit maximizing, which may or may not be an accurate assumption for Ecuadorian intermediaries in their role as providers of directed credit. Since we use the real value of fixed assets as a proxy for size, the positive effect of size also reflects the fact that the availability of collateral is a prerequisite for obtaining long-term debt. Finally, the positive effect of the size variable could capture the greater economic and political bargaining power of large firms in obtaining long-term directed credit. Other than size, no firm characteristic has a statistically significant coefficient at conventional levels in the fixed-effects logit equation, although the direction of the effects is identical to that in the other models. The lack of precision in the coefficient estimates in the fixed-effects logit model should not be too surprising; estimation of the conditional likelihood function implies a loss of efficiency because many observations drop out from the (conditional) likelihood (see Chamberlain [1980] for details). Given size, past operating profits as a proportion of total assets (fixed capital plus inventories plus liquid assets) do not have a statistically significant effect on access to long-term debt in any equation (actually the point estimate is always negative). The larger firms in the panel are more prof-

LESSONS FROM ECUADOR

215

itable than the smaller ones; however, it is somewhat worrisome from the point of view of the allocation of directed credit that, conditional on size, profits do not matter in determining access to long-term credit. Association with a business group with bank links, captured by a dummy variable, is not a significant determinant of access. This is somewhat surprising because members of business groups may be thought to have superior clout in accessing financial resources, in addition to being informationally less disadvantaged. Similarly, all else being equal, the age of the firm is not a significant determinant of access. The explanation for both these results may be that the effect of bank association or age is basically subsumed by the size variable, given the high probability that larger firms are group members and, at the same time, older. The overall past degree of leverage is positively related to access to long-term debt. Past access to both short-term and long-term debt may work as a predictor of the ability to obtain long-term debt. The initial stock of liquid assets does not play a statistically significant role. Table 6.10 reports the results for the SCI sample from estimating a sample selection model for the length of debt maturity using standard twostep procedures. The dependent variable is long-term debt as a proportion of total debt including trade debt. The results reported are those obtained when either a probit model or a logit model is used in the first step. The coefficient of cash flow in the maturity equation is negative, which emphasizes the concerns expressed above regarding criteria for the allocation of long-term directed credit in Ecuador. Paralleling the results for access, the length of maturity is positively and significantly related to lagged leverage. This latter result may reflect the fact that having obtained debt in the past is an indication of the ability to obtain long-term debt in the future. It is also consistent with the idea that higher leverage increases the risk of liquidation and makes long-term debt more attractive for firms. The length of maturity is also positively associated with size, but the association is not very significant. There is evidence of a strong and positive association between lagged asset maturity (proxied here by the ratio between fixed capital and the sum of fixed capital (estimated) inventories and liquid assets). This is consistent with the idea that firms tend to match the maturity structure of assets and liabilities, as implied by the conventional wisdom and predicted by Hart and Moore's (1994) more formal model. It is also consistent with the hypothesis that fixed assets may be a better form of collateral for long-term debt, so that

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

216

JARAMILLO AND SCHIANTARELLI

Variable
Age + 2a Age + 3a Age + 4a

Two-stage heckit, probit selection

Two-stage heckit, logit selection

-0.01
(-0.33)

-0.02
(-0.94)

-0.04
(-1 .00)

-0.04
(-1.65)

-0.05
(-1.11)

-0.04
(-1.56)

Bank association11 Asset maturity lagged0 Leverage lagged Operating profits laggedd Growth of real sales Firm size, capital stock6 Initial stock of liquid assets

-0.01
(-0.18)

0.01
(0.46)

0.29
(5.91)

0.30
(8.03)

0.12
(2.07)

0.05
(1.93)

-0.20
(-2.14)

-0.19
(-3.27)

0.29
(1.26)

0.03
(1.67)

0.04
(1.26)

0.01
(0.43)

0.28
(1.99)

0.27
(3.08)

X
F-statistic (p-value) Observations (positive) Firms
a b

0.50
(2.41)

0.18
(3.05)

10.14
(0.00)

10.02
(0.00)
1,140

1,140
731

731

Dummy variable for the age of the firm in deviation from the youngest firms. Dummy variable for association with a business group with bank links. c Proxied by the ratio between fixed capital and the sum of fixed capital (estimated) inventories and liquid assets. d Proportion of total assets (fixed capital plus inventories plus liquid assets). e Size proxied by the logarithm of the real capital stock at the beginning of the period. Note: The dependent variable is the probability of obtaining long-term credit. Year and industry dummies are included. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.10. Length of Maturity Equation, SC1 Sample

LESSONS FROM ECUADOR

217

In order to investigate the effect of financial liberalization, it would have been desirable to be able to link the two samples in order to cover a longer data period, both pre and post liberalization, but this was not possible. 8 Although financial liberalization started in 1984, it is a lengthy process that has included setbacks. Similar results were obtained with different breaking points.

7

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

their availability is associated with longer maturity of debt. Finally, the growth of real sales has a positive but not significant effect on maturity. One important issue that deserves to be investigated is whether changes have occurred over time in the determinants of access to long-term credit and, conditional on having access, on the maturity structure of debt. In particular, it would be useful to know whether financial liberalization has introduced any change in the allocation mechanism of long-term debt. The longer SC2 sample provides some evidence on this issue.7 Table 6.11 reports the results obtained from estimating the sample selection model for the SC2 sample. After some experimentation, we allowed the coefficients on lagged operating profits as a proportion of total assets and the lagged value of the logarithm of real assets to vary before and after liberalization, in both the access and maturity equations (table 6.11 contains a dummy variable that equals 1 from 1989 onward).8 The results suggest that the probability of having access to long-term debt before financial liberalization is positively related to size and leverage and negatively related to profits. These results are similar to those obtained for the SCI sample. The main difference is that the profit variable is now negative and significant (it was not significant for the SCI sample), which heightens concerns regarding the criteria used to allocate long-term directed credit. The coefficient on profits increases significantly after liberalization, but it remains negative. The increase in the value of the coefficient may reflect the fact that financial intermediaries start paying more attention to accounting measures of firms' credit rating after financial liberalization. This would also be confirmed by the fact that the (positive) coefficient on the log of total real assets is significantly and substantially larger after financial reform, which is consistent with a greater importance of collateral. The negativity of the coefficient in the post-reform period could be explained by the fact that better (more profitable) firms prefer to use shortterm credit in order to take advantage of future disclosure of positive information, as Diamond (1991b) suggests for the firms at the upper end of the quality spectrum. However, there is another possible explanation for this

218

JARAMILLO AND SCHIANTARELLI

Variable
Age + 2
a

Probit

Two-stage heckit

-0.13
(-3.57)

-0.02
(-1.73)

Age + 3a Age + 4a

-0.01
(-0.16)

0.04
(2.54)

-0.22
(-1.83)

0.02
(0.95)

Operating profits lagged (proportion of total assets) Interaction: operating profits and dummy for 1989 or later Real assets lagged (log) Interaction: real assets and dummy for 1989 or later Leverage lagged Asset maturity laggedb

-2.23
(-6.74)

-0.82
(-8.30)

1.22
(3.03) 0.064 (7.89)

0.67
(6.10)

-0.00
(-1.74)

0.24
(16.17)

0.00
(0.22)

0.73
(9.52)

0.26
(11.97)

0.28
(14.19)

X
Chi2 F-statistic (p-value) Observations Observations (positive) Firms
a b

0.40
(9.08) 1115.48 (0.00) 30.77 (0.00) 8,060
6,113
731 731

Dummy variable for the age of the firm in deviation from the youngest firms. Proxied by the ratio between fixed capital and the sum of fixed capital (estimated) inventories and liquid assets. Note: Year and industry dummies are included. Source: Authors' calculations.

result. Although the real supply of long-term directed credit continued to decrease in 1991 and 1992, market-provided long-term credit decreased even faster. As a result, the share of directed credit in total long-term credit provided to firms increased, and it is possible that the allocation of this portion of long-term credit has continued to be problematic in more recent years.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.1 1 . Sample Selection Model for Access to Long-Term Debt and the Maturity Equation, SC2 Sample

LESSONS FROM ECUADOR

219

Maturity and Performance Two main issues are involved in the relationship between the maturity structure of debt and firms' performance. The first is whether the availability of long-term finance allows firms to improve their productivity. The second is whether the availability of long-term finance stimulates capital accumulation by firms. There are at least two reasons why access to longterm debt may improve firms' productivity. On the one hand, it may allow firms access to better and more productive technologies, which the firm may be reluctant to finance with short-term debt because of fears of liquidation. On the other hand, lack of availability of long-term finance may put a squeeze on working capital, and this may have adverse consequences on productivity. The other side of the coin is that short-term debt, if it entails more continuous monitoring, may force firms to reduce inefficiencies and increase productivity at each level of measurable inputs (capital stock, number of workers, and materials). Ultimately, the issue is an empirical one. In table 6.12 a standard Cobb-Douglas production function is estimated, with capital, labor, and materials, for the SCI sample, the only one for which the necessary data are available. The logarithm of the real value of sales is a proxy for output. Other variables include the log of employment, the log of the real value of fixed assets, and the real value of materials used in production. In addition, total factor productivity depends on the maturity structure of debt and the overall degree of leverage. One potential reason for the inclusion of leverage is that financial pressure may force the firm and its managers to be more efficient.9 However, it is possible that with more leveraging, controlling shareholders may have a smaller incentive to strive for efficiency because they reap a smaller fraction of the rewards.
9

See also Nickell and Nicolitsas (1995) for an analysis using panel data from the United Kingdom.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

The sign and magnitude of the profit coefficients in the second-stage maturity equation parallel those in the probit equation. Size does not play a significant role in the maturity equation, while the maturity composition of assets and the degree of leverage both have a significantly positive effect on the length of the maturity structure of debt.

220

JARAMILLO AND SCHIANTARELLI

Variable
Constant Real value of assets lagged (log) Real value of material used in production Real value of material used in production lagged Real value of fixed assets (log) Real value of fixed assets lagged (log) Employment (log) Employment lagged (log) Maturity Maturity lagged Leverage Leverage lagged

(D
-0.03
(-3.65)

(2)
-0.04
(-4.08)

(3)
-0.04
(-4.57)

(4)
-0.04
(-2.60)

0.41
(3.23)

0.46
(8.61)

0.38
(6.64)

0.39
(6.96)

0.36
(5.51)

-0.15
(-1.98)

0.08
(2.05)

0.05
(1.00)

0.05
(1.07)

0.15
(2.18)

-0.11
(-1.55)

-0.08
(-1.75)

0.42
(6.64)

0.45
(6.24)

0.46
(6.87)

0.42
(3.33)

-0.03
(-1.16)

0.16
(1.70)

0.18
(2.28)

0.35
(2.04) -0.096 (-1.58)

0.01
(0.22) -0.005 (-0.24)

0.02
(0.198) -0.001 (-0.03)

D87 D88
Wald [df] (p-value) Sargan [df] (p-value)

-0.03
(-3.77)

-0.03
(-3.01)

-0.03
(-3.18)

-0.016 (-0.93) -0.024 (-1.23) 172.50(11] 0.000 19.90(25] 0.752 -4.642 0.000 0.089 0.929

-0.05
(-5.52) 206.30(5] 0.000 64.82(31] 0.000 -3.543 0.000 -1.277 0.202

-0.04
(-3.25) 169.66[5] 0.000 56.89(31] 0.003 -2.714 0.007 -0.766 0.444

-0.04
(-3.44) 183.52(4] 0.000 59.20(32] 0.002 -3.488 0.000 -0.761 0.447

M1
(p-value)

M2
(p-value)

Note: The dependent variable is the logarithm of the real value of sales, a proxy for output. The models are estimated using GMM first differences. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.12. Production Function, SC1 Sample

LESSONS FROM ECUADOR

221

10

Arellano and Bond's (1988) Dynamic Panel Data program was used for estimation.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Different specifications of the equations are estimated in terms of their dynamic structure. All equations are estimated by GMM after taking first differences (see Arellano and Bond 1991).10 The first-difference transformation removes the firm-specific and time-invariant components of the error term. Removal of the firm-specific component of the error term is important in order to prevent the coefficient of maturity from merely capturing the fact that better firms may simply receive more long-term debt. Lagged values (two or more periods) of the regressors and of the dependent variable are used as instruments to account for potential endogeneity of the regressors, either because the variables are decided jointly with production or because there are measurement errors. The equations also contain year dummies. The table reports the test of overidentifying restrictions (denoted here as the Sargan test), distributed as chi-squared, and tests for first and second-order serial correlation, distributed as a standardized normal. The results suggest that, when beginning-of-period maturity and beginning-of-period leverage are added to the static version of the production function, there is no statistically significant effect on productivity. When maturity and leverage are entered as end-of-period variables, the effect of maturity is positive and almost significant, while the leverage effect is virtually zero. When the leverage variable is excluded from the equation, the effect of maturity becomes significant. Still, the test of overidentifying restrictions of all the specifications illustrated so far suggests that there is some form of misspecification. We explored dynamic misspecification, reestimating the production function including the lagged value of the dependent variable and contemporaneous and lagged values of all the regressors (financial and real). This model can be interpreted as the unrestricted version of a model in which the dynamics are generated by an autoregressive error term of order one. The results are reported in column (4) in table 6.12. Now the equation passes the test of overidentifying restrictions. Again, contemporaneous maturity has a positive effect on productivity, while the leverage effects are insignificant. What is the impact of the maturity structure of debt and fixed capital accumulation? This issue is investigated by estimating an augmented accelerator type of investment function, where the investment rate is a function of its own lagged value, the contemporaneous and once-lagged rate of

222

JARAMILLO AND SCHIANTARELLI

Conclusions Several lessons emerge from the empirical analysis of the maturity structure of debt in Ecuador. The most striking fact is the unequal distribution of the maturity structure of debt. This is summarized by the fact that, at one end of the spectrum, almost 30 percent of all firms never have access to longterm credit during the period covered by the richer panel presented here. At the other end, almost 30 percent of all firms always have some long-term debt among their liabilities. The main determinant of the probability of obtaining long-term credit is a firm's size (proxied by the real value of the fixed assets). This positive association is consistent with several explanations. One is simply that the availability of collateral is a prerequisite for obtaining longterm credit. Moreover, since larger firms in Ecuador tend to be more profitable, this result could also reflect the positive association between firm quality and access to long-term credit. Larger firms are likely to have better bargaining power and greater political influence in obtaining long-term financial resources. One disturbing additional result is that, conditional on size, operating profits either do not increase the probability of receiving long-term credit or may actually decrease it. Moreover, conditional on having obtained access, they are negatively correlated with the length of the maturity structure of
11 Firms

are classified as large if their fixed capital stock exceeds $6 million in 1975 prices.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

growth of real sales, past cash flow (net of interest rate payments), leverage, and maturity. All the coefficients were allowed to differ across small and large firms.11 Table 6.13 reports the results obtained using the GMM estimator. As expected, if capital market imperfections were more important for smaller firms than for larger firms, the coefficient is greater and more significant for the former. The other financial variables, leverage and maturity, do not appear to play an important role, and are not significant at conventional levels, whether they are included contemporaneously or once lagged. When their contemporaneous values are used as regressors, there is some weak evidence of a positive association between maturity and investment, but only for large firms (t = 1.58), while for small firms the association is actually negative (t = -1.73).

LESSONS FROM ECUADOR

223

Variable
Constant Investment rate lagged, small firms Investment rate lagged, large firms Past cash flow lagged, small firms3 Past cash flow lagged, large firms3 Rate of growth in real sales lagged, small firms Rate of growth in real sales lagged, large firms Rate of growth in real sales, small firms Rate of growth in real sales, large firms Maturity, small firms Maturity, large firms Maturity lagged, small firms Maturity lagged, large firms Leverage lagged, small firms Leverage lagged, large firms
D87 D88

(D
0.02 (2.18) 0.11 (2.25) 0.07 (1.67) 0.24 (2.37) 0.14 (1.69) 0.04 (1.88)

(2)

-0.01

(-0.82)

0.02 (1.34) 0.09 (1.75) 0.08 (1.52) 0.24 (2.33) 0.17 (1.85) 0.03 (1.59) -0.005 (-0.23) 0.05 (0.65)

-0.14

(-1.55)

-0.20

-0.09
(-1.09) 0.05 (1.04) -0.0007 (-0.01) -0.002 (-0.05)

(-1.73) 0.15 (1.58)

-0.02
(-0.34)

0.03
(0.72) -0.003 (-0.156) -0.035 (-2.22) 33.48[12] 0.001

-0.01
-0.05

-(0.58) (-3.2) 29.40[10] 0.001 64.49[50] 0.082 -9.325 0.000 0.712 0.477

Wald [df] (p-value) Sargan [df] (p-value) M1 (p-value) M2 (p-value)
a

54.70[48]
0.227 -8.725 0.000 0.596 0.551

Net of interest rate payments. Note: The models are estimated using GMM first differences. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 6.13. Investment Function, SC1 Sample

224

JARAMILLO AND SCHIANTARELLI

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

debt. This raises some questions on the mechanism used in allocating longterm financial resources in Ecuador during the period under examination. It is interesting to note that the negative effect of profits is greater before financial liberalization, suggesting that the allocation problem was particularly severe for directed credit. After financial liberalization, the coefficient on profit increases, but not quite enough to make it positive. The increase is consistent with the presence of greater incentives for banks to pay more attention to accounting measures of firms' credit ratings. This would also be confirmed by the fact that the (positive) coefficient on the log of total real assets is significantly and substantially larger after financial reform, which is consistent with a greater importance of collateral. The negativity of the profit coefficient in the post-reform period could be explained by the fact that better (more profitable) firms prefer to use short-term credit. Alternatively, it could be due to the fact that allocation problems still remained in the early 1990s for long-term directed credit, which, in spite of its real contraction, increased as a share of total long-term credit in 1991 and 1992 due to the even faster real decrease in the supply of market-provided credit. The data also suggest that there is a strong positive association between asset maturity and debt maturity. This matching of assets and liabilities confirms both the conventional wisdom and the theoretical models that can be used to rationalize it. Does the availability of long-term finance make a difference to a firm's performance, in terms of either productivity or capital accumulation? With respect to productivity, does long-term credit facilitate access to more productive technologies or does the less intense monitoring and the lesser fear of liquidation associated with long-term debt actually reduce productivity? The results obtained from estimating an augmented production function are unequivocal in suggesting that shorter maturity is not conducive to greater productivity. Moreover, there is some evidence that long-term debt may actually lead to productivity improvements. Although these results suggest that long-term debt may have a positive impact on the quality of capital accumulation, estimation of an investment equation does not show that the maturity structure of debt has a large and significant impact on the amount of fixed investment.

Internal Capital Markets and the Financing Choices of Mexican Firms, 1995-2000
Gonzalo Castaneda
Since the seminal paper by Fazzari, Hubbard, and Petersen (1988), many empirical studies linking investment and financial market imperfections have analyzed whether a firm's financial structure has real effects. The literature on investment and financial markets up to now has not rejected the postulated hypothesis when the estimated value of the cash flow sensitivity of investment is larger for the subsample of liquidity-constrained firms after controlling for growth opportunities. The most commonly used criteria for identifying the subset of cash-constrained firms are dividend payout behavior, firm size and age, tangibility of assets, credit ratings, variations over time in the tightness of financial constraints, ownership concentration, bank linkages, and group membership (Schiantarelli 1996). With regard to the last classification criterion, it is presumed that financing bottlenecks faced by divisions of a conglomerate and by firms belonging to a business network are loosened mainly for two reasons. First, the existence of an internal capital market helps provide retained earnings to cash-constrained member firms that exhibit growth potential. Second, member firms share risk, collateral, and reputation, which help them avoid being rationed out of capital markets. One of the advantages of using this criterion to partition the sample is that there is little possibility of an endogeneity problem, which could introduce a bias in the estimation of the cash flow coefficient. This is so because membership in a particular group is generally a stable component of corporate governance. In other words, in the short and medium terms, it
Gonzalo Castaneda is a professor of economics at the Universidad de las Americas, Puebla, Mexico.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 7

226

CASTANEDA

One year of information is lost when lagging fixed assets, and two more years are lost when applying the dynamic panel procedure.

1

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

is not likely that firms will change this feature of their corporate structure on the grounds of financial considerations. The existing econometric evidence supports the view that internal and external sources of financing are not perfect substitutes because the investment variable is more sensitive to cash flow (or cash stock) variations in independent firms than in group firms. Some of the empirical papers that have used this sorting criterion are Hoshi, Kashyap, and Scharfstein (1991) for Japan; Schaller (1993) and Chirinko and Schaller (1995) for Canada; Cho (1995) and Shin and Park (1999) for Korea; Elston and Albach (1995) for Germany; Lament (1997) for the United States; Perotti and Gelfer (1998) for Russia; Babatz (1998) for Mexico; Schiantarelli and Sembenelli (2000) for Italy; and Gallego and Loayza (2000) for Chile. This chapter presents econometric evidence on this issue for Mexico in 1990-2000. The late 1990s in Mexico were characterized by a bank crisis and limited new issues of financial instruments through domestic money and capital markets. Thus, the period under study provides an excellent natural experiment for analyzing firms' investment behavior under conditions of severe market failure. Econometric results show a decreased sensitivity of investment to cash stock after 1994, contrary to prior expectations. Moreover, despite the fact that group membership helped in removing financial bottlenecks in 1993-94, no empirical evidence is found to support the influence of internal capital markets on the observed weaker financial constraints since 1995.1 Accordingly, a more detailed database is required in order to test the hypothesis that group membership may have reduced financial constraints faced by large Mexican firms during the financial paralysis years (1995-2000) to a larger extent than during the financial liberalization years (1990-94). Nevertheless, the results are consistent with the paradoxical Mexican growth observed during 1996-2000, showing no financial constraints for financially healthy firms in a setting of bank crisis, yet no evidence is offered on how this growth really happened. The main econometric findings of this chapter, derived from a database of large firms listed on the Mexican Securities Market (known by its Spanish acronym, BMV, for Bolsa Mexicana de Valores), are the following:

FINANCING CHOICES OF MEXICAN FIRMS

227

Database and Methodology Database The database contains a panel of nonfinancial firms listed on the BMV. The raw data, collected by the BMV, are available on microfilm for the early years of the sample and through two electronic systems for firms that are currently listed on the market: Integral System of Automated Securities and Infosel-Financiero. The database has information on balance sheets and income statements for an unbalanced panel of 176 firms, and it allows for building a balanced panel of 69 firms for 1990-2000. Each firm in the database presents at least four years of information; this is necessary to provide an adequate lag structure for the explanatory variables and their instruments. In some years, a subset of firms was not quoted on the stock exchange (although their information is public since they issued bonds or commercial paper in BMV); consequently, Tobin's Q cannot be calculated for the entire unbalanced panel. The sample covers

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

(i) network firms and firms with bank ties show, as expected, a smaller cash stock coefficient in 1993-94 than independent firms; (ii) sensitivity of investment to cash stock decreases during the financial paralysis period for all types of listed firms except those with bank linkages; and (iii) for network firms and firms with bank ties, investment is positively related to the group's pooled cash stock only before 1995. Other studies have shown that, at least for small firms, the cash flow sensitivity of investment falls when moving from financial repression to liberalization. For instance, Harris, Schiantarelli, and Siregar (1994) for Indonesia, Gelos and Werner (1998) for Mexico, and Gallego and Loayza (2000) for Chile obtain these results. Given that background, this chapter shows an asymmetry in investment behavior in Mexico, in that financial paralysis is associated with weaker financial bottlenecks. It also proposes, in intuitive terms, a possible micro foundation for this paradoxical result. In particular, the network structure might have reduced agency problems and helped firms retain their access to external sources of financing. Moreover, in the new macroeconomic setting, firms belonging to a network might have had an incentive to use their internal capital markets more heavily.

228

CASTANEDA

two contrasting periods: financial liberalization (1990-94) and financial paralysis (1995-2000); this allows testing for structural change during the bank crisis of 1995. One advantage of the unbalanced panel, with some firms delisted before 2000 and others included only after 1990, resides in that it incorporates richer information. The unbalanced panel precludes the introduction of a survivor bias that might handicap the models' statistical inferences if the analysis focused only on the firms that are present for the entire period. It is possible that some firms were delisted because of financial difficulties; hence, by removing these firms from the data set, the econometrician might artificially diminish the impact of cash constraints on investment. Likewise, by incorporating latecomers to the market, the econometrician is also taking into account changes in the macroeconomic context that induced the listing of new firms. Removing from the database this type of information would also bias the statistical results. Moreover, it is important to emphasize that firms included in the sample, either independent or members of a group, are large by Mexican standards and, hence, are the least likely to miss profitable investment opportunities due to lack of external financing. Undoubtedly, this last feature of the data set will work in favor of rejecting the null hypothesis. Consequently, a rejection of the presence of financing constraints using this data set might be reversed with another data set that includes small and medium firms. On the contrary, if there is evidence that financing constraints matter for large firms, the results will be very robust. The database contains financial information used in the different models employed: fixed assets, net sales, export sales, depreciation, inventories, cash flow, and stock. Additional microeconomic data of a qualitative nature are also included: group membership, bank links, and export orientation. The first two variables are constructed from the list of boards of directors presented in Annual Financial Facts and Figures, published by BMV. All monetary variables are presented in real terms; the consumer price index to adjust for inflation is available on the Web pages of the Institute Nacional de Estadistica, Geografia y Informatica and Banco de Mexico. In the database, dummy variables are created for group membership, bank linkages, and export orientation. Network membership is present when at least two board members of a particular firm are sitting on the board of at least another listed firm, whatever their position. A bank tie exists when at least one of the firm's board members belongs to the directorate of one or

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCING CHOICES OF MEXICAN FIRMS

229

Descriptive Statistics The description of investment patterns is made by dividing the sample into two periods: financial liberalization (1991-94) and financial paralysis (19952000). The latter period is, in turn, split into periods of bank crisis (19951996) and steady recovery (1997-2000). The sample is also divided into two categories: network and independent firms. It is expected that the analysis of averages and medians will offer a meaningful first view of the dynamics of investment and its relation to firm structure. Table 7.1 shows that the investment rate fell in both types of firms after 1994, although investment rates were particularly hard hit during the bank crisis years. The table also shows that disparities in investment rates within each category were much smaller during 1995-2000 than before. The mean cash flow ratio is close to 3.9 times larger in network firms than in independent firms, and this gap widened during the period of steady recovery. In contrast to the financial liberalization period, network firms' cash flow ratios surpassed their investment rates as of 1995; on the contrary, independent firms never had enough internal funds to cover their investment needs. There is no clear pattern of how investment rates vary across firm categories, except that they are somewhat higher among independent firms in terms of their means.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

more banks. A firm is defined as export oriented if its export ratio is more than 50 percent. The definition of a business group deserves some clarification. It is well known that many of the firms listed on the BMV belong to business groups; however, lack of information about ownership structure and nonpublic firms precludes the possibility of a more precise categorization of groups. Nonetheless, a simple definition based on the interlocking of directorates is helpful in distinguishing between firms composed of at most a narrow network of divisions—here referred to as independent—and firms with extended connections far beyond the legal concept of holding. In other words, independent firms consist of either truly independent firms or holding companies with some affiliates, while extended groups are formed by legally independent firms, which may or may not be constituted as holdings. This classification will allow testing whether firms that are part of extended business groups face weaker financial bottlenecks than the rest of the listed firms.

230

CASTANEDA

Variable
Network firms Number of observations Mean Cash flow ratio Investment rate Production ratio Median Cash flow ratio Investment rate Production ratio Independent firms Number of observations Mean Cash flow ratio Investment rate Production ratio Median Cash flow ratio Investment rate Production ratio

Financial liberalization (1991-94)
349

Financial paralysis (1995-2000)
513

Bank crisis (1995-96)
164

Steady recovery (1997-2000)
349

0.134
(0.653)

0.131
(0.672)

0.068 (0.887) 0.038 (0.228) 2.303 (5.482) 0.059 -0.001

0.160
(0.542)
0.152

0.441
(4.621) 2.532 (7.502)

0.116
(0.546)

(0.641)
2.902 (5.770)
0.131

2.710
(5.680)

0.101 0.144 1.311

0.108
0.058

0.072
1.394

1.264

1.069

126

284

95

189

0.035 (0.330) 0.500 (2.330) 2.702 (3.945) 0.068

0.033 (0.797)

0.099 (0.631) 0.053 (0.232)

-.0004 (0.869)

0.135
(0.472) 3.236 (5.141) 0.079 0.060

0.177
(0.551)

2.081
(2.875) 0.053 0.009

3.816
(5.885)

0.100
0.082
1.673

0.147 1.451

1.551

1.255

Note: Standard deviations are in parentheses. The investment rate is the current gross acquisition of fixed assets divided by one-period lagged net fixed assets. The production ratio or value of capital is defined as the current value of production (net sales minus inventories) divided by one-period lagged net fixed assets. The cash flow ratio is defined as the one-period lagged ratio of cash flow to net fixed assets. Data for 1990 are not included because a year of observations is lost when lagging net fixed assets. Calculations were made with an unbalanced panel of firms. The group definition considered here is based on the interlocking of directorates. Source: Author's calculations using data from the Mexican Securities Market and Infosel.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 7.1. Descriptive Statistics for the Investment Pattern of Firms Listed on the Mexican Securities Market, 1991-2000

FINANCING CHOICES OF MEXICAN FIRMS

231

The mean cash flow ratio of nonmember firms was high during the bank crisis and close to zero during recovery. This strange result seems to be caused by the presence of some extremely negative observations in the 19972000 period, as becomes evident when analyzing the medians, whose behavior is closer to expectations. This observation could explain why the recovery in average investment rate among independent firms after 1996 was not accompanied by a similar change in mean cash flows. Thus, it makes sense in the econometric analysis to run the regressions with a reduced sample that excludes seriously financially distressed firms. From all these facts, five conclusions emerge. First, the financial paralysis years might have affected the growth potential of the sample firms by reducing investment levels. Second, the crisis made the growth rates of the capital stock more uniform, probably because the macroeconomic context precluded outstanding performances, and some severely distressed cases disappeared from the sample. Third, network firms seem to have more internal resources, some of which might have ended up invested in affiliates not included in the sample. In fact, the cash flow of network firms could have been enough to cover the investment undertaken during the financial paralysis period. Fourth, the financial situation of some listed firms might have been seriously affected by the bank crisis. And fifth, at least in terms of means, corporate structure per se does not seem to be a key factor in explaining the level of investment and its dynamics. The fifth conclusion is an apparent contradiction of the idea that network firms had more chances to deal with the financial paralysis period due to the existence of financial cushions. However, the strength of the link between investment and cash flow must be investigated more formally in a multivariate econometric setting. Two caveats are in order. First, many independent firms in the sample are in fact business groups with a holding company structure. That is, the two categories used here may be more precisely defined as narrow and extended groups. Second, financial constraints are only one part of the story; investment is also determined by growth opportunities, among other things, as will be analyzed with the regression model. In fact, as can be seen from the mean and the median of the production-to-capital ratio in table 7.1, independent firms had, in general, better growth opportunities than network firms with the exception of the bank crisis years. Notice also that when the 1995-2000 period is split, the investment rate seems to vary with the production-to-capital ratio for both categories of firms.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

232

CASTANEDA

Econometric Techniques and Software We used Hansen's (1982) generalized method of moments (GMM) to estimate the models, but we used a system specification in which equations in levels and differences are jointly estimated, as suggested by Arellano and Bover (1995) for dynamic panel models. The econometric literature recognizes the existence of an endogeneity bias in the estimated coefficients when the explanatory variables are simultaneously determined with the dependent variable or when there is a two-way causality relationship. This joint endogeneity calls for an instrumental variable procedure to obtain consistent estimates. However, the use of weak instruments might result in biased estimators as well, as shown by Staiger and Stock (1997) and other authors. Therefore, a dynamic GMM technique is attractive since the panel nature of the data allows for the use of lagged values of the endogenous variables as instruments, as suggested by Arellano and Bond (1991). For this method to work, it is necessary to assume that explanatory variables are weakly exogenous, that is, they are not correlated with future innovations of the dependent variable. Furthermore, the panel data make it possible to address the issue of firm-specific components of the error term. In particular, such components are removed when taking first differences in the regression equation expressed in levels. Moreover, if the original error term is serially uncorrelated or follows a moving average process of finite order, then, under weak exogeneity, lagged values of the dependent and explanatory variables in levels are valid instruments for the equation in differences. In other words, when the error term in the level equation is serially uncorrelated, then the equation in differences presents a first-order moving average error term; thus, endogenous variables lagged two or more periods are appropriate instruments.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Moreover, although the sample firms' investment rate during the financial paralysis years did not recover to the levels attained in the early 1990s, net sales growth showed great dynamism after 1996. In particular, median growth rates increased between financial liberalization and steady recovery from 4.2 percent to 4.6 percent, and from 5.4 percent to 8.7 percent for network and independent firms, respectively.

FINANCING CHOICES OF MEXICAN FIRMS

233

Investment Behavior and Internal Capital Markets Undoubtedly, the paralysis the Mexican financial system experienced in 1995-2000 had consequences on the real side of the economy. In particular, it can be argued that investment decisions at the firm level were modified during this period. The Mexican economy was able to grow regardless of the financial paralysis in formal domestic markets; hence, financing for real investment—and production—had to be channeled through alternative mechanisms, such as the international capital markets or business groups' internal capital markets. Therefore, the role of asymmetric information in financial constraints and investment might have changed in a context of bank crisis and its aftermath. In a scenario of financial collapse and total disruption of economic activity, a traditional investment equation can hardly be estimated. However, in the natural experiment offered by the recent Mexican experience, investment was in general positive during the years of financial paralysis, and especially during the steady recovery of the economy. Consequently, the econometric models test for a subsample of the database whether asymmetric
2 Ox Professional and GiveWin are distributed by Timberlake Consultants (http://www.timberlake.co.uk), and the DPD package is available for downloading at http://www.nuff.ox.ac. uk/Users/Doornik/.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

According to Blundell and Bond (1998), the difference estimator has statistical problems when the dependent and explanatory variables are very persistent over time, in the sense that these variables are weak instruments for the equation in differences. In this scenario, the system estimator of Arellano and Dover (1995) can be implemented. An efficient GMM estimator can be achieved when lagged differences of the endogenous variables are used to instrument the equation in levels in combination with the level instruments suggested above for the equation in differences. We estimated the GMM models using the Dynamic Panel Data (DPD) routines (version 1.0) for the Ox Professional package (Version 2.2). The DPD software was written by Doornik, Arellano, and Bond, and thus their statistical procedures replicate the results of Arellano and Bond (1991). Likewise, DPD was used interactively with GiveWin (Version 1.2), a menudriven program.2

234

CASTANEDA

Methodology for Hypothesis Testing The traditional way to analyze the asymmetric information theory of investment is to test whether investment in those firms a priori considered less affected by asymmetric information problems is indeed less sensitive to variations in cash flows. In this exercise, firms associated with networks are assumed to face weaker financial constraints due to the presence of an internal capital market. In addition, a smaller coefficient for the investment-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

information conditions varied between the financial liberalization and paralysis periods. The failure of many banking institutions and the implosion of the formal capital market might have caused a widening in the information gap between borrowers and lenders, aggravating this asymmetry. However, access to international capital markets and the rise of a network source of financing might have curtailed some of these information problems, at least for firms that are export oriented or belong to business groups. For those firms that belong to a network, it is common to find interlocking directorates, overlapping majority shareholders, and the presence of holding structures; hence, when financial resources move throughout the network's internal capital market, information flows more freely and agency costs are reduced. Consequently, because of a change in firms' financial structure from traditional liabilities to network financing, it might be expected that investment should have been less sensitive to cash flow and collateral after 1994 in an environment where the real side of the economy kept growing until 2000. In the database, the ratio of suppliers' credit to fixed assets increased from 0.226 to 0.336 as the economy moved from financial liberalization to financial paralysis. Given the typical network structure of Mexican firms, these averages reinforce the hypothesis that the financial relaxation was due, at least to some extent, to the workings of internal capital markets. Presumably, the availability of financial resources before the crisis was high enough to give some independence to the different affiliates of a business group. In fact, as will be discussed below, these affiliates may be more efficient under stable market conditions, which would allow them to operate as autonomous profit centers, although this would come at the cost of facing more stringent financial constraints.

FINANCING CHOICES OF MEXICAN FIRMS

235

The Econometric Models The neoclassical micro foundation of the investment model is well known in the literature. For example, Love (2000) and Laeven (2000) present a model in which firms face adjustment costs in fixed capital acquisition distributed over time and an imperfect capital market characterized by financial rationing. These two features justify the introduction of lagged values of investment and cash stock as independent variables. Furthermore, as shown in Gilchrist and Himmelberg (1998), the marginal profitability of capital equals the ratio of sales to capital stock (up to a scalar parameter) under a Cobb-Douglas production function; thus, a sales or production ratio can be used in the regression model to capture growth
Supposedly, the estimations from an Euler equation represent a manager's rational investment decisions. However, it is still not clear that the typical characterization of the maximization problem is flawless. Is it true that decisionmaking is made by professional managers taking care of fragmented shareholders? Does each firm perform as an autonomous profit center, irrespective of the possible overlapping of majority shareholders in different firms? Is a firm's profitability the only concern of maximizing managers, or are they also preoccupied by prestige and relative profitability? Is there a learning process and myopic behavior in decisionmaking? These are just a few of the issues that deserve further exploration before discarding any econometric work not based on standard micro foundations. This chapter follows a more modest approach by estimating an ad hoc regression equation, which, in any case, is conventionally used in the literature.
3

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

cash stock (flow) relationship during 1995-2000 would be evidence of a relaxation of financial constraints for firms. The investment equations estimated in this section are not explicitly derived from an Euler equation and mathematical micro foundations; that is, they do not come from the first-order conditions of a manager's intertemporal maximization problem taking into consideration a nonnegativity constraint on dividends. However, the equation models presented here are, under certain assumptions and a first-order Taylor approximation, close to other models derived under those conditions.3 As suggested above, the main hypothesis to be tested with the first model is that financial constraints may have been relaxed during 1995-2000 for large network firms. However, it is not possible to assume from the model whether this is so because of the existence of an internal capital market or because these firms have better access to foreign capital markets.

236

CASTANEDA

Traditional Model Although a network structure is not assumed in the neoclassical investment literature, the econometric model estimated here is adjusted in that regard. Financial restrictions are captured through firms' cash flow or stock. Dummy variables are used to estimate shifts in the investment-financial restriction link defined according to the nature of firms and periods. Investment behavior in a traditional model is given by:

where J!;t is gross investment in fixed assets, Ki>t,i is the stock of fixed assets at the beginning of the period, Yi>t is firm production (or net sales), FRi>t-i indicates financial restrictions (or the lack thereof) at the beginning of the period, £ is the firm fixed-effects variable, dt is the time fixed-effects variable, [i;)f is the error term, and DU^ is a dummy variable used to capture variations in the impact of financial restrictions for specific firm-year observations. In particular, DUi>t takes the value 1 if the firm-year observation is a priori financially restricted and zero otherwise. The financial restrictions variable (FR^-i) is cash stock (or cash flow). It measures internal funds available to the firm, which in principle can be used for financing the firm's investment projects. This variable is considered at the beginning of the period because current year projects are financed with resources accumulated in previous years. Furthermore, it is normalized by the stock of fixed assets at the beginning of the period.4

)

4

Some authors argue that cash flow measures investment opportunities rather than the availability of internal funds. Myers and Majluf (1984) present a theoretical justification that the cash stock can be interpreted as the "cash on hand" to be used to finance firms' investment projects.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

opportunities. This option is preferred to the use of Tobin's Q, which is not available for all the sampled firms. Moreover, the adequacy of the latter variable is questionable because of the small turnover rate of most of the stocks in the Mexican market, which reduces the possibility of price efficiency.

FINANCING CHOICES OF MEXICAN FIRMS

237

Cross-Financing Model The previous model suffers from an identification problem with respect to the ultimate facts that help reduce financial bottlenecks for network firms. Even if the empirical evidence shows that in the post-bank crisis years financial restrictions on investment were less severe, this might be exclusively associated with foreign investment inflows, and not necessarily with an increased use of financing through internal capital markets. Financing from abroad could also result from repatriated capital that left the country when domestic investors panicked due to the peso and bank crisis.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

The model uses a time dummy variable to test whether 1995 was in fact a threshold year when financial structure changed; dummy variables are also used to sort the sample into independent and network firms, independent and bank-linked firms, and nonexport and export-oriented firms. With these distinct variables, it is possible to analyze different implications of financial restrictions, depending on the time period and nature of the firm. Finally, the production (net sales) ratio is included as a proxy for the firm's expected marginal profitability of capital and growth opportunities. The model uses current and lagged values of the production rate. According to theory, the coefficients associated with cash stock and the current production ratio should be positive; that is, investment should respond positively to the availability of internal funds and growth opportunities. However, the coefficient associated with the interaction term should be negative when the less financially constrained firm-year observations have a value of 1 for the dummy. In an analysis across firms, when the dummy variable specifies group membership (bank linkage or export orientation) and the sum of the two coefficients associated with cash stock is zero, then it can be asserted that the network structure (bank tie or international scope) removes the financial restrictions caused by asymmetric information. In an analysis across periods, when the dummy has a time dimension and the sum of coefficients associated with the financial constraint variables is close to zero, it can be argued that during the financial paralysis period the change in financial structure helped to overcome bottlenecks. The coefficient for lagged investment is expected to be positive but less than 1, reflecting the inertia behind adjustment costs in the capital stock.

238

CASTANEDA

where AFR^ is the group-level financial variable, which in this case is the pooled cash flow (or cash stock) for each group at the beginning of the period (APR/,,..! = AFRkt-i if k and i belong to the same group and AFR^i = 0 if the firm is classified as independent). Moreover, pooled cash stock is defined by grouping all listed firms connected with the same banks.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Attending to this criticism, a second equation is formulated to incorporate explicitly the functioning of a network capital market. This equation is an extension of the models presented in Lament (1997) and Shin and Park (1999). If indeed the lower cash stock sensitivity of investment for member firms is explained by transfers within the network, then investment in associate firms should be positively related to the conglomerate's aggregate resources, and especially to those of cash-rich affiliates. Undoubtedly, it is not an easy task to specify the nature of this crossfinancing. In a more detailed model, it would be necessary to define a priori the channels used to transfer resources within these networks. On the one hand, it might be useful to classify firms within the group into cash-rich and liquidity-constrained categories. On the other hand, it might be important to estimate the pool of funds that were in fact transferred to constrained firms as well as the mechanism used for such a transfer. As a first approximation of the problem, in the more simplified model estimated below, all member firms are considered constrained, and the sum of cash flow (stock) from all associate firms included in the database is assumed to be a potential source of funding. From this perspective, a group's cash stock is a cash pool that can be transferred toward investment projects in financially constrained firms. Moreover, this consolidated cash flow (stock) also works as a backup in case the internally generated cash in each firm is not enough to service debt obligations. That is, the group's cash might function as virtual collateral for member firms, increasing in that way the willingness of lenders to grant additional credit. Investment behavior with cross-financing is given by:

FINANCING CHOICES OF MEXICAN FIRMS

239

Estimation Results Due to the panel nature of the database, we estimated models (7.1) and (7.2) by the GMM, using the dynamic approach suggested by Arellano and Bover (1995). Thus, the explanatory variables are instrumented with their lagged values, either in levels or in differences. In particular, we estimated both models using the GMM system, where the level instruments for the difference equation presents two, three, and four lags, while the equation in levels presents only one lagged value for the instruments expressed in differences. Both models were estimated with a refined database. We removed firm-year observations with a negative cash flow ratio and those reporting a zero annual depreciation or an investment ratio below zero or above 0.75.5 Once these deletion criteria were applied, the database diminished from 1,096 to only 499 or 383 observations, depending on whether
The upper limit was set to exclude those firm-year observations where mergers and acquisitions might have taken place, and which cannot be explained with the traditional investment model. On the contrary, the lower limit refers to those firms where divestment in fixed assets is taking place. In the study period, there were 28 cases of mergers and acquisitions for the firms included in the sample according to news found in different issues of the magazine Expansion.
5

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Additional extensions to the model are implemented using dummy variables for time period, group membership, and bank linkage as in model (1), which allows building new interaction variables with both FR and APR. Therefore, the influence of group membership (bank tie) on the investmentcash stock sensitivity for own and pooled resources, before and after the bank crisis, can be tested. If the group's pooled cash stock is empirically related to individual investment in member firms for the entire sample period, the model with cross-financing helps solve the stated identification problem; thus, the existence of financial networks in the Mexican economy is not rejected. This result would not imply that foreign sources of financing were not relevant for loosening financial constraints during the paralysis period. It only provides evidence that large Mexican firms used their networks in the 1990s. Furthermore, it is possible that part of the private sector foreign financing might have been relocated through internal capital markets more intensively after 1994.

240

CASTANEDA

Traditional Model With regard to equation (7.1), the GMM estimation results presented in table 7.2 have p-values that suggest absence of misspecification for the Sargan test of overidentifying restrictions, which tests the validity of instruments.7 Furthermore, there is no persistent serial correlation and only first-order serial correlation is not rejected; hence, it can be stated that the models are properly specified. Results shown in table 7.2 come from the one-step estimation, which yields reliable standard errors. All models were run with the one-year lagged cash flow ratio and the one-year lagged cash stock ratio as proxies for the financial restriction variable; however, the latter ratio showed a better fit according to the estimated coefficients' p-values. Therefore, only estimations with cash stock are presented. For comparison, column (5) estimates equation (7.1) as in column (2), using the cash flow ratio instead. In the model presented in column (1) in table 7.2, the dummy in the interaction term has a time dimension, which makes a dynamic interpretation possible. Notice that all the coefficients are statistically significant and the signs are as suggested by the hypotheses stated above. The main result from this estimation is that not only did the 1995 bank crisis not exacerbate financial constraints for the average firm listed on the BMV, but in addition these constraints were removed by the change in financial structure as suggested by the Wald test. Thus, this test does not reject the proposition that during 1995-2000 there was no relation between cash stock and investment. As discussed in more detail below, this paradoxical result can be explained by the existence of an internal capital market. We argue that control rights exerted by the parent company or surplus affiliates diminish conflicts of interest in a lender-borrower relationship, and hence in a network structure information asymmetries are less stringent. Accordingly, the investment6

It is important to recall that when more stringent deletion criteria are applied, some additional observations are removed from the unbalanced panel when constructing GMM instruments. 7 Only the Sargan test based on the two-step GMM estimator is heteroskedasticity-consistent, as pointed out by Arellano and Bond (1991).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

cash stock or cash flow was used in the regressions. This drastic reduction was primarily caused by the influence of the 1995-96 crisis on the financial health of some firms.6

FINANCING CHOICES OF MEXICAN FIRMS

241

When the model in column (2) in table 7.2 was run using the complete database, all firms before 1995 were financially constrained; moreover, the point estimation for independent firms' cash stock coefficient is higher than 1 (3.455), that is, an increase in cash stock had a multiplier effect. Presumably, firms in financial distress decided to reduce their operations and sell physical assets, either because cash was needed to finance working capital and financial obligations or because it was simply decided to reduce the profile and size of the company. The multiplier effect in this case implies that a reduction of 1 peso in cash stock is associated with a divestment larger than 1. This can be caused by the lumpiness of fixed assets, so that owners are forced to sell assets with a value higher than the financial need. More generally, a firm may decide to sell sizable physical assets and reduce operations, perhaps induced by the need to liquidate outstanding debt.

8

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

cash stock sensitivity might have been reduced because listed firms decided to use more actively their internal capital market after 1995. In order to provide a more rigorous test for this statement, the model is reformulated in column (2) in table 7.2 by allowing the interaction term of the financial restriction to vary across time and across firms. Notice that, indeed, group membership made a difference before 1995 since this type of firm did not seem to be financially constrained according to the Wald test. However, similar Wald tests show that financial bottlenecks were removed for both types of firms, member and nonmember, once the bank crisis hit the economy.8 These striking econometric results are robust to the earlier finding that the domestic financial collapse did not paralyze large Mexican firms' investment. Nevertheless, these results do not directly support the theory that internal capital markets were used more actively after 1995. Perhaps the observed change in financial structure is explained by the fact that many firms had the opportunity to tap international financial markets due to the rapid increase in manufacturing exports. If the latter statement is true, it may be expected that export firms were capable of overcoming the negative effects of the financial paralysis. Furthermore, once fresh capital entered the economy, the internal capital market of business groups helped redistribute financial resources, which, in turn, diminished financial constraints even in nonexpert firms affiliated with groups. This type of analysis is presented in column (3) in table 7.2, where the financial constraint is interacted with an export-orientation dummy variable. Contrary to a priori expectations, export orientation increased investment-cash stock sensitivity before 1995. Nonetheless, once domestic financial markets were paralyzed, export and nonexpert firms did not present financial constraints according to their respective Wald

Table 7.2. Estimation Results for the Traditional Investment Equation with Financial Constraints, Mexico, 1991-2000
(D
(2) (3) (4) (5)

Variable 0.167 0.132 0.162 (0.042) 0.013 (0.063) -0.009 (0.163) 0.339 (0.013) -0.327 (0.017) 0.020 (0.011) -0.014 (0.025) 0.885 (0.038) -0.862 (0.036) -0.626 (0.207) 0.517 (0.289) 0.509 (0.000) -0.390 (0.012) -0.307 (0.053) 0.432 (0.008) 0.112 (0.000) 0.164 (0.000) 0.113 (0.000) 0.153 (0.000) 0.140 (0.000) 0.062 (0.469) (0.094) 0.020 (0.015) -0.016 (0.038) 0.317 (0.003) -0.294 (0.005) (0.024) (0.028) 0.020 (0.002) -0.014 (0.014) 0.325 (0.080) -0.308 (0.099) -0.343 (0.093) 0.405 (0.032) 0.021 (0.010) -0.017 (0.027) 0.295 (0.004) -0.272 (0.007) 0.187

Lagged investment rate

Production ratio

Lagged production ratio

Lagged cash stock (flow) ratio, 01

Time* lagged cash stock (flow) ratio, 02

Group*lagged cash stock (flow) ratio, 03

Time*group*lagged cash stock (flow) ratio, 04

Export* lagged cash stock ratio, 03

Time*export*lagged cash stock (flow) ratio, 04

Bank tie*lagged cash stock ratio, 03

Time*bank tie*lagged cash stock ratio, 04

Constant

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Number of observations
499 499 120 120 120 120 499 499 383

Number of firms

97
(0.010)

Wald-tests (P-value) (0.000) (0.002) (0.000) (1.000) (0.000) (0.000) (0.523) (0.460)
Ho:

Joint (Chi2) (0.000) (1.000) (0.815) (0.999) (0.000) (0.487) (0.305) (0.814)
Ho:

Specification tests (P-values)

Sargan test (Chi2) (0.000) (0.316) (0.898) (0.449)
Ho: Ho:

(1 .000)
(0.000) (0.893) (0.949)
Ho:

Serial correlation

First order

Second order

Third order

Wald test (P values) linear restrictions P1+p2=0 P1+p2=0 P1+P2=0 (0.287)
Ho:

P1+p2=0 (0.260)
Ho:

P1+p2=0 (0.784)
Ho:

(0.291)
Ho:

(0.291) P1+P3=0 (0.915)
Ho: Ho:

P1+P3=0 (0.000) P1+P2+ P3+P4=0 (0.125)

P1+p3=0 (0.809)
Ho:

P1+P3=0 (0.223)
Ho:

P1+P2+ P3+P4=0 (0.332)

P1+P2+ P3+P4=0 (0.002)

P1+P2+ P3+P4=0 (0.159)

Nofe: Values are from GMM system estimation; the dependent variable is the ratio of gross investment to lagged net fixed assets. Numerical results come from the onestep covariance estimators, except the p-value of the Sargan test, which corresponds to second-step estimates. Heteroskedasticity-corrected standard deviations are used to calculate the p-values presented in parentheses. All models use the cash stock ratio except the model in column (5), which uses the cash flow ratio. Time fixed effects (not shown) were estimated when most coefficients were significant, as in columns (2) and (4). Instruments for the difference equation are included if the variable is present in the model equations. Level instruments are all variables dated t-2, t-4. Instruments for the level equation are dummies. Instruments in differences are all variables dated t-1. The series period is 1993-2000; the longest time series is 8 and the shortest is 1. Source: Author's calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

244

CASTANEDA

tests. That is, although we did not explicitly test for this, it is plausible that listed firms did not exhibit financial bottlenecks in 1995-2000 either because they obtained foreign funding directly in international capital markets or because they received financial resources from their linkages with firms that issued bonds or equity abroad. However, the fact that before 1995 export orientation was not connected with foreign financing, and hence with lower financial constraints, is consistent with the suggestion that internal capital markets were not very active in that period. Another explanation could be that cash-rich firms within a group, presumably export firms, "subsidized" nonexpert firms in times of limited foreign financing. This scenario reduces the investment sensitivity to cash stock for nonexport firms as long as both types of firms belong to the same business group, but not for the export firms that had to rely on their own cash stock. The importance of membership before the crisis is also evident in column (4) in table 7.2, where the presence of bank linkages is used as the group criterion. Although a policy of financial liberalization was implemented during the sample period, firms linked to banks through interlocking directorates were much less financially constrained than independent firms. In addition, the sum of the corresponding coefficients was not statistically different from zero according to the Wald test. Moreover, additional Wald tests show that this situation was reversed for 1995-2000. Although independent firms did not have to rely any longer on retained earnings for their investment projects, bank-linked firms maintained dependence on cash stock since the point estimate of 0.137 is statistically different from zero. These econometric results are in line with the presumption that the bank crisis harmed the financial health of firms with bank ties. Firms without bank ties removed financial constraints in 1995 onward by taking advantage of international financing; firms with bank ties had to rely more on internal resources. A tentative explanation is that for the latter firms, access to international financing was somewhat limited because the market took into consideration the troublesome bank connection. Column (5) in table 7.2 gives estimation results using the cash flow ratio instead of the cash stock ratio. Although three of the individual coefficients are not statistically different from zero, the Wald tests offer similar conclusions to those in column (2). Again, group membership removes financial bottlenecks during the financial liberalization period.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCING CHOICES OF MEXICAN FIRMS

245

Cross-Financing Model The significance of the interaction term with the group dummy in the first half of the 1990s only implies that firms with a network structure were less financially constrained. It is not possible to tell whether this result is explained by the existence of an internal capital market or because of the fact that those firms had better access to foreign financing sources. Therefore, a more detailed analysis of the workings of internal capital markets is needed. With that purpose in mind, table 7.3 presents the GMM system estimation results for investment equation (7.2). The model's new feature is the introduction of the pooled cash stock (or flow) for each group—lagged one period—as a proxy for the influence of the internal capital market on the member firms' investment. In columns (1), (2), (4), and (6), pooled cash stock (or flow) is standardized with the sum of the pooled firms' capital stock at the beginning of the period. In columns (3) and (5), the sum of pooled cash stock is standardized with the firm's own capital stock at the beginning of the period. This last specification assumes that the pool of financial resources available in the internal capital market should have more influence on the firm's investment when that pool is larger relative to the size of the firm's physical assets. Notice that the six sets of estimations presented in table 7.3 are well specified according to the Sargan and autocorrelation tests, where again only first-order serial correlation is not rejected. Moreover, most coefficients have the expected sign for the variables considered in the previous model (lagged investment, production, and the financial constraint variable introduced by itself and interacted with the time, group, and time-group

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

In summary, group membership and bank ties were important in reducing financial constraints before the beginning of the bank and currency crisis. These constraints were on average fully removed for all types of listed Mexican firms (export, nonexport, affiliated, and independent) except for firms with bank ties, where the investment-cash stock sensitivity became positive and statistically different from zero. A tentative explanation is that large Mexican firms may have been able to tap the international capital market during 1995-2000 as long as they were financially sound. It is possible that, in addition, active internal capital markets moved resources from export to nonexport firms, although no econometric evidence is presented here to support this possibility.

Table 7.3. Estimation Results for an Investment Equation with Cross-Financing
Standardization

Pooled fixed assets 0.153 (0.037)

Variable

(1)

Pooled fixed assets (2)
Own fixed assets (3) Own fixed assets (5)

Pooled fixed assets (4) Pooled fixed assets (6)

Lagged investment rate

Production ratio

Lagged production ratio

Lagged cash stock (flow) ratio -0.571 (0.000) (0.078) 0.444 (0.021) 0.002 (0.103) -0.003 (0.052) -1.627 (0.045) 1.027 (0.195) 0.129 (0.007) 0.749 (0.000) 0.015 (0.000) 0.004 (0.907) -0.373

<PD

-0.044 (0.711) 0.032 (0.000) -0.029 (0.000) 0.453 (0.001) 0.009 (0.031) -0.001 (0.000) 0.465 (0.006) 0.021 (0.002) -0.017 (0.006) 0.638 (0.000) 0.015 (0.864) 0.018 (0.053) -0.015 (0.137) 1.521 (0.072) 0.129 (0.088) 0.011 (0.027) -0.006 (0.316) 0.342 (0.008)

-0.080 (0.487) 0.029 (0.115) -0.025 (0.153) 0.003 (0.984) 0.003 (0.984) -0.019 (0.888) 0.215 (0.334) -0.448 (0.091) -0.310 (0.068) 0.415 (0.018) -0.000 (0.658)

Group* lagged cash stock (P3) (flow) ratio

-0.620 (0.000)

Time* group*lagged cash stock (P4) (flow) ratio Lagged pooled-cash stock (flow) ratio by group (P5) Time * lagged pooled-cash stock (flow) ratio by group (06) Bank* lagged cash stock (p3) ratio Time* bank *lagged cash stock (P4) ratio Lagged pooled-cash stock ratio by bank tie (P5)

0.529 (0.079) -0.001 (0.959) -0.009 (0.531)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Time * lagged pooled-cash stock -0.176 0.000 (0.935)
0.160 0.201 0.212 0.104

ratio by bank tie (P6) (0.000) 0.131 (0.000)
322 499 120 235

Constant (0.000)
417 100 120 498

0.241
(0.000) (0.000)
63 87

(0.000)

(0.000)

Number of observations

Number of firms (0.000) (0.000)
(1 .000)

287 75
(0.000) (0.000) (1.000) (0.000) (0.613) (0.450) Ho:p1+p2=0 (0.292) Ho: P1+P3=0 (0.825) Ho:p1+p2+ p3+p4=0 P3+P4=0 (0.000) Ho: P5+P6=0 (0.276) (0.000) Ho: p5+p6=0 (0.203)
(1 .000)

Wald-tests (P-value) (1.000) (1.000) (0.000) (0.892) (0.916) Ho:p1+P2=0 (0.392) Ho: P1+p3=0 (0.984) Ho:p1+p2+ P3+P4=0 (0.393) Ho: p5+p6=0 (0.002)

Joint (Chi2) (1.000) (0.000) (0.387) (0.864) (0.783) Ho: Pl+p2=0 (0.102) Ho: P1+P3=0 (0.449) Ho:p1+p2 Ho:p1+p2=0 (0.533) (0.272) Ho:p1+p3=0 (0.052) Ho: p1+p2+ P3+P4=0 (0.300) Ho: p5+p6=0 (0.803) Ho:p1+p3=0 (0.310) Ho:p1+p2+ P3+P4=0 (0.206) Ho: p5+p6=0 (0.595) Ho: P1+P2=0 (0.311) (0.648) (0.730) (0.000) (0.000)

(0.000)

Specification tests (P-values)

Sargan test:

(1.000)

Serial correlation

First order Second order

(0.000)

(0.512)

Third order

(0.548)

Wald test

Ho:p1+p2=0

Linear restriction

(0.574)

(P-value)

Ho:p1+p3=0

(0.252)

Ho: P1+P2+

p3+p4=0

(0.177)

Ho: P5+P6=0

(0.474)

Note: Values are from GMM system one-step estimation; the dependent variable is the ratio of gross investment to lagged net fixed assets. Numerical results come from the one-step covariance estimators, except the p-value of the Sargan test, which corresponds to second-step estimates. Heteroskedasticity-corrected standard deviations are used to calculate the p-values presented in parentheses. All models use the cash stock ratio except the model in column (6), which uses the cash flow ratio instead. Time fixed effects (not shown) were estimated when most coefficients were significant, as in columns (1), (2), and (5). Instruments for the difference equation (the instruments are included if the variable is present in the model equations) are level instruments: all variables dated t-2 and t-4. Instruments for the level equation are level instruments, dummies. Instruments in differences are all variables dated t-1. The series period is 1993-2000; the longest time series is 8, the shortest is 1. Source: Author's calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

248

CASTANEDA

dummy variables). The regressions with cash stock have the best fit in terms of p-values for the individual coefficients and the joint significance of the model. In fact, as can be seen in column (6), most of the coefficients have the unexpected sign when cash flow is used; moreover, all but two coefficients in this column are not statistically different from zero. The coefficient's sign for the pooled cash stock variable is positive, as expected from theory, in four of six columns in table 7.3. Likewise, in half of the regressions it is also statistically significant in the one-step estimation. Thus, the cash stock of associated firms spurred investment for the average member firm during the financial liberalization period. In other words, this model presents empirical evidence that validates the hypothesis of financial relaxation in network firms due to the workings of an internal capital market. The only difference between columns (1) and (2) in table 7.3 is the lagged investment ratio variable, which is not statistically significant in the first regression and removed in the second regression equation. Once this is done, the coefficient for cross-financing is positive and statistically significant for the financial liberalization period. Although the individual coefficient for the pooled cash stock variable interacted with time is not statistically significant, the corresponding Wald test indicates that pooled cash stock for 1995-2000 is not a relevant variable. In other words, the data do not show econometric evidence of a working internal capital market for the financial paralysis period. Moreover, the remaining Wald tests provide similar conclusions to those found when estimating equation model (7.1). In column (3) in table 7.3, when the criteria for the standardization of cash stock are modified, similar results are reached, as can be seen from the Wald tests. Thus, there is econometric evidence of an internal capital market operating in the first half of the 1990s, not only because member firms are less financially constrained than independent firms, but also because the individual investment rate of firms is positively related to pooled cash stock. This relation disappears during the financial paralysis period. In columns (4) and (5), cross-financing is theoretically associated with firms' bank links for both standardization criteria. Only the first of these regression equations provides evidence of a working internal capital market for the financial liberalization period. Furthermore, as in the previous regressions, the Wald test rejects the existence of this form of cross-financing for the financial paralysis period. Notice also that in the latter period, firms

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCING CHOICES OF MEXICAN FIRMS

249

with bank ties show a positive and statistically significant investment-cash stock sensitivity, as in the estimations for equation model (7.1). It is striking that large independent firms not linked to a bank do not show financial constraints once the Mexican banks collapsed and financing through the domestic capital market practically came to a halt. However, the results presented here seem to be robust because the same story holds when the time dummy variable is interacted with either cash stock or pooled cash stock. It is important to emphasize that these findings do not indicate that the internal capital market ceased to exist during the crisis years. It is possible that export firms were issuing international bonds in order to finance their own investment as well as other firms' investments within the same group. Even if this feature were true, it may not necessarily appear in the econometrics because yearly data and one-year lagged cash stock might not capture a dynamic internal market. A plausible explanation could be as follows. The opportunity cost of money was relatively low during the financial liberalization period; hence, large and healthy firms were willing to hoard cash in their treasuries even for one-year periods before using it to cross-finance firms within an internal capital market. However, money became relatively expensive once the bank crisis emerged. From that moment onward, firms did not accumulate cash; instead, they decided to use idle resources to pay back debt. Furthermore, many of these firms had obtained access to international capital markets by 1995, and thus they preferred to use this cheaper source of financing as their working and fixed capital. Therefore, pooled funds in this period came from retained earnings and foreign sources. Consequently, internal capital markets were more active in terms of their volume of operations, and more efficient in terms of their speed in channeling resources from one firm to another. All in all, the regression results show only that there was a change in the financial structure in 1995 in comparison with the financial liberalization period, and that this change was not in favor of more internal financing but less. This finding is consistent with the swift macroeconomic recovery of the Mexican economy since 1996. However, the source of the firms' liquidity remains an open question. Four alternatives seem feasible: international capital markets, internal capital markets, suppliers' credit, or a combination of strategies. The evidence that internal capital markets operated before the crisis and that even nonexpert firms were not financially

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

250

CASTASEDA

The Mexican Paradox Macroeconomic Context9 In the second half of the 1990s, the Mexican economy experienced a severe financial crisis. After a badly managed financial liberalization and a disruptive overshooting of the exchange rate, many banks became bankrupt during 1995 and 1996, and the entire banking community was overburdened by massive defaults on loans. In the first two years of the financial crisis, the Mexican government implemented a wide variety of bailout programs, which entailed heavy fiscal costs. Despite these efforts, the high ratio of nonperforming loans to outstanding debt created extreme liquidity problems and new lending was practically interrupted. In the first year of the crisis, aggregate demand sharply contracted and the annual GDP growth rate fell to -10 percent by mid-1995. The depressed demand levels in nontradable activities and the financial distress experienced by most nonfinancial firms contributed to the paralysis of the Mexican financial system. Over the years, external sources of financing were sharply reduced, at least through the traditional channels of bank, money, and capital markets. Real outstanding debt granted by commercial banks to the nonfinancial private sector diminished by 72 percent between 1995 and the first half of 2000. Likewise, the net flow of financing channeled through the Mexican securities market fell from an annual average of US$6.23 billion in 1991-94, to $1.96 billion in 1996-99 (Castaneda 2001a). However, as the initial panic came to an end, the Mexican economy started to show promising signs of recovery. Not only did the economy rebound within a year, it grew steadily, averaging an annual rate slightly above

9

It is beyond the scope of this chapter to empirically test the moral hazard theory stated in this section. We lay out the theoretical argument to give an explanation of some of the coefficient signs in the investment equation, which at the same time is consistent with the macroeconomic stylized facts presented here.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

constrained after the crisis makes a strong case in favor of the dynamic internal capital market story. Nonetheless, further research is needed with a more detailed data set to provide sound econometric evidence.

FINANCING CHOICES OF MEXICAN FIRMS

251

10 Lederman and others (2000) provide evidence that aggregate investment in Mexico during 1980-99 was linked to the performance of the tradable sector. These authors find that the multiplier effect of tradable output on growth of fixed investment is two times larger than the multiplier effect of nontradable output.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

5 percent during 1996-2000. Moreover, other macroeconomic indicators improved: internal savings as a share of GDP increased from 14.7 percent in 1994 to 20.3 percent in 1999; the current account deficit decreased from 7 percent of GDP in 1994 to 2.9 percent in 1999; and inflation decreased from 52 percent in 1995 to less than 10 percent in 2000. From the demand side, the main engine of this noninflationary growth was undoubtedly the export sector, which increased at impressive rates after trade liberalization. In addition, the positive effect of the rapid growth in exports was reinforced by the swift recovery of fixed investment. The ratio of exports to GDP increased from 15.2 percent in 1993 to 32.7 percent in 1999. Mexico is the eighth-largest exporter in the world and the second trading partner of the United States. The tradable sector was important in spurring growth, not only because of the dynamism of export demand, but also because fixed investment in this sector was heavily stimulated. The strength of fixed investment was increased by the initial sharp real depreciation, which drastically decreased the amount of U.S. dollars paid per man/hour in 1995. While the annual average rate of export growth between 1996 and 1999 was close to 13 percent, fixed investment grew on average one percentage point higher.10 Moreover, the recovery spread to nontradable activities and, as soon as 1997, the economy was experiencing the highest growth in two decades. Hence, the macroeconomic upturn is a striking phenomenon that deserves further explanation. A plain export-led-growth argument does not seem to capture the whole story in a context of financial disarray. It is not enough to argue that demand-side multiplier effects were able to pull the rest of the economy. Such outstanding performance in the real sector would not have been possible without financial flows moving from the booming export sector to the nontradable sector. Export firms, besides being the best candidates for obtaining domestic financing, have issued bonds and equity in international markets, supported by their dollar-denominated income flows. Several analysts, including Lederman and others (2000) and Krueger and Tornell (1999), have argued that access to U.S. financial markets by

252

CASTANEDA

Some interesting findings from this survey are the following: (i) suppliers' credit was large during 1998-2001, representing 40 to 50 percent of total financing; (ii) credit granted by commercial banks has been less than half the credit provided by suppliers since 1999; (iii) credit from parent companies or some other companies in the group oscillated around 15 percent in the four years of the sample; (iv) suppliers' credit is larger for nonexpert firms; for instance, it was 54.7 percent for nonexpert firms in 1999/1, which was eight points above the percentage observed among export firms; (v) the importance of suppliers' credit varies inversely with the size of the firm; for instance, 57.7 percent of funding was for the smallest firms and 22.2 percent for the largest in 1999/1; (vi) size is a key determinant for having access to funds from foreign banks: the largest firms obtained 29.6 percent of their credit from this source in 1999/1, which is much larger than the 2.3 percent obtained by small firms; and (vii) about 60 percent of the surveyed firms did not received any bank credit in the last four years. 12 Shin and Park's (1999) econometric results do not reject the existence of an internal capital market for the Korean business groups known as chaebols.

11

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Mexican firms producing tradable goods was a key factor in explaining the recovery after 1995. The former study shows that aggregate investment reacted to U.S. real interest rates during this period. However, sustained aggregate growth also required that the surplus cash flow of booming firms be channeled to the rest of the economy. In other words, increased demand for nontradable goods could be met only if financing were available to the producer. In fact, the data show a swift recovery of domestic sales of durable and nondurable goods during the period, creating not only a demand for working capital financing but for capital spending as well. According to a survey of 500 firms carried out by Mexico's central bank for 1998-2001, the financial structure of Mexican firms relies more on trade credit from suppliers than on bank credit or any other source of external financing.11 Consequently, a change in financial structure might well be an important ingredient in explaining a recovery. Trade credit is commonly observed in countries that have asymmetric information problems that hamper the functioning of external capital markets. Biais and Collier (1997) and Petersen and Rajan (1997) present a formal explanation of trade credit based on adverse selection considerations. However, trade credit is not the only missing link in the export-led-growth theory; the existence of business groups contributed to the formation of a strong internal capital market, thus making possible the speedy recovery of the Mexican economy.12 As in many developing economies, Mexico's industrial structure is characterized by networks of firms tightly controlled by closed groups of

FINANCING CHOICES OF MEXICAN FIRMS

253

13

For a description of Mexican groups, see Castaneda (1998, 1999).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

owners, usually members of the same family. These business groups tend to be vertically integrated and widely diversified (Khanna 2000). Khanna analyzes the two sides of the debate on the presence of business groups in an economy: either groups substitute for missing outside institutions and have positive implications for society at large, or they produce exploitation and rent seeking by majority shareholders.13 Because of the lack of a well-functioning capital market, perhaps due to institutional links with the social and corporate governance arena, manager/owners prefer to set up large conglomerates, which tend to stabilize aggregate profits (Aoki 2001). Moreover, the largest shareholders of these networks typically own a financial group and/or a bank, which allows them to avoid being rationed out from the use of scarce savings. The empirical literature on unrelated business diversification of U.S. conglomerates, summarized in Lang and Stultz (1993) and Montgomery (1994), shows that the performance of affiliated firms (or divisions) is poorer than that observed in independent firms in the same industries. However, there are theoretical reasons to expect an inverted relationship in emerging markets, as suggested by Khanna and Palepu (1997). The absence of intermediaries, limited protection offered by property rights, and weak enforcement of law create large agency costs that handicap the functioning of a formal capital market. Thus, according to Leff (1976) and Khanna and Palepu (1998), firms are encouraged to build networks where an internal capital market arises. Through this internal market, groups diversified across unrelated business activities smooth out income flows, overcoming financial constraints for some of their affiliates. This argument suggests that, in the case of Mexico, the structure of business groups might have contributed to the economy's recovery, especially in cash-constrained firms with no direct access to international capital markets. The banking system's collapse and the interruption of financing flows through the domestic financial system were overcome by a change in the firms' capital structure. Many firms became more dependent on trade credit, and the business groups' internal capital markets created a financial cushion that kept the economy working. Through this internal capital market, cash-rich firms were able to channel cash flow to liquidity-constrained

254

CASTANEDA

An Intuitive Story for Solving the Paradox A tentative explanation for some of this chapter's econometric results is that internal capital markets became very dynamic in Mexico after traditional financing was interrupted.14 This interpretation is consistent not only with the lack of sensitivity of investment to cash stock after 1994, but also with the fact that aggregate production was not permanently halted due to a lack of bank financing during the financial paralysis period. Thus, the argument is that resources coming from foreign sales, international issues of financial assets, repatriation of capital flight, and proceeds from the bailout of bank debts and the sales of companies were allocated through internal capital markets. At the beginning of the crisis, depressed aggregate demand levels caused a severe moral hazard problem. Thus, individuals decided to stop investing through external markets in firms with a bleak perspective, mainly oriented to the domestic economy. At the same time, investors increased their lending to booming firms, which were mainly in the export sector. These firms used the financial resources not only for productive activities, but also to offer some financing. The recipients of credit were bank-rationed firms that were financially sound and had network connections with the lender. Consequently, cash-rich firms started to offer trade and direct credit to affiliates even in nonexpert sectors of the economy; this allowed the former firms to obtain valuable inputs for their regular production and to make additional profits from financial activities. External investors were aware of the existence of internal capital markets, yet they willingly lent money because the business group structure guaranteed the expected return. That is, parent firms reduced the opportunistic behavior of borrowing firms, and hence the initial moral hazard problem was attenuated. This story is a possible explanation of why, in a context of financial disruption, certain sectors did not enter into a severe recession, and why
For an overview of the efficiency of internal capital markets and the nature of their agency costs, see Gertner, Scharfstein, and Stein (1994) and Scharfstein and Stein (1997).
14

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

network affiliates. Obviously, in this setting, firms belonging to a business network or supplying to export firms have better chances of surviving.

FINANCING CHOICES OF MEXICAN FIRMS

255

the demand spillover from a booming sector led to higher output despite the presence of a bank crisis. The essence of the microeconomic argument is that although a firm's expected production may reach a low level, investors may still have an incentive to keep lending to that firm as long as they have control of the borrower's decisions, as is the case in network transactions. The argument is presented formally in Castaneda (200Ib). It is important to clarify two key elements in this line of reasoning: first, the nature of the affiliates' choice of financial structure before the crisis, and second, the incentives of outside investors to provide funding despite the diversion of resources through internal capital markets. With regard to the first element, the story assumes that, under stable market conditions, network firms might select a rather loose governance structure to increase market pressure on divisions. This strategy would reduce transaction costs presented in large conglomerates with long hierarchies and centralized decisionmaking. Under a loose structure, affiliates operate as profit centers, with a large degree of leeway, in particular on financing issues. At the same time, the affiliates endure the financial cost of being partially rationed in credit markets. In other words, affiliates under an autonomous structure not only need to rely more on their internal cash flow, they are also precluded from extensively using cross-financing with other network firms. For this to be a rational decision, it has to be the case that the benefit from the low transaction costs more than offsets the financial cost of being cash-constrained. Hence, in an investment equation model, this strategy might produce sensitivity of investment to cash stock, irrespective of whether the economy is experiencing financial liberalization. It can be assumed that as the economy entered into a more uncertain environment in 1995, the relative importance of these two costs was reversed. In the new macroeconomic scenario, network firms were more interested in coping with financial constraints, in spite of experiencing a loss in efficiency due to the introduction of a tight decisionmaking process. Under this setting, the cash stock coefficient in an investment equation would tend to be close to zero, given the cross-financing taking place among the network's affiliates, irrespective of whether there is paralysis in external markets. With respect to lending incentives, the second key element, outside investors are aware that they do not have control rights in an arms-length relationship with the borrowing firm. Therefore, only under appropriate

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

256

CASTANEDA

Conclusions The econometric results presented in this chapter coincide with previous findings in the literature that support the view that investment behavior is conditioned by the presence of asymmetric information in financial markets. In particular, the estimations for 1993-94 are consistent with those studies in which the firm's liquidity is defined by whether it belongs to a business network. Firms listed on the Mexican Securities Market were classified as either

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

market conditions would they be willing to offer credit, namely, only if the borrowing firm had the incentive to pay back the loan. However, even if this condition were not present for some firms, outsiders might still be willing to lend if a third party would guarantee repayment. When the third party belongs to a network and has controlling rights in the cash-constrained affiliate firm, such as in the case of a parent company, outsider money might come in. On the one hand, the third party might be able to exert control over the affiliate in a business network where cross-shareholdings and interlocking of directorates prevail. On the other hand, if it experienced a high demand for its product, the third party would have an incentive to pay back the borrowed funds, including those that were diverted through the internal capital market. In this fashion, the initial moral hazard problem of an arms-length transaction with a fragile borrower would be removed by combining an arms-length transaction involving a cash-rich borrower with a network transaction between such a borrower and a cash-constrained associate. Although this story obviously oversimplifies the Mexican experience, it brings together five important stylized facts of the country's recent macroeconomic paradox: (i) a drastic reduction in financing through domestic credit markets in the context of a currency and bank crisis; (ii) a booming foreign market that expands the demand for goods produced by some large firms; (iii) international issues of financial assets by export firms; (iv) increased relevance of suppliers' credit and internal capital markets; and (v) the rapid recovery of the Mexican economy since 1996, including the nontradable sector. It is important to emphasize that we have not tested this theory through the econometric models presented in the previous sections. It only offers a micro foundation to help explain some of the estimated results.

FINANCING CHOICES OF MEXICAN FIRMS

257

independent or associated with extended business groups, as the available data allowed. The latter definition corresponds to those firms with at least two board members sitting on the board of at least one other firm. The main results derived from the two regression models are the following: (i) independent firms' investment was limited by liquidity constraints only during the financial liberalization period; (ii) there was no statistical connection between cash stock and firms' investment during the financial paralysis period (1995-2000) for independent or network firms; (iii) healthy firms with bank ties were not financially constrained during the financial liberalization period, yet the situation was reversed in 1995; (iv) pooled cash stock was a determinant of network firms' individual investment rate only in the early period of the sample; and (v) investment by export-oriented firms exhibited a larger cash stock sensitivity than that of nonexport firms during financial liberalization, but the relationship was statistically close to zero during financial paralysis for both types of firms. In the analysis across firms, results (i) and (iv) suggest the existence of an internal capital market within each business group, in which affiliate firms, despite being rationed out from the external capital market, can still have access to liquidity through the use of cross-financing. These resources are either a direct source of financing or they work like shared collateral, which allows cash-constrained members to get better credit conditions from external markets. Moreover, result (iii) indicates that when the market assigned financial resources, it discounted the connection between firms and troubled banks after the bank crisis of 1995. Finally, if the treasuries of export-oriented firms are used to cross-finance nonexport but affiliated firms, a preliminary explanation can also be offered for unexpected result (v). This reasoning is valid in a context of limited access to foreign financing, which indeed was the case even for export-oriented firms in the first half of the 1990s. In the analysis across periods, result (ii) shows, somewhat paradoxically, that the period of bank paralysis was not a severe financial bottleneck, at least for listed and nondistressed firms. Supposedly, alternative means of financing, such as suppliers' credit, were used more intensively after 1994. That is, the finding is consistent with the presence of internal capital markets where borrowers and lenders have intertwined control rights; however, this explanation is not thoroughly validated by the econometrics presented here. Theoretically, the empirical evidence can be rationalized by a change in in-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

258

CASTANEDA

15 For an empirical analysis of the implications of business groups in Asia, see Claessens and others (1999).

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

centives once stable market conditions disappeared in 1995. Presumably, between 1995 and 2000, firms might have preferred to deal with the transactions costs of having a large hierarchical organization rather than being financially constrained. This feature allowed moving from cash-constrained firms organized as autonomous profit centers to tightly managed affiliates enjoying the benefits of internal capital markets. In addition, from a macroeconomic perspective, the existence of these internal capital markets was probably helpful. They might have worked as financial buffers to avoid a prolonged recession and even contributed in the recovery of the economy, given that the banking system had stopped lending throughout the late 1990s. The Mexican episode represents an interesting case of a positive externality of business networks.15 In the theoretical literature on comparative corporate governance, group structure is usually considered the result of market failures. In particular, lack of intermediaries and endemic information asymmetries are partially offset by the formation of business groups and internal capital markets, which reduce opportunistic behavior and make economic activity possible. The Mexican experience analyzed here provides a case where this type of governance structure may have fulfilled a crucial role in financial activities.

Investment and Financial Restrictions at the Firm Level in Uruguay
Julio de Brun, Nestor Gandelman, and Eduardo Barbieri
Development of the capital market in Uruguay did not accompany the deepening of the banking sector after the financial liberalization that took place in the 1970s. Instead, Uruguay can be considered a typical case of a bankbased financial system, as opposed to a market-based system in which the capital market plays an important role in determining firms' financial structure. The main objective of this chapter is to evaluate the effect of financial restrictions on firms' investment in the context of a system based on shortterm credit by banks. Recent expansion of the use of corporate bonds as a financing instrument raises several issues of interest. This instrument allowed some firms to finance restructuring projects that were extremely difficult to fund through the banking sector. However, recent default (or near-default) episodes are reminiscent of agency problems and opportunistic behavior by shareholders in relation to bondholders in cases of long-term debt. Those cases of default have discouraged investors from participating in the incipient capital market since 1998, at least with volumes similar to those registered in 1996-97. Macroeconomic shocks, like the Russian crisis of 1998 and the devaluation of the Brazilian real in 1999, have caused recent contractions in credit and corporate bond issues to the market. A second objective of this chapter is to evaluate the impact of these changes in economic and credit conditions on the investment decisions of firms.

Julio de Brun is president of the Central Bank of Uruguay, Nestor Gandelman is a researcher at the Centro de Estudios de la Realidad Economica y Social and professor of economics at Universidad ORT Uruguay, and Eduardo Barbieri is a professor of economics at Universidad ORT Uruguay.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CHAPTER 8

260

DE BRUN, GANDELMAN, AND BARBIERI

The Macroeconomic Environment and the Financial Market Uruguay has experienced an intense and long depression since 1999. Gross domestic product (GDP) growth has been mildly negative during this period. As shown in figure 8.1, private investment dropped 14 percent in 1999 and another 14 percent in 2000. Nevertheless, the government has been successful in keeping the fiscal deficit under control, benefiting the country's image in international capital markets and contributing to a low country risk premium relative to the region. Uruguay has a relatively developed financial system, a "European" system characterized by concentrated public banking and a few local banks. The country's stability, especially compared with Argentina and Brazil, makes Uruguay's financial markets a safe haven in relative terms. This situation has led to a large amount of total deposits, especially by non-Uruguayan account holders. Overall, the financial sector has exhibited stable behavior and is currently undergoing a slow process of consolidation and mergers, as well as a

Figure 8.1. Gross Capital Investment in Uruguay, 1988-2000
(Thousands of 1983 Uruguayan pesos)

Source: Central Bank of Uruguay.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

An additional area of interest is to evaluate how firms' characteristics impact their access to credit. The chapter's third objective is to test whether firm attributes such as size and ownership structure (foreign versus domestically owned) matter in determining the availability of financial resources.

FINANCIAL RESTRICTIONS IN URUGUAY

261

• Ratios of overhead costs and bank net interest margins are lower than those of Argentina and Brazil. • The ratio of liquid liabilities to GDP is higher than it is in Argentina and Brazil. • The ratio of bank assets to GDP and the ratio of claims of deposit money banks on the private sector to GDP are higher than in Argentina, and similar to those in Brazil. But when the indicators of the Uruguayan banking system are considered in a wider environment, there seems to be a long road ahead to achieve a well-developed financial system. The indicators of banking development for Uruguay in the Demirgii^-Kunt and Levine database are below world averages, and so it is considered, in this classification, as an underdeveloped banking system. Whether this restricts firms' ability to invest is the main issue to be examined in this chapter. The Uruguayan capital market is even less developed than the banking system. Market capitalization as percentage of GDP is less than 1 percent, total value traded is almost insignificant, and the turnover ratio (the value of stock transactions relative to the size of the market) is around 3 percent. World averages, according to the Demirgu^-Kunt and Levine database, are 39 percent, 17 percent, and 35 percent, respectively. Given these figures, Uruguay can be classified as an underdeveloped, bank-based economy, as opposed to a market-based system. Correspondingly, capital markets have played a minimal role in the development of the financial sources of funds. In 1990, the market value of private sector securities accounted for only US$0.8 million of the $721 million securities market. The first figure represents the value of stock outstanding, constituting

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

restructuring of its institutions to accommodate the number of firms to the size of the market. In 2001, actions were taken to ensure the sale and restructuring of troubled institutions, and banking institutions are developing a range of new products in accordance with global trends. Given these adjustments, average profitability has been acceptable despite the recession. The Demirgiic-Kunt and Levine (1999) database on financial systems development confirms the favorable position of the Uruguayan banking system in comparison with the region. Uruguay's banking system has the following characteristics:

262

DE BRUN, GANDELMAN, AND BARBIERI

the only private sector securities at that time. In comparison, total bank deposits in both local and foreign currency accounted for $10,747 million, while total loans accounted for $7,401 million in 1995 on average. Uruguay experienced a period of strong growth between 1996 and 1998 and, as is usual, during that period the credit of the banking sector experienced a procyclical evolution. The credit of the private banking sector (the state-owned Banco de la Republica is excluded because its behavior cannot be characterized as profit maximizing) to the domestic private nonfinancial sector grew at an average rate of 18 percent a year in real terms between December 1995 and December 1998. After the central bank introduced prudential restrictions for the banking sector at the end of 1998, the growth rate of private banks' credits to the domestic nonfinancial sector (excluding the government) declined to 3.8 percent a year in 1999 and 2000. In addition to the impact that macroeconomic conditions may have had on investment, another issue worth addressing is whether this tightening in financial conditions has affected firms' decisions to invest. Its small size notwithstanding, the Uruguayan capital market also experienced an expansion cycle, which was supported by institutional changes, followed by a contraction phase. Starting in 1996, new legislation and rulings regarding private sector debt (obligaciones negotiables) were introduced. Before that, in 1994, the maximum amount that could be issued by each firm was set at the equivalent of 50 percent of capital. In 1996, the Securities Markets Law (Law No. 16.749) was passed. This law established the issuing procedure and the role that the central bank had to play with regard to transparency and investor information (among other requirements), and helped with the organization of formal markets. The market then grew rapidly until 1999. Other legislation organized and favored the securities market, such as the Mutual Funds Law (Law No. 16.774) and the Pension Funds Law (Law No. 16.713). The Mutual Funds Law set up and regulated mutual funds and how they could invest in each type of security. The Pension Funds Law set the percentage of private sector securities that each pension fund could have in its portfolio. With respect to private bonds and commercial paper, the new law of 1996 created a more adequate regulatory framework for private paper. The results were immediate, and new private issues accounted for about 20 per-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCIAL RESTRICTIONS IN URUGUAY

263

Table 8.1. Total Outstanding Debt in the Financial Market, 1990-2000 Year 1990-92 1993 1994 1995 1996 1997 1998 1999 2000
Source: Central Bank of Uruguay.

Issuers

Value (millions of U.S. dollars)

0 1 11 17 27 43 45 43 39

0 4.0 56.3 160.5 309.8 544.7 602.7 759.5 688.0

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

cent of capitalization in 1997 and 1998. Around that time, most sizable firms were family owned and managed, which in some instances resulted in undesirable consequences in capital markets. Unfortunately, the level of leverage for private issuing firms was far from desirable, and this adversely affected the market and its credibility when the first default became public. Worse still, the case proved to be a conspicuous scam: the resulting scandal affected all private issues and had a chilling effect on new issues. This brought about a series of new regulations regarding credit rating and information disclosure. The year 2000 witnessed a fall in total financial market debt outstanding for the first time in seven years. Table 8.1 shows that total security debt fell by nearly $70 million, or about 9 percent. Nevertheless, a closer look at the figures raises additional concerns. First, there has been a significant change in the sector distribution of private sector debt, as can be seen in tables 8.2 and 8.3. In 2000, 75 percent of the new issues were from financial firms, compared with 40 percent in 1996. There has been a steep fall not only in total issues, but also in issues by nonfinancial firms. These developments suggest an interesting situation for analysis in relation to the behavior of investment and the conditions of financial markets. Uruguay experienced soft credit conditions until 1999, both because of the expansion of domestic credit and the appearance of new financial instruments in the capital market, despite its small size. After the Brazilian crisis and other external shocks since the end of 1998, along with problems involving domestic firms that went to the market in previous years and experienced default, near default, or default problems, conditions in the financial

264

DE BRUN, GANDELMAN, AND BARBIERI

(Percent) Sector Financial sector Nonfinandal sector Tourism Services Chemistry Industry Agro-industry - Not available. Source: Central Bank of Uruguay.

1996
35 65
— — — — —

2000
66 34 15
5 2 4 8

market became more stringent. Whether the decrease in investment in 1999 and 2000 was related to more severe credit conditions or the consequence of a demand slowdown is a question to be addressed later in the chapter. First we lay out the theoretical framework for the analysis.

The Theoretical Model We use two theoretical approaches as a framework for the empirical analysis. The first approach derives the optimal level of the capital stock as a function of output and the user cost of capital. This approach assumes a specific mechanism of adjustment between desired and actual levels of capital to obtain an investment equation. The second approach explicitly introduces financial constraints in the investment decision process of a profit-maximizing firm. The first-order conditions of the optimization process help to derive an Euler equation that relates the investment ratio of the firm to financial variables and other determinants.
Table 8.3. Sector New issues (number of issues) Financial sector (percent) Nonfinancial sector (percent) Source: Central Bank of Uruguay. Composition of New Issues, 1996 and 2000 1996 19 40 60 2000 9 75 25

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 8.2.

Debt Capitalization, 1996 and 2000

FINANCIAL RESTRICTIONS IN URUGUAY

265

This first approach is more flexible (and ad hoc) than the second approach. The first approach assumes that firms are profit maximizers, but does not explicitly introduce financial constraints or model the decision to pay dividends or issue new shares. Therefore, a firm's problem is to maximize:

subject to:

where E is the conditional expectations operator, p is output price, F(«) is a gross value-added function, K is the capital stock, N is labor, / is gross investment, pK is the price of capital goods, and 8 is the depreciation rate. If the price of firm output and the price of the investment goods are constant over time, the steady-state solution is:

where r denotes the interest rate and uc the user cost of capital. If prices are allowed to vary over time, the solution has an extra term:

The extra term on the right-hand side of equation (8.4) reflects the capital gain (or loss) due to a change in prices. Assuming a functional form for the production function, it is possible to obtain the basic relationship between capital and output. Under a CobbDouglas specification, Yt = F(KnNthjfhj) - AtK"Nf, where Y denotes outp equation (8.4) can be rewritten as:

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Accelerator Model of Investment with Error Correction

266

DE BRUN, GANDELMAN, AND BARBIERI

Under a more general CES production function such as

the relation between capital and output is:

and v are the elasticities of substitution and scale, respectively. Both equations (8.6) and (8.6') imply that the long-run capital level is proportional to output (and more generally to sales), a term reflecting the user cost of capital, and the parameters of the production function.1 Consider now a specific firm. Since adjustment is not instantaneous, the following dynamic adjustment specification between capital and output, provided by sales, is explored:

Rewriting in error correction form, we have

In addition to the ht term, the growth rate of capital depends on the growth rate of sales, an error correction term, and a scale factor. The estimation of equation (8.8) assumes that the term reflecting the user cost of capital and parameters of the production function can be controlled
1 Note that equations is a particular case of eation when q-cse of equation tion (8.6')6') when qqqq== 1.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

and, taking logs,

FINANCIAL RESTRICTIONS IN URUGUAY

267

—— is used to proxy the growth rate of capital. Equation (8.8) is augmented Kt-i with the profit-to-capital ratio to control for financial constraints in this relationship. Summing up, the equation to be estimated is:

where T)!f = e,( + a, + dt and Hit represents the profits of firm i at time t. The error correction coefficient yi - 1 is expected to be negative, implying that when the capital level is above the desired level, investment will be lower. The scale coefficient YI - 1 + PO + Pi is expected not to be statistically different from 0. If the profit terms capture only transitory effects, the sum of the coefficients on profits cp = q>0 + (pi would not be significant. Another possible specification of the investment equation is the traditional accelerator profit model that is derived by differencing equations like (8.7), removing the possibility of an equilibrium relationship of the variables in levels.2 Using the investment rate as a proxy for net growth in capital stock, the following expression is obtained:

In this specification, the implied long-run parameter

characterizesizess

the long-run relationship between the variables in differences (sales growth and investment) and not in levels (sales and capital).
The error correction specification is a consequence of a long-run relationship between the variables in levels. If the error correction is not a good description of the data process, then the variables in levels are not cointegrated and there is no long-run relationship between them.
2

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

by including year-specific and firm-specific effects. The investment ratio

268

DE BRUN, GANDELMAN, AND BARBIERI

Here we consider a model of the investment and financing decisions of firms that explicitly introduces restrictions on debt-equity financing. These restrictions can be tested using an Euler equation approach, as in Jaramillo, Schiantarelli, and Weiss (1996), or by a more ad hoc regression strategy (introduced above), as is more common in the literature. Firms are assumed to maximize their value for shareholders, given by:

subject to:

where Wis the value of the firm, D is dividends (or cash flow), Q" is the nominal value of new shares, E is the conditional expectations operator, T is the corporate tax rate, p is output price, F(-) is a gross value-added function, G(-) is a convex adjustment cost function (both assumed to be linearly homogeneous), K can be interpreted as total assets or capital stock, N is labor, / is gross investment, pK is the price of capital goods, v is the fiscal impact on the firm's flow of funds of the tax deductions from depreciation of the stock of capital, B is total debt, i is the risk-free rate of interest, and p(-) is the premium per unit of debt, which can be expressed

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Euler Equation Approach

FINANCIAL RESTRICTIONS IN URUGUAY

269

From the first-order conditions for K, I, N, B, and Q" of the problem of maximizing equation (8.11) subject to restrictions (8.12a) through (8.12e) when equation (8.13) is not binding, we derive the following Euler equation for the case of no financial constraints:4

Equation (8.14) determines the (incremental) investment-to-capital ratio as a function of the product-capital ratio (or the sales-to-capital ratio) and two terms that reflect the opportunity cost of capital.
As the literature of agency costs on financing shows, problems of asymmetric information may introduce conflicts of interest between bondholders and stockholders, resulting in a higher premium on the risk-free rate of interest as the leverage ratio (debt to total assets) increases. Other explanations rely on reasons not related to agency problems. Caprio and Demirgiic-Kunt (1997) survey theoretical arguments and provide evidence on the intention of firms to match the maturity of assets and debts. Demirgiic-Kunt and Maksimovic (1994) also argue that firm size may be relevant for financing choices because access to financial markets may be a function of size. 4 For details on this procedure, see Jaramillo, Schiantarelli, and Weiss (1996).
3

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

as a function of leverage (B and K at the end of the previous period) and size (K).3 In addition to restrictions (12a) through (12e), the firm maybe subject to financial constraints that could affect the leverage ratio. As in Jaramillo, Schiantarelli, and Weiss (1996), a restriction indicating an upper limit to the leverage ratio is introduced in the maximization process:

270

DE BRUN, GANDELMAN, AND BARBIERI

Equation (8.15) is similar to the one obtained for the case with no financial constraints, equation (8.4), except for the inclusion of the last two terms. These terms incorporate the effect of the leverage ceiling on the investment ratio. Making a few assumptions from the expectations terms and after an appropriate parameterization of the adjustment cost function G(-) and the premium term p(')> equation (8.5) can be estimated, as will be done in the following section. The assumptions are as follows: • Rational expectations. This assumption implies that Et(Xt+l) = Xt+l + «t+1, u being a white noise expectation error not correlated with any period t information. • Adjustment cost function. It is assumed that G(') can be written as

• Premium. The premium function p(-) is supposed to be linear in the debt-to-capital ratio, that is,

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

If constraint (8.13) is binding, the Euler equation in the presence of financial constraints will be:

FINANCIAL RESTRICTIONS IN URUGUAY

271

Equation (8.14') is the specification under the hypothesis of nonbinding financial constraints and (8.15') is the alternative specification when financial constraint (8.3) is binding. Note that subscript i denotes firms. In these equations, /is a fixed, firm-specific effect and year is a time-specific effect.

Empirical Tests Data Set The estimation uses annual account data extracted from two main sources. The first source consists of financial statements of firms that report their accounting data to the Superintendencia del Mercado de Valores at the Central Bank of Uruguay (the regulatory agency for the Uruguayan capital market). These are publicly traded firms and (for the most part) companies that issue corporate bonds through the capital market. The second

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Substituting these expressions for G(*) and p(-) in equations (8.14) and (8.15), and applying the rational expectations operator, we have the following parameterized Euler equations:

272

DE BRUN, GANDELMAN, AND BARBIERI

source is the financial statements of firms that provide accounting data to the Liga de Defensa Comercial, a nonprofit organization devoted to promoting transparency in the credit market. This organization usually prepares reports on the credit history of its affiliates, sometimes on the basis of financial statements. Although companies that go public and issue of shares or corporate bonds are compelled to report their financial accounts to the central bank, the information provided to the Liga de Defensa Comercial is voluntary. We collected accounting data from 54 companies that report information to the central bank and 100 firms that do the same with the Liga de Defensa Comercial. But most of that data were not useful in implementing the estimation of the models presented in the previous section because the information was incomplete in many cases, and most firms only occasionally report financial statements to the Liga de Defensa Comercial. We made two types of adjustments to the raw data. First, all the financial statements were expressed in terms of December 2000 pesos. With the second adjustment, we tried to estimate the result of having nominal accounts exposed to inflation in the financial statements. This correction typically affects the financial results of the firm and attempts to remove inflationary distortions from net income. Although most of the firms make adjustments for valuation of fixed assets and they are usually presented as values at the end of the fiscal year, the other items in the financial statements are presented as historical values. This causes a problem when the financial results are presented because interest on loans denominated in local currency is expressed in nominal terms, while loans denominated in foreign currencies explicitly introduce an adjustment for devaluation. However, the correction of financial statements for inflation must express interest payments in real terms. We also collected information about the ownership structure of firms, identifying those companies controlled by foreign entities or with significant foreign participation (including in this definition all firms in which foreign shareholders possess at least 30 percent of total shares) and those that belong to national investors (the rest). We constructed an unbalanced panel of 56 firms with at least five consecutive years of data for all the variables included in the models to be estimated. Table 8.4 presents summary information on the financial data for 1995-2000.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCIAL RESTRICTIONS IN URUGUAY

273

Indicators of investment show a picture roughly similar to that described for the country as a whole in the first section. Both gross and net investment grew until 1998, then suffered a strong contraction in 1999, when Uruguay confronted the macroeconomic problems associated with the devaluation of the Brazilian real at the beginning of the year. During 2000, the investment ratios remained low, with a notable decrease in investment by new firms, which showed the higher investment rates during the period considered. When firms are classified according to their size based on the value of total assets, investment ratios tend to be higher in larger firms. When companies are classified by taking sales into account, the investment ratios have a similar evolution in small and large firms, which suggests some kind of proportionality between investment and sales. Indicators of activity also provide a good synthesis of the scenario that Uruguay confronted during the second half of the 1990s. Sales growth increased until 1997 and then began to decelerate. Asset turnover (sales/total assets) also slowed after 1997, causing deterioration in profitability ratios, as will be shown below. The asset turnover ratio tends to be higher in larger firms when they are defined according to the value of their assets. Again, the distinction between large and small firms is not so clear when they are classified by sales volume. Profitability ratios show a steady decline during the period considered, with notable reductions in 1999 and 2000. Net operating income (income before financial charges and taxes) became almost zero in 1999 and 2000, while net income became negative in the last two years of the period. There is no clear distinction in the relative performance of large and small firms during the recent crisis. When size is defined according to total assets, small firms performed better in 1999, but when total sales is used as an indicator, the picture is the opposite. On a different dimension, the recession seems to have more substantially affected the results of new firms, defined as those created during the 1990s. The ratios of indebtedness were remarkably stable during the period under study, and were similar for all types of firms. Old firms tend to have larger leverage ratios than new firms (consistent with the hypothesis that stresses the importance of age in access to credit), and large firms tend to be more indebted than small ones (which is also consistent with the theories that emphasize the importance of collateral in the determination of finan-

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 8.4. Summary of Indicators, 1995-2000 (Percent unless otherwise noted)
1996 1997
31.4 34.9 32.4 59.8 37.9 27.9 30.1 39.7 21.6 18.4 54.5 24.1 15.6 17.6 24.3 14.5
6.5

Indicator
1998
23.6 21.3 42.0 22.9 25.3 24.8 19.5 12.6
9.6

1995
1999
20.1 22.6
9.1 7.4

2000

Investment

Gross investment/fixed assets

14.1 11.7 45.9
3.0

22.4 19.8 53.2 19.8 35.5 24.4 32.1 30.4 14.5 10.4 51.1 16.3 11.3 15.5 12.5 36.2 23.5 165.5 50.0 13.0 52.0
8.1

Old firms

28.1 61.1

New firms

Smaller than average (total assets)

Bigger than average (total assets)

21.4 12.3 15.8
4.7 2.4 4.1 6.9

26.4 18.8 24.2

41.3
5.9

Smaller than average (total sales)

Bigger than average (total sales)

35.1 13.3 15.1 35.8 11.4 15.2 14.0
7.6 1.2 5.3 0.7

Net investment/fixed assets

Old firms

New firms

34.9 -8.1
2.5

40.3 13.7
3.8 7.5 4.3 3.3

Smaller than average (total assets)

Bigger than average (total assets)
6.0 3.0

12.4

34.2 -0.3 27.1
9.1

Smaller than average (total sales)

Bigger than average (total sales)

Activity and growth
-1.3 117.6
6.5

Sales growth

Old firms

10.3 26.2
1.9 2.7 3.2

New firms
0.1

51.8 10.0 -0.2
9.0

Smaller than average (total assets)

Bigger than average (total assets)

31.7 14.9 13.7

-0.4
0.7 2.1

19.9 -0.5 21.5

Smaller than average (total sales)

Bigger than average (total sales)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Asset turnover (sales/total assets) 132.3 142.8 132.5 109.5 61.8 71.8 24.6 64.6 57.0 42.7 93.8
0.1 5.6 1.0

133.1 145.1 153.5 143.2 122.3 25.7 124.4 66.0 109.3 127.1
4.4

Old firms 34.8 43.4 42.2 28.4 152.1 79.7 133.3 141.5 10.6
9.7

138.4 154.0 166.7 84.4 144.7 160.0 11.1 10.7 15.6 22.5 11.8
7.5

New firms 70.3 79.8 129.5 157.4 10.6 10.3 13.8 11.4 11.2 10.9 11.0 10.4 11.0 10.6 11.0
7.1 9.0

Smaller than average (total assets)

163.6

Bigger than average (total assets)

Smaller than average (total sales)

136.1

Bigger than average (total sales) 13.0 12.7 28.5 13.4 12.1 13.7 10.0 11.8 11.3 13.1 13.7
7.8

146.7

Profitability

Net operating income/sales

Old firms

New firms

-8.5
5.9 0.2 2.1 8.8 7.8 8.4 3.9

-4.4 -1.4
1.7

Smaller than average (total assets)

Bigger than average (total assets) 12.0 16.8 14.2 15.0
8.8

Smaller than average (total sales)

-5.6
7.5 2.5 3.3

Bigger than average (total sales) 17.2
9.6

Net operating income/total assets

Old firms 20.8 16.9 14.3 10.3 13.9 13.5
6.9 6.0 7.7 8.5

New firms

-0.2 12.0
7.1 8.8 5.0 1.8 3.7

Smaller than average (total assets) 19.4 15.1 14.9 13.3 10.0 47.3 13.5 15.1 20.0 20.0 22.8 13.8

Bigger than average (total assets)

Smaller than average (total sales)

10.4 12.1
5.4 4.7

6.7

-0.6 11.0 -6.7 -0.3 17.2
7.3 6.6

Bigger than average (total sales)

Net income/sales

-9.4 -7.0 14.0
6.6

Old firms

New firms

-74.6
1.7

-22.3 -11.7 (continued)

Smaller than average (total assets)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 8.4 (Continued)

Indicator

1995 33.9 13.0
6.1 6.4 5.0 2.2

1996 1997 1998 1999
-30.8 -11.1
1.8 0.7
1.6

2000
-6.6 -18.7
2.7

Bigger than average (total assets)

Smaller than average (total sales)

25.8 15.4
9.6 7.9 6.2

Bigger than average (total sales)

11.4 44.5 34.1 29.8 19.1 19.2 18.6 23.6
7.1

Net income/net worth

-8.9 -8.6 -10.1 -14.4

Old firms

45.2 34.8 29.9 28.1 36.0 14.9 32.4 22.9 12.0 60.7 61.3 54.4 63.5 52.9 63.3 55.4 38.6 37.4 50.2 35.8 45.5 47.3 57.6 40.3 39.4 49.1 39.9 41.4 22.1 58.7 58.2 64.0 60.1 55.4 59.4 -6.2
1.5

New firms

24.3 28.8 40.6 19.1 40.1 25.6 58.3 60.0 46.5 59.3 55.9 61.3 55.7 44.6 45.8 35.7 43.5 50.8 31.2 59.0 24.7 55.0 54.5 55.7 53.6 55.9 53.8

Smaller than average (total assets)

Bigger than average (total assets)

-1.8 -4.9
9.6

-0.1 -22.1
5.7

Smaller than average (total sales)

Bigger than average (total sales)

Leverage and liquidity

Total debt/total assets

57.1 58.3 49.0 57.2 56.7 57.4 62.3 42.7 42.3 45.0 40.4 49.3

59.5 59.8 58.4 61.8 55.6 63.0 60.6 53.0 52.2 55.8 55.0 49.5

Old firms

New firms

64.6

Smaller than average (total assets)

Bigger than average (total assets)

Smaller than average (total sales)

Bigger than average (total sales)

Financial debt/net assets

39.6

Old firms

38.6

New firms

56.6

Smaller than average (total assets)

37.1

Bigger than average (total assets)

44.6

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Smaller than average (total sales)
39.3 47.7 41.3 40.0 41.8 47.4 51.3 39.1 37.5 40.1 42.4 58.3

Bigger than average (total sales) 0.529 0.507
0.681 1.510 1.314 1.117 0.813 1.586 1.142 0.841 164.3 165.1 157.4 178.9 121.0 189.3 127.6 122.0 1.448 0.716

Long-term debt/net worth (times) 0.943 0.943
0.713 3.771 1.912

0.351

0.793 0.977
1.010

1.362

Old firms 0.547 0.484
1.397

0.350 0.543 0.602
1.085

New firms

0.368

Smaller than average (total assets) 0.699 0.250
143.5 144.2 138.9 157.3 110.1 149.9 135.1 139.4 134.4 132.5 198.9 216.9 189.1 216.9 162.6 190.3 174.6 192.9 173.6

0.305 0.872 0.447 0.767

Bigger than average (total assets)

0.442

0.399
2.117 0.619 116.8 111.1 141.2 107.1 133.0 118.7 110.1

Smaller than average (total sales)

0.384

Bigger than average (total sales)

0.309

Current ratio

153.9

Old firms

153.7

New firms

158.2

Smaller than average (total assets) 208.0

172.5

Bigger than average (total assets)

115.8

Smaller than average (total sales)

158.9

Bigger than average (total sales)

140.4

Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

278

DE BRUN, GANDELMAN, AND BARBIERI

Estimation Issues Two broad strategies can be followed to test whether constraints on financing decisions are relevant to the investment process of firms. One approach uses the Euler equation derived from the maximization problem outlined above. As Mairesse, Hall, and Mulkay (1999) point out, the Euler equation method allows the analyst to deal with the problem of expectations regarding future profitability of investment. Average profitability, the product-to-capital ratio, and the one-period-ahead expected change in the adjustment costs of investment are all that is needed to describe the change in expectations of the future profitability of investment. The Euler equation approach does not require computing Tobin's Q ratio, which is difficult in markets like that of Uruguay, where only a few firms trade their shares on a stock exchange. To allow for potential endogeneity of the regressors in the Euler equation, we use the generalized method of moments (GMM) estimator (seeHansen [1982] and Arellano and Bond [1991] on the application of the GMM method in panel data, and Hansen and Singleton [1982] on application of the GMM method to first-order conditions). The significance of the additional parameters included in the binding restriction model make it possible to draw conclusions about the relevance of financial constraints in

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

cial structure). But all types of companies analyzed had leverage ratios around 60 percent during the last five years. Although profitability declined, financial debts increased their importance as the funding instrument for working capital, as shown by the ratio of financial debt to net assets. This trend is visible for all types of companies described. Moreover, as profits decreased and eventually became negative, the net worth of firms grew less than total assets, so long-term debt increased its importance as a means of long-term financing. The stability of the leverage ratio notwithstanding, the current ratio (current assets/current liabilities) experienced a pronounced decline during the last two years, indicating a relative increase in short-term financing instruments as a source of funds with respect to other mechanisms, including the internal generation of resources. The latter source of funds, which is related to the capacity of the firms to obtain profits, has been negatively affected by the current economic recession.

FINANCIAL RESTRICTIONS IN URUGUAY

279

in which i = 1 . . . N, t = 1 ... T, yit is the variable to be explained, xit is the vector of explanatory variables (including lagged yit), and oc, and dt are firmspecific and time-specific effects, respectively. The year specificities can be addressed by year dummies, but possible problems arise from the a,, which if ignored can introduce persistent serial correlation of the residuals. The within transformation eliminates a,, but does not make it possible to obtain consistent estimates if the variables on the right-hand side are endogenous or predetermined. We obtain this problem by first-differencing, in which case equation (8.16) becomes:

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

the investment decision. The only difference between equations (8.14') and (8.15') is the sign of the debt term and the appearance of a term with the square of debt in the restricted case. Nevertheless, it is necessary to warn that the results from an Euler equation approach are sensitive to the specification of the model. In the present case, the specification of the p(-) and G(*) functions can introduce an arbitrary structure in the equation to be estimated. For this reason, the results obtained from Euler equations are supplemented with the use of an error correction formulation and the traditional accelerator profit model, as in equations (8.9) and (8.10). In fact, the error correction equation (8.9) and the accelerator profit model (8.10) are not very different from the Euler equation (8.15')> so the two methodologies can shed light on the same phenomenon despite the different macroeconomic foundations. The estimation procedure for equations of the type (8.9), (8.10), and (8.150 using panel data has traditionally dealt with the "permanent unobservable differences" across firms included in the error term by one of two methods: (i) carrying out within-firm transformation (subtracting from each variable the time average over the sample period) or (ii) firstdifferencing the data. Consider the usual linear regression model written for panel data with firm and year effects:

280

DE BRUN, GANDELMAN, AND BARBIERI

Results Accelerator Model of Investment with Error Correction: Traditional Approach For purposes of comparison, table 8.5 shows the within regression results for equation (8.9). As in all the regressions, the estimates are for 1997-2000 and dummies for each year are included (but their estimates are not reported). Although data for 1995-2000 are available, the first two years allow for lags of order two to serve as instruments in estimation. Table 8.5 presents the results for the basic equation as well as for several augmented equations that attempt to capture the effects of firm size and type of ownership. A priori, larger and foreign-owned firms are expected to suffer less from financial constraints. Conventional wisdom sees the acquisition of a foreign partner as a sign of strength for a firm. This strength can be the result of brand name reputation, the inflow of new capital, more efficient internal organization, or new (and probably more aggressive) strategic competition. In any event, foreign participation should ease the fundraising efforts of firms. Larger firms are expected to experience fewer financial restrictions than smaller firms. Public information about smaller firms in general is worse than information about larger firms, and there are no easy ways to solve this asymmetry in order to attract investors. Moreover, larger firms may be able to offer larger collateral, which in many cases is required for longer-term loans. Firm size is defined on the basis of average capital. The dummy variable size takes the value 1 for all firms with above-average capital. There

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

These first-difference estimates are free from potential correlated effects due to unobserved firm-specific factors that are constant over time. In the present model, it is probable that the dependent variable and some of the explanatory variables are simultaneously determined, introducing biases into the estimations. To deal with this problem, we obtained GMM estimates based on instrumenting equation (8.17) by the lagged level values of the variables. We used the Dynamic Panel Data statistical package for Ox to perform the estimations (Arellano, Bond, and Doornik 1997).

FINANCIAL RESTRICTIONS IN URUGUAY

281

The proxy for cash inflows is the contribution margin minus interest payments. The contribution margin is calculated as net operating income (earnings before interest and taxes) plus capital depreciation.

5

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

is also a dummy variable taking the value 1 for all firms with more than 70 percent national ownership (that is, origin=l is the approximation for a national firm). Also of interest is interaction of relevant variables with the effects of the 1999-2000 crisis. Several alternatives are explored, but only two are reported here: a dummy for post-crisis years 1999 and 2000, and a dummy for the spread between interest rates denominated in foreign and domestic currency. This spread reflects the expected devaluation of the local currency, given the high dollarization of the Uruguayan economy, and higher expected devaluation implies higher firm liabilities and lower access to funds. In all specifications, the error correction term has the correct sign and is significant. But the lagged sales term is also significant and its coefficient is similar to the error correction term. Implicitly, the two sales terms cancel and the investment-capital ratio is linked to the lagged log of capital. In all specifications, the coefficient estimates of the current and lagged terms of the proxy for cash inflows are positive and significant.5 This is consistent with the presence of financial restrictions. Moreover, liquidity constraints are not temporary and are not eliminated in the following periods. We explored other proxies for liquidity constraints, such as net income, net operating income, and a cash approximation calculated as net income plus depreciation of capital, all scaled by the capital stock. The results do not differ much, with all the coefficients being of similar order. Net investment (also scaled by capital) is the dependent variable in all the error correction and accelerator equations. We also used gross investment, but the results were robust to these modifications. The within results in column (1) in table 8.5 are consistent with the existence of financial constraints on investment in the period under consideration. More convincing information is contained in the other columns in the table. The regression in column (2) allows for interaction with size, and columns (3) and (4) allow for interaction with the crisis variables. Column (5) splits by ownership, column (6) includes the three interactions together, and column (7) presents what happened with small firms after the crisis. The estimates of the product of the size dummy and the cash proxy are negative, suggesting that larger firms experience lower or no financial con-

Table 8.5. Error Correction Model for Investment/Capital within Estimates, 1997-2000
(2)
(4) 213 213 213 213 213

Variable
(3)
0.233*** 0.239** (0.125) -0.867*** (0.120) -0.933*** (0.111) 0.151** (0.093) 0.234*** (0.020) 0.385 (0.124) (0.126) -0.842*** (0.121) -0.901*** (0.112) 0.104*** (0.028)
0.110

(D
0.271*** 0.245*** (0.124) -0.880*** (0.119) -0.946*** (0.109) 0.050 (0.126) 0.256*** (0.026) 0.306 -0.415*** (0.212) 0.094** (0.057) 0.380 (1.319) 4.085* (2.905)
0.081

(5) (6)

(7) 213

Usable observations

213

Dy(t)
-0.882*** (0.119) -0.945*** (0.110) 0.020 (0.056) 0.259*** (0.025) (0.092)
0.214

0.242***

0.255***

0.240*** (0.125) -0.884*** (0.119) -0.941*** (0.110)
0.014

(0.125)

(0.124)

k-y(t-1)

-0.869***

0.869***

(0.120)

(0.118)

y(t-D

-0.932***

-0.935***

(0.110)

(0.109)

C/K (t)

0.108***

0.110***

(0.016)

(0.016)

(0.056) 0.261*** (0.025) 0.274

C/K(t-1) 0.309

0.234***

0.233***

(0.020)

(0.020)

Sum of C coef.

0.342

0.343

Size*C/K (t)

-0.438***

(0.210)

Crisis*C/K (t)

(0.060)

Crisis*C/K(t-1)

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Origin*C/K (t) -0.042 (0.091) 0.101** (0.057) 0.793
539 535 528 550

-0.016 (0.094)

(1-size)*crisis*C/K(t) 0.792 0.789 0.798 0.793
541

R2
546

0.789

0.795

Wald

530

* Significant at 15 percent. ** Significant at 10 percent. *** Significant at 5 percent. Note: I/K is net investment over capital at the beginning of the period; Dy, first difference in log of sales; k, log of capital; y, log of sales; C/K, contribution margin (net operating income plus capital depreciation) over capital at the beginning of the period; size, a dummy with value 1 for firms with capital greater than average; crisis in columns (3), (6), and (7) is a dummy with value 1 for years 1999 and 2000, in column (4) it is the spread between interest rates denominated in foreign and domestic currency; and origin in columns (5) and (6) has a value 1 for firms with less than 30 percent foreign ownership. Standard errors are in parentheses. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

284

DE BRUN, GANDELMAN, AND BARBIERI

6

Double and triple interactions between size, crisis, and origin were also explored, but are not reported here. These results are available from the authors on request.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

straints. The coefficient estimates of both crisis proxies have the expected sign, suggesting that financial constraints on Uruguayan firms have increased during the recent recession. However, their significance level is not high (10 percent and 15 percent, respectively). The coefficient of the interaction with type of ownership is not statistically or economically significant. Given the significance of size and crisis, we explored the effects of the crisis for small firms.6 The results in column (7) in table 8.5 imply that small firms were especially adversely affected after the crisis. Given the strong significance of the lagged capital sales difference and lagged sales, it seems appropriate to experiment with a different specification, like the accelerator model of equation (8.10). As Mairesse, Hall, and Mulkay (1999) point out, the error correction specification (8.9) can be seen as a reparameterization of the basic equation (8.7) in levels, while the simple accelerator model of equation (8.10) is derived from first-differencing equation (8.7). If the permanent unobservable differences are present only in the equation in levels, they are still present in the error correction specification and the within estimates are the appropriate procedure. But in the case of the accelerator specification, the firm-specific effects are removed when equation (8.7) is first-differenced. Therefore, column (la) in table 8.6 presents total estimates of equation (8.10). However, to account for the possibility of different trends in capital and output growth rates at the firm level, the within estimator is also computed and included in column (Ib). As in the case of the error correction formulation, proxies to account for size, origin, and the 1999-2000 crisis are included. Both the individual coefficient and the sum of the coefficients on the liquidity variable are positive, with the same implication as in the error correction specification. Qualitatively, the accelerator model reflects the same conclusions as the error correction formulation. Column (2) in table 8.6 shows the negative coefficient estimate for the product of size and cash flow, indicating that larger firms suffer less from financial constraints. Moreover, the absolute value magnitude of this estimate is basically the same as the sum of the cash variables. Thus, for large firms, there are no financial constraints. Column (4) shows that the coefficient of the interaction of cash flow with the spread between interest rates denominated in domestic and foreign currency lagged

Table 8.6. Accelerator Model for Investment/Capital, Total and within Estimates, 1997-2000
Within
(2) (3) (5) (4) (7) (6)

Total

Variable

da)

db)

Usable observations l/K(t-1)

Dy(t)

213 -0.034** (0.020) 0.459***

C/K (t)

C/K(t-1)

Sum of C coef. Size*C/K (t) 0.073 (0.070)

213 0.004 (0.024) 0.619*** (0.122) -0.055*** (0.017) 0.172*** (0.020) 0.116

(0.116) 0.136*** (0.019) 0.309*** (0.021) 0.445

213 -0.036** (0.020) 0.475*** (0.116) 0.138*** (0.019) 0.309*** (0.021) 0.447 -0.444** (0.258)

213 -0.037** (0.021) 0.455*** (0.116) 0.068 (0.069) 0.330*** (0.029) 0.398

213 -0.026 (0.021) 0.487*** (0.116) 0.129*** (0.033) 0.084*** (0.113) 0.213

213 -0.034** (0.020) 0.458*** (0.117) 0.159* (0.112) 0.309*** (0.021) 0.468

213 -0.038** (0.021) 0.458*** (0.116) 0.051 (0.069) 0.336*** (0.029) 0.387

Crisis*C/K (t) 0.618 (0.699) 7.280*** (3.529)

213 -0.038** (0.205) 0.471*** (0.117) 0.111 (0.155) 0.326*** (0.030) 0.437 -0.431** (0.261) 0.058 (0.074)

Crisis*C/K(t-1)

Origin*C/K (t)

-0.042 (0.110)

-0.027 (0.116)

(1-size)*crisis*C/K(t) 0.690 313 0.686 308 0.695 319 0.684 305 0.691 312

R2 Wald

0.378 117

0.683 306

0.090 (0.070) 0.687 309

* Significant at 15 percent. ** Significant at 10 percent. *** Significant at 5 percent. /Vote: For variable definitions, see note in table 8.6. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

286

DE BRUN, GANDELMAN, AND BARBIERI

GMM Estimates of the Error Correction and Accelerator Models The within estimates cannot deal with problems of bias caused by random measurement errors in the right-hand-side variables, simultaneity between contemporaneous right-hand variables and the disturbance, or the predetermined nature of some of the right-hand-side variables. To deal with these problems, an instrumental variable method of estimation is needed. In the case of correlation of the disturbances over time, the strategy usually implemented is the GMM estimation applied to the model in first differences. The first problem with GMM estimates is the selection of the instruments for the differenced variables. If residuals are not serially correlated, the lagged levels of the variables included in the regression starting from the second lag are candidates for instruments. Table 8.7 presents the first-difference GMM estimates for the error correction model and table 8.8 presents the estimates for the simple accelerator model.7 The cash flow coefficient tends to be smaller compared with the within estimates in tables 8.5 and 8.6. The signs of all the relevant coefficients are similar to the ones reported there. The cash flow coefficients are significantly positive. Large firms have a smaller cash flow coefficient, but the difference between large and small firms is not statistically significant. Financial constraints are greater in the crisis period, when smaller firms suffered significantly more from financial restrictions compared with larger firms. The interaction with the ownership dummy is not statistically significant. In the accelerator model, the cash flow coefficient for smaller firms is significantly larger. The interaction with the crisis dummy also confirms the previous results. However, the coefficient of the triple interaction size-crisis-cash flow is not significant. Finally, the Sargan test for overidentification has acceptable values in all specifications.

For all GMM estimations, the first-step first-difference GMM estimators are reported using robust standard errors. The Sargan test is from the two-step estimation. The instruments are lags two and three of net income and all the independent variables in the model.

7

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

one period has the expected sign and is significant. Finally, foreign or national ownership does not seem to play a relevant role in explaining financial constraints on investment.

Table 8.7.
(2) 157

Error-Correction Model for Investment/Capital GMM Estimates, 1997-2000
(3)
(7)

Variable
(4) (5) (6)

(D

Usable observations Dy(t)

k-y(t-1)

y(t-D

157 0.053 (0.426) -1.290*** (0.325) -1.311*** (0.257)

157 -0.033 (0.306) -1.645*** (0.424) -1.736***

C/K (t)

C/K(t-1)

Sum of C coef. Size*C/K (t) 0.070 (0.077) 1.637** (0.925)

157 0.092 (0.418) -1.268*** (0.330) -1.311*** (0.249) 0.073*** (0.035) 0.204*** (0.029) 0.277

0.101 -(0.419) -1.271*** (0.330) -1.299*** -(0.249) 0.074*** -(0.034) 0.203*** (0.029) 0.277 -0.151 (0.463) 0.013 (0.073) 0.224*** (0.040) 0.237 0.176

157 0.096 -(1.256) -1.256*** (0.333) -1.300*** (0.247) 0.055*** (0.024) 0.209*** (0.031) 0.264 157 0.082 (0.426) -1.253*** (0.337) -1.310*** (0.250) -0.029 (0.284) 0.205*** (0.030) 157 0.060 (0.435) -1.290*** (0.330) -1.290*** (0.269) -0.038 (0.214) 0.226*** (0.041)

(0.41 5) -0.013 (0.053) 0.213*** (0.045) 0.200

Crisis* C/K (t)

Origin*C/K (t)

0.103 (0.279)

0.188 -0.287 (0.540) 0.078 (0.077) 0.046 (0.204)

(1-size)*crisis*C/K(t)

R2 Wald

0.810 885
18.611

0.810 935

0.813 3,011

0.816 2,393
20.813

0.806 1,078 21.555

0.811 3,719
16.464

0.110** (0.057) 0.784 10,833 13.512

Sargan

22.385

21.579

* Significant at 1 5 percent. ** Significant at 10 percent. *** Significant at 5 percent. Note: For variable definitions, see note in table 8.6. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 8.8. Accelerator Model for Investment/Capital GMM Estimates, 1997-2000
(2) 157 157 157 157 157 157

Variable
(4) (6)

(D

(3)
-0.061 -0.139* (0.095) -0.181 (0.523) (0.053) (0.053) (0.053) (0.057) -0.054 -0.048 -0.060

(5) (7)

Usable observations

157

l/K(t-1)

-0.055

-0.105**

(0.052)

(0.064)

Dy(t)
(0.429) (0.422) 0.093* (0.015) 0.319*** 0.315*** (0.019) 0.297 (0.016) (0.049) 0.566 -4.007* (2.666) (0.502) (0.414) 0.401*** -0.018 0.035 (0.138) 0.351*** (0.037) 0.386 (0.436)

0.168 0.165

-0.080

0.126 0.215 0.140

0.132
(0.429) 0.045 (0.142) 0.347*** (0.039) 0.392

(0.427)

(0.485)

C/K (t)

0.140***

0.145***

(0.011)

(0.011)

C/K(t-1)

0.315***

0.316***

(0.018)

(0.021)

Sum of C coef .

0.455

0.461 0.412

Size*C/K (t)

-3.343*

(2.194)

Crisis*C/K (t) (0.138) (0.649)

0.112
2.784***

0.264* (0.177)

Origin*C/K (t)

0.160
(0.500)

-0.270 (0.510)

(1-size)*crisis*C/K(t) 0.642 7,304 18.457 0.650 3,398 15.578 0.636 2,557 17.285 0.295 4,845 13.741

0.101
(0.143) 0.642 6,900 18.378
863

R2

0.641

0.434

Wald

2,037

Sargan

17.454

17.048

* Significant at 15 percent. ** Significant at 10 percent. *** Significant at 5 percent. Note: For variable definitions, see note in table 8.6. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

FINANCIAL RESTRICTIONS IN URUGUAY

289

In order to implement equation (8.15'), we ran the following regression:

with the contribution margin (calculated as net operating p K Y income8 plus capital depreciation) and _ with the sales-to-capital ratio. If K the specification is correct, (Xo, CX2, and oc4 are expected to be positive and a l5 oc3, and oc5 are expected to be negative. Recall also that the differences between the Euler equation for the case of no constraints (equation (8.14')) and the financially constrained Euler equation are the signs on the two debt terms. Therefore, finding (X3 < 0 and a 4 > 0 and significant is evidence that firms have hit the leverage ceiling. Alternatively, finding a3 = 0 and oc4 < 0 is consistent with the existence of a premium for debt and a nonbinding debt ceiling. Column (1) in table 8.9 presents the econometric results of estimating equation (8.16) by GMM. Several augmented versions of the same regression were also considered to test its robustness and goodness of fit. The results are convincing. Most of the signs have the expected pattern and are significant. More relevant to the scope of this chapter, the coefficient estimate for the squared debt-to-capital ratio is significant and negative (the expected sign for the firms where the leverage ceiling is binding). This suggests the relevance of the extra financial constraints in equation (8.13) in the firm maximization problem shown in equations (8.11) and (8.12). The only reason for caution is that the debt-to-capital ratio has a negative sign but is not significant. The sign and significance of the coefficients of the interactions of the leverage terms with the various dummies do not lead to clear conclusions. The Sargan test of overidentifying restrictions does not reject the specification in any of the models.
8

proxying

71

Net operating income is denned as earnings before interest and taxes.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Euler Equation

290

DE BRUN, GANDELMAN, AND BARBIERI

Variable Usable observations GI/K(t-1) GI/K(t-1) squared Y/K(t-1) B/K(t-1) B/K(t-1) squared MC/K(t-1) Size*B/K (t) Size*B squared/K (t) Crisis*B/K (t) Crisis*B squared/K (t) Origin*B/K (t) Origin*B squared/K (t)

(1) 157
0.319** (0.017) -0.015*** (0.007) 0.025*** (0.011) -0.078 (0.188) 0.028*** (0.007) -0.304**

(2) 157
0.336*** (0.168) -0.015*** (0.007) 0.023*** (0.010) -0.039 (0.188) 0.029*** (0.006) -0.354***

(3) 157

(4) 157

0.336*** (0.172) -0.015*** (0.007) 0.002 (0.055) -0.233** (0.143) 0.029*** (0.007) -0.154
0.155

0.339*** (0.171) -0.016*** (0.008) 0.026 (0.023) -0.177 (0.208) 0.028*** (0.007) -0.245*
0.168

0.166

0.168
2.039 (1.495)

1.372

(1 .690)
3.082 (5.022) -0.031 (0.40) 0.139* (0.097) -0.005*** (0.002) 0.536 15,045 26.638 0.527 14,631 24.397 0.579 37,617 41.631 0.533 16,520 26.406

R2
Wald Sargan

* Significant at 15 percent. ** Significant at 10 percent. *** Significant at 5 percent. Note: GI/K is the dependent variable; GI/K, gross investment over capital at the beginning of the period; Y/K, sales over capital at the beginning of the period; B/K, total debt over capital at the beginning of the period; MC/K, contribution margin over capital at the beginning of the period; and crisis, the spread between interest rates denominated in foreign and domestic currency. Source: Authors' calculations.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Table 8.9. Euler Equation GMM Estimates, 1997-2000

FINANCIAL RESTRICTIONS IN URUGUAY

291

The chapter has presented results for testing three alternative specifications of an investment equation using panel data of Uruguayan firms: a traditional accelerator model of investment, an error correction version of that accelerator model, and an Euler equation approach. We used these models of investment to test for the existence of financial constraints in the investment decision process. The more flexible (and ad hoc) error correction and accelerator models suggest that cash flow plays an important role in investment decisions. Moreover, there is evidence that financial restrictions are more severe for smaller firms. Additional evidence is found that financial constraints were tighter during the crisis years of 1999-2000, and this was particularly true for smaller firms. So there is evidence that the decrease in investment in 1999-2000 was associated with more severe credit conditions. The analysis also explored the effects of ownership type on access to financial resources. Although we conjectured that foreign-owned firms would suffer less severe restrictions, there was no supporting evidence to this effect. This is probably due to the fact that most national firms in the database used in this chapter are well-established, mature firms with respectable brand names. Taking into consideration the relative robustness of the results of the accelerator and error correction models, the estimates confirm the presence of financial restrictions on investment decisions of Uruguayan firms in the period under consideration (all estimates correspond to 1997-2000). Finally, the estimates of the Euler equation model confirm the importance of financial constraints for Uruguayan firms.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Concluding Remarks

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

This page intentionally left blank

Akerlof, G. 1970. "The Market for Lemons: Quality Uncertainty and the Market Mechanism." Quarterly Journal of Economics 113: 79-119. Albuquerque, R., and H. Hopenhayn. 2001. "Optimal Lending Contracts and Firm Dynamics." Rochester, NY. Unpublished. Alesina, A., A. Carrasquilla, and R. Steiner. 2000. The Central Bank in Colombia. Documento de trabajo no. 13. Bogota, Colombia: FEDESARROLLO. Allen, F. 2001. "Do Financial Institutions Matter?" Journal of Finance 56: 1165-75. Alvarez, F., and U. Jermann. 2001. "Quantitative Asset Pricing Implications of Endogenous Solvency Constraints." Review of Financial Studies 14(4): 1117-51. Aoki, M. 2001. Toward a Comparative Institutional Analysis. Cambridge, MA: MIT Press. Arbelaez, M. A., and J. J. Echavarria. 2001. "The Colombian Financial Sector: Still Very Behind?" Bogota, Colombia. Unpublished. Arellano, M., and O. Bover. 1995. "Another Look at the Instrumental Variable Estimation of Error-Components Models." Journal of Econometrics 68: 29-51. Arellano, M., and S. Bond. 1988. Dynamic Panel Data Estimation UsingDPD:A Guide for Users. Working Paper 88/15. Institute for Fiscal Studies, London, England. . 1991. "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations." Journal of Economic Studies 58: 277-97. Arellano, M., S. Bond, and}. A. Doornik. 1997. Dynamic Panel Data Estimation Using DPD for Ox. Oxford, UK. Unpublished. Auerbach, A. J., and K. Hassett. 1992. "Tax Policy and Business Fixed Investment in the United States." Journal of Public Economics 47: 141-70. Babatz, G. 1998. "The Effects of Financial Liberalization on the Capital Structure and Investment Decisions of Firms: Evidence from Mexican Panel Data." Secretaria de Hacienda y Credito Publico, Mexico. Unpublished. Banco Central de Costa Rica (BCCR). Indicadores economicos. San Jose, Costa Rica: BCCR. http://www.bccr.fi.cr. Banco Central de la Republica Argentina. September 2000. Informacion de entidadesfinanderas. Buenos Aires, Argentina: Banco Central de la Republica Argentina.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Bibliography

294

BIBLIOGRAPHY

Banco Central del Ecuador. Various years. Information estadistica mensual, various issues. Banerjee, A., and A. Neumann. 1991. "Risk-Bearing and the Theory of Income Distribution." Review of Economic Studies 58(2): 211-35. Barajas, A., R. Steiner, and N. Salazar. 1999. Foreign Investment in Colombia's Financial Sector. IMF Working Paper WP/99/150. International Monetary Fund, Washington, D.C. Barclay, M. L, and C. W. Smith, Jr. 1995. "The Maturity Structure of Corporate Debt." Journal of Finance 50: 609-31. Bebczuk, R. 2000. "Corporate Saving and Financing Decisions in Argentina." Paper presented at the Quintas Jornadas de Economia Monetaria e Internacional, Universidad Nacional de La Plata, La Plata, Argentina. http://www.depeco.econo.unlp.edu.ar/jemi/2000/trabajo9.pdf. Beck, T., A. Demirguc-Kunt, and R. Levine. 2002. Law, Endowments and Finance. NBER Working Paper 9089. National Bureau of Economic Research, Cambridge, MA. Beck, T., and R. Levine. 2002. "Industry Growth and Capital Allocation: Does Having a Market- or Bank-Based System Matter?" Journal of Financial Economics 64(2): 147-80. Beck, T., R. Levine, and N. Loayza. 2000. "Finance and the Sources of Growth." Journal of Financial Economics 58: 261-300. Becker, G. 1964. Human Capital. New York, NY: Columbia University Press. Berger, A., and G. Udell. 1995. "Relationship Lending and Lines of Credit in Small Firm Finance." Journal of Business 68(3): 351-81. Berger, A., L. Klapper, and G. Udell 2000. "The Ability of Banks to Lend to Informationally Opaque Small Businesses." Washington, D.C. Unpublished. Bernanke, B. 1983. "Non-Monetary Effects of the Financial Crisis in the Propagation of the Great Depression." American Economic Review 73(3): 257-76. Bernanke, B. S., and A. Blinder. 1988. "Credit, Money and Aggregate Demand." American Economic Review 78(2): 435-39. Bernanke, B., and M. Gertler. 1989. "Agency Costs, Net Worth, and Business Fluctuations." American Economic Review 79:14-31. . 1990. "Financial Fragility and Economic Performance." Quarterly Journal of Economics 105(1): 87-114.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

295

Bernanke, B., M. Gertler, and S. Gilchrist. 1994. The Financial Accelerator and the Flight to Quality. NBER Working Paper 4789. National Bureau of Economic Research, Cambridge, MA. . 1996. "The Financial Accelerator and the Flight to Quality." Review of Economics and Statistics 78: 1-15. . 1999. "The Financial Accelerator in a Quantitative Business Cycle Framework." In Handbook of Macroeconomics, eds. J. Taylor and M. Woodford. Volume 1C. Amsterdam, The Netherlands: Elsevier Science, North-Holland. Bernhardt, D., and H. Lloyd-Ellis. 2000. "Enterprise, Inequality and Economic Development." Review of Economic Studies 67: 147-68. Bester, H. 1985. "Screening vs. Rationing in Credit Markets with Imperfect Information." American Economic Review 75: 850-55. Biais, B., and G. Gollie. 1997. "Trade Credit and Credit Rationing." Review of Financial Studies 10(4): 903-37. Blanchard, O., and S. Fischer. 1989. Lectures on Macroeconomics. Cambridge, MA: MIT Press. Blundell, R., and S. R. Bond. 1998. "Initial Conditions and Moment Restrictions in Dynamic Panel Data Models." Journal of Econometrics 87: 115-43. Bolanos, R., and F. Gutierrez. 1999. Financiamiento de lapequenay mediana empresa en Costa Rica. San Jose, Costa Rica: Institute Centroamericano de Administracion de Empresas (INCAE). Bond, P., and R. Townsend. 1996. "Formal and Informal Financing in a Chicago Ethnic Neighborhood." Economic Perspectives (Federal Reserve Bank of Chicago) 20(4): 3-27. Bond, S. 2002. "Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice." Portuguese Economic Journal 1(2): 141-62. Bond, S., J. Elston, J. Mairesse, and B. Mulkay. 1997. Financial Factors and Investment in Belgium, France, Germany and the UK: A Comparison Using Company Panel Data. NBER Working Paper 5900. National Bureau of Economic Research, Cambridge, MA. Booth, L, B. Aivazian, A. Demirgiic-Kunt, and V. Maksimovic. 2001. "Capital Structure in Developing Countries." Journal of Finance 56(1): 87-130. . 2000. "Capital Structures in Developing Countries." Toronto, Canada: University of Toronto. Unpublished, http://www.rotman.utoronto.ca/ -booth/capitals.pdf.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

296

BIBLIOGRAPHY

Burdisso, T., L. D'Amato, G. Escude, and G. McCandless. 2001. "How Much Do SMEs Borrow from the Banking System in Argentina?" Banco Central de la Republica Argentina, Buenos Aires, Argentina. Unpublished. Butler, J., and R. Moffit. 1982. "A Computationally Efficient Quadrature Procedure for the One Factor Multinomial Probit Model." Econometrica 50(3): 761-64. Caballero, R. 2000. "Macroeconomic Volatility in Latin America: A Conceptual Framework and Three Case Studies." Economia 1: 31-107. Caballero, R., and A. Krishnamurthy. 1999. "Emerging Markets Crises: An Asset Market Perspective." MIT, Cambridge, MA. Unpublished. Calomiris, C. W., and C. Himmelberg. 1995. Government Credit Policy and Industrial Performance: Japanese Machine Tool Producers 1963-1991. Policy Research Working Paper no. 1434. World Bank, Washington, D.C. Calomiris, C., and G. Hubbard. 1990. "Firm Heterogeneity, Internal Finance, and 'Credit Rationing.'" Economic Journal 100: 90-104. . 1995. "Internal Finance and Investment: Evidence from the Undistributed Profits Tax of 1936-37." Journal of Business 68: 443-82. Calomiris, C. W., and C. M. Kahn. 1991. "The Role of Demandable Debt in Structuring Optimal Banking Arrangements." American Economic Review 81: 497-513. Calvo, G. A. 2000. "Balance-of-Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign Governments." In Currency Crises, ed. P. Krugman. Chicago, IL: National Bureau of Economic Research/University of Chicago Press. Calvo, G., A. Izquierdo, and E. Talvi. 2002. Sudden Stops, the Real Exchange Rate and Fiscal Sustainability: Lessons from Argentina. Research Department Working Paper 469. Inter-American Development Bank, Washington, D.C. Calvo, G. A., L. Leiderman, and C. Reinhart. 1995. "Capital Inflows to Latin America with Reference to the Asian Experience." In Capital Controls, Exchange Rates and Monetary Policy in the World Economy, ed. S. Edwards. Cambridge, UK: Cambridge University Press. Caprio, Jr., G., and A. Demirguc-Kunt. 1997. The Role of Long-Term Finance: Theory and Evidence. Policy Research Department Working Paper 1746. World Bank, Washington, D.C. Carrasquilla, A., and M. A. Arbelaez. 2000. La politica financiera entre 1998 y 2000: Su impacto sobre las entidades de credito. Documento CEDE 2000-09. Universidad de los Andes y CEDE, Bogota, Colombia.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

297

Carrasquilla, A., and J. P. Zarate. 1997. "Reforma del sector financiero en Colombia: ^Donde estamos?" Paper presented at the XV Simposio del Mercado de Capitales, Bogota, Colombia. Castaneda, G. 1998. La empresa mexicanaysugobierno corporative: Antecedentes y desafiospara el siglo XXI. Puebla, Mexico: Universidad de las AmericasPuebla/Alter Ego Editores. . 1999. "Governance of Large Corporations in Mexico and Productivity Implications." Studies in Business Management 3(1): 57-90. . 200la. "Los grupos economicos y el credito comercial como estabilizadores financieros de la economia mexicana (1996-2000)." Universidad de las Americas-Puebla, Mexico. Unpublished. . 200Ib. "Internal Capital Markets and Financial Crisis. An Explanation of the Mexican Recovery after the 1995 Crisis." Universidad de las Americas-Puebla, Mexico. Unpublished. Chamberlain, G. 1980. "Analysis of Covariance with Qualitative Data." Review of Economic Studies 47: 225-38. Chirinko, R. S., and H. Schaller. 1995. "Why Does Liquidity Matter in Investment Equations?" Journal of Money, Credit and Banking 27: 527-48. Cho, Y. D. 1995. "Financial Factors and Corporate Investment: A Microeconometric Analysis of Manufacturing Firms in Korea." Doctoral dissertation, Oxford University, Oxford, UK. Claessens, S., A. Demirgiic-Kunt, and H. Huizinga. 2001. "How Does Foreign Entry Affect Domestic Banking Markets?" Journal of Banking and Finance 25:891-911. Claessens, S., S. Djankov, J. P. H. Fan, and L. H. P. Lang. 1999. "The Rationale for Groups: Evidence from East Asia." World Bank, Washington, D.C. Unpublished. Claessens, S., and L. Laeven. 2002. "Financial Development, Property Rights, and Growth." World Bank, Washington, D.C. Unpublished. Cole, R. A. 1998. "The Importance of Relationships to the Availability of Credit." Journal of Banking and Finance 22: 959-77. Corbo, V., and L. Hernandez. 2000. "Private Capital Inflows and the Role of Economic Fundamentals." In Capital Flows, Capital Controls, and Currency Crises, ed. F. Larrain. Ann Arbor, MI: University of Michigan Press. D'Auria, C., A. Foglia, and P. Marullo-Reedtz. 1999. "Bank Interest Rates and Credit Relationships in Italy." Journal of Banking and Finance 23: 1067-93.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

298

BIBLIOGRAPHY

De Gregorio, J., and J. W. Lee. 1999. Economic Growth in Latin America: Sources and Prospects. Documento de trabajo 66. Universidad de Chile, Centre de Economia Aplicada, Santiago, Chile, and Global Development Network, Washington, D.C. http://www.gdnet.org/pdf74 ll_Degregorio-lee.PDF. Demirgu9-Kunt, A., and E. Detragiache. 1999. "Financial Liberalization and Financial Fragility." Annual World Bank Conference on Development Economics, 1998. Washington, D.C.: World Bank. Demirguc-Kunt, A., and R. Levine. 1996. "Stock Market Development and Financial Intermediaries: Stylized Facts." World Bank Economic Review 10: 291-321. . 1999. Bank-Based and Market-Based Financial Systems: Cross-Country Comparisons. Working Paper 2143. World Bank, Washington, D.C. Demirguc-Kunt, A., and R. Levine, eds. 2001. Financial Structure and Economic Growth. Cambridge, MA: MIT Press. Demirguc-Kunt, A., and V. Maksimovic. 1994. Capital Structures in Developing Countries: Evidence from Ten Countries. Policy Research Working Paper 1320. World Bank, Washington, D.C. . 1998. "Law, Finance, and Firm Growth." Journal of Finance 53: 2107-37. . 1999. "Institutions, Financial Markets, and Firms' Debt Maturity." Journal of Financial Economics 54(3): 295-336. Devereux, M., and F. Schiantarelli. 1989. Investment, Financial Factors and Cash Flow: Evidence from UK Panel Data. NBER Working Paper 3116. National Bureau of Economic Research, Cambridge, MA. Diamond, D. W. 1984. "Financial Intermediation and Delegated Monitoring." Review of Economic Studies 51: 393-414. . 199la. "Debt Maturity Structure and Liquidity Risk." Quarterly Journal of Economics 106: 710-37. . 1991b. "Monitoring and Reputation: the Choice between Bank Loans and Directly Placed Debt." Journal of Political Economy 9(9): 688-721. . 1993. "Seniority and Maturity of Debt Contracts." Journal of Financial Economics 33: 341-68. Domac, I., and G. Ferri. 1999. "Did Financial Shocks Disproportionately Hit Small Business in Asia? Evidence from Malaysia and the Republic of Korea." Paper prepared for the conference, "The Credit Crunch in East Asia: What Do We Know? What Do We Need to Know?" World Bank, Washington, D.C. http://www.worldbank.org/research/interest/confs/past/pstconfs.htm.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

299

Dunning, J. H., and J. R. Dilyard. 1999. "Towards a General Paradigm of Foreign Direct and Foreign Portfolio Investment." Transnational Corporations 8: 1-52. Easterly, W., R. Islam, and J. Stiglitz. 2001. "Shaken and Stirred: Explaining Growth Volatility." In Annual Bank Conference on Development Economics 2000, eds. B. Pleskovic and J. Stiglitz. Washington, D.C.: World Bank. Echavarria, J. J., and F. Tenjo. 1993. "Inversion, liquidez y fuentes de financiacion en la industria colombiana." Coyuntura Economica 13: 103-34. Echavarria, J. J., and S. Zuluaga. 2001. "Multinacionales, exportaciones y aprendizaje en la industria colombiana." Bogota, Colombia: FEDESARROLLO. Unpublished. Echeverry, J. C., and N. Salazar. 1999. "Hay un estancamiento en la oferta de credito?" Archives de macroeconomia 118. Bogota, Colombia: Departamento Nacional de Planeacion. Edwards, S. 1995. "Poverty, Income Distribution, and Human Resources." In Crisis and Reform in Latin America, ed. S. Edwards. Washington, D.C.: World Bank. Elston, J. A., and H. Albach. 1995. "Firm Ownership Structure and Investment: Evidence from German Manufacturing." IFO Studien Zeitschrift Empirische Wirtschaftsforschung.g. Heft 1. Evans, D., and B. Jovanovic. 1989. "An Estimated Model of Entrepreneurial Choice under Liquidity Constraints." Journal of Political Economy 97(4): 808-27. Fanelli, J. M., and M. Damill. 1988. Decisiones de cartera y transferencias de riqueza en unperiodo de inestabilidad finandera. Serie Seminarios. Institute Torcuato Di Telia, Buenos Aires, Argentina. Fanelli, J. M., and S. Keifman. 2002. "Finance and Changing Trade Patterns in Developing Countries." In Finance and Competitiveness in Developing Countries, eds. J. M. Fanelli and R. Medhora. London, UK: Routledge. Fazzari, D., G. Hubbard, and B. Petersen. 1988. "Financing Constraints and Corporate Investment." Brookings Papers on Economic Activity 1988(1): 141-95. Fischer, S. 1988. "International Reserves." In The Open Economy: Tools for Policymakers in Developing Countries, ed. R. Dornbusch. Oxford: Oxford University Press. Freixas, X., and J. C. Rochet. 1998. Microeconomics of Banking. Cambridge, MA: MIT Press.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

300

BIBLIOGRAPHY

Galindo, A., F. Schiantarelli, and A. Weiss. 2002. Does Financial Liberalization Improve the Allocation of Investment? Micro Evidence from Developing Countries. Research Department Working Paper 467. Inter-American Development Bank, Washington, D.C. Gallego, F., and N. Loayza. 2000. Financial Structure in Chile: Macroeconomic Developments and Microeconomic Effects. Working Paper 75. Central Bank of Chile, Santiago. . 2001. "Financial Structure in Chile: Macroeconomic Developments and Microeconomic Effects." In Financial Structure and Economic Growth, eds. A. Demirgiic-Kunt and R. Levine. Cambridge, MA: MIT Press. Gelos, G., and A. Werner. 1998. La inversion fija en el sector manufacturer mexicano 1985-94: el rol de los factores financieros y el impacto de la liberalizacion financiera. Manuscript no. 9805. Banco de Mexico, Mexico City, Mexico. Gelos, G. R., and A. M. Werner. 1999. "Financial Liberalization, Credit Constraints, and Collateral: Investment in the Mexican Manufacturing Sector." International Monetary Fund, Washington, D.C. Unpublished. Gertler, M. 1992. "Financial Capacity and Output Fluctuations in an Economy with Multi-Period Financial Relationships." Review of Economic Studies 59: 455-72. Gertler, M., and G. Hubbard. 1988. Financial Factors in Business Fluctuations. NBER Working Paper 2758. National Bureau of Economic Research, Cambridge, MA. Gertner, R., D. Scharfstein, and J. Stein. 1994. "Internal versus External Capital Markets." Quarterly Journal of Economics 109: 1211-30. Gilchrist, S., and C. Himmelberg. 1998. Investment, Fundamentals and Finance. NBER Working Paper 6652. National Bureau of Economic Research, Cambridge, MA. Government of Ecuador. Various years. Superintendencia de Bancos, MemoHas, various issues. Greene, W. 1992. "A Statistical Model for Credit Scoring." New York University, New York. Unpublished. . 1997. Econometric Analysis. Third edition. Upper Saddle River, NJ: Prentice Hall. Greenwald, B., and J. Stiglitz. 1988a. Financial Market Imperfections and Business Cycles. NBER Working Paper 2494. National Bureau of Economic Research, Cambridge, MA.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

301

—. 1988b. "Information, Finance Constraints, and Investment Fluctuations." In Finance Constraints, Expectations, and Economic Activity, eds. M. Kohn and S. C. Tsaing. Oxford, UK: Oxford University Press. —. 1993. "Financial Market Imperfections and Business Cycles." Quarterly Journal of Economics 108: 77-114. Hall, L., and D. Lopez. 2000. Factores cualitativosy cuantitativos en la morosidad de la cartera de comerdo, demanda de mercado y legislacion del sector. Institute de Investigaciones en Ciencias Economicas, San Jose, Costa Rica. Unpublished. Hall, L., and R. Madrigal. 2000. "Diseno de un sistema parametrizado de calificacion de credit o (credit scoring)." San Jose, Costa Rica: Institute de Investigaciones en Ciencias Economicas. Unpublished. Hansen, L. P. 1982. "Large Sample Properties of Generalized Method of Moment Estimators." Econometrica 50(4): 1029-54. Hansen, L. P., and K. J. Singleton. 1982. "Generalized Instrumental Variables Estimation of Non-Linear Rational Expectations Models." Econometrica 50: 1269-86. Harris, J. R., F. Schiantarelli, and M. G. Siregar. 1994. "The Effect of Financial Liberalization on the Capital Structure and Investment Decisions of Indonesian Manufacturing Establishments." World Bank Economic Review 8(1): 17-47. Harris, M., and A. Raviv. 1990. "Capital Structure and the Informational Role of Debt." Journal of Finance 45: 321-49. Harrison, A., and M. McMillan. 2002. "Does Direct Foreign Investment Affect Domestic Firms' Credit Constraints?" University of California, Berkeley, and Tufts University, Medford, MA. Unpublished. Harrison, A., I. Love, andM. McMillan. 2001. Global Capital Flows and Financing Constraints. NBER Working Paper 8887. National Bureau of Economic Research, Cambridge, MA. Hart, O. 1995. Firms, Contracts, and Financial Structure. London, UK: Oxford University Press. Hart, O., and J. Moore. 1994. "A Theory of Debt Based on the Inalienability of Human Capital." Quarterly Journal of Economics 109(4): 841-79. . 1995. "Debt and Seniority: An Analysis of the Role of Hard Claims in Constraining Management." American Economic Review (June): 567-85. Hayashi, F., and T. Inoue. 1991. "The Relationship between Firm Growth and Q with Multiple Capital Goods: Theory and Evidence from Panel Data on Japanese Firms." Econometrica 59(3): 731-53.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

302

BIBLIOGRAPHY

Heckman, J. 1974. "Heckman Shadow Prices, Market Wages, and Labor Supply." Econometrica 42(4): 679-94. . 1979. "Sample Selection Bias as a Specification Error." Econometrica 47(1): 153-62. Hernandez, A., and J. Tolosa. 2001. "La politica monetaria en Colombia en la segunda mitad de los anos noventa." Borradores de Economia 172. Heshmati, A. 1994. "On the Single and Multiple Trends Representation of Technical Change." Applied Economics Letters 3:495-99. Hommes, R. 1987. "Refbrma financiera." In Evolution crisisy reforma del sistema financiero colombiano, ed. M. Avella. Bogota, Colombia: Universidad Nacional, Centro de Investigaciones para el Desarrollo. Hoshi, T., A. Kashyap, and D. Scharfstein. 1990a. "The Role of Banks in Reducing the Costs of Financial Distress in Japan." Journal of Financial Economics 27: 67-88. . 1990b. Asymmetric Information, Corporate Finance, and Investment. Chicago, IL: University of Chicago Press. . 1991. "Corporate Structure, Liquidity, and Investment: Evidence from Japanese Industrial Groups." Quarterly Journal of Economics 106: 33-60. Hsiao, C. 1986. Analysis of Panel Data. Cambridge, UK: Cambridge University Press. Hsiao, C., and A. K. Tahmiscioglu. 1997. "A Panel Analysis of Liquidity Constraints and Firm Investment." Journal of the American Statistical Association 92: 455-65. Hubbard, G. 1994. 7s There a Credit Channel in Monetary Policy? NBER Working Paper 4977. National Bureau of Economic Research, Cambridge, MA. . 1998. "Capital Market Imperfections and Investment." Journal of Economic Literature 36: 193-225. Hubbard, G., A. Kashyap, and T. M. Whited. 1995. "International Finance and Firm Investment." Journal of Money, Credit, and Banking. 27: 683-701. Huck, P., S. Rhine, P. Bond, et al. 1999. "A Comparison of Small Business Finance in Two Chicago Minority Neighborhoods." Economic Perspectives 23(2): 46-62. Institute Nacional de Aprendizaje (INA) and Caja Costarricense del Seguro Social (CCSS). 2000. Registro sobre Empresas en el Regimen del Seguro Social. San Jose, Costa Rica: INA and CCSS.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

303

icy Research Working Paper 1092. World Bank, Washington, D.C. . 1996. "Capital Market Imperfections before and after Financial Liberalization: An Euler Equation Approach to Panel Data for Ecuadorian Firms." Journal of Development Economics 51: 367-86. Jimenez, R., and T. Miller. 1993. "Vision panoramica de los programas privados de apoyo a la microempresa en Costa Rica." In Financiamiento y apoyo a la microempresa, eds. C. Gonzales and T. Miller. Second edition. San Jose, Costa Rica: Academia de Centroamerica. Kale, J. R., and T. H. Noe. 1990. "Risky Debt Maturity Choice in a Sequential Equilibrium." Journal of Financial Research 13: 155-65. Kashyap, A., and J. Stein. 1994. The Impact of Monetary Policy on Bank Balance Sheets. NBER Working Paper 4821. National Bureau of Economic Research, Cambridge, MA. Kehoe, P., and F. Perri. 2000. "International Business Cycles with Endogenous Incomplete Markets." Forthcoming in Econometrica. Kehoe, T., and D. Levine. 1993. "Debt-Constrained Asset Markets." Review of Economic Studies 60(4): 865-88. Khanna, T. 2000. "Business Groups and Social Welfare in Emerging Markets: Existing Evidence and Unanswered Questions." European Economic Review 44: 748-61. Khanna, T., and K. Palepu. 1997. "Why Focused Strategies May be Wrong for Emerging Markets". Harvard Business Review 75(4): 41-51. . 1998. The Future of Business Groups in Emerging Markets. Long Run Evidence from Chile. Harvard University Business School, Cambridge, MA. Unpublished.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Inter-American Development Bank. 2000. "Terms of Reference: Determinants and Consequences of Financial Constraints Facing Firms in Latin America and the Caribbean." Inter-American Development Bank, Washington, D.C. Japelli, T., and M. Pagano. 2001. "Information Sharing in Credit Markets: International Evidence." In Defusing Default: Incentives and Institutions, ed. M. Pagano. Washington, D.C.: Inter-American Development Bank. Jaramillo, F., and F. Schiantarelli. 1997. Access to Long-Term Debt and Effects on Firm Performance: Lessons from Ecuador. Policy Research Working Paper 1275. World Bank, Washington, D.C. Jaramillo, F., F. Schiantarelli, and A. Weiss. 1993. The Effect of Financial Liberalization on the Allocation of Credit: Panel Data Evidence for Ecuador. Pol-

304

BIBLIOGRAPHY

Kindleberger, C. P. 1978. Manias, Panics and Crashes. New York: Basic Books. King, R. G., and R. Levine. 1993. "Finance, Entrepreneurship and Growth." Journal of Monetary Economics 32: 513-42. . 2001. "Finance and Growth: Schumpeter Might be Right." Quarterly Journal of Economics 108: 717-37. Kiyotaki, N., and J. Moore. 1997. "Credit Cycles." Journal of Political Economy 105:211-48. Krueger, A., and A. Tornell. 1999. The Role of Bank Restructuring in Recovering from Crisis: Mexico 1995-98. NBER Working Paper 7042. National Bureau of Economic Research, Cambridge, MA. Krueger, D., and F. Perri. 2001. Risk Sharing: Private Insurance Markets orRedistributive Taxes? Stanford and New York: Stanford University and New York University. La Porta, R., F. L6pez-de-Silanes, and A. Shleifer. 1998. "Law and Finance." Journal of Political Economy 106: 1113-55. La Porta, R., F. L6pez-de-Silanes, A. Shleifer, and R. W. Vishny. 1997. "Legal Determinants of External Finance." Journal of Finance 52: 1131-50. Laeven, L. 2000. "Financial Liberalization and Financing Constraints: Evidence from Panel Data on Emerging Economies." World Bank, Washington, D.C. Unpublished. . 2001. Financial Liberalization and Financing Constraints: Evidence from Panel Data on Emerging Economies. Washington, D.C.: World Bank. Lamont, O. 1996. Cash Flow and Investment: Evidence from Internal Capital Markets. NBER Working Paper 5499. National Bureau of Economic Research, Cambridge, MA. . 1997. "Cash Flow and Investment: Evidence from Internal Capital Markets." Journal of Finance 52(1): 83-109. Lang, L., and R. Stulz. 1993. "Tobin's Q, Corporate Diversification, and Firm Performance." Journal of Political Economy 102: 142-74. Lederman, D., A. M. Mendez, G. Perry, and J. Stiglitz. 2000. "Mexico: Five Years after the Crisis." Paper presented at the ABCDE Conference of the World Bank, Washington, D.C. Leff, N. 1976. "Capital Markets in the Less-Developed Countries: The Group Principle." In Money and Finance in Economic Growth and Development, ed. R. McKinnon. New York: Marcel Dekker. Lehnert, A. 1998. "Credit Constraints." Doctoral dissertation, University of Chicago, IL.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

305

Lehnert, A., E. Ligon, and R. Townsend. 1999. "Liquidity Constraints and Incentive Contracts." Macroeconomic Dynamics 3(1): 1-47. Levine, R. 1997. "Financial Development and Economic Growth: Views and Agenda." Journal of Economic Literature (June): 688-726. . 1998. "The Legal Environment, Banks, and Long-Run Economic Growth." Journal of Money, Credit, and Banking 30: 596-613. . 2002. "Bank-Based or Market-Based Financial Systems: Which is Better?" Forthcoming in Journal of Financial Intermediation. Levine, R., N. Loayza, and T. Beck. 2000. "Financial Intermediation and Growth: Causality and Causes." Journal of Monetary Economics 46: 31-77. Ligon, E., T. Thomas, and T. Worrall. 2001. Informal Insurance Arrangements in Village Economies. CUDARE Working Paper Series 825. University of California, Berkeley, CA. Lochner, L., and A. Monge. 2002. Human Capital Formation with Endogenous Credit Constraints. NBER Working Paper 8815. National Bureau of Economic Research, Cambridge, MA. Lora, E. 2001. Structural Reform in Latin America: What Has Been Reformed and How to Measure It. Research Department Working Paper 466. InterAmerican Development Bank, Washington, D.C. Lora, E., and F. Barrera. 1997. A Decade of Structural Reform in Latin America: Measurement and Growth Effects. Research Department Working Paper 350. Inter-American Development Bank, Washington, D.C. Lora, E., P. Cortes, and A. M. Herrera. 2001. Los obstdculos al desarrollo empresarial y el tamano de las firmas en America Latina. Research Department Working Paper 447. Inter-American Development Bank, Washington, D.C. Love, 1.2000. "Financial Development and Financing Constraints: International Evidence from the Structural Investment Model." Columbia University, New York. Unpublished. Forthcoming in Review of Financial Studies. Machauer, A., and M. Weber. 2000. "Number of Bank Relationships: An Indicator of Competition, Borrower Quality, or Just Size?" Johann Wolfgang Goethe-Universitat, Frankfurt, Germany. Unpublished. Mairesse, J., B. H. Hall, and B. Mulkay. 1999. Firm-Level Investment in France and the United States: An Exploration of What We Have Learned in Twenty Years. NBER Working Paper 7437. National Bureau of Economic Research, Cambridge, MA.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

306

BIBLIOGRAPHY

Manski, C. 1995. Identification Problems in the Social Sciences. Cambridge, MA: Harvard University Press. Manski, C., and J. Horowitz. 1995. "Identification and Robustness with Contaminated and Corrupted Data." Econometrica 63(2): 281-302. Mayer, C. 1990. "Financial Systems, Corporate Finance, and Economic Development." In Asymmetric Information, Corporate Finance, and Investment, ed. G. Hubbard. Chicago, IL: University of Chicago Press. . 1994. "The Assessment: Money and Banking: Theory and Evidence." Oxford Review of Economic Policy 10:1-13. Mazzoli, M. 1998. Credit, Investments and the Macroeconomy. Cambridge, UK: Cambridge University Press. Melnik, A., and S. Plaut. 1986. "Loan Commitment Contracts, Terms of Lending, and Credit Allocation." Journal of Finance 41:425-35. Modigliani, F., and M. H. Miller. 1958. "The Cost of Capital, Corporation Finance, and the Theory of Investment." American Economic Review 48(3): 261-97. Monge, A. 2001. "Long-Term Relationships, Creation and Destruction of Firms and Aggregate Dynamics." Northwestern University, Evanston, IL. Unpublished. Monge, A., J. Cascante, and L. Hall. 2001. "Enforcement, Contract Design and Default: Exploring the Financial Markets of Costa Rica." In Defusing Default, ed. M. Pagano. Washington, D.C.: Inter-American Development Bank. Monge-Naranjo, A., and L. Hall. 2002. "Access to Credit and the Effect of Credit Constraints on Costa Rican Manufacturing Firms." Inter-American Development Bank, Washington, D.C. Unpublished. Montgomery, C. 1994. "Corporate Diversification." Journal of Economic Perspectives 8(3): 163-78. Morduch, J. 1995. "Income Smoothing and Consumption Smoothing." Journal of Economic Perspectives 9(3): 103-14. Morley, S. A., J. J. Machado, and J. J. Pettinato. 1999. Indexes of Structural Reform in Latin America. Santiago, Chile: Economic Commission for Latin America and the Caribbean, Economic Development Division. Myers, S. C. 1977. "Determinants of Corporate Borrowing." Journal of Financial Economics 5: 147-75. Myers, S., and N. Majluf. 1984. "Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have." Journal of Financial Economics 13(2): 187-221.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

307

Nawata, K., and N. Nagase. 1996. "Estimation of Sample Selection Bias Models." Econometric Reviews 15: 387-400. Nickell, S. J. 1978. The Investment Decisions of Firms. Oxford, UK: Cambridge University Press. Nickell, S. J., and D. Nicolitsas. 1995. How Does Financial Pressure Affect Firms? Applied Economics Discussion Paper Series 170. Oxford, UK: Oxford University. Padilla, A. J., and A. Requejo. 2001. "The Costs and Benefits of the Strict Protection of Creditor Rights: Theory and Evidence." In Defusing Default: Incentives and Institutions, ed. M. Pagano. Washington, D.C.: InterAmerican Development Bank. Pagano, M., ed. 2001. Defusing Default: Incentives and Institutions. Washington, D.C.: Inter-American Development Bank. Pagano, M., and T. Jappelli. 1993. "Information Sharing in Credit Markets." Journal of Finance 43: 1693-1718. Perotti, E., and S. Gelfer. 1998. "Investment Financing in Russian Industrial Groups." London, UK: Centre for Economic Policy Research. Unpublished. Petersen, M., and R. Rajan. 1994. "The Benefits of Lending Relationships: Evidence from Small Business Data." Journal of Finance 49: 3-37. . 1997. "Trade Credit: Theories and Evidence." Review of Financial Studies 10(3): 661-91. Rajan, R. 1992. "Insiders and Outsiders: The Choice between Relationsip and Arm's-Length Debt." Journal of Finance 47: 1367-1400. Schaller, H. 1993. "Asymmetric Information, Liquidity Constraints, and Canadian Investment." Canadian Journal of Economics 26: 552-74. Scharfstein, D., and J. Stein. 1997. The Dark Side of Internal Capital Markets: Divisional Rent Seeking and Inefficient Investment. NBER Working Paper 5969. National Bureau of Economic Research, Cambridge, MA. Schiantarelli, F. 1996. "Financial Constraints and Investment: Methodological Issues and International Evidence." Oxford Review of Economic Policy 12(2): 70-89. Schiantarelli, F., and A. Sembenelli. 1995. Forms of Ownership and Financial Constraints: Panel Data Evidence from Leverage and Investment Equations. Chestnut Hill, MA: Boston College. . 1996. "Determinants and Consequences of the Maturity Structure of Firms' Debt: The Case of the UK and Italy." Unpublished. . 2000. "Forms of Ownership and Financial Constraints." Empirica 27: 175-92.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

308

BIBLIOGRAPHY

Stulz, R. M. 2000. Does Financial Structure Matter for Economic Growth? A Corporate Finance Perspective. Columbus, Ohio: Ohio State University. Summers, L. H. 1981. "Taxation and Corporation Investment: A Q-Theory Approach." Brookings Papers on Economic Activity 1: 67-127. Superintendencia de Sociedades. 1978. Los conglomerados de sociedades en Colombia. Bogota, Colombia: Superintendencia de Sociedades. . 2000. "Analisis de los estados financieros consolidados de los grupos empresariales y situaciones de control 2000." Superintendencia de Sociedades, Bogota, Colombia. Unpublished.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Schiantarelli, F., and V. Srivastava. 1996. Debt Maturity and Firm Performance: A Panel Study of Indian Public Limited Companies. Policy Research Department Working Paper 1724. World Bank, Washington, D.C. Schmukler, S., and E. Vesperoni. 2001. "Does Integration with Global Markets Affect Firms' Financing Choices?" In Financial Structure and Economic Growth, eds. A. Demirgiic-Kunt and R. Levine. Cambridge, MA: MIT Press. Schumpeter, J. 1934. The Theory of Economic Development. Cambridge, MA: Harvard University Press. Seth, R., D. E. Nolle, and S. K. Mohanty. 1998. "Do Banks Follow Their Customers Abroad?" Financial Markets, Institutions and Instruments 7:1-25. Sharpe, S. 1990. "Asymmetric Information, Bank Lending and Implicit Contracts: A Stylized Model of Customer Relationships." Journal of Finance 45: 1069-87. Shin, H. H., and Y. S. Park. 1999. "Financing Constraints and Internal Capital Markets: Evidence from Korean Chaebols." Journal of Corporate Finance 5: 169-91. Staiger, D., and J. H. Stock. 1997. "Instrumental Variables Regression with Weak Instruments." Econometrica 65: 557-86. Stiglitz, J. E., and A. Weiss. 1981. "Credit Rationing in Markets with Imperfect Information." American Economic Review 71(3): 393-410. . 1986. "Credit Rationing and Collateral." In Recent Developments in Corporate Finance, eds. J. Edwards, J. Franks, J. Mayer, et al. Cambridge, UK: Cambridge University Press. Streb, J., J. Bolzico, P. Druck, et al. 2002. Bank Relationships: Effect on the Availability and Marginal Cost of Credit for Firms in Argentina. Latin American Research Network Working Paper R-444. Inter-American Development Bank, Washington, D.C.

CREDIT CONSTRAINTS AND INVESTMENT IN LATIN AMERICA

309

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Tybout, J. 1983. "Credit Rationing and Investment Behaviour in a Developing Country." Review of Economics and Statistics 65: 598-607. Ulate, A. 2000. "El rol de las MIPYMES en Costa Rica: Generar empleo o acumular valor?" Institute de Investigaciones en Ciencias Economicas, San Jose, Costa Rica. Unpublished. Villalobos, V. 1996. "Mitos y realidades sobre el financiamiento a las microempresas: Algunas implicaciones para las ONGs." Revista Parlamentaria (Costa Rica) 4(2). Whited, T. M. 1998. "Why Do Investment Euler Equations Fail?" Journal of Business and Economic Statistics 16: 478-88. . 2001. "Debt, Liquidity Constraints, and Corporate Investment: Evidence from Panel Data." Journal of Finance 47: 1425-60. Wilcox, R. 1997. Introduction to Robust Estimation and Hypothesis Testing. New York: Academic Press. Williamson, S. D. 1986. "Costly Monitoring, Financial Intermediation, and Equilibrium Credit Rationing." Journal of Monetary Economics 18: 159-79. . 1987. "Costly Monitoring, Loan Contracts, and Equilibrium Credit Rationing." Quarterly Journal of Economics 102: 35-145. World Bank. 2001. Finance for Growth: Policy Choices in a Volatile World. World Bank Policy Research Report. World Bank/Oxford University Press, New York. . Various years. World Development Indicators. World Bank, Washington, D.C. Wurgler, J. 2000. "Financial Markets and the Allocation of Capital." Journal of Financial Economics 58: 187-214. Yong, M. 1988. "Estructura, desempeno, situacion actual y perspectivas de politica para las pequenas y medianas empresas industriales de Costa Rica." Proyecto BT-HOL 7093-1700. Unidad Conjunta CEPAL/ ONUDI, Santiago, Chile. Unpublished.

Copyright © by the Inter-American Development Bank. All rights reserved. For more information visit our website: www.iadb.org/pub

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close