WPS6163 Policy Research Working Paper 6163 How Bank Competition Affects Firms’ Access to Finance Inessa Love María Soledad Martínez Pería The World Bank Development Research Group Finance and Private Sector Development Team August 2012 Policy Research Working Paper 6163 Abstract Combining multi-year, firm-level surveys with country- of competition on access to finance depends on the level panel data for 53 countries, the authors explore the environment that banks operate in. Some features of the impact of bank competition on firms’ access to finance. environment, such as greater financial development and They find that low competition, as measured by high better credit information, can mitigate the damaging values of the Lerner index, diminishes firms’ access to impact of low competition. But other characteristics, finance, while commonly-used bank concentration such as high government bank ownership, can exacerbate measures are not robust predictors of firms’ access the negative effect. to finance. In addition, they find that the impact This paper is a product of the Finance and Private Sector Development Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at mmartinezperia@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team How Bank Competition Affects Firms’ Access to Finance Inessa Love and María Soledad Martínez Pería Keywords: bank competition, bank concentration, access to finance JEL: G20, L1 Sector Board: FSE  Inessa Love is an Assistant Professor at the University of Hawaii and María Soledad Martínez Pería is a Lead Economist in the Finance and Private Sector Development Team in the Development Research Group of the World Bank. We thank Sandeep Singh and Diego Anzoategui for excellent research assistance. We are also grateful for funding received for this project from the Knowledge for Change Program. The paper’s findings and conclusions are those of the authors and do not necessarily reflect the views of The World Bank. Corresponding author: Maria S. Martinez Peria, 1818 St. N.W., MSN MC 3-307, Washignton, D.C. 20433. mmartinezperia@worldbank.org. Phone: (202)458-7341. 1 1. Introduction The impact of bank competition on financial markets and firms is an important topic for policymakers and researchers alike.1 Interest on this subject intensified during the recent global financial crisis, as many questioned whether high competition was partly to blame.2 At the same time, the downfall of some institutions as a result of the crisis and the emergency measures taken by some governments to deal with this episode - such as mergers, bailouts, recapitalizations, and extension of guarantees- have led to concerns about the future of bank competition and its potential implication for access to bank finance.3 How bank competition affects firms’ access to finance is in itself a much debated question in the economic literature and in policy circles and, as we discuss below, the theoretical predictions and the empirical evidence on this subject are mixed. Combining multi-year firm- level surveys with panel country-level data on bank competition for 53 countries, this paper offers new evidence on the link between competition and firms’ access to finance. In particular, our paper evaluates whether competition improves access to finance and analyzes the extent to which different features of the environment in which banks operate affect the link between competition and access. Theory provides ambiguous predictions regarding the impact of competition on access to finance. The conventional market power hypothesis argues that competition in the banking system reduces the cost of finance and increases the availability of credit. On the other hand, the information hypothesis argues that, in the presence of information asymmetries and agency costs, competition can reduce access by making it more difficult for banks to internalize the benefits of investing in building lending relationships, in particular, with opaque clients (Petersen and Rajan, 1995; Marquez, 2002; Hausewald and Marquez, 2006). 1 The Economist magazine hosted a virtual debate on this topic on June 1 st, 2011. See http://www.economist.com/debate/days/view/706 2 For example, Dell’Ariccia, Igan, and Laeven (2008) document that the rapid growth of credit in U.S. mortgage markets in the run up to the crisis was accompanied by a reduction in lending standards (lower loan application denial rates), which they argue was in part explained by the entry of new and large lending institutions. 3 See OECD (2009, 2010). 2 Most of the existing empirical studies on the link between competition and access to finance use concentration measures as proxies for competition and yield mixed results. Using data from the US, Petersen and Rajan (1994, 1995) find that SMEs are more likely to obtain financing when credit markets are concentrated.4 Similarly, using a survey dataset of German manufacturing firms, Fischer (2000) finds that more concentration leads to more information acquisition and greater credit availability. On the other hand, using enterprise survey data for 74 countries, Beck, Demirguc-Kunt, and Maksimovic (2004) find that in more concentrated banking sectors firms of all sizes face higher financing obstacles and the impact of concentration decreases with firm size. Chong, Lu, and Ongena (2012) also find a positive association between concentration and credit constraints, using a survey on the financing of Chinese SMEs combined with detailed bank branch information.5 In contrast to previous studies that equate high concentration with lack of competition, recent papers that use direct measures of banks’ pricing behavior provide less ambiguous findings on the link between competition and access to finance. Using the Panzar and Rosse H- statistic (1982,1987), which captures the elasticity of bank revenues to input prices, Claessens and Laeven (2005) find that competition is positively associated with countries’ industrial growth in a sample of 16 countries over the period 1980-1990.6 The authors argue that this suggests that more competitive banking sectors are better at providing financing to financially dependent firms. Exploiting a very rich dataset on Spanish SMEs and using the Lerner (1934) index – the difference between banks’ prices and marginal costs relative to prices - as a measure of competition, Carbó-Valverde, Rodriguez Fernandez, and Udell (2009) also find evidence that competition promotes access to finance.7 At the same time, the authors find that their results for 4 A related literature for the US has examined the impact of bank deregulation on access to finance. Zarutskie (2006) finds that deregulation in the US, which increased the competitiveness of US banking markets, caused newly formed firms to use significantly less external debt, consistent with the notion that competition exacerbates credit constrains. Rice and Strahan (2010) exploit the geographical variation in branching restrictions across US states and find that in states more open to branching, small firms are more likely to borrow and do so at lower rates. However, the authors find that there no effects on the amount that small firms can borrow. 5 Relatedly, there are studies on the impact of bank mergers on SME lending that find that these firms can be hurt by mergers (see Peek and Rosengren, 1996; Berger et al., 1998; Sapienza, 2002; Bonaccorsi di Patti and Gobbi, 2007; and Degryse, Masschelein and Mitchell, 2010 among others). Erel (2011) shows that while, in general, bank mergers can benefit borrowers through lower interest rates, if the geographical overlap between merging banks is so extensive as to significantly increase concentration in banking markets, then spreads increase after mergers. 6 Higher values of the H-statistic are associated with more competitive banking systems. 7 Higher values of the Lerner index denote higher markups and lower levels of bank competition. 3 the Lerner index are not consistent with results using concentration measures as proxies for competition. They conclude that concentration is not a good measure of competition. As described above, existing studies on the link between competition and financial access either analyze cross-sectional data or are only able to look at multi-year data for a single country. In contrast, our paper combines multi-year firm-level data with panel country-level data on bank competition. One advantage of our dataset is that it contains repeated cross-sections of firms for the countries in our sample. This allows us to control for unobserved differences between countries, using country fixed effects in our estimations. Such unobserved differences may be correlated with both access to finance and the extent of competition. Thus, our methodology isolates within country variation in competition and access. This is an improvement over previous cross-country studies, such as Beck, Demirguc-Kunt, and Maksimovic (2004). Furthermore, in contrast to US and cross-country studies that measure competition with concentration indicators, we also present results using the Lerner index, a direct measure of bank pricing behavior. Several studies have argued theoretically and empirically that pricing behavior measures such as the Lerner index are superior to concentration measures as indicators of competition.8 Concentration is a measure of market structure, while competition is a measure of market conduct. There can be competition in concentrated markets, if there is a credible threat of entry and exit (i.e., if markets are contestable). A contribution of our paper is the ability to distinguish the impact of concentration and competition in a multi-country setting. Also, the fact that we offer cross-country evidence using the Lerner index allows for more general results relative to those that focus on individual countries. We find that low competition, as proxied by high levels of the Lerner index, is associated with diminished access to finance by firms, while concentration has a less robust relationship with access. We use different weighting schemes to account for differences in the number of firms across countries and the variance of the estimated Lerner index. Overall, our results support the market power hypothesis. Furthermore, our results confirm that concentration measures are not reliable predictors of firms’ access to finance, which is in line with previous contradictory evidence. 8 See among others Cetorelli (1999), Claessens and Laeven (2004), Demirguc-Kunt et al. (2004), and Carbo-Verde et al. (20009). 4 In addition, we explore whether the characteristics of the environment in which banks operate affect the impact of competition on access to finance.9 To do that, we interact our measures of competition with country-level measures of financial development, the availability of credit information, and government bank ownership. We find that countries with higher levels of financial development and better information availability experience a less pronounced decline in access to finance as a result of low levels of competition (high values of the Lerner index). The flip side of this finding is that low competition is more detrimental for firms operating in countries with low levels of financial development or lacking credit information. In addition, we find that significant government bank ownership exacerbates the damaging impact of low bank competition. The rest of the paper is organized as follows. Section 2 introduces our multiple datasets and presents summary statistics. Section 3 outlines our regression model. Section 4 presents our baseline results. Section 5 discusses the results interacting the competition measures with different aspects of the environment in which banks operate. Section 6 concludes. Appendices A1 and A2 contain detailed descriptions of the construction of the firm-level measure of access to finance and the estimation method for the Lerner index, respectively. 2. Data We combine firm-, bank- and country-level data from various sources. Table 1, Panel A gives a list of all the variables used in the paper and details their sources. The firm-level data come from World Bank Enterprise Surveys.10 The data are collected in several waves and contain repeated cross-sections for the countries in our sample. Because our goal is to isolate within country variation in competition across time, we only focus on countries that have survey data for at least two years. We use firm survey data to construct our measure of access to finance and several control variables. Access to finance is an indicator variable that equals one when a firm has a loan, overdraft, or line of credit, and zero otherwise. We prefer to use this objective measure of access 9 Beck et al. (2004) analyze how different aspects of the institutional and regulatory environment, as well as the ownership structure of the banking system, affect the impact of bank concentration on firms’ perceptions of financing obstacles. 10 The data are available at www.enterprisesurveys.org 5 to credit, rather than subjective measures of financing obstacles, because the former is more comparable across countries and it does not depend on cultural biases that might influence individuals’ perceptions; plus it is more reliable and easier to interpret. Appendix A1 gives a detailed description of the process used to construct our measure of access to finance. We also include several firm-level variables that may influence the extent of firms’ access to finance, such as firm size, measured as log of the number of employees, a dummy for manufacturing industry (the omitted category is service and other industries), a dummy for exporting firms, a dummy for foreign-owned firms, a dummy for government-owned firms, and the log of firm age in years. The bank-level data come from Bankscope, a commercial database by Bureau Van Dijk including annual balance sheet and income statement information for banks across the world. Only banks classified as commercial, cooperative, Islamic, savings, and bank holding companies are considered in the analysis. We leave out central banks and investment banks, because they are not directly involved in providing loans to firms. We use bank-level data to construct the Lerner index, a direct measure of pricing behavior by banks, which captures the markup in prices – i.e., the difference between prices and marginal costs, measured as a ratio of prices. Higher values of the index indicate higher markups or lower levels of competition.11 We start with annual bank-level data and estimate a translog cost function using all available data for each country. We then calculate the marginal cost equation (by taking the derivative of the translog cost equation) and finally the Lerner index for each bank, which we then average for each country and year. Appendix A2 describes in detail the process we use to calculate the Lerner index. We also use bank-level data to construct two commonly used measures of concentration: Concentration 3 is the share of banking system assets held by the three largest banks and 11 There is an extensive literature measuring bank competition using the Lerner index. See Fernandez de Guevara et al. (2005, 2007), Berger, Klapper and Turk-Ariss (2009), Carbó et al. (2009), Turk-Ariss (2009), Anzoategui, Martinez-Pería and Rocha (2010), Beck, de Jonghe, and Schepens (2011), Anzoategui, Martinez-Pería and Melecky (2012), and Delis (2012), among others. 6 Herfindahl index is the sum of the squared market share of each bank.12 In both cases, higher values indicate more concentration. Our final dataset is limited to countries which have both firm-level data on access to finance from the Enterprise Surveys and bank-level data from Bankscope to calculate the Lerner index and the concentration measures. This dataset contains information on 53 countries for the period 2002-2010.13 Table A1 gives a list of the countries and years included in the final dataset. Some countries have had only two surveys in our time frame, while others have had three or more, with a maximum of five surveys for Bulgaria. The coverage of firms varies by country. For example, India has more than 4,000 firms covered in two surveys, while Malawi has fewer than 300 firms covered in two surveys. Because of this variation, we test the robustness of our results to weighting our regressions by the inverse of the square root of the number of firms in the survey, so that each country carries the same importance in our estimations. Our bank-level sample contains data on 3,409 banks and over 16,000 bank-year observations. Table A2 gives a list of countries in our sample along with the value of the Lerner index for each country over time.14 Finally, we supplement our dataset with annual country-level data from several sources. We obtain data on private credit to GDP and inflation from the World Bank World Development Indicators database. Data on the quality of credit information come from the World Bank Doing Business dataset15 and information on the share of assets held by government-owned banks comes from the World Bank Survey of Bank Regulation and Supervision.16 Table 1, Panel B reports basic summary statistics for the firm survey variables and country-level variables. More than half of the firms in our sample have access to finance; in other words, they use at least one of three credit products such as loans, lines of credit, or 12 The Herfindahl index, calculated as gives a greater weight to larger banks. 13 At the firm-level, we have over 68,000 observations. 14 Because we use one year lagged values of the Lerner index in our regression, we report bank data for years 2001- 2009. 15 Data available at www.doingbusiness.org 16 Data available at http://go.worldbank.org/SNUSW978P0 7 overdrafts.17 The average size of our firms is about 100 employees, and it varies from one to over 1,700, with the median firm size of 25 employees. Thus, most of the firms in our sample are small and medium-sized enterprises. The median firm age is 12 years, while the average is almost 18. Firm age varies from one year old startups to close to 200 years old firms. In our sample, 62 percent of firms are in manufacturing (the rest are in services, retail or construction), 23 percent are exporters (classified as such if they export at least 10 percent of their total output), about 10 percent are foreign-owned, and about 5 percent of firms are considered government owned. The Lerner index has an average of 0.25, a median of 0.23 and a standard deviation of about 0.07. The range is between 0.07, which indicates very low markups and, hence, high competition and 0.43, which implies very high markups and, therefore, low competition. Average concentration is high, with the top three banks comprising close to 60 percent of total bank assets. The lowest share of assets held by the top three banks is about 28 percent, while the highest is over 98 percent. The Herfindahl index varies between 0.05 and 0.74. 3. Regression Model Our goal is to evaluate the impact of bank competition on firms’ access to finance. To do that, we estimate the following simple model: Access ict =ac + b1 Competitionct + b2Fict + b3Xct +e ict (1) where Access is the indicator variable for whether firm i in country c at time t has a bank loan, line of credit, or overdraft; Competition refers to either the Lerner index or to two measures of concentration: the share of assets held by the top three banks and the Herfindahl index. F and X represent firm-level (e.g., size, manufacturing, exporter, etc.) and country-level (e.g., inflation and financial development as proxied by private credit to GDP) control variables, respectively, described in the data section. We capture unobservable differences between countries by including country fixed effects (represented in equation 1 by ac) and we cluster errors at the 17 While in principle it would be interesting to distinguish between firms that have access to each of these types of products, unfortunately, the design of the Enterprise Surveys does not allow for this possibility. 8 country-year level.18 Thus, our estimates represent within country variation in the relationship between competition and access to finance. We assume that country-level measures of competition are exogenous to the firm-level measure of access to finance. In other words, each individual firm is small enough to affect country-level measures of bank competition. However, to further mitigate any possible reverse causality concerns, we use one year lagged values for competition, as well as for the other country-level control variables. We use several weighting schemes in our estimations. First, because the Lerner index is an estimated variable, we weight our regression by the inverse of its standard deviation. This takes into account the precision with which the Lerner index is estimated and gives less weight to those observations that are estimated with less precision (i.e., that have larger standard errors). Second, because the number of firms varies for different surveys, we weight our regressions by the inverse of the square root of the number of observations (i.e., firms) in each country-year. This gives relatively less weight to countries with a large number of observations, which otherwise will be overrepresented in the sample. Third, we combine the two weighting factors in a product form (i.e., the weight equals the product of the inverse of the Lerner’s standard deviation and the inverse of the square root of the number of firms in the country and year). Finally, we also report regressions without any weights for comparison. 4. Baseline Results Table 2 reports our baseline results for the estimation of equation (1). The estimations shown in column (b) are weighted by the inverse of the standard deviation of the Lerner index, while other regressions are not weighted. We observe that the Lerner index is significant in all specifications, while the measures of concentration (Concentration 3 and Herfindahl index) are not significant. This establishes our first main result that low competition (i.e., a high value of the Lerner index) is associated with lower access to finance, while the link between concentration and access to finance is not significant. Weighting regressions by the inverse of the standard deviation for the Lerner index does not materially alter our results (compare column (b) to column (a) in Table 2). 18 Because of the inclusion of country fixed effects and to avoid an incidental parameters problem, we report our results using a linear probability model, however, our results are robust to using a fixed effects logit model. 9 Since we estimate the regressions using a linear probability model, we can interpret the coefficient on the Lerner index as an increase in the probability of access to finance. In our sample, one standard deviation of the Lerner index is about 0.07. Using the estimated coefficient of about 0.75 (from model (b) in Table 2, which corrects for the variance of the Lerner index by appropriate weighting), we obtain that a one standard deviation change in the Lerner index results in an approximately 5 percentage points change in the probability of having access to finance. In our sample, the average access to finance dummy equals one for about 60 percent of all firms, with a standard deviation of about 49 percent. A 5 percentage points change is modest for an average country, but it is more economically important for a country with low access to start with. Most of the control variables have the predicted signs. Larger and older firms are more likely to have access to finance. Manufacturing firms are more likely than service and retail enterprises to have access to finance, because they have more collateral which helps them obtain financing. Exporters also are more likely to have access to bank finance, however, foreign- owned firms are less likely. This might be because foreign firms can obtain financing from their parent company and, thus do not need to borrow from local banks. At the same time, government-owned firms appear less likely to have access to bank finance, which is a little surprising. We find that private credit, which is our measure of financial development, is associated with a higher likelihood of access to finance. The inflation rate has a negative association with our measure of access to finance, which is not surprising. Note that because of the country fixed effects, these variables capture the impact of the variation in inflation and private credit from the long-run country average. In Table 3 we present our robustness checks with regressions weighted by the inverse of the square root of the number of firms in the country-year survey. Because there is a large variation in the number of firms in each survey and our variables of interest are measured at the country-year level, the regressions reported so far give more weight to surveys with a larger number of firms, which will have a disproportional impact on the estimated coefficients. We find that weighting by the inverse of the number of firms has no impact on the significance of the Lerner index. We continue to find that low competition has a negative impact on firms’ access to finance. 10 5. Interaction Results The damaging impact of low levels of competition on access could be either mitigated or exacerbated by certain features of the environment in which banks operate. Here we investigate three such factors. First, we consider the extent of financial development in a country, measured by the ratio of private credit to GDP. This measure implicitly captures the institutional factors that determine the level of financial development, such as property rights protection or contract enforcement, as well as the actual use of finance by the private sector. We investigate whether the impact of competition varies depending on countries’ levels of financial development. Our hypothesis is that financial development will mitigate the damaging impact of lack of competition. As we discussed above, the market power hypothesis argues that competition in the banking market reduces the cost of finance and increases the availability of credit. Financial development is commonly associated with reduced cost of finance and wider availability of finance (see King and Levine, 1993 and Love, 2003). Therefore, in an environment with relatively low cost of finance, the marginal negative impact of low competition on costs is likely to be less damaging than in an environment where the costs are high to start with. Similarly, in an environment with wider availability of finance, the reduction in credit availability that is due to lack of competition is likely to be less damaging than in an environment with more scarcely available finance. Therefore, we anticipate that financial development will lessen the damaging impact of low competition on access. Our second variable of interest is credit information. Availability of credit information is directly linked to the impact of competition through the information hypothesis discussed in the introduction. The hypothesis states that in the presence of information asymmetries and agency costs, competition can reduce access to finance by making it more difficult for banks to internalize the benefits of investing in building lending relationships, in particular, with opaque clients (Petersen and Rajan, 1995; Marquez, 2002). Therefore, with improvement in the availability of information through public or private credit registries, the information asymmetries are reduced and, thus, the impact of low competition on access to finance through its impact on information production is likely to be reduced. 11 Finally, we consider how the extent of government bank ownership affects the link between competition and access. Often government-owned banks have a mandate to promote financial access, so in theory we would expect them to alleviate financing constraints. However, the empirical evidence suggests that government ownership is commonly associated with low bank efficiency and ineffective allocation of resources, including political lending (see La Porta et al., 2002; Iannotta, Nocera and Sironi, 2007; Micco, Panizza, and Yañez, 2007; Berger, Hasan, and Zhou, 2009; and Farazi, Feyen and Rocha, 2011, among others). In competitive environments the inefficiencies of government ownership could be mitigated by the pressure from competitors to government-owned banks. On the flip side, in countries with high government bank ownership lack of competition can be especially damaging as the checks and balances introduced by market mechanisms may be weak or absent. Thus, we expect that high government bank ownership will exacerbate the negative impact of low competition on access. To investigate whether the impact of competition measures on access to finance varies as a function of country-level characteristics such as financial development, credit information, and government bank ownership, we interact each of these country-level variables with our measure of competition, while simultaneously adding the same country-levels variable by themselves in the regression. The three measures and their sources are described in Table 1A. Table 4 reports our results for the interaction of financial development and competition. We observe a negative sign on the Lerner, as before, and a positive interaction with the measure of financial development. For both concentration measures we observe similar patterns – negative for concentration and positive for the interaction term. The magnitude of the coefficients in model (a) implies that for a country with an average level of financial development in our sample (which equals to 0.4), the marginal impact of the Lerner index is about 0.6; for a country with low financial development (such as one standard deviation below the average), the marginal impact is twice as large and the magnitude is about 1.2; while for a country with a high level of financial development (such as one standard deviation above the average) the marginal impact of the Lerner index is about zero. Translating these effects into changes in the probability of access to finance, we find that a one standard deviation increase in the Lerner results in a drop of almost 8 percentage points in the likelihood 12 of access to finance, when financial development is low. The same change in the Lerner leads to a 4 percentage points decline in the probability of access to finance for firms in a country with an average level of financial development, and no change in the likelihood of access to finance for those firms in a country with high levels of financial development. Table 5 presents the interactions with the country-level credit information index.19 Once again, we observe that the Lerner index is negative and highly significant, while the interaction with credit information is positive and highly significant. The results for the Herfindahl index are similar, while in the case of the share of assets held by the top 3 banks neither the measure by itself nor the interaction are significant. The credit information index varies from zero to six, with an average of about 3 and a standard deviation of 2. Using coefficients estimated in model (a), our results suggest that for a country with average credit information, the Lerner has a negative impact of 0.85, while for a country with low credit information (one standard deviation below the average), the impact more than doubles to 1.96. On the other hand, for countries with a high level of credit information (one standard deviation above the average), the impact is close to 0.27. Translating these into a probability of access to finance, we find that in a country with low credit information availability, a one standard deviation increase in the Lerner index results in about a 13 percentage point decrease in the probability of using a financial product such as a loan, line of credit or an overdraft. However, in a country with high credit information availability there is practically no impact. Thus, better credit information significantly mitigates the negative impact of low competition or high concentration. Finally, in Table 6 we show interactions with the country average share of government bank ownership. Here, we observe that the interactions are negative and significant, suggesting that higher government bank ownership is associated with a more damaging impact of low competition and high concentration. The Lerner index by itself is not significant, which suggests that when government ownership of banks in the country is zero, the impact of competition or 19 Because credit information index has very low variability from year to year in our sample and is not available for earlier years in our sample, we use country average credit information in our interaction term. The average itself is than subsumed into the country fixed effects and does not enter on its own. 13 concentration on access to finance is also zero. However, in countries with high government ownership, the negative impact is significant. The average government ownership in our sample is 0.25 with a standard deviation of 0.23. These numbers suggest that in a country with average government ownership the impact of Lerner is 0.79, while in a country that is one standard deviation above the average the impact is about twice as large, at 1.5. This translates into a decline in the probability of access to finance of about 10 percentage points as a result of a one standard deviation increase in the Lerner index in a country with high government bank ownership. The same impact is only 5 percent in a country with average government bank ownership and close to zero in a country without any government bank ownership. To summarize, we find that the impact of competition and concentration depends on the environment in which banks operate. This may explain the contradictory results observed in the previous literature, as in some countries the negative impact of low levels of competition may be mitigated by some positive factors such as availability of credit information or the overall level of financial development, while in other countries the impact may be exacerbated by factors such as high government bank ownership. 6. Conclusions The theory on the impact of bank competition on access to finance offers conflicting predictions and the empirical literature provides mixed results and suffers from a number of limitations. Combining multi-year firm-level data on access to finance with panel country-level data on bank competition, this paper offers new evidence on the link between competition and access to finance. One advantage of our dataset is that it allows us to control for unobserved differences between countries, using country fixed effects. Thus, we are able to isolate within country variation in competition and access to finance. Also, contrary to other studies that equate concentration with competition, we conduct estimations using direct measures of banks’ pricing behavior. Our results indicate that higher bank competition, as measured by lower levels of the Lerner index, increases firms’ access to finance, while commonly used concentration measures 14 are not reliable or robust predictors of financial access. In addition, we find that the impact of competition on access to finance depends on the environment in which banks operate and some features of the environment, such as higher levels of financial development and better credit information, can mitigate the damaging impact of low competition, while other characteristics, such as high government bank ownership, can exacerbate the negative impact. Overall, our results suggest that there are benefits to promoting bank competition. We leave for future research an analysis of the specific policies (e.g., adopting low barriers to bank entry and exit, fostering competitive pressures from non-bank competitors, facilitating access to credit information, implementing measures to ensure consumer protection, etc.) that regulators can implement to increase competition in the banking sector. 15 References Anzoategui, D., Martinez Peria, M., Rocha, R., 2010. “Bank competition in the Middle East And Northern Africa Region,� Review of Middle East Economics and Finance Vol. 6 : No. 2, Article 2, 2010. Anzoategui, D., Martinez Peria, M., Melecky, M., 2012. “Bank competition in Russia: An examination at different levels of aggregation.� Emerging Markets Review 13(1): 42-57, 2012. Beck, T., Demirgüç-Kunt, A., Maksimovic, V., 2004. “Bank competition and access to finance: international evidence.� Journal of Money, Credit, and Banking 36, 627–48. Berger, A.N., Hasan, I., Zhou, M., 2009. “Bank ownership and efficiency in China: What will happen in the world’s largest nation?� Journal of Banking & Finance 33, 113-130. Berger, A.N., Saunders, A., Scalise, J., Udell, G., 1998. “The effects of bank mergers and acquisitions on small business lending.� Journal of Financial Economics 50, 187-229. Bonaccorsi di Patti, E., Gobbi, G., 2007. “Winners or losers? The effects of banking consolidation on borrowers.� Journal of Finance 62, 669-96. Carbó, S., Humphrey, D., Maudos, J., Molyneux P., 2009. “Cross-country comparisons of competition and pricing power in European banking.� Journal of International Money and Finance 28, 115-134. Carbó-Valverde, S., Rodriguez-Fernandez, F., Udell, G., 2009. “Bank market power and SME financing constraints.� Review of Finance 13, 309-340. Cetorelli, N., 1999. “Competitive analysis in banking: appraisal of the methodologies,� Federal Reserve Bank of Chicago Economic Perspectives, first quarter, 2-15. Chong, T.T.L., Lu, L., Ongena, S., 2012. “Does banking competition alleviate or worsen credit constraints faced by small and medium enterprises? Evidence from China.� CentER Discussion Paper, 2012-013. Claessens, S., Laeven, L., 2004. “What drives bank competition? Some international evidence.� Journal of Money, Credit and Banking 36(3), Part 2, 563-583. Claessens, S., Laeven, L., 2005. “Financial dependence, banking sector competition, and economic growth.� Journal of the European Economic Association 3(1), 179-207, 03. Delis, M.D., 2012. “Bank competition, financial reform, and institutions: The importance of being developed.� Journal of Development Economics 97(2), pp. 450–465. 16 Dell’Ariccia, G., Igan, D., Laeven, L., 2008. “Credit booms and lending standards: evidence from the subprime mortgage market.� IMF working paper 08/106. Degryse, H., Masschelein, N., Mitchell, J., 2011. “Staying, dropping, or switching: The impact of bank merges on small firms.� Journal of Finance 24(4), 1102-1140. Demirgüç-Kunt, A., Laeven, L., Levine, R., 2004. “Regulations, market structure, institutions and the cost of financial intermediation.� Journal of Money, Credit and Banking 36, 593- 622. Erel, I., 2011. “The effect of bank mergers on loan prices: Evidence from the United States.� Review of Financial Studies, 24(4), 1068-1101. Farazi, S., Feyen, E., Rocha, R., 2011. “Bank ownership and performance in the Middle East and North Africa Region.� The World Bank Policy Research Working Paper 5620, April. Fernandez de Guevara, J., Maudos, J, Perez, F., (2005). “Market Power in European Banking Sectors.� Journal of Financial Services Research 27, 109-137. Fernandez de Guevara, J., Maudos, J, Perez, F., (2007). “Integration and Competition in the European Financial Markets,� Journal of International Money and Finance 26, 26-45. Fischer, K-H., 2000. “Acquisition of information in loan markets and bank market power An empirical investigation.� Working Paper, Goethe University, Frankfurt. Hauswald, R., Marquez, R., 2006. “Competition and strategic information acquisition in credit markets.� Review of Financial Studies 19(3), 967-1000. Iannotta, G., Nocera, G., Sironi, A., 2007. “Ownership structure, risk and performance in the European banking industry.� Journal of Banking & Finance 31, 2127-2149. King, R., Levine, R., 1993. “Finance and growth: Schumpeter might be right.� Quarterly Journal of Economics 153, 717-738. LaPorta, R., López-de-Silanes, F., Shleifer, A., 2002. “Government ownership of banks.� Journal of Finance 57, 265-301. Lerner, A., 1934. “The concept of monopoly and the measurement of monopoly power.� Review of Economic Studies 1, 157–175. Love, I., 2003. “Financial development and financing constraints: International evidence from the structural investment model.� Review of Financial Studies 16(3), 765-791. Marquez, R., 2002. “Competition, adverse selection, and information dispersion in the banking industry.� Review of Financial Studies 15(3), 901-926. 17 Micco, A., Panizza, U., Yañez, M., 2007. “Bank ownership and performance. Does politics matter?� Journal of Banking & Finance 31, 219-241. OECD, 2009. The financial crisis: reforms and exit strategies. http://www.oecd.org/dataoecd/55/47/43091457.pdf OECD, 2010. Competition, concentration and stability in the banking sector. OECD Policy Roundtables. http://www.oecd.org/dataoecd/52/46/46040053.pdf Panzar, J.C., Rosse, J.N., 1982. “Structure, conduct and comparative statistics.� Bell Laboratories Economic Discussion Paper. Panzar, J.C., Rosse, J. N., 1987. “Testing for `monopoly' equilibrium.� Journal of Industrial Economics 35, 443-456. Peek, J., Rosegren, E., 1996. “Small business credit availability: how important is size of lender.� In A. Saunders and I. Walter (eds.), Financial system design: The case for universal banking. Homewood, IL, Irwin Publishing. Petersen, M. A., Rajan, R., 1994. “The benefits of lending relationships: evidence from small business data.� Journal of Finance 49, 3-37. Petersen, M. A., Rajan, R., 1995. “The effect of credit market competition on lending relationships.� The Quarterly Journal of Economics 110(2), 407-43. Rice, T., Strahan, P., 2010. “Does credit competition affect small-firm finance?� Journal of Finance 65(3), 861-889. Ruckes, M., 2004. “Bank competition and credit standards.� Review of Financial Studies 17(4), 1073-1103. Sapienza, P., 2002. “The effects of banking mergers on loan contracts.� Journal of Finance 57, 329-68. Turk-Ariss, R., 2009. “Competitive behavior in Middle East and North Africa banking systems.� The Quarterly Review of Economics and Finance 49, 693-710. Zarutskie, R., 2006. “Evidence on the effects of bank competition on firm borrowing and investment.� Journal of Financial Economics 81(3), 503-37. 18 Table 1. Panel A: Variable Description and Data Sources Variable Description and Data Source Firm-Level Variables Dummy variable equal to 1 if the firm has access to bank finance Access to finance (loan, overdraft or line of credit) from World Bank Enterprise Surveys. Number of permanent full-time employees from World Bank Firm size (employees) Enterprise Surveys. Dummy variable equal to 1 if the firm is in the manufacturing sector Manufacturing from World Bank Enterprise Surveys. Dummy variable equal to 1 if 10% or more of sales are exported Exporter directly or indirectly by the firm from World Bank Enterprise Surveys. Dummy variable equal to 1 if 50% or more of the firm is owned by Foreign-owned foreign organizations from World Bank Enterprise Surveys. Dummy variable equal to 1 if 10% or more of the firm is owned by the Government-owned government from World Bank Enterprise Surveys. Firm age Age of the firm in years from World Bank Enterprise Surveys. Country-Level Variables Lerner index constructed using variables from Bankscope (see Lerner index Appendix 2). Concentration 3 Fraction of total assets held by top 3 banks from Bankscope. Herfindahl index calculated as the sum over all banks in the country of Herfindahl index the squared market share (in terms of assets) of each bank from Bankscope. Inflation calculated as the annual change in the GDP deflator from the Inflation rate World Development Indicators (WDI) World Bank. Financial development Domestic credit to the private sector (fraction of GDP) from WDI. Depth of credit information index is a measure of the coverage, scope and accessibility of credit information available through either a public Credit information index credit registry or a private credit bureau, both by law, and in practice. (0-6) obtained from Doing Business Indicators. Fraction of banking system's assets in banks that are 50% or more Share of government-owned banks government-owned from the World Bank Regulation and Supervision Survey. 19 Table 1. Panel B: Summary Statistics Standard Variable Obs Mean Median Min Max Deviation Firm-Level Variables Access to finance 68353 0.602 1 0.489 0 1 Firm size (employees) 68353 100.622 25 207.429 1 1755 Manufacturing 68353 0.623 1 0.485 0 1 Exporter 68353 0.235 0 0.424 0 1 Foreign-owned 68353 0.095 0 0.293 0 1 Government-owned 68353 0.049 0 0.217 0 1 Firm age 68353 17.729 12 16.983 1 193 Country-Level Variables Lerner index 68353 0.251 0.235 0.067 0.066 0.437 Concentration 3 67720 0.589 0.567 0.172 0.279 0.985 Herfindahl index 68353 0.188 0.142 0.132 0.050 0.714 Inflation rate 68353 0.086 0.062 0.088 -0.074 0.795 Financial development 68353 0.403 0.295 0.300 0.019 1.571 Credit information index 38240 3.438 4 2.008 0 6 Share of government-owned banks 61693 0.250 .17 0.232 0 .752 20 Table 2: Baseline Regressions The regressions below are estimated using country fixed effects and robust standard errors clustered at the country-year level. The dependent variable Access to finance is a dummy variable that indicates whether the firm has access to a loan, overdraft, or a line of credit. The Lerner index is a measure of competition (higher values imply lower levels of competition). The Herfindahl index (HI) and Concentration 3 are measures of concentration. Log firm size is the logarithm of the firm’s number of permanent employees. Log firm age is the logarithm of the firm’s age in years. Government-owned and Foreign-owned are dummy variables that equal one if the firm has government or foreign ownership, respectively. Exporter is a dummy variable that indicates if the firm is an exporting firm. Manufacturing is a dummy variable that takes value 1 if the firm is in the manufacturing industry. Financial development is measured as domestic credit to the private sector as a fraction of GDP. The inflation rate is measured as the growth rate of the GDP deflator (annual). Results in column (b) are weighted by the inverse of the standard deviation of the Lerner Index. *** p<0.01, ** p<0.05, * p<0.1 Access to finance Variables (a) (b) (c) (d) Lerner index -0.615** -0.749** [0.269] [0.335] Concentration 3 -0.129 [0.142] Herfindahl index -0.033 [0.148] Log firm size 0.086*** 0.088*** 0.087*** 0.087*** [0.004] [0.004] [0.004] [0.004] Manufacturing 0.028** 0.028** 0.031*** 0.032*** [0.011] [0.011] [0.011] [0.011] Exporter 0.032*** 0.028*** 0.032*** 0.031*** [0.008] [0.010] [0.008] [0.008] Foreign-owned -0.080*** -0.092*** -0.083*** -0.081*** [0.011] [0.012] [0.011] [0.011] Government-owned -0.136*** -0.164*** -0.137*** -0.141*** [0.031] [0.035] [0.031] [0.032] Log firm age 0.013*** 0.016*** 0.013*** 0.012*** [0.004] [0.004] [0.004] [0.004] Financial development 0.315*** 0.331*** 0.219** 0.273*** [0.090] [0.061] [0.105] [0.087] Inflation rate -0.357** -0.450*** -0.347** -0.359** [0.138] [0.158] [0.136] [0.140] Constant 0.486*** 0.538*** 0.525** 0.324** [0.153] [0.162] [0.202] [0.148] Observations 68,353 68,353 67,270 68,353 R-squared 0.211 0.187 0.213 0.209 21 Table 3: Regressions Correcting for Differences in the Number of Firms Across Countries The regressions below are estimated using country fixed effects and robust standard errors clustered at the country-year level. The dependent variable Access to finance is a dummy variable that indicates whether the firm has access to a loan, overdraft, or a line of credit. The Lerner index is a measure of competition (higher values imply lower levels of competition). The Herfindahl index (HI) and Concentration 3 are measures of concentration. Log firm size is the logarithm of the number of permanent employees. Log firm age is the logarithm of the firm’s age in years. Government-owned and Foreign-owned are dummy variables that equal one if the firm has government or foreign ownership, respectively. Exporter is a dummy variable that indicates if the firm is an exporting firm. Manufacturing is a dummy variable that takes value 1 if the firm is in the manufacturing industry. Financial development is measured as domestic credit to the private sector as a fraction of GDP. The inflation rate is measured as the growth rate of the GDP deflator (annual). Results in columns (a), (c), and (d) are weighted by the inverse of the square root of the number of firms; those in column (b) are weighted by the inverse of (the square root of the number of firms  the inverse of the standard deviation of the Lerner Index). *** p<0.01, ** p<0.05, * p<0.1 Access to finance Variables (a) (b) (c) (d) Lerner index -0.434** -0.387** [0.200] [0.195] Concentration 3 -0.000 [0.122] Herfindahl index 0.034 [0.138] Log firm size 0.089*** 0.089*** 0.089*** 0.089*** [0.003] [0.004] [0.003] [0.003] Manufacturing 0.019* 0.018* 0.020* 0.020** [0.010] [0.011] [0.010] [0.010] Exporter 0.044*** 0.042*** 0.042*** 0.044*** [0.008] [0.012] [0.008] [0.008] Foreign-owned -0.074*** -0.086*** -0.076*** -0.075*** [0.012] [0.013] [0.012] [0.011] Government-owned -0.198*** -0.232*** -0.197*** -0.200*** [0.022] [0.037] [0.023] [0.022] Log firm age 0.013** 0.018*** 0.013** 0.013** [0.005] [0.007] [0.006] [0.005] Financial development 0.357*** 0.332*** 0.352*** 0.369*** [0.056] [0.043] [0.060] [0.052] Inflation rate -0.160 -0.212* -0.152 -0.156 [0.099] [0.116] [0.098] [0.099] Constant 0.369*** 0.382*** 0.341* 0.227* [0.135] [0.138] [0.186] [0.130] Observations 68,353 68,353 67,270 68,353 R-squared 0.199 0.181 0.201 0.198 22 Table 4: Regressions Including the Interaction of Competition and Concentration with Financial Development The regressions below are estimated using country fixed effects and robust standard errors clustered at the country-year level. The dependent variable Access to finance is a dummy variable that indicates whether the firm has access to a loan, overdraft, or a line of credit. The Lerner index is a measure of competition (higher values imply lower levels of competition). The Herfindahl index (HI) and Concentration 3 are measures of concentration. Log firm size is the logarithm the number of permanent employees. Log firm age is the logarithm of the fir m’s age in years. Government-owned and Foreign-owned are dummy variables that equal one if the firm has government or foreign ownership, respectively. Exporter is a dummy variable that indicates if the firm is an exporting firm. Manufacturing is a dummy variable that takes value 1 if the firm is in the manufacturing industry. Financial development is measured as domestic credit to the private sector as a fraction of GDP. The inflation rate is measured as the growth rate of the GDP deflator (annual). Results in column (b) are weighted by the inverse of the standard deviation of the Lerner Index. *** p<0.01, ** p<0.05, * p<0.1 Variables Access to finance (a) (b) (c) (d) Lerner index -1.354*** -1.386*** [0.368] [0.429] Concentration 3 -0.521*** [0.166] Herfindahl index (HI) -0.677*** [0.158] Financial development -0.160 -0.047 -0.223 0.130 [0.226] [0.171] [0.151] [0.088] Lerner  Financial development 1.943** 1.779** [0.777] [0.696] Concentration 3  Financial development 0.898*** [0.177] HI Index  Financial development 0.993*** [0.197] Log firm size 0.086*** 0.088*** 0.086*** 0.086*** [0.004] [0.004] [0.004] [0.004] Manufacturing 0.027** 0.027** 0.029*** 0.029*** [0.010] [0.011] [0.011] [0.011] Exporter 0.033*** 0.030*** 0.032*** 0.031*** [0.008] [0.010] [0.008] [0.008] Foreign-owned -0.078*** -0.090*** -0.080*** -0.080*** [0.011] [0.012] [0.011] [0.011] Government-owned -0.136*** -0.164*** -0.128*** -0.129*** [0.030] [0.035] [0.030] [0.030] Log firm age 0.012*** 0.016*** 0.012*** 0.011*** [0.004] [0.004] [0.004] [0.004] Inflation rate -0.317** -0.416** -0.325** -0.350** [0.145] [0.160] [0.135] [0.134] Constant 0.692*** 0.696*** 0.816*** 0.513*** [0.167] [0.176] [0.203] [0.142] Observations 68,353 68,353 67,270 68,353 R-squared 0.213 0.189 0.217 0.214 23 Table 5: Regressions Including the Interaction of Competition and Concentration with Credit Information The regressions below are estimated using country fixed effects and robust standard errors clustered at the country-year level. The dependent variable Access to finance is a dummy variable that indicates whether the firm has access to a loan, overdraft, or a line of credit. The Lerner index is a measure of competition (higher values imply lower levels of competition). The Herfindahl index (HI) and Concentration 3 are measures of concentration. Log firm size is the logarithm of the number of permanent employees. Log firm age is the logarithm of the firm’s age in years. Government-owned and Foreign-owned are dummy variables that equal one if the firm has government or foreign ownership, respectively. Exporter is a dummy variable that indicates if the firm is an exporting firm. Manufacturing is a dummy variable that takes value 1 if the firm is in the manufacturing industry. Financial development is measured as domestic credit to the private sector as a fraction of GDP. The inflation rate is measured as the growth rate of the GDP deflator (annual). Credit information is the country average credit information index that measures the coverage, scope and accessibility of credit information available through either a public credit registry or a private credit bureau. Results in column (b) are weighted by the inverse of the standard deviation of the Lerner Index. *** p<0.01, ** p<0.05, * p<0.1 Access to finance Variables (a) (b) (c) (d) Lerner index -2.758*** -3.219*** [0.568] [0.476] Concentration 3 -0.333 [0.341] Herfindahl index -0.729* [0.424] Lerner  Credit information 0.557*** 0.642*** [0.132] [0.104] Concentration 3  Credit information 0.059 [0.076] H Index  Credit information 0.184* [0.101] Log firm size 0.087*** 0.088*** 0.089*** 0.089*** [0.004] [0.004] [0.004] [0.004] Manufacturing 0.019** 0.015* 0.024** 0.025** [0.009] [0.009] [0.009] [0.010] Exporter 0.033*** 0.031*** 0.031*** 0.031*** [0.008] [0.010] [0.008] [0.008] Foreign owned -0.080*** -0.092*** -0.084*** -0.082*** [0.011] [0.012] [0.011] [0.011] Government owned -0.175*** -0.190*** -0.177*** -0.176*** [0.027] [0.033] [0.028] [0.028] Log firm age 0.011*** 0.015*** 0.011*** 0.010** [0.004] [0.004] [0.004] [0.004] Financial development 0.244*** 0.271*** 0.211** 0.283*** [0.080] [0.056] [0.085] [0.069] Inflation rate -0.346** -0.432*** -0.368*** -0.409*** [0.144] [0.164] [0.140] [0.151] Constant 1.119*** 1.225*** 0.711** 0.553*** [0.226] [0.203] [0.346] [0.195] Observations 65,428 65,428 64,345 65,428 R-squared 0.224 0.199 0.219 0.217 24 Table 6: Regressions Including the Interaction of Competition and Concentration with Government Bank Share The regressions below are estimated using country fixed effects and robust standard errors clustered at the country-year level. The dependent variable Access to finance is a dummy variable that indicates whether the firm has access to a loan, overdraft, or a line of credit. The Lerner index is a measure of competition (higher values imply lower levels of competition). The Herfindahl index (HI) and Concentration 3 are measures of concentration. Log firm size is the logarithm of the number of permanent employees. Log firm age is the logarithm of the firm’s age in years. Government-owned and Foreign-owned are dummy variables that equal one if the firm has government or foreign ownership, respectively. Exporter is a dummy variable that indicates if the firm is an exporting firm. Manufacturing is a dummy variable that takes value 1 if the firm is in the manufacturing industry. Financial development is measured as domestic credit to the private sector as a fraction of GDP. The inflation rate is measured as the growth rate of the GDP deflator (annual). Share of government banks is the fraction of banking system's assets in banks that are 50% or more government-owned. Results in column (b) are weighted by the inverse of the standard deviation of the Lerner Index. *** p<0.01, ** p<0.05, * p<0.1 Variables Access to finance (a) (b) (c) (d) Lerner index -0.018 0.045 [0.294] [0.261] Concentration 3 0.180 [0.198] Herfindahl index 0.338* [0.183] Lerner  Share government banks -3.068** -4.131*** [1.281] [1.123] Concentration 3  Share government banks -1.491* [0.860] H Index  Share of government banks -2.452** [0.940] Log firm size 0.086*** 0.088*** 0.088*** 0.088*** [0.004] [0.004] [0.004] [0.004] Manufacturing 0.015 0.011 0.019** 0.019** [0.009] [0.009] [0.010] [0.009] Exporter 0.036*** 0.034*** 0.036*** 0.036*** [0.008] [0.010] [0.008] [0.008] Foreign-owned -0.076*** -0.088*** -0.080*** -0.078*** [0.011] [0.012] [0.011] [0.011] Government-owned -0.204*** -0.220*** -0.209*** -0.208*** [0.019] [0.026] [0.019] [0.018] Log firm age 0.010*** 0.014*** 0.008** 0.007* [0.004] [0.004] [0.004] [0.004] Financial development 0.354*** 0.355*** 0.266*** 0.276*** [0.085] [0.059] [0.078] [0.076] Inflation rate -0.357** -0.428** -0.309** -0.316** [0.150] [0.184] [0.150] [0.156] Constant 0.797*** 0.937*** 0.966*** 0.643*** [0.214] [0.206] [0.336] [0.193] Observations 62,143 62,143 61,060 62,143 R-squared 0.222 0.192 0.222 0.219 25 Table A1. Number of Firms, by Country and Year of Survey Survey Year # of 2002 2003 2004 2005 2006 2007 2008 2009 2010 Total Surveys Albania 167 202 276 645 3 Angola 421 306 727 2 Argentina 1019 1017 2036 2 Armenia 169 351 356 876 3 Azerbaijan 167 345 291 803 3 Bangladesh 948 1471 2419 2 Belarus 247 325 240 812 3 Benin 182 143 325 2 Bolivia 603 346 949 2 Bosnia-Herzegovina 200 343 543 2 Botswana 340 260 600 2 Brazil 1619 1170 2789 2 Bulgaria 242 492 298 1008 269 2309 5 Burkina Faso 138 357 495 2 Cameroon 168 351 519 2 Chile 941 989 1003 2933 3 China 1353 1572 2925 2 Colombia 990 934 1924 2 Congo 338 334 672 2 Croatia 169 227 615 1011 3 Czech Republic 258 334 223 815 3 El Salvador 463 681 1144 2 Estonia 164 216 264 644 3 Georgia 172 198 334 704 3 Guatemala 431 503 934 2 Honduras 446 421 867 2 Hungary 243 605 283 1131 3 India 1461 3086 4547 2 Indonesia 644 1313 1957 2 Kazakhstan 246 582 464 1292 3 Kenya 211 653 864 2 26 Table A1. (continued) Survey Year # of 2002 2003 2004 2005 2006 2007 2008 2009 2010 Total Surveys Latvia 170 203 262 635 3 Lithuania 197 228 205 267 897 4 Macedonia 165 199 354 718 3 Malawi 146 145 291 2 Mali 131 490 295 916 3 Mauritius 159 374 533 2 Moldova 173 103 349 354 979 4 Nicaragua 450 466 916 2 Peru 120 626 980 1726 3 Philippines 600 1093 1693 2 Poland 493 104 971 402 1970 4 Romania 250 594 472 1316 3 Russian Federation 489 593 903 1985 3 Serbia & Montenegro 298 482 780 2 Slovakia 158 212 252 622 3 Slovenia 185 221 271 677 3 South Africa 424 929 1353 2 Turkey 503 550 1276 1083 3412 4 Ukraine 446 588 756 1790 3 Uruguay 583 584 1167 2 Vietnam 1080 1014 2094 2 Zambia 190 482 672 2 Total 9545 7198 2393 10977 11372 5924 2413 12472 6059 68353 137 27 Table A2. Lerner Index by Country and Year Lerner Index Countries 2001 2002 2003 2004 2005 2006 2007 2008 2009 Albania 0.382 0.207 0.269 0.246 0.296 0.260 0.265 0.254 0.272 Algeria 0.255 0.362 0.404 0.452 0.510 0.509 0.447 0.504 0.550 Angola 0.449 0.463 0.616 0.537 0.437 0.399 0.368 0.396 0.430 Argentina 0.184 1.385 0.499 0.282 0.286 0.274 0.249 0.255 0.290 Armenia 0.261 0.276 0.370 0.346 0.311 0.343 0.309 0.307 0.213 Azerbaijan 0.429 0.277 0.348 0.349 0.338 0.277 0.274 0.281 0.267 Bangladesh 0.247 0.213 0.227 1.021 0.273 0.281 0.249 0.270 0.291 Belarus 0.254 0.198 0.189 0.199 0.203 0.244 0.294 0.275 0.287 Benin 0.295 0.292 0.267 0.248 0.232 0.172 0.213 0.272 0.314 Bolivia 0.177 0.247 0.181 0.160 0.167 0.194 0.235 0.303 0.230 Bosnia-Herzegovina 0.162 0.200 0.194 0.230 0.177 0.199 Botswana 0.227 0.207 0.227 0.273 0.252 0.303 0.207 0.281 0.185 Brazil 0.174 0.181 0.226 0.191 0.229 0.239 0.256 0.357 0.239 Bulgaria 0.290 0.206 0.245 0.225 0.237 0.284 0.285 0.254 0.268 Burkina Faso 0.310 0.375 0.343 0.307 0.337 0.298 0.314 0.212 0.258 Cameroon 0.471 0.478 0.451 0.450 0.421 0.417 0.406 0.380 0.221 Chile 0.317 0.288 0.279 0.209 0.300 0.347 0.383 0.231 0.377 China 0.224 0.295 0.319 0.314 0.336 0.354 0.351 0.336 0.331 Colombia 0.126 0.124 0.200 0.217 0.267 0.232 0.270 0.244 0.325 Congo 0.250 0.134 0.243 0.258 0.163 0.210 0.168 0.138 0.182 Costa Rica 0.107 0.109 0.199 0.204 0.195 0.186 0.201 0.157 0.135 Croatia 0.205 0.196 0.188 0.172 0.190 0.183 0.184 0.257 0.163 Czech Republic 0.100 0.127 0.184 0.196 0.177 0.176 0.222 0.221 0.301 Dominican Republic 0.141 0.129 0.134 0.190 0.141 0.135 0.117 0.134 0.148 Egypt 0.182 0.172 0.229 0.188 0.217 0.190 0.260 0.222 0.220 El Salvador 0.266 0.301 0.256 0.296 0.332 0.314 0.327 0.333 0.328 Estonia 0.066 0.129 0.171 0.176 0.257 0.344 0.252 0.161 0.160 Ethiopia 0.451 0.349 0.446 0.514 0.556 0.609 0.516 0.539 0.511 Georgia 0.405 0.349 0.341 0.332 0.403 0.374 0.239 0.272 0.276 Ghana 0.401 0.296 0.313 0.214 0.220 0.193 0.257 0.225 Guatemala 0.097 0.107 0.163 0.160 0.193 0.198 0.206 0.236 0.213 28 Lerner Index Countries 2001 2002 2003 2004 2005 2006 2007 2008 2009 Honduras 0.136 0.198 0.269 0.183 0.188 0.211 0.217 0.259 0.228 Hungary 0.176 0.183 0.197 0.222 0.224 0.233 0.228 0.192 0.226 India 0.153 0.207 0.228 0.276 0.232 0.226 0.224 0.199 0.213 Indonesia 0.194 0.194 0.207 0.243 0.226 0.227 0.229 0.230 0.214 Ivory Coast 0.269 0.259 0.229 0.251 0.274 0.242 0.256 0.235 0.294 Jamaica 0.202 0.202 0.225 0.217 0.236 0.230 0.221 0.278 0.285 Jordan 0.194 0.217 0.286 0.307 0.417 0.333 0.287 0.289 0.286 Kazakhstan 0.338 0.323 0.321 0.317 0.320 0.274 0.348 0.313 0.318 Kenya 0.192 0.209 0.261 0.246 0.268 0.277 0.290 0.246 0.245 Korea 0.228 0.241 0.262 0.279 0.271 0.261 0.248 0.213 0.223 Latvia 0.240 0.294 0.322 0.309 0.321 0.309 0.280 0.281 0.222 Lebanon 0.130 0.152 0.178 0.145 0.165 0.150 0.141 0.167 0.203 Lithuania 0.117 0.183 0.175 0.172 0.206 0.199 0.203 0.180 0.115 Macedonia 0.384 0.311 0.300 0.337 0.396 0.367 0.310 0.214 0.233 Malawi 0.322 0.365 0.309 0.280 0.370 0.350 0.377 0.353 0.392 Malaysia 0.302 0.326 0.339 0.334 0.313 0.303 0.301 0.293 0.326 Mali 0.293 0.284 0.312 0.291 0.293 0.323 0.244 0.221 0.241 Mauritania 0.274 0.288 0.300 0.250 0.362 0.398 0.309 0.396 0.407 Mauritius 0.289 0.394 0.238 0.285 0.283 0.288 0.272 0.294 0.263 Mexico 0.132 0.136 0.164 0.175 0.186 0.185 0.204 0.222 0.248 Moldova 0.365 0.388 0.340 0.349 0.268 0.288 0.298 0.243 0.290 Mozambique 0.291 0.197 0.164 0.198 0.247 0.243 0.221 0.253 0.315 Nepal 0.293 0.260 0.238 0.298 0.324 0.312 0.303 0.284 0.286 Nicaragua 0.193 0.190 0.271 0.374 0.372 0.355 0.396 0.398 0.407 Nigeria 0.290 0.266 0.233 0.233 0.270 0.320 0.323 0.365 0.331 Oman 0.279 0.404 0.409 0.437 0.435 0.430 0.365 0.411 0.434 Pakistan 0.129 0.173 0.280 0.292 0.332 0.302 0.238 0.240 0.218 Panama 0.226 0.272 0.309 0.302 0.307 0.298 0.270 0.289 0.292 Paraguay 0.090 0.056 0.042 0.072 0.124 0.094 0.108 0.079 0.096 Peru 0.170 0.224 0.217 0.714 0.321 0.316 0.339 0.352 0.354 Philippines 0.055 0.187 0.190 0.227 0.227 0.232 0.228 0.182 0.225 Poland 0.172 0.201 0.215 0.225 0.221 0.227 0.262 0.240 0.270 Romania 0.215 0.216 0.183 0.179 0.211 0.181 0.181 0.195 0.183 29 Lerner Index Countries 2001 2002 2003 2004 2005 2006 2007 2008 2009 Russian Federation 0.319 0.259 0.258 0.235 0.245 0.244 0.251 0.253 0.104 Serbia 0.376 0.350 0.338 0.333 0.272 0.207 0.168 0.180 Sierra Leone 0.495 0.487 0.462 0.443 0.372 0.336 0.189 0.135 0.149 Slovakia 0.112 0.125 0.163 0.116 0.138 0.194 0.179 0.210 0.221 Slovenia 0.183 0.200 0.193 0.223 0.199 0.193 0.220 0.147 0.208 South Africa 0.347 0.374 0.240 0.154 0.244 0.293 0.319 0.292 0.272 Sri Lanka 0.174 0.183 0.227 0.240 0.220 0.251 0.228 0.215 0.232 Tanzania 0.468 0.429 0.284 0.319 0.303 0.325 0.273 Thailand 0.194 0.220 0.301 0.344 0.340 0.290 0.289 0.291 0.350 Turkey 0.229 0.199 0.263 0.196 0.217 0.207 0.209 0.201 0.265 Ukraine 0.223 0.208 0.216 0.234 0.207 0.245 0.216 0.269 0.262 Uruguay 0.001 0.092 0.033 0.151 0.236 0.193 0.256 0.231 0.229 Uzbekistan 0.580 0.373 0.293 0.290 0.318 0.350 0.357 0.310 0.286 Venezuela 0.238 0.279 0.299 0.314 0.267 0.305 0.280 0.263 0.300 Vietnam 0.263 0.246 0.203 0.213 0.231 0.237 0.246 0.214 0.243 Zambia 0.353 0.200 0.075 0.241 0.307 0.348 0.364 0.376 0.346 30 Appendix 1: Constructing the measure of access to finance We use the World Bank Enterprise Surveys dataset (www.enterprisesurveys.org) assembled with a module of identical questions included in all questionnaires. The common framework of the questionnaire enables cross- country analyses using variables specified in the core module. A complication in constructing the measure of access stems from changes in the core survey modules made for surveys administered after 2005. That is, the variables required to construct a measure of access are defined differently in the old (2002-2005) and new (2006-2010) core modules.. From the old surveys, we consider the following questions:  “Do you have an overdraft facility or line of credit?�: Yes/No  “For the most recent loan or overdraft�: o When was this financing approved (year)? o Did the financing require collateral or a deposit? o If yes, what share of collateral was:  Land and buildings?  Machinery?  Intangible assets (accounts receivable, inventory)?  Personal assets of owner/manager (e.g. house)? o What was the approximate value of collateral required as a percentage of the loan value? o What is the loan's approximate annual cost/ rate of interest? o What is the duration (term) of the loan? From the new surveys, we consider the following questions:  “At this time, does this establishment have an overdraft facility?�: Yes/No  “At this time, does this establishment have a line of credit or loan from a financial institution?�: Yes/No 31 Given the nature of differences in the questionnaires, overdraft facility, line of credit, and loan are impossible to identify separately. Instead, we define Access to finance as having access to any one of the three credit facilities. The dependent variable, Access to finance, is a dummy variable that takes value 1 if the firm responds “yes� to either of the two questions, and 0 if “no� to both. A further obstacle arises due to the loan or overdraft question in the old surveys not being a dichotomous yes/no query. We assume that firms answering any further questions about their most recent loan or overdraft facility have access to at least one of the two types of financing. 32 Appendix 2: Constructing the Lerner Index20 The Lerner Index is defined as the difference between banking output prices and marginal costs (relative to prices). It is calculated as: , where P is the price of outputs and MC is the marginal cost. Price is calculated as the total gross revenue of the bank divided by the total assets. We compute the marginal costs by taking the derivative with respect to total assets from the following empirical specification of the translog cost function: ln(Cit) = α0i + β0ln(Qit) + β10.5[ln(Qit)]2 + α1ln(W1it)+ α2ln(W2it) + α3ln(W3it)+ β2ln(Qit)*ln(W1it) + β3ln(Qit)*ln(W2it) + β4ln(Qit)*ln(W3it) + α4ln(W1it)*ln(W2it) + α5ln(W1it)*ln(W3it) + α6ln(W2it)*ln(W3it) + α70.5[ln(W1it)] 2+ α80.5[ln(W2it)] 2 + α90.5[ln(W3it)] 2+ α10 ln(Equity)it+ α11ln(Net Loans)it + Fi + Yt + eit where Cit is total operating plus financial costs for bank i in time period t, Q is total assets, W 1 is the ratio of interest expenses to total deposits, W2 is the ratio of personnel expenses to total assets, W3 is the ratio of other operating and administrative expenses to total assets, Equity is the ratio of firm equity to total assets, Net Loans is the ratio of net loans to total assets, Fi are firm fixed effects and Yt are the full set of year dummies. A constrained fixed effects regression with time dummies is estimated under restrictions of symmetry and homogeneity of degree one in the price of inputs. We use bank-level data from Bankscope to calculate the Lerner Index. Only banks classified as commercial, cooperative, Islamic, savings, and bank holding companies are considered in the analysis. 20 We follow the literature on the Lerner index mentioned in footnote 11. 33 Within each country, we omit outlying observations that are in the top and bottom 1 percentile of the distribution for ln(W1), ln(W2), ln(W3) and their interaction with each other, ln(Equity) and ln(Net Loans). Under the assumption that the slope of the cost function within a country is constant through time, we calculate the marginal costs (MC) for all banks in a country from a single translog cost function regression over the entire range of available years from 1996-2010. Using the estimated coefficients, MC is calculated as: MCit = (β0+ β 1*ln(Qit)+ β2*ln(W1it)+ β3*ln(W2it)+ β4*ln(W3it))*( Cit /Total Assets) Variations in bank-level Lerner within a country are, thus, a result of variations in Q, W1, W2, W3, C, total assets and P. When the degrees of freedom in the fixed effects regression for a country are less than 20, we do not compute the Lerner Index due to low precision. In addition, the computed Lerner levels for banks within a country falling in the top and bottom 1 percentiles of the distribution are omitted as outliers. The Lerner Index for a country in a particular year is the average of all bank-level Lerners for that year. 34