Finance in Lower Income Countries
An Empirical Exploration

This paper considers how a comprehensive set of factors relates to financial sector performance in low-income countries (LICs). It finds that corruption and inflation are associated with a shallower and less efficient financial system, while legal origin and characteristics of the supervisory and regulatory framework have no significant relationship with performance. Moreover, better contract enforcement and information about borrowers are associated with more private sector credit. Some results are surprising. Countries with more foreign bank penetration seem to have shallower and not necessarily more efficient financial sectors, while a larger presence of state-owned banks is correlated with more bank deposits and lower overhead costs, even after controlling for market size and concentration. Although these relationships are robust, more research is needed to ascertain the direction of causality and identify channels of transmission before deriving policy implications.


This paper considers how a comprehensive set of factors relates to financial sector performance in low-income countries (LICs). It finds that corruption and inflation are associated with a shallower and less efficient financial system, while legal origin and characteristics of the supervisory and regulatory framework have no significant relationship with performance. Moreover, better contract enforcement and information about borrowers are associated with more private sector credit. Some results are surprising. Countries with more foreign bank penetration seem to have shallower and not necessarily more efficient financial sectors, while a larger presence of state-owned banks is correlated with more bank deposits and lower overhead costs, even after controlling for market size and concentration. Although these relationships are robust, more research is needed to ascertain the direction of causality and identify channels of transmission before deriving policy implications.

I. Introduction

Although complex, diversified, and low-cost financial systems are often considered a prerogative of advanced countries, some lower-income countries (LICs) manage to do much better than others developing and maintaining these.2 Measures of depth and cost efficiency of banking systems vary greatly within this group (Figures 1-4). Furthermore, differences in income per capita (which are also substantial) explain only some of the variation (Figure 5). What explains the rest? This paper analyzes indicators of financial sector development and performance in LICs and relates them to a comprehensive set of potential explanatory variables.

Figure 1.
Figure 1.

Bank Deposits in LICs by Region

(Percent of GDP)

Citation: IMF Working Papers 2005, 167; 10.5089/9781451861860.001.A001

Source: IMF. International Financial Statistics.
Figure 2.
Figure 2.

Bank Credit to the Private Sector in LICs by Region

(Percent of GDP)

Citation: IMF Working Papers 2005, 167; 10.5089/9781451861860.001.A001

Source: IMF, International Financial Statistics.
Figure 3.
Figure 3.

Banking Sector Overhead Costs in LICs by Region

(Percent of assets)

Citation: IMF Working Papers 2005, 167; 10.5089/9781451861860.001.A001

Source: Fitch, Bankscope database.
Figure 4.
Figure 4.

Banking System Net Interest Margin in LICs by Region

(Percent of interest earning assets)

Citation: IMF Working Papers 2005, 167; 10.5089/9781451861860.001.A001

Source: Fitch, Bankscope database.
Figure 5.
Figure 5.

Financial Sector Depth and GDP Per Capita

Citation: IMF Working Papers 2005, 167; 10.5089/9781451861860.001.A001

Source: IMF, International Financial Statistics; and authors’ calculations.

What do we know so far? A large literature has drawn on aggregate and bank-level data to uncover the determinants of financial sector development and performance in broad cross-sections of countries. Institutions (broadly defined) have been identified as a key element in financial sector performance. Institutions, in turn, have been traced back to differences in legal origin (La Porta, and others, 1998), geographical conditions at the time of colonization (Acemoglu, Johnson, and Robinson, 2001), and cultural factors (Stulz and Williamson, 2003). Other studies have focused on the role of state banks (La Porta and others, 2002; Micco, Panizza, and Yañez, 2004), foreign banks (Claessens, Demirgüç-Kunt, Huizinga, 2001), and inflation (Boyd, Levine, and Smith, 2001). In addition, regulations restricting bank activities have been found to hinder financial sector performance, while those encouraging private sector monitoring of banks appear to help (Barth, Caprio, and Levine, 2004). Recent work has also uncovered that compliance with international standards of good regulation and supervision is associated with healthier banking systems (Das, Quintyn, and Chenard, 2004; Podpiera, 2004), and better creditor protection and information access increase credit to the private sector (Djankov, McLeish, and Shleifer, 2005).

Although research in this field has progressed enormously, our knowledge of the factors associated with financial sector development and performance in LICs is still sparse. In general, existing studies have either not looked at LICs or grouped them together with developed and middle-income countries. Using a broad sample increases degrees of freedom, but it may also introduce unwanted heterogeneity if the factors that explain financial sector performance differ across country groups.3 From the point of view of designing a financial sector strategy in LICs, it may be more useful to understand why the financial sector in the Philippines—say—works better than its counterpart in Zambia rather than why Denmark is more financially developed than Sri Lanka.

In this paper, we re-examine the relationship between financial sector performance and many of the factors highlighted in the literature focusing exclusively on LICs. In addition, we try to be comprehensive, considering all possible correlates and searching for robust relationships. Performance is defined to include the depth of the banking system, both in terms of bank deposit generation and credit issued to the private sector, and cost-efficiency measures such as overhead costs and interest margins.4 Ideally, we would have liked to study other performance measures as well, such as access (especially by the poor) or how much long-term lending is provided, but there is no cross-country database available on these dimensions. We would have also liked to analyze fragility, but good cross-country indicators of fragility are difficult to find.5

The explanatory variables considered include geographic characteristics (some of which are used as proxies for institutional quality), legal/colonial origin, political and macroeconomic factors, the structure of the banking system, and features of the regulatory and business environment. First, we look at bivariate correlations among these variables and the performance indicators, and then we examine multivariate correlations with specifications including a gradually expanding set of regressors. Introducing more variables in the specification helps reduce omitted variables concerns, but it inevitably results in the introduction of endogenous regressors. When possible, we try to mitigate the problem by measuring Right Hand Side (RHS) variables as long-term averages of years preceding those in which the Left Hans Side (LHS) variables are measured, but this does not allay all concerns. Accordingly, we do not purport to uncover causation, but simply aim at identifying robust correlations that indicate directions for future work.

To summarize the results, first we find considerable regional differences in performance. LICs in South and East Asia and in the Middle East and North Africa (MENA) are relatively more financially developed; Latin America is in the middle; and African countries and the transition economies are the least financially developed regions.6 This pattern holds even after controlling for differences in per capita income.

Turning to the bivariate and multivariate correlation results, corruption and inflation are, not surprisingly, associated with a shallower and more inefficient financial system. In contrast, in LICs we find that legal origin has no significant bearing on financial sector performance. More foreign bank penetration is associated with a shallower financial sector and is not significantly associated with efficiency. This result may reflect the more cautious behavior of foreign banks when extending credit to the private sector in environments with high information asymmetries and contract-enforcement problems. The interpretation is ambiguous however, since foreign banks may be more likely to enter markets that are “underbanked”. For instance, using bank-level data for several Latin American countries, Clarke and others (2005) show that, on average, foreign banks seem to lend less to informationaly opaque small businesses.7 Perhaps surprisingly, banking systems with more state-owned banks appear to be better at deposit mobilization and have lower overhead costs.8 This is consistent with the regional differences described above: South and East Asia and the Middle East and North Africa regions have more efficient and deeper banking systems and have more state-owned banks and a smaller foreign bank presence.

Finally, characteristics of the regulatory and supervisory system—such as disclosure requirements, auditing requirements, and supervisory powers to discipline banks—are not significantly related to financial performance in LICs.

The paper is structured as follows: Section II provides a description of the methodology. Section III contains an overview of the data. Section IV presents the main regression results. Section V concludes.

II. Methodology

Since our focus is on lower-income countries, we restrict our analysis to the countries defined by the World Bank as low-income and lower-middle-income countries (Table A1).9 This group is large and heterogeneous, both geographically and in terms of income per capita. It includes the poorest countries in the world as well as a few relatively sophisticated emerging markets, such as Russia, Brazil, and Thailand. The total number of countries is 89, but the sample used in the regressions is smaller and varies across specifications depending on availability of data. Three countries (China, Jordan, and Eritrea) are excluded from the regressions because they are outliers with respect to the depth variables

Because many of the explanatory variables we want to consider are not available in time series, we confine our investigation to cross-sectional correlations and regressions. In addition, we restrict attention to the banking system, which is where the bulk of financial activity in LICs is concentrated. Within the banking system, we focus on commercial banks, neglecting other types of banking institutions, such as development banks or microfinance institutions, because consistent data for these entities are not available.10

A. Dependent Variables

To measure financial sector performance, we focus on five indicators, which are also used as the dependent variables in our regressions. The deposit-to-GDP ratio measures the ability of banks to attract financial savings and provide a liquid store of value. The ratio of private sector credit to GDP captures the extent to which the private sector relies on banks to finance consumption, working capital, and investment. The third indicator is the loan-to-asset ratio, which measures the proportion of bank funds allocated to private sector loans rather than government securities, liquid reserves, foreign assets, or other assets. This is a measure of how much intermediation is performed by the banking system.

The last two indicators of financial performance are the ratio of overhead costs to total assets (OH) and the ratio of the net interest margin to interest-earning assets (NIM). They are alternative measures of the cost efficiency of the banking system. OH includes all costs incurred by banks except for the interest paid on liabilities, while NIM is the difference between interest earned on assets and interest paid on liabilities. Banking systems with high operating costs must earn high net interest margins to recover such costs, so NIM is used in the literature as a measure of efficiency.11 However, margins can be high also when banks earn monopolistic profits. Conversely, the NIM may be low where banks take little or no risks, or when they are forced to lend to priority sectors at subsidized interest rates; in these cases, a low NIM may not indicate high efficiency. With this caveat, we will follow the literature and consider low net interest margins to be associated with a more efficient financial system.

Data to compute the depth indicators are obtained from International Financial Statistics, while the efficiency measures are computed from bank level data from Fitch’s Bankscope database. Since we are interested in the efficiency of the financial sector as a whole, the ratios are constructed using country aggregates over all commercial banks available in the sample. Of course, we have to assume that the Bankscope sample is representative of the universe of banks. Country coverage is quite good, although less comprehensive than for depth.12

Summary statistics for all the depth and performance variables are in Table 2. Not surprisingly, there is a strong correlations among measures of financial depth. In addition, deeper financial systems tend to be more efficient, though the correlation is far from perfect and is not significantly different from zero for the loan-to-asset ratio.

Table 1.

Summary Statistics and Regional Means of Selected Explanatory Variables

article image
Notes: MENA denotes Middle East and North Africa. ***, **, *: significant, respectively at the 1 percent, 5 percent and 10 percent levels.

Test of the difference between the sample and the regional means.

Table 2.

Financial Sector Performance Indicators: Summary Statistics

article image
Sources: IMF, International Financial Statistics; Bankscope.

B. Explanatory Variables

Economic theory and existing empirical research point to a very broad set of potential determinants of financial sector performance. We try to be as comprehensive as possible in our approach and consider all the relevant variables for which we can find information for LICs. We group potential explanatory variables into six categories, roughly ordered (based on theoretical considerations) from the most to the least exogenous to the banking system: the geographic and legal environment; the political environment; the macro economy; business environment; banking market structure; and regulation and supervision.13

Geography, endowments, and legal origin

The costs of providing financial services is likely to be affected by the geographic and institutional characteristics of the country. Among the former, the density of the rural population should capture geographical barriers to the delivery of financial services. A country’s latitude might also affect financial development through its effects on institutions via the country’s colonial history.

The law and finance literature, recently surveyed by Beck and Levine (2003), has emphasized the linkage between legal tradition, institutional characteristics (especially the protection of private property rights and the ability of the legal system to adapt to changes in the economic environment), and financial development. La Porta and others (1998) find countries with English legal origin to have deeper financial markets (measured by private credit to GDP and indicators of securities market depth) than other countries, while the French legal tradition appears to hinder financial development.

Another variable that might capture institutional features relevant to financial development is settlers’ mortality. According to Acemoglu, Johnson, and Robinson (2001), in countries where geographic conditions discouraged settlement by Europeans, colonizers aimed mainly at extracting natural resources, and created institutions to suit that purpose. Such institutions were less conducive to business and financial development than those of settlement colonies. After independence, post-colonial governments did not alter the institutional landscape much, and institutions in “extractive” colonies continued to hinder financial development. Beck, Demirgüç-Kunt, and Levine (2003) find empirical evidence that settlers’ mortality is negatively correlated with financial development, while Acemoglu and Johnson (2004) find that the effect goes through “property rights institutions” protecting citizens against the risk of expropriation.

Finally, ethnic fractionalization may also be a proxy for the exogenous determinants of institutions, under the theory that in more ethnically diverse countries consensus to support the provision of public goods, such as institutions, is difficult to achieve (Easterly and Levine, 1997).

Political environment

Even if deep determinants of institutions and geographical factors are favorable to financial development, political instability may be a deterrent. Political turmoil may bring macroeconomic instability and a deterioration in business conditions. Civil strife and outright war can destroy capital and infrastructure. Expropriation may follow revolutions or coups d’etat. In addition, corruption may increase the cost of doing business and create uncertainty about property rights. We use measures of political stability, internal conflict, military control of the government, and freedom from corruption to proxy for the political environment.

Macroeconomic variables

Theoretical models (Huybens and Smith, 1998, 1999) suggest that inflation may aggravate asymmetries of information in credit markets, reducing the real rate of return and the volume of credit. Consistent with these theories, Boyd, Levine, and Smith (2001) find inflation to be negatively associated with measures of financial depth. This negative effect, however, peters out at relatively moderate rates of inflation (15 percent).

The fiscal situation may also affect financial sector performance. If there is a large fiscal deficit that cannot be financed through borrowing, the government might resort to both the inflation tax and financial repression (for instance, forcing banks to hold large unremunerated reserves). We use the fiscal balance as a regressor to capture this channel. The fiscal position may also influence financial development in other ways. The opportunity to invest in government securities, if these securities pay competitive interest rates, may give banks an attractive instrument to manage their liquidity as well as a relatively safe investment opportunity. Since data on stocks of domestic government debt are not available for most countries in our sample, we proxy government debt with interest payments on government debt from the Fund’s World Economic Outlook (WEO) database.

In a number of LICs migrant remittances constitute a large financial flow. To the extent that they are intermediated through the formal financial system, they may spur financial development.

Bank ownership and market structure

Three variables capture aspects of bank ownership and market structure in our regressions: the market share of state banks, the market share of foreign banks, and market concentration (measured by the market share of the largest five banks).

There has been a long-standing debate on whether government ownership of banks plays a useful developmental role or is just an instrument for corruption and political patronage, leading to inefficiencies, misallocation of resources, and instability. In recent years, the latter view has gained increasing support, resulting in a trend toward privatization around the world. In LICs, however, bank privatization has been uneven, and state-owned banks remain dominant in several countries.14 In a recent study, La Porta and others. (2002) find that government ownership of banks is negatively associated with financial development. Specifically, these authors regress growth in private credit between 1960 and 1995 on the share of bank assets held by state banks (measured in 1970), and obtain a negative and significant coefficient, suggesting that state ownership of banks is detrimental to financial development.

Another lively debate surrounds the desirability of entry by foreign banks in developing country markets.15 Proponents of foreign banks claim that subsidiaries of large international banks can achieve better economies of scale and risk diversifications in small markets, introduce more advanced technology (especially risk management) and better supervision and regulation (since subsidiaries are regulated by the home country), and increase competition in cartelized markets. Critics point out that foreign banks lack the local market knowledge to lend to small and medium size borrowers and just serve the safest customers, such as multinational corporations or large domestic firms.16 Domestic banks, unable to compete in the high quality market, may be forced to close down, leaving lower quality customers without credit. Another criticism is that foreign banks may withdraw from the market too quickly in periods of crisis.17

Concerning the relationship between bank ownership and efficiency, bank level data suggest that, in developing countries, foreign-owned banks have lower operating costs and higher profitability than private domestic banks, while state-owned banks have higher costs and lower profitability than the other two categories (Micco, Panizza, and Yañez, 2004). Foreign bank entry also seems to increases competition in developing countries, lowering interest margins and profitability (Claessens, Demirgüç-Kunt, and Huizinga, 2001; Micco, Panizza and Yañez, 2004)). A recent study of Latin America, however, finds the opposite to be true (Levy-Yeyati and Micco, 2003).

A third market structure characteristic is concentration. In canonical economic models, more market concentration should lead to oligopolistic behavior, resulting in higher prices and smaller output than perfect competition. From this perspective, more concentrated banking markets should be shallower. On the other hand, banking theories highlighting the role of banks as producers of information suggest that the opposite may be true: the expectation of enjoying ex post rents (thanks to limited competition) may encourage banks to produce more information and lend more ex ante, especially to more opaque clients, such as new firms, small firms, or firms with fewer tangible assets (Petersen and Rajan, 1995; Marquez, 2000).

Investment climate indicators

It is widely acknowledged that “market infrastructure” is important to financial sector performance. We derive several measures of the quality of infrastructure from the World Bank Business Environment Survey (WBES). This database provides a comprehensive new set of measures of administrative and regulatory obstacles to business activity for a large group of countries. One advantage of these indicators is that they directly measure quantifiable aspects of the business environment, rather than reflecting broad judgments by market participants. Some of the indicators are directly related to banking, as they measure the cost of establishing collateral and recovering defaulted loans, and the availability of information on potential borrowers (through credit registries and other sources).

Supervisory and regulatory framework

The supervisory and regulatory framework is another key component of the financial sector infrastructure. This area has received increasing attention following the emerging market financial crises of the 1990s, which were attributed in part to gaps in financial supervision. Reflecting these concerns, beginning in 1999 the IMF and the World Bank have been devoting substantial resources to the evaluation of regulatory and supervisory frameworks and the dissemination of international best practices in the field through the Financial Sector Assessment Program (FSAP).

To study the relationship between regulation and supervision and financial depth, we rely on the 1999 version of the World Bank survey of bank regulators and supervisors, described by Barth and others (2001). Following the structure of the survey, we group system characteristics in six categories: restrictions on the scope of bank activities; disclosure requirements; the powers of supervisors to discipline banks; accounting standards; and auditing requirements. For each category, we identify survey questions that can be characterized as more stringent regulation, and code the answer as a zero if the regulation is absent and as a one if it is present. We then sum the values in each category and divide by the number of questions covered, so that we obtain an index that varies between zero and one.18 A seventh dimension of the regulatory framework is the presence of an explicit deposit insurance scheme, a zero-one dummy.

C. Empirical Model

To get a sense for the patterns in the data, we start by examining bivariate correlations among the five financial sector performance indicators and the explanatory variables. Then we estimate various multivariate regressions as follows: after controlling for the overall level of development through GDP per capita, we introduce in the regression the first category of variables (geography and legal). After examining various indicators, we move to the next group (the political variables) while keeping as controls in the regression variables from the first group that appear significant and robust (if any). We continue in this fashion until we examine all variables of interest. As degrees of freedom erode, we also replicate the regressions using more parsimonious specifications to gauge robustness. Finally, to simplify the presentation we use the same set of specifications for the first three indicators (deposits-to-GDP, private credit-to-GDP, and the loan-to-assets ratio). Slightly different specifications are used to study the determinants of OH and NIM. Notably, for efficiency measures we control for the size of the financial sector in order to control for scale effects.

The regressions are estimated using OLS with robust standard errors. All dependent variables are measured as averages over 1999-2001 to reduce the effects of short-term economic fluctuations. Whenever possible, we measure right hand side variables as averages over 1991-98 to reduce joint endogeneity problems. However, for some of the variables (for example, the business environment indicators), we only have observations contemporaneous to the indicators of performance.

III. Overview of Data

A. Regional Patterns

The largest share of the countries in our sample is in Sub-Saharan Africa (42 percent), followed by the transition countries of Europe and Central Asia (18 percent), East and South Asia (16 percent), Latin America and the Caribbean (15 percent), and the MENA region (9 percent). While there is considerable intra-regional variation in financial performance (Figures 1-4), on average banking systems in Asia and in MENA are deeper and more efficient. At the other extreme, financial development remains limited in most transition countries and in Africa, while Latin America is somewhere in the middle. In Latin America, however, a larger share of bank assets goes to finance the private sector compared with other regions.19

Several regional differences in the explanatory variables are worth noting (Table 1). For instance, state-owned banks hold the largest share of bank assets in the MENA region and in Asia, while Sub-Saharan Africa has the smallest presence of state banks and the largest presence of foreign banks. The quality of information about borrowers is the best in Latin America, while political risk and the risk of internal conflicts are highest in Sub-Saharan Africa and in transition countries.

The various performance indicators are all significantly correlated among one another, suggesting that banking systems that are more efficient are also better at deposit mobilization and intermediation to the private sector (Table 2). The strongest relationship between deposits and private credit (around 80 percent). Correlations between depth and efficiency measures, while significant, are not very strong, suggesting that it is important to distinguish between these two dimensions.

B. Bivariate Correlations

Several country characteristics are strongly correlated with depth and efficiency of the financial sector (Tables 3 and 4). Countries with a higher income per capita have a deeper financial sector and a lower interest margin, while there is no significant correlation with overhead costs. Political stability and inflation are also strongly correlated with depth and efficiency, while legal origin is not. Less concentrated banking systems are deeper and allocate a larger share of assets to private sector credit.

Table 3.

Financial Depth: Bivariate Correlations

article image
Note: p- values appear in brackets.
Table 4.

Financial Sector Efficiency: Bivariate Correlations

article image
Note: p- values appear in brackets.

Surprisingly, countries with a larger share of bank assets held by state-owned banks seem to have a more efficient banking system and more bank deposits, but a lower loan-to-asset ratio. Conversely, countries with a greater foreign bank presence have shallower banking systems, both in terms of deposits and credit to the private sector. Among business environment indicators, credit information sharing is significantly positively correlated with depth and efficiency (as measured by net interest margins), and the speed of contract enforcement is positively correlated with credit to the private sector. Finally, regulation and supervision variables are not significantly correlated with the cross-section of financial depth.

IV. Results from Multivariate Regressions

A. Financial Depth

Table 5 contains the first set of regressions, in which we study how geographic and institutional features affect financial depth.20 In each regression, we include GDP per capita and a dummy for transition countries as basic control variables. GDP per capita controls for any country characteristic associated with the level of development.21 The transition dummy controls for the special circumstances of countries emerging from central planning, with little or no experience of market-based financial intermediation.22 As expected, we find that countries with a more sparse rural population have a shallower banking sector; the effect is particularly pronounced for bank deposits. Another geographical variable, latitude, is not significant (results not reported).

Table 5.

Financial Depth: Geography, Institution, and Political Variables

article image
Notes: Robust t statistics in brackets. Regressions include a constant, which is not reported.* significant at 10%; ** significant at 5%; *** significant at 1%.

Turning to institutions, in our sample of LICs there are no countries with German or Scandinavian legal origin. In addition, Soviet legal origin is captured by the transition dummy, so to test the legal origin theory we introduce only a dummy variable for French legal origin, while English legal origin is the residual category. This dummy has a negative coefficient in the deposits and private credit regressions, but the coefficient is not significant. In the loan/asset ratio regression, the coefficient of French legal origin is actually positive and (marginally) significant. This suggests that the theoretical predictions and the empirical findings of La Porta and others (1998) do not apply to LICs. As we shall see, legal origin variables continue to explain little as the specification is altered.23

In contrast, our regressions support the theory that settlers’ mortality captures country characteristics strongly associated with financial development. When this variable is introduced in the regression, its coefficient is negative and strongly significant for both deposits and private credit. Moreover, rural density becomes insignificant, suggesting that settlers’ mortality better captures fundamental country characteristics relevant to the financial development process. Unfortunately, this variable is available for only 52 countries. To avoid losing a substantial fraction of our sample, we exclude settlers’ mortality from the benchmark specification. As a robustness test, we have replicated all the regressions for a smaller sample including settlers’ mortality as a control, and find that the results reported in the rest of this section remain broadly unchanged.24 Another proxy for the exogenous determinants of institutions, ethnic fractionalization, does not have any explanatory power.

Political instability and internal conflict are associated with a shallower financial system. An even clearer association is with an index of corruption: not surprisingly, more corrupt countries have lower bank deposits and less credit to the private sector. In addition, when corruption is controlled for, measures of political instability tend to lose significance (results not reported), so we keep the corruption index in the benchmark specification. The variables introduced so far explain about 40 percent of the variation in deposits and private credit, but only 10 percent of that in the loan-to-asset ratio.

Among the macroeconomic variables, inflation is negatively and significantly correlated with credit to the private sector and the loan-to-asset ratio, though not with deposits (Table 2). The performance of the regressions improves quite a bit when inflation is introduced; particularly, the R-squared of the loan-to-asset ratio regression now reaches 30 percent. If we allow for threshold effects at inflation above 15 percent, we find that disruptions to the credit market become more severe for higher levels of inflation, in contrast with the findings of Boyd, Levine, and Smith (2001).25

Another interpretation of the negative correlation between inflation and financial development is that it is driven by an omitted variable, the fiscal balance, as countries in a difficult fiscal position might resort to both inflation financing and financial repression. When we control for the fiscal balance, however, the relationship between inflation and bank credit remains unaltered (not reported). This suggests that an asymmetric information interpretation of such relationship may be more appropriate (Huybens and Smith, 1998, 1999). A proxy for domestic government borrowing, interest payments on government debt scaled by GDP, has a positive sign, but is not significant in this specification. Likewise, migrant remittances enter with the expected positive sign, but are not statistically significant.

Turning to bank ownership, consistent with the bivariate correlations, the share of bank assets controlled by the government is positively and significantly correlated with bank deposits and negatively but insignificantly correlated with private credit and the loan-to-asset ratio. This suggests that financial systems with more state banks are more successful at deposit mobilization, though they find it difficult to convert these funds into loans to the private sector. Since state banks are likely to be more prevalent where there are more obstacles to private sector development, causality is unlikely to go from deposit mobilization to the share of public banks. On the other hand, reverse causality may be responsible for the negative coefficient in the private credit and loan-to-asset ratio regressions. The coefficient of deposit mobilization, however, is no longer significant if a regional dummy (for Africa or Asia) is introduced, so this result may reflect unexplained regional differences.

The multivariate regressions also confirm the negative association between financial development and foreign bank penetration found in the bivariate correlations. Countries with a larger share of foreign banks have less deposits and less private sector credit. The coefficient of foreign banks is also negative in the loan-to-asset ratio regression, but is only significant (and marginally so) in some specifications.26 However, the interpretation of this coefficient is ambiguous, as foreign banks may be more likely to enter markets that are “underbanked.” Further research is necessary to identify the direction of causality in the relationship between foreign bank presence and financial depth.

Concentration—measured as the share of bank assets held by the largest five banks—is negatively correlated with private credit and the loan-to-asset ratio, suggesting that the predictions of standard economic theory are more relevant than information-based theories of banking. But the result does not seem to be robust to alternative measures of concentration.27 Also, the share of bank assets controlled by the largest five banks may not be a good measure of concentration when comparing countries with large differences in the total number of banks, or where banks have local monopoly power.

All in all, market characteristics seem to be very important at explaining variation in financial depth measures. The R-squared of the regressions improve markedly when these variables are introduced.

Among the many business environment indicators in the WBES database, measures of the time required to enforce contracts, the availability of credit information, and coverage of credit registries have explanatory power in the financial depth regressions, particularly for private credit and the loan-to-asset ratio. This is true even after controlling for geographic, institutional, macroeconomic, and market structure characteristics. The signs indicate that better access to information and speedier enforcement of contracts are associated with deeper credit markets, as expected. Other business environment characteristics, including measures of the costs of establishing collateral, and starting or closing a business, do not seem to be significantly correlated with financial sector depth in our sample. As in the case of foreign banks, the direction of causality is ambiguous, as countries with more developed credit market may also have more incentives to modernize the judiciary, reduced legal delays, and introduce an efficient system of information sharing among banks.

Finally, once other factors are controlled for, differences in the regulatory and supervisory framework do not add much explanatory power to the regressions. This remains true in more parsimonious specifications, in which regressors that are not significant are excluded. In the two cases in which the relationship is significant (stronger supervisory powers to discipline banks in the loan-to-assets ratio regression and auditing requirements in the deposit regression), more stringent requirements seem to reduce depth rather than increase it. This result is particularly interesting because the possible endogeneity of supervision and regulation should bias the coefficient in the positive direction, i.e., in more financially developed countries there should be more incentives to set up a strong supervisory and regulatory framework (Barth and others, 2003).

There are two, not mutually exclusive, interpretations of this finding. First, the crude indicators employed here may not adequately reflect the regulatory and supervisory framework, especially since they capture whether regulations are on the books rather than how they are implemented in practice. A second interpretation is that in LICs the obstacles to financial development are so pervasive that differences in regulation and supervision have only second order effects.

Another result is that the presence of an explicit deposit insurance scheme does not lead to more deposit mobilization in LICs; in fact, the coefficient of this variable is negative and marginally significant in the regression of bank deposits. This may be because deposit insurance is not fully credible in countries where the fiscal position is often precarious and political instability undermines government credibility. Deposit insurance may also increase the instability of the banking system in countries with a weak institutional environment (Demirgüç-Kunt and Detragiache, 2002). Instability, in turn, may deter depositors even if they are partially insured.28

B. Financial Sector Efficiency

To study the covariates of the two financial sector efficiency indicators, NIM and OH, we proceed along similar lines as in the previous section. First, we introduce in the regression the size of the economy (measured by the logarithm of GDP) to capture scale economies.29 Indeed, banks seem to be more efficient in large economies. The next step is to introduce geographic and institutional characteristics. These variables do not seem to have much explanatory power for bank efficiency. Even settlers’ mortality, which was strongly negatively correlated with depth, is only marginally significant here. In contrast, the political environment has important effects on bank efficiency. Corruption is particularly detrimental, but so are political instability and political risk. More political instability may mean that it is harder to enforce property rights, that the bureaucracy is more inefficient, or that the regulatory framework is more uncertain, which may all translate into higher costs for banks.

Turning to the macroeconomic variables, not surprisingly inflation is associated with higher NIM and OH, consistent with theoretical papers suggesting that informational frictions increase with uncertainty (Huybens and others, 1998 and 1999). This continues to be the case after controlling for fiscal variables. So keeping inflation in check seems to have beneficial effects for financial sector performance both in terms of depth and efficiency.30 Contrary to what we found for depth, however, the marginal effect of inflation seems to be more marked at lower rates of inflation.

Concerning bank ownership, foreign bank penetration has no significant relationship with efficiency, while countries with a larger presence of state banks have significantly lower OHs. This relationship, which also emerged from the bivariate correlations, is robust to altering the set of control variables and the sample. The correlation is also robust to controlling for the loan-to-asset ratio, so differences in the share of government assets in bank portfolios cannot explain it. In addition, if state banks are more prevalent in countries with more difficult conditions for private sector development, then reverse causality should bias the coefficient upwards not downwards. In contrast, La Porta and others (2002) find OH (measured in 1999) to be positively correlated with the share of state banks (measured in 1970), while Barth and others (2004) report a positive but insignificant relationship between OH and NIM and state bank penetration after controlling for legal origin and prudential supervision and regulation. These studies, however, use samples including both developing and developed countries, so the difference may arise because we are considering only LICs.

In a panel of bank level data, Micco, Panizza, and Yañez (2004) find that, after controlling for country and time fixed effects and for bank characteristics, state banks have higher overhead costs than private banks or foreign banks in developing countries and not in developed countries. How can these findings be reconciled with ours? Part of the explanation is that state banks are larger than private domestic banks, and larger banks are more efficient. If bank size is omitted in the regressions of Micco, Panizza, and Yañez, the coefficient of state banks is no longer significant. Another possibility may be that LICs with a smaller presence of state banks are countries that, having an especially inefficient state sector, were forced to privatize rapidly during the 1990s. If conditions in these countries make it particularly difficult for banks (including private banks) to operate efficiently, and these conditions are omitted in our regressions, then the puzzle might be explained. Obviously, additional research is needed to interpret this finding.

The third variable capturing market structure is concentration. In our sample, concentration is positively correlated with efficiency, even after controlling for the overall size of the market, the business environment, and supervision and regulation.31 In contrast, for a sample including also developed countries Demirgüç-Kunt and others (2003) find the relationship to be negative, though not statistically significant when features of the institutional and regulatory environment are included in the regression.32 Less concentrated markets may be markets in which less efficient, marginal banks have not gone out of business even though technological changes would make consolidation beneficial. These banks may continue to operate because they have market power in local markets or in particular market segments, or because there are regulatory barriers to consolidation. This may results in more deposits and private credit but at the cost of higher overheads and interest margins.

Finally, there does not seem to be a robust link between efficiency and the business environment nor between efficiency and regulation and supervision. Regulatory indexes always enter with a positive sign, suggesting that tighter regulation tends to increase costs, but the coefficient is significant only for disclosure and accounting rules. Neither, however, is robust to changes in the specification. It should be pointed out that when these variables are introduced the sample size drops to about 40 countries, so there is not much information to draw inference from.

C. Magnitude of Effects

In presenting the results, we have emphasized the statistical significance of the coefficients of the various explanatory variables. An equally important question is the economic significance of the relationships studied. For illustrative purpose we have computed predicted changes in the dependent variables resulting from a one standard deviation increase in the value of each explanatory variable using the last specification in Table 5 (for depth) and Table 11 (for efficiency).

Table 6.

Financial Depth: Macroeconomic Variables

article image
Notes: Robust t statistics in brackets. Regressions include a constant, which is not reported.* significant at 10%; ** significant at 5%; *** significant at 1%.
Table 7.

Financial Depth: Bank Ownership and Concentration

article image
Notes: Robust t statistics in brackets. Regressions include a constant, which is not reported.* significant at 10%; ** significant at 5%; *** significant at 1%.
Table 8.

Financial Depth and Business Environment

article image
Notes: Robust t statistics in brackets. Regressions include a constant, which is not reported.* significant at 10%; ** significant at 5%; *** significant at 1%.
Table 9.

Supervision and Regulation

article image
article image
Notes: Robust t statistics in brackets.* significant at 10%; ** significant at 5%; *** significant at 1%.
Table 10.

Efficiency: Geography, Legal Origin, and Political Variables

article image
Notes: Robust t statistics in brackets. Regressions include a constant, which is not reported.* significant at 10%; ** significant at 5%; *** significant at 1%.
Table 11.

Bank Efficiency: Macroeconomic Variables and Market Structure

article image
Notes: Robust t statistics in brackets. Regressions include a constant, which is not reported.* significant at 10%; ** significant at 5%; *** significant at 1%.