77292 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 2: 379-408 Determinants of Commercial Bank Interest Margins and Profitability: Some International Evidence Ash Demirgiic,-Kunt and Harry Huizinga Using bank-level data for 80 countries in the years 1988-9S, this article shows that differences in interest margins and bank profitability reflect a variety of determinants: bank characteristics, macroeconomic conditions, explicit and implicit bank taxation, deposit insurance regulation, overall financial structure, and underlying legal and insti- tutional indicators. A larger ratio of bank assets to gross domestic product and a lower market concentration ratio lead to lower margins and profits, controlling for differences in bank activity, leverage, and the macroeconomic environment. Foreign banks have higher margins and profits than domestic banks in developing countries, while the op- posite holds in industrial countries. Also, there is evidence that the corporate tax burden is fully passed onto bank customers, while higher reserve requirements are not, espe- cially in developing countries. As financial intermediaries, banks play a crucial role in the operation of most economies. Recent research, as surveyed by Levine (1997), shows that the effi- cacy of financial intermediation can affect economic growth. Crucially, financial intermediation affects the net return to savings and the gross return to invest- ment. The spread between these two returns mirrors bank interest margins, in addition to transaction costs and taxes borne directly by savers and investors. Thus bank interest spreads could be interpreted as an indicator of the efficiency of the banking system. In this article we investigate how bank interest spreads are affected by taxation, the structure of the financial system, and financial regula- tions, such as deposit insurance. A comprehensive review of the determinants of interest spreads is offered by Hanson and Rocha (1986), who summarize the role that implicit and explicit taxes play in raising spreads and discuss some of the determinants of bank costs and profits, such as inflation, scale economies, and market structure. Using ag- gregate interest data for 29 countries in the years 1975-83, the authors find a positive correlation between interest margins and inflation. Recently, several studies have examined the impact of international differ- ences in bank regulation using cross-country data. Analyzing interest rates in 13 Organisation for Economic Co-operation and Development (OECD) countries in Ash Demirgiic.-Kunt is with the Development Research Group at the World Bank, and Harry Huizinga is with the Department of Economics, Tilburg University, and the Centre for Economic Policy Research in London. The authors gratefully acknowledge comments by three anonymous referees. They also thank Anqing Shi for excellent research assistance and Paulina Sintim-Aboagye for help with the manuscript. © 1999 The International Bank for Reconstruction and Development/THE WORLD BANK 379 380 THE WORLD BANK ECONOMIC REVIEW, VOL 13, NO. 2 the years 1985-90, Bartholdy, Boyle, and Stover (1997) find that the existence of explicit deposit insurance lowers the deposit interest rate by 25 basis points. Barth, Nolle, and Rice (1997) use 1993 data from 19 industrial countries to further examine the impact of banking powers on bank return on equity, con- trolling for several bank and market characteristics. They find that variations in banking powers, bank concentration, and the existence of explicit deposit insur- ance do not significantly affect the return on bank equity. In this article we extend the existing literature in several ways. First, we use bank-level data for 80 industrial and developing countries in 1988-95 to provide summary statistics on the size and decomposition of bank interest margins and profitability. Second, we use regression analysis to examine the underlying deter- minants of interest spreads and bank profitability. This empirical work enables us to infer the extent of taxation and regulation on bank customers and on banks themselves. Apart from covering many banks in many countries, this study is unique in its coverage of the determinants of interest margins and profitability. These determi- nants include a comprehensive set of bank characteristics (such as size, leverage, type of business, foreign or domestic ownership), macroeconomic indicators, taxa- tion and regulatory variables, financial structure variables, and legal and institu- tional indexes. Among these, the ownership variable, the tax variables, some of the financial structure variables, and the legal and institutional indicators have not been included in any previous study in this area. To check whether some of these determinants affect banking differently in developing and industrial countries, we interact these variables with the country's gross domestic product (GDP) per capita. I. BANK INTEREST SPREADS AND PROFITABILITY We can measure the efficiency of bank intermediation using both ex ante and ex post spreads. The ex ante spread is the difference between the contractual rates charged on loans and rates paid on deposits. The ex post spread is the difference between banks' actual interest revenues and their actual interest ex- penses. The ex post spread differs from the ex ante spread by the amount of loan defaults. The ex post spread is a more useful measure because it controls for the fact that banks with high-yield, risky credits are likely to face more defaults. An additional problem with using the ex ante spread is that data are generally avail- able at the aggregate industry level and are put together from a variety of sources. Thus they are not completely consistent. For these reasons we focus on ex post interest spreads in this article. There is, however, a problem with ex post spreads, in that the interest income and loan loss reserving associated with a particular loan tend to materialize in different time periods. As a measure of what we call bank "efficiency," we consider the accounting value of a bank's net interest income divided by total assets (TA), or the net interest margin {NIM). Bank "profitability" is a bank's before-tax profits {BTP) divided by total assets. Profitability could also be measured by the return on Demirgiif-Kunt and Huizinga 381 equity as opposed to the return on assets. It is well known that, ceteris paribus, a bank with a higher equity ratio will have a higher return on assets and a lower return on equity than a bank with a lower equity ratio. The problem in some developing countries is that banks operate with extremely low equity capital, often supported by implicit state guarantees, which inflates their return on eq- uity. Using unadjusted returns on equity may then be more distortionary than using returns on assets. Ideally, we should use risk-adjusted returns on equity, but since these are not available, we analyze returns on assets after controlling for the banks' equity ratio. We do this by entering the equity ratio as an indepen- dent variable in the profit regression. Thus, by straightforward accounting, BTA = ATP^ TX_ ( ' TA ~ TA+TA where ATP is after-tax profits. From the bank's income statement, before-tax profits divided by total assets further satisfies the following accounting identity: BTA n\ KJTAA^NI1 OV LLP TA TA TA TA where Nil is noninterest income, OV is overhead, and LLP is loan loss provi- sioning. NII/TA reflects the fact that many banks also engage in nonlending ac- tivities, such as investment banking and brokerage services, OV/TA accounts for the bank's entire overhead associated with all of its activities, and LLPFT A mea- sures actual provisioning for bad debts. Although the net interest margin can be interpreted as a rough index of bank efficiency, this does not mean that its reduction always signals improved effi- ciency. A reduction in the net interest margin can, for example, reflect a reduc- tion in bank taxation or, alternatively, a higher loan default rate. In the first instance the reduction in the net interest margin may reflect the improved func- tioning of the banking system, while in the second case the opposite may be true. Also, variation in an accounting ratio, such as the net interest margin, may reflect differences in net interest income (the numerator) or differences in, say, nonlending assets (a component of the denominator). In the data set the accounting data are organized so as to be comparable inter- nationally. However, differences in accounting conventions regarding the valua- tion of assets, loan loss provisioning, hidden reserves, and so on may remain. Vittas (1991) reviews the pitfalls in interpreting bank operating ratios. Account- ing data also tend to reflect economic realities with a long lag so that they are not able to flag pending banking crises, such as those that have recently occurred in Southeast Asia. This article focuses on accounting measures of income and profitability as investors equalize (risk-adjusted) financial returns on bank stocks in the absence of prohibitive barriers. Gorton and Rosen (1995) and Schranz (1993) also focus 382 THE WORLD BANK ECONOMIC REVIEW, VOL 13, NO. 2 on accounting measures of profitability when examining managerial entrench- ment and bank takeovers. The above accounting identity (equation 2) suggests a useful decomposition of the realized interest spread—the net interest margin—into its constituent parts: noninterest income, overhead, taxes, loan loss provisions, and after-tax bank profits. Hanson and Rocha (1986) take this approach, with some modifications. As a first step to analyzing the data, in section HI we provide an accounting breakdown of the net interest margin for individual countries and for selected aggregates. Although it may be misleading to compare accounting ratios without controlling for differences in the macroeconomic environment in which banks operate and the differences in their business, product mix, and leverage, these breakdowns are still a useful initial indicator of differences across countries. Next, controlling for bank characteristics and the macroeconomic environ- ment, we provide an economic analysis of the determinants of the interest and profitability variables—the net interest margin and before-tax profits divided by total assets. This empirical work offers insights into how bank customers and banks themselves are affected by these variables. The net interest margin regres- sions tell us how the spread determinants affect the combined welfare of deposi- tors and lenders. The relationship between the interest spread and a bank's cor- porate taxes, for instance, reveals the extent to which a bank is able to shift its tax bill forward to its depositors and lenders. Generally, taxes and other variables can affect interest rates as well as the volume of loans and deposits. In the short term the major effects may come through pricing changes, in which case the net interest margin and before-tax profits as a share of total assets immediately reveal easily interpreted welfare consequences for banks and their customers. With market imperfections in the form of credit rationing or imperfect competition in credit markets, changes in quantities generally have first-order welfare implications independent of changes in prices. We do not, however, evaluate changes in quantities in this article. Lastly, the before-tax profit regressions show how spread determinants affect bank shareholders. The regression analysis starts from the following equation: (3) lijt = ct0 + a, Biit + pVX,, + y,T, + 8,0, + e* where lijt is the dependent variable (either the NIM or BTP/TA) for bank i in country / at time t, Bijt are characteristics of bank i in country / at time f, X,-, are characteristics of country / at time f, T, and C, are time and country dummy variables, and £y, is a white-noise error term. We estimate several specifications of equation 2 including different bank and country variables. n. DATA In this study we use income statement and balance sheet data of commercial banks from the BankScope database provided by IBCA. IBCA'S coverage is compre- Demirgtif-Kunt and Huizinga 383 hensive in most countries, accounting for 90 percent of all bank assets. We begin with all commercial banks worldwide, with the exception of France, Germany, and the United States, for which we include only several hundred commercial banks listed as "large." To ensure reasonable coverage for individual countries, we in- clude only countries with at least three banks for a given year. We end up with a data set that includes 80 countries during the years 1988-95, with about 7,900 individual commercial bank observations. This data set includes all OECD coun- tries, as well as many developing countries and transition economies (table 1). Several countries, such as Luxembourg, the Netherlands, and Egypt, have a net interest margin close to 1 percent (column 2 of table 1). This is the low end of the distribution. Egypt's low net interest margin can be explained by a predomi- nance of low-interest directed credits by the large state banking sector. Gener- ally, developing countries, and especially Latin American countries, such as Ar- gentina, Brazil, Costa Rica, Ecuador, and Jamaica, have relatively large spreads. This is also true for certain Eastern European countries, such as Lithuania and Romania. Columns 3-6 in table 1 break down the net interest income into its four com- ponents: overhead minus noninterest income, taxes, loan loss provisioning, and net profits. Taxes as a share of net interest income (column 4) reflect the explicit taxes that banks pay (mostly corporate income taxes). Banks also face implicit taxation because of reserve and liquidity requirements and other restrictions on lending that come through directed or subsidized credit policies. These indirect forms of taxation directly lower the net interest income rather than the tax vari- able. Nonetheless, the tax variable indicates that there is considerable interna- tional variation in the explicit taxation of commercial banks. Several countries in Eastern Europe impose high explicit taxes on banking. (For example, taxes as a percentage of net interest income are only 17.5 in Lithuania and 13.7 in Hungary compared with 26.2 in Romania, 83.3 in Russia, and 23.2 in the Czech Repub- lic.) The lowest share of taxes in net interest income is 0 for Qatar, where there is no significant taxation of banking. For some countries, such as Norway, Swe- den, and Costa Rica, low tax shares reflect the tax deductibility of bad debts, which are plentiful. Loan loss provisioning as a share of net interest income is a direct measure of differences in credit quality across countries (column 5). It also reflects differ- ences in provisioning regulations. This variable is high for some Eastern Euro- pean countries. It is also high for some industrial countries, such as France and the Nordic countries. The fourth component of net interest income is net profits (column 6). As a residual, net profits as a share of net interest income reflect the extent to which the net interest margin translates into net-of-tax profitability. The remaining columns in table 1 tabulate the various accounting ratios (rela- tive to total assets) in the accounting identity (equation 2). Noninterest income as a share of total assets reveals the importance of fee-based services for banks in different countries (column 7). Banks in Eastern Europe—for example, those in Estonia, Hungary, and Russia—seem to rely heavily on fee-based operations. Table 1. Bank Interest Spreads and Profitability: Economy Averages, 1988-95 (percent) Net interest income* J Overhead minus Loan loss Net Noninterest Loan loss Net Net interest noninterest income Taxes provisions . profits income Overhead Taxes provisions profits Economy margin* (as a percentage of net interest income) (as a percentage of total assets) Argentina 7.3 35.6 5.5 28.5 30.3 6.3 9.4 0.4 1.8 2.0 Australia 3.0 32.8 8.5 28.1 33.2 1.3 2.8 0.3 0.7 0.6 Austria 1.7 54.9 6.8 25.7 24.8 0.5 1.5 0.1 0.5 0.3 Bahrain 2.2 30.8 2.7 32.9 43.3 0.8 1.4 0.0 0.6 1.1 Belgium 2.0 48.7 10.2 20.4 24.0 0.9 2.1 0.2 0.4 0.4 Bolivia 3.1 73.7 1.9 12.6 12.6 2.0 5.2 0.6 0.6 -1.1 Botswana 6.0 43.7 11.8 12.9 31.5 2.8 5.4 0.7 0.7 1.9 Brazil 8.9 60.5 11.6 13.8 17.5 4.5 10.2 1.1 1.3 1.4 Canada 2.9 47.3 12.7 21.5 19.0 1.2 2.5 0.4 0.6 0.6 Chile 4.3 71.7 2.4 13.8 14.2 -0.1 3.0 0.1 0.6 0.5 China 2.1 30.0 15.9 — 54.2 1.0 1.6 0.3 — 1.2 2 Colombia 6.0 53.8 10.3 12.9 27.7 5.8 8.3 0.7 0.9 2.2 Costa Rica 13.6 40.7 4.7 57.6 9.9 3.5 8.1 0.8 5.7 3.5 Cyprus 1.0 59.1 11.1 15.8 22.0 3.1 3.2. 0.3 0.3 0.5 Czech Rep. 3.3 13.5 23.2 53.4 13.6 1.5 2.1 0.6 2.0 0.3 Denmark 4.8 52.9 5.4 33.3 8.6 1.0 3.7 0.3 1.6 0.3 Dominican Rep. 6.6 52.8 8.6 9.1 30.9 3.1 6.3 0.6 0.5 2.3 Ecuador 7.7 52.8 4.7 12.8 34.9 3.8 8.1 0.4 1.0 2.5 Egypt 1.4 -32.7 11.2 62.6 63.5 2.1 1.4 0.3 0.7 1.2 El Salvador 3.2 34.3 1.9 14.0 49.8 1.6 2.9 0.1 0.4 1.5 Estonia 4.7 -35.9 24.1 — 111.7 8.7 7.0 1.1 — 5.3 Finland 1.8 50.1 9.4 55.6 -10.7 1.2 2.1 0.2 0.8 -0.1 France 2.4 48.3 7.0 50.8 -1.7 1.4 2.6 0.2 1.0 0.1 Germany 2.0 51.6 12.3 29.5 12.6 1.1 2.1 0.3 0.6 0.3 Greece 3.0 33.8 12.7 25.7 29.7 2.2 3.4 0.4 0.6 1.0 Guatemala 5.6 80.5 3.6 16.0 1.4 5.7 0.2 1.1 Haiti 2.8 53.8 7.3 12.4 26.5 2.8 4.2 0.2 0.4 0.8 Honduras 4.3 72.3 9.8 — 17.9 0.9 4.0 0.4 — 0.8 Hong Kong, China 2.5 17.1 10.2 6.0 67.8 1.3 1.4 0.3 0.2 2.0 Hungary 4.7 17.5 13.7 68.8 29.9 5.8 7.0 0.6 2.7 1.4 India 4.0 18.2 12.4 19.3 50.2 1.6 2.0 0.6 0.7 2.3 Indonesia 3.6 47.5 10.9 17.8 26.2 1.2 2.9 0.4 0.7 0.9 Israel 2.8 41.9 17.1 23.6 17.3 1.8 3.2 0.4 0.7 0.4 Italy 3.4 56.5 14.3 17.4 11.9 1.4 3.3 0.5 0.5 0.4 Jamaica 10.5 33.9 21.2 2.2 43.1 2.8 6.3 2.2 0.3 4.5 Japan 1.6 61.9 16.2 10.0 12.1 0.2 1.3 0.2 0.1 0.2 Jordan 2.1 48.0 10.4 24.1 24.5 1.4 2.4 0.2 0.5 0.5 Korea, Rep. of 1.8 36.4 12.5 34.0 29.9 1.5 2.2 0.2 0.5 0.5 Lebanon 2.7 45.6 9.2 13.7 35.3 0.9 2.1 0.3 0.5 0.9 Lithuania 10.6 29.8 17.5 81.7 -22.2 5.0 _ _ _ Luxembourg 0.8 -11.5 28.2 52.7 46.0 0.9 1.0 0.2 0.3 0.3 Malaysia 2.7 40.0 15.9 17.3 29.2 0.8 1.9 0.4 0.4 0.8 Malta 2.4 37.6 18.0 6.2 39.1 1.1 2.0 0.4 0.1 0.9 Mexico 4.6 40.8 6.1 42.2 15.4 2.1 4.5 0.3 1.1 0.9 Morocco 3.4 66.8 13.6 0.1 21.9 1.3 3.5 0.5 0.0 0.8 Nepal 3.6 10.5 25.3 16.1 48.1 2.1 2.4 1.0 0.5 1.8 00 VI Netherlands 1.4 43.1 9.7 21.4 26.1 1.0 1.7 0.1 0.3 0.4 Nicaragua 4.4 85.2 8.0 18.5 -10.4 3.3 6.3 0.3 0.9 0.2 Nigeria 5.3 -29.3 13.1 88.3 27.8 5.8 7.0 0.7 1.6 1.8 Norway 3.2 51.6 4.6 44.3 3.2 1.2 2.8 0.1 1.4 0.2 Oman 4.1 43.1 5.5 15.2 36.2 1.4 3.3 0.2 0.6 1.4 Pakistan 2.8 38.8 28.6 32.6 1.8 2.9 0.9 _ 0.8 Panama 2.1 29.9 4.3 20.3 46.4 1.4 2.0 0.1 0.4 1.0 Papua New Guinea 3.2 -2.6 20.2 40.8 45.8 4.2 5.0 0.4 0.9 1.1 Paraguay 5.9 63.5 5.5 11.4 23.3 2.5 6.2 0.4 0.7 1.5 Peru 6.5 43.8 14.3 47.0 12.1 5.7 9.6 0.7 1.7 0.8 Philippines 4.1 29.8 6.6 10.3 55.0 3.0 4.3 0.3 0.4 2.2 Poland 6.1 16.8 27.9 23.3 34.9 2.4 3.6 1.6 1.3 2.1 Portugal 3.3 45.9 8.0 25.5 23.7 1.0 2.5 0.3 0.9 0.7 (continued on following page.) Table 1. (continued) (percent) Net interest income* Overhead minmt Loan loss Net Noninterest Loan loss Net noninterest income Taxes provisions profits income Overhead Taxes provisions profits Net interest Economy margin* (as a percentage of net interest income) (as a percentage of total assets) Qatar 1.9 6.6 0.0 15.0 85.2 1.1 1.3 0.0 0.2 1.6 Romania 9.7 1.9 26.2 36.8 44.3 2.4 2.8 2.3 3.7 4.3 Russia 4.7 -5.0 33.3 47.2 37.1 10.9 7.0 1.9 2.6 4.7 Singapore 2.2 20.7 21.6 8.7 56.4 1.0 1.4 0.5 0.1 1.3 South Africa 3.9 45.1 11.8 16.1 29.0 1.9 3.6 0.5 0.7 1.1 Spain 3.6 60.3 10.2 17.7 12.7 1.2 3.2 0.4 0.6 0.7 Sri Lanka 3.7 31.8 11.1 9.7 52.5 2.0 3.0 0.5 0.4 2.1 W Swaziland 5.4 52.1 16.3 2.8 30.9 2.7 5.5 0.9 0.2 1.7 00 Sweden 3.1 26.3 1.9 64.6 11.2 1.5 2.5 0.1 1.9 0.3 Taiwan (China) 2.0 34.6 10.1 10.8 45.5 1.0 1.6 0.2 0.2 1.0 Tunisia 2.3 31.4 9.9 56.1 48.0 2.2 3.1 0.2 1.1 0.8 Turkey 6.3 11.7 10.0 32.9 47.2 4.0 5.4 0.8 0.8 3.3 United Kingdom 2.3 18.4 20.6 29.8 40.9 2.3 3.0 0.4 0.7 0.8 United States 3.9 47.6 12.5 15.2 25.8 1.8 3.6 0.5 0.7 1.0 Venezuela 7.2 49.9 2.7 16.7 30.6 2.8 6.4 0.2 1.0 2.5 Yemen 4.0 48.8 14.1 2.6 34.6 -0.5 1.4 0.6 0.1 1.4 Zambia -4.7 186.1 -6.6 -49.1 -30.4 9.5 0.4 0.3 2.4 1.7 — Not available. Note: Ratios are calculated for each bank in each country and then averaged over the country's sample period. a. The net interest margin is defined as net interest income divided by total assets. b. Columns 3 though 6 show the shares of the four components of net interest income. These shares add to 100 percent except for cases where information on loan loss provisioning is missing. Source: Authors' calculations based on data from the BankScope database of the IBCA. Demirguf-Kunt and Huizmga 387 This is also the case in some Latin American countries, such as Argentina, Brazil, Colombia, and Peru, and in a few African countries, such as Nigeria and Zambia. Overhead as a share of total assets reveals variations in operating costs across banking systems (column 8). This variable reflects variations in employment and in wage levels. Despite high wages, overhead as a share of total assets appears to be lowest at around 1 percent for high-income countries, such as Japan and Lux- embourg. It is notably high at 3.6 percent for the United States, perhaps reflect- ing the proliferation of banks and bank branches because of banking restrictions. Jamaica, Lithuania, and Romania stand out with high tax-to-asset ratios of around 2 percent (column 9). Loan loss provisioning, proxied by loan loss provi- sioning as a share of total assets, is equally high in Eastern Europe and in some developing countries (column 10). Finally, net profits divided by total assets also tend to be relatively high in developing countries (column 11). Table 2 presents statistics on interest spreads and profitability for selected aggregates. The first breakdown is by ownership; a bank is said to be foreign- owned if 50 percent or more of its shares are owned by foreign residents. There is only a small difference in the net interest margin for domestic banks (3.7 percent) and foreign banks (2.9 percent). This small difference, however, masks the fact that foreign banks tend to achieve higher interest margins in developing coun- tries and lower interest margins in industrial countries.1 This may reflect the fact that foreign banks are less subject to credit allocation rules and have technical advantages (in developing countries), but also have distinct informational disad- vantages relative to domestic banks (everywhere). Foreign banks pay somewhat lower taxes than domestic banks (column 4). This gap may reflect differences in the tax rules governing domestic and foreign banks, as well as the ability of foreign banks to shift profits internationally to minimize their global tax bill. Foreign banks also have relatively low provision- ing, as indicated by loan loss provisioning as a share of total assets, which is consistent with the view that foreign banks generally do not engage in retail banking operations. The next breakdown is by bank size. For countries with at least 20 banks, large banks are defined as the 10 largest banks according to the value of their assets. Large banks tend to have lower margins and profits and smaller overheads. They also pay relatively low direct taxes and have lower loan loss provisioning. Table 2 also considers bank groupings by national income levels and loca- tion.2 Breaking down the data into four income levels, we see that the net interest margin is highest for countries in the middle-income groups. Banks operating in middle-income countries also have the highest values for overhead, taxes, and loan loss provisioning as shares of total assets. Net profits as a share of total 1. See Claessens, Demirguc.-Kunt, and Huizinga (1997) for more detailed information on the average spreads of domestic and foreign banks for different groupings of countries by income. That article also considers how entry by foreign banks affects the interest spreads and operating costs of domestic banks. 2. For country groupings by income, see World Bank (1996). Table 2. Bank Interest Spreads and Profitability, Selected Aggregates, 1988-95 (percent) Net interest income* Overhead minus Loan loss Net Noninterest Loan loss Net Net interest noninterest income Taxes provisions profits income Overhead Taxes provisions profits Bank groupings margin* (as a percentage of net interest income) (as a percentage of total assets) i All banks 3.5 43.1 11.5 24.8 20.6 1.6 3.2 0.3 0.8 0.8 Bank ownership Domestic 3.7 46.2 11.1 22.8 19.9 1.6 3.3 0.4 0.8 0.8 Foreign' 2.9 29.0 13.1 33.5 24.4 1.6 2.8 0.3 0.7 0.8 Bank size6 Large 2.6 35.5 13.1 27.5 23.9 1.2 2.5 0.3 0.6 0.5 Small 3.4 48.0 11.9 22.0 18.2 1.5 3.1 0.4 0.7 0.7 Country income Low 2.8 37.9 11.3 20.0 30.8 3.2 3.1 0.5 0.8 1.5 Lower-middle 5.7 36.8 11.0 24.9 27.2 3.2 5.1 0.7 1.3 1.8 GO 00 Upper-middle 4.1 32.7 11.2 27.3 28.8 2.1 3.8 0.4 1.0 0.9 High 2.6 30.0 10.3 31.8 27.9 1.2 2.3 0.2 0.7 0.5 Region Africa 3.3 59.2 9.6 14.2 16.9 4.5 4.4 0.6 1.1 1.6 Asia 3.0 20.1 14.7 17.3 47.9 1.8 2.4 0.4 0.5 1.5 Latin America 6.2 48.7 6.8 21.1 23.4 3.1 6.2 0.5 1.1 1.5 Middle East and North Africa 2.9 26.1 8.5 23.4 41:9 1.6 2.6 0.3 0.5 1.1 Transition economies' 6.4 13.2 21.8 51.9 13.1 4.4 4.5 1.4 3.0 1.9 Industrial economies 2.7 32.9 10.4 34.7 21.9 1.2 2.5 0.3 0.8 0.4 Note: The data by income group and by region are means of country averages. Income and region classifications follow World Bank definitions as published in World Bank (1996). a. The net interest margin is defined as net interest income divide by total assets. b. Columns 3 though 6 show the shares of the four components of net interest income. These shares add to 100 percent. c. A foreign bank is defined as having at least 50 percent foreign ownership. d. Large includes the largest 10 banks; the remaining banks are classified as small. The large versus small distinction is made only if there are more than 20 banks in a given year. e. The transition economies are China, Czech Republic, Estonia, Hungary, Lithuania, Poland, Romania, Russia, and Slovenia. Source: Authors' calculations based on data from the BankScope database of the IBCA. Demtrgiif-Kunt and Huizmga 389 assets tends to be highest for banks operating in lower-income countries. Banks operating in higher-income countries, instead, achieve the lowest net interest margins, and they face the lowest values of overhead, taxes, loan loss provision- ing, and net profits as shares of total assets. The breakdown by region reveals that the net interest margin is highest for banks operating in the transition economies at 6.4 percent and is also high in Latin America at 6.2 percent. It is lowest for banks operating in industrial coun- tries at 2.7 percent. The transition economies further stand out with high values of overhead, taxes, loan loss provisioning, and net profits as shares of total as- sets. Banks operating in industrial countries have the lowest ratio of net profits to total assets at 0.4 percent, probably because of the high level of competition in banking services. Table 3 provides information on some of the macroeconomic and institutional variables used in the regression analysis. The data are for 1995 or the most recent year available. The tax rate is computed on a bank-by-bank basis as taxes paid divided by before-tax profits. The figure reported in the table is the average for all banks in the country in 1995. Reserves divided by deposits are the banking system's aggregate central bank reserves divided by aggregate banking system deposits. Actual reserve holdings reflect required and excess reserves. Reserves are generally remunerated at less-than-market rates, and therefore actual reserves may be a reasonable proxy for required reserves, as averaged over the different deposit categories. For several developing countries—Botswana, Costa Rica, Greece, and Jordan—the reserve ratio is above 40 percent, indicating substantial financial repression. In contrast, this ratio is low in Belgium, France, and Luxem- bourg at 0.01. The deposit insurance variable is a dummy variable that takes on a value of 1 if there is an explicit deposit insurance scheme (with defined insurance premia and insurance coverage) and a value of 0 otherwise. Even if there is an explicit deposit insurance scheme, however, the ex post insurance coverage may prove to be higher than the de jure coverage, if the deposit insurance agency chooses to guarantee all depositors. With a value of 0 there is no explicit deposit insurance, even if the authorities offer some type of implicit insurance. Next, table 3 presents some indicators of the structure of financial markets. The concentration variable is defined as the ratio of the assets of the three largest banks to the assets of the total banking sector. As is well known, the concentra- tion of the U.S. banking market is low, at 16 percent, compared with values of about 50 percent for France and Germany. Note, however, that the U.S. figure may understate the concentration ratio in individual banking markets, which are protected from outside competition by interstate banking and branching restric- tions. The number of banks in the table reflects the number of banks in the data set with complete information. The ratio of bank assets to GDP is defined as the total assets of the deposit-money banks divided by GDP. This ratio reflects the banking sector's overall level of development. The ratio of stock market capitali- zation to GDP measures the extent of stock market development. Developing coun- Table 3. Economic and Institutional Indicators GDP per capita Inflation' Taxrattf' Reserves/ Deposit Market Number of Bank assets! Stock market Law and Economy (U.S. dollars) (percent) (percent) deposits insurance' concentration* banks' GDP1 capital/GDP order* Argentina . 3,825 0.02 0.15 0.06 1 0.48 11 0.23 0.13 5 Australia 14,542 0.03 0.47 0.01 0 0.45 44 0.77 0.70 6 Austria 16,947 0.02 0.23 0.04 1 0.75 12 1.27 0.14 6 Bahrain 7,902 0.02 0.01 0.10 — 0.94 7 0.49 — 5 Belgium 16,197 0.03 0.27 0.01 1 0.46 49 1.53 0.39 6 Bolivia 665 0.11 0.10 0 0.57 11 0.43 0.01 3 Botswana 1,844 0.18 0.32 0.41 — 0.94 5 0.14 0.09 — Brazil 2,113 22.95 0.38 0.13 0 0.43 56 0.32 0.21 3 Canada 16,091 0.02 0.35 0.01 1 0.56 72 0.72 0.65 6 Chile 2,481 0.12 0.07 0.08 1 0.40 23 0.45 1.10 5 China 468 0.12 0.21 0 0.99 5 0.80 0.06 — Colombia 1,445 0.23 0.18 0.23 1 0.35 28 0.20 0.22 2 Costa Rica 1,936 0.23 0.09 0.59 0 0.76 22 0.15 0.07 — Cyprus 7,500 0.03 0.34 0.16 — 0.75 9 0.91 0.30 5 Czech Rep. 3,165 0.15 2.01 0.18 1 0.76 15 0.87 0.33 — Denmark 22,386 0.02 0.16 0.07 1 0.77 56 0.55 0.33 6 Dominican Rep. 829 0.13 0.19 0.25 1 0.62 13 0.17 — 4 Ecuador 1,243 0.23 0.12 0.10 0 0.89 6 0.24 0.15 4 Egypt 709 0.12 0.25 0.20 0 0.89 9 0.65 0.13 4 El Salvador 994 0.12 0.04 0.32 1 0.86 4 0.28 0.67 3 Estonia 2,820 0.29 0.20 _ 0 7 — — — Finland 18,275 0.03 0.48 0.15 1 0.70 12 0.70 0.35 6 France 18,128 0.01 0.26 0.01 1 0.48 98 0.99 0.34 6 Germany 16,572 0.02 0.56 0.04 1 0.50 82 1.19 0.24 6 Greece 5,140 0.10 0.21 0.46 1 0.70 16 0.40 0.15 6 Guatemala 898 0.10 0.10 0.32 0 0.29 24 0.16 — 3 Haiti 230 0.72 0.25 0.35 1.00 3 0.10 — 3 Honduras 900 0.25 0.33 0.14 0 1.00 3 0.22 0.09 3 Hong Kong, China 11,911 0.02 0.13 — 0 0.44 35 — 2.17 6 Hungary 2,330 0.20 0.13 — 1 0.40 22 0.40 0.04 — India 423 0.06 0.04 0.16 1 0.90 5 0.35 0.38 4 Indonesia 718 0.13 0.30 — 0 0.38 21 0.41 0.33 5 Ireland 13,653 0.01 0.26 0.05 1 0.75 12 0.46 0.42 6 Israel 10,515 0.10 0.53 0.07 0 0.48 26 0.91 0.42 5 Italy 15,491 0.05 0.48 — 1 0.27 66 0.62 0.11 6 Jamaica 1,573 0.33 0.24 0.33 0 0.52 10 0.30 0.41 3 Japan 23,960 0.02 0.57 0.01 1 0.21 81 1.32 0.72 6 Jordan 1,263 0.02 0.31 0.49 0 0.93 7 0.70 0.70 5 Korea, Rep. of 5,663 0.06 0.26 0.10 0 0.17 43 0.55 0.40 5 Lebanon 1,800 — 0.24 0.16 1 0.61 6 0.79 — Lithuania 1,233 0.36 0.37 0.14 0 0.76 8 0.17 0.03 Luxembourg 21,433 0.07 0.45 — 1 0.30 108 0.41 0.14 6 Malaysia 3,108 0.06 0.32 0.12 0 0.31 49 0.84 2.82 5 Malta 6,102 0.04 0.29 0.08 — 0.69 7 0.75 — Mexico 1,749 0.45 0.20 0.23 1 0.59 20 0.35 0.32 3 Morocco 853 0.07 0.34 0.07 0.63 8 0.46 0.18 6 Nepal 203 0.07 0.35 — — 1.00 3 0.22 0.06 Netherlands 17,187 0.02 0.21 0.01 1 0.84 25 1.14 0.90 6 New Zealand 12,008 0.08 0.28 0.03 0 0.52 8 0.87 0.53 6 Nicaragua 786 0.09 0.22 0.27 — 0.63 13 0.32 — Nigeria 339 0.65 0.06 0.14 1 0.87 9 0.13 0.03 3 Norway 23,083 0.03 0.16 0.01 1 0.52 27 0.68 0.31 6 Oman 5,696 0.04 0.12 0.05 0 0.69 6 0.29 0.15 5 Pakistan 377 0.14 0.53 0.19 0 0.73 15 0.37 0.16 2 Panama 2,435 0.05 0.08 — — 0.54 9 0.69 0.10 3 Papua New Guinea 1,104 0.03 0.23 0.03 — 0.78 5 0.31 3 Paraguay 1,049 0.13 0.17 0.33 0 0.35 23 0.20 — 4 Peru 1,046 0.13 0.43 0.32 1 0.65 22 0.13 0.20 3 Philippines 615 0.07 0.12 0.12 1 0.44 21 0.43 0.79 4 Poland 1,903 4.65 0.40 0.10 1 0.45 32 0.29 0.04 Portugal 5,199 0.11 0.14 0.03 1 0.32 38 0.88 0.19 6 Qatar 12,820 — 0.01 0.04 — 1.00 2 0.71 — Romania 1,341 2.16 0.30 0.33 0 0.70 7 0.15 Russia 1,989 1.90 0.46 0.19 0 0.44 18 0.12 0.05 (Table continued on following page.) Table 3. (continued) GDP per capita Inflation* Tax rate* Reserves! Deposit Market Number of Bank assets/ Stock market Law and Economy (U.S. dollars) (percent) (percent) deposits insurance* concentration* banks' GDP1 capital/GDP order* South Africa 2,176 0.09 0.23 0.04 0 0.71 15 0.67 2.09 4 Saudi Arabia 5,316 0.04 0.00 0 0.96 4 0.41 0.33 5 Singapore 13,436 0.04 0.29 0.08 0 0.48 19 0.96 1.74 6 Spain 9,137 0.05 0.26 0.08 1 0.50 49 1.01 0.35 4 Sri Lanka 640 0.11 0.23 0 0.63 7 — — — Swaziland 787 0.47 0.16 0.21 — 4 0.23 0.30 — Sweden 19,387 0.04 -4.91 — 0 0.41 18 0.62 0.78 6 Taiwan (China) 7,268 0.02 0.18 — 1 0.40 25 — — 5 Thailand 1,807 0.04 0.31 0.04 0 0.49 14 0.98 0.86 5 Tunisia 1,464 0.05 0.20 0.04 — 0.55 8 0.55 0.22 — Turkey 1,848 1.06 0.14 0.27 1 0.43 29 0.17 0.17 5 United Kingdom 13,478 0.02 0.36 — 1 0.39 71 1.12 1.27 6 United States 20,931 0.01 0.27 0.03 1 0.16 372 0.48 0.95 6 u> Venezuela 2,651 0.51 0.06 0.25 1 0.46 17 0.13 0.05 4 Yemen 280 0.51 1.00 3 Zambia 247 0.55 0.34 0.08 0 1.00 3 0.10 — 4 — Not available. Note: For all variables 1995 figures were reported, if available. Otherwise figures are for the most recent year available. a. Inflation is the annual inflation of the GDP deflator. b. The tax rate is defined as total taxes paid by banks divided by before-tax profits. c. Deposit insurance is a dummy variable that takes the value 1 if there is an explicit deposit insurance scheme and 0 otherwise. d. Market concentration is defined as the ratio of the assets of the largest three banks to total banking assets. e. This value is the number of banks in the data set with complete information. f. Bank assets include the total assets of the deposit money banks. g. The law and order indicator reflects the degree to which the citizens of a country are willing to accept the established institutions to make and implement laws and adjudicate disputes. It is scored 0-6 with higher scores indicating sound political institutions and a strong court system. Lower scores indicate a tradition of depending on physical force or illegal means to settle claims. Source: GDP per capita and inflation are from World Bank National Accounts. The tax rate, market concentration, and number of banks are from IBCA'S BankScope database. Reserves/deposits and bank assets/GDP are from the International Monetary Fund, International Financial Statistics. Deposit insurance is compiled from Kyei (1995) and Talley and Mas (1990). Stock market data are from International Finance Corporation's emerging market database. The law and order indicator is produced by the International County Risk Rating Agency. Demirgiif-Kunt and Huizinga 393 tries tend to have lower bank-to-GDP and capitalization-to-GDP ratios, with some notable exceptions. Malaysia, South Africa, and Thailand, for instance, have relatively high ratios for both variables. The final column in the table provides an index of law and order, which is one of the institutional variables used in the regression analysis. This variable is scaled from 0 to 6, with higher scores indicating sound political institutions and a strong court system. Lower scores reflect a tradition in which physical force or illegal means are used to settle claims. The table shows considerable variation among countries in the sample. HI. EMPIRICAL RESULTS Tables 4 and 5 report the results of regressions of the net interest margin and before-tax profits as a share of total assets, respectively. Measuring profitability using the return on equity (as opposed to using the return on assets and control- ling for equity ratios as we do here) does not lead to significantly different results and thus is not reported. All regressions include country and year fixed effects. The tables report several specifications, the basic one including a set of bank and macroeconomic indicators as regressors. These are important control variables to which we subsequently add the taxation variables, the deposit insurance in- dex, the financial structure variables, and the legal and institutional indicators. We drop some variables from these two regressions because we want to ensure that banks from a reasonable number of countries are included. The estimation technique is weighted least squares, with the weight being the inverse of the num- ber of banks for the country in a given year. This weighting corrects for the fact that the number of banks varies considerably across countries. Bank Characteristics and Macroeconomic Indicators The first bank characteristic is the book value of equity divided by total assets lagged one period (E/TA M ). We lag total assets by one period to correct for the fact that profits, if not paid out in dividends, have a contemporaneous impact on bank equity. Buser, Chen, and Kane (1981) examine the theoretical relationship between bank profitability and bank capitalization. They find that banks gener- ally have an interior optimal capitalization ratio in the presence of deposit insur- ance. Banks with a high franchise value, reflecting costly bank entry, have incen- tives to remain well-capitalized and to engage in prudent lending behavior (see Caprio and Summers 1993 and Stiglitz and Uy 1996). Berger (1995b) provides empirical evidence that U.S. banks show a positive relationship between bank profitability and capitalization, although the evidence is not conclusive. The au- thor notes that well-capitalized firms face lower expected bankruptcy costs for themselves and their customers, thereby reducing their cost of funding. The basic specification (column 1 in tables 4 and 5) confirms that there is a positive relationship between E/TA^ and net interest income and bank profitabil- ity. In the regressions this variable is also interacted with GDP per capita (measured 394 THE WORLD BANK ECONOMIC REVIEW, VOL. 13. NO. 2 Table 4. Determinants of Net Interest Margins Regression results Independent variable (1) (2) (3) (4) (5) Bank characteristics Equity/lagged total assets (ETA,.,) 0.046* • • 0.047* ** 0.044* ** 0.064* ** 0 . 0 6 3 " * (0.007) (0.007) (0.007) (0.007) (0.006) Equity/lagged total assets interacted with GDP per capita -0.001 0.000 -0.001 -0.002* * -0.002* ** (0.001) (0.001) (0.001) (0.001) (0.001) Loans/total assets 0.017*" 0.008" 0.012* ** 0.022*" 0.019"* (0.004) (0.004) (0.004) (0.004) (0.004) Loans/total assets interacted with GDP per capita -0.000 0.001 *** 0.001* ** 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) Non-interest earning assets/total assets -0.016" -0.020* »* -0.021 **» -0.011 -0.020* ** (0.007) (0.007) (0.008) (0.007) (0.007) Non-interest earning assets/ total assets interacted with GDP per capita -0.001* -0.001 0.000 -0.002" -0.001 (0.001) (0.001) (0.001) (0.001) (0.001) Customer and short-term funding/total assets -0.007 0.003 0.004 -0.000 -0.004 (0.005) (0.005) (0.006) (0.005) (0.005) Customer and short-term funding/total assets interacted with GDP per capita 0.000 -0.000 0.000 -0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.001) Overhead/total assets 0.173"* 0.025*" 0.213*" 0.141*" 0.310"* (0.022) (0.019) (0.019) (0.018) (0.019) Overhead/total assets interacted with GDP per capita 0.002* ** 0.004* 0.004* 0.009* ** 0.005*" (0.002) (0.002) (0.002) (0.002) (0.002) Foreign ownership dummy 0.004* ** 0.003" 0.004* ** 0.004* *» 0.003* ** (0.001) (0.001) (0.001) (0.001) (0.001) Foreign ownership dummy interacted with GDP per capita - 0 . 0 0 1 " * -0.001*" -0.001* ** -0.000* ** -0.000* ** (0.000) (0.000) (0.000) (0.000) (0.000) Macroeconotnic indicators GDP per capita 0.000 0.000 0.000 0.000 -0.011*" (0.001) (0.001) (0.001) (0.001) (0.002) Growth rate 0.004 0.005 0.006 -0.011 -0.020)* * (0.008) (0.008) (0.008) (0.008) (0.007 Inflation rate 0.021 **• 0.026* ** 0.025* ** 0.020* ** 0.003 (0.006) (0.006) (0.006) (0.006) (0.005) Real interest rate 0.044"* 0.060* ** 0.058*" 0.051*" 0.025* ** (0.007) (0.007) (0.007) (0.007) (0.006) Real interest rate interacted with GDP per capita 0.001 -0.004 -0.003* -O.005*" -0.000 (0.002) (0.002)" (0.002) (0.002) (0.002) Demirgiic-Kunt and Huizmga 395 Table 4. (continued) Regression results Independent variable (V (2) (3) (4) (S) Taxation Reserves -0.076*** - 0 . 0 7 6 * " -0.024* -0.104*" (0.015) (0.015) (0.016) (0.016) Reserves interacted with GDP per capita 0.011"* 0.011*" 0.009*** 0.004 (0.003) (0.003) (0.003) (0.004) Tax rate 0.016"* 0.015"* 0.017*" 0.017*" (0.002) (0.002) (0.002) (0.002) Tax rate interacted with GDP per capita -0.001 • • • -0.001* »* -0.001 *** - 0 . 0 0 1 * " (0.000) (0.000) (0.000) (0.000) Deposit insurance Deposit insurance dummy -0.009*** (0.003) Financial structure Bank assets/GDP -0.024* » (0.010) Bank assets/GDP interacted with GDP per capita 0.001* (0.001) Stock market capitalization/GDP 0.016*" (0.005) Stock market capitalization/GDP interacted with GDP per capita -0.002*** (0.001) Stock market capitalization/bank assets -0.013** * (0.003) Stock market capitalization/bank assets interacted with GDP per capita 0.001** (0.001) Number of banks -0.001 (0.015) Market concentration 0.004 (0.005) Total assets (U.S. dollars) 0.003* *» (0.000) Legal and institutional indicators Contract enforcement dummy -0.042*" (0.007) Contract enforcement dummy interacted with GDP per capita 0.003*" (0.001) Law and order index -0.003* ** (0.001) Law and order index interacted with GDP per capita -O.000*" (0.000) Corruption -0.009* ** (0.001) (Table continued on following page.) 396 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 2 Table 4. (continued) Regression results Independent variable (V (2) (3) (4) (5) Corruption interacted with GDP per capita 0.001 »** (0.000) Adjusted R2 0.50 0.51 0.50 0.58 0.63 Number of observations 5,841 5,276 5,212 5,054 4,497 * Significant at the 10 percent level. ** Significant at the 5 percent level. *** Significant at the 1 percent level. Note: The regressions are estimated using weighted least squares pooling bank-level data across 80 countries for the 1988-95 time period. The number of banks in each period is used to weight the observations. The regressions abo include country and time dummy variables that are not reported. The dependent variable is the net interest margin defined as net interest income divided by total assets. Standard errors are given in parentheses. Source: Authors' calculations. Table 5. Determinants of Bank Profitability Regression results Independent variable (1) (2) (3) (4) (S) Bank characteristics Equity/lagged total assets 0.047* ** 0.051"* 0.055* ** 0.058** 0.015*" (0.009) (0.009) (0.009) (0.010) (0.006) Equity/lagged total assets interacted with GDP per capita 0.002 0.002* *» 0.003*** 0.002* ** 0.003* ** (0.001) (0.001) (0.001) (0.001) (0.001) Loans/total assets -0.013*** -0.024* ** -0.023** -0.015*** -0.018*" (0.005) (0.005) (0.005) (0.005) (0.004) Loans/total assets interacted with GDP per capita 0.001*" 0.003*** 0.003* *» 0.003* ** 0.001* ** (0.000) (0.000) (0.000) (0.000) (0.000) Non-interest earning assets/ total assets -0.005 -0.010 -0.011 -0.014 -0.033* ** (0.010) (0.010) (0.010) (0.010) (0.007) Non-interest earning assets/ total assets interacted with GDP per capita -O.007*" -0.007*** -0.007* ** -0.008*** 0.002* * (0.001) (0.001) (0.001) (0.001) (0.001) Customer and short-term funding/total assets -0.029* ** -0.017** -0.014*" -0.031"* -0.051*** (0.006) (0.007) (0.008) (0.001) (0.005) Customer and short-term funding/total assets interacted with GDP per capita 0.002* *» -0.000 -0.000 0.001 0.002*" (0.000) (0.001) (0.001) (0.001) (0.000) Overhead/total assets -0.023 -0.006 -0.004 -0.024 -0.114*" (0.025) (0.026) (0.026) (0.026) (0.019) Overhead/total assets interacted with GDP per capita -0.030*** -0.049* »* -0.049* ** -0.048*** 0.007* ** (0.003) (0.003) (0.003) (0.003) (0.002) Demhguc-Kunt and Huizmga 397 Table 5. (continued) Regression results Independent variable (1) (3) (4) (S) Foreign ownership dummy 0.005*" 0.006*" 0.006*" 0.006*" 0.006*" (0.001) (0.001) (0.001) (0.001) (0.001) Foreign ownership dummy interacted with GDP per capita -0.001*** - 0 . 0 0 1 " * -0.001*" - 0 . 0 0 1 * " 0.000*»' (0.000) (0.000) (0.000) (0.000) (0.000) Macroeconomic indicators GDP per capita 0.008* *• 0.008*" 0.008*" 0.007* 0.000 (0.001) (0.001) (0.001) (0.002) (0.002) Growth rate 0.002 -0.006 -0.007 -0.019 0.004 (0.010) (0.011) (0.011) (0.011) (0.007) Inflation rate 0.011 0.015* 0.014* 0.009 0.011* (0.008) (0.008) (0.008) (0.008) (0.005) Real interest rate 0.023"* 0.029* ** 0.029*" 0.023*1 0.026* ** (0.009) (0.010) (0.010) (0.009) (0.006) Real interest rate interacted with GDP per capita -0.000 -0.001 -0.001 -0.000 -0.003 ** (0.002) (0.002) (0.002) (0.002) (0.002) Taxation Reserves -0.126*" -0.129"* -O.106"* - 0 . 0 9 1 " * (0.021) (0.021) (0.023) (0.016) Reserves interacted with GDP per capita 0.029* ** 0.031"* 0.032*" 0.005*" (0.004) (0.004) (0.004) (0.004) Tax rate 0.022"* 0.022*** 0.021*** 0.017*** (0.003) (0.003) (0.003) (0.002) Tax rate interacted with GDP per capita -0.000 -0.000* * -0.003" 0.000*" (0.000) (0.000) (0.000) (0.000) Deposit insurance Deposit insurance dummy -0.005 (0.004) Financial structure Bank assets/GDP -0.028* (0.014) Bank assets/GDP interacted with GDP per capita 0.002* (0.001) Stock market capitalization/GDP 0.010 (0.007) Stock market capitalization/ GDP interacted with GDP per capita 0.000 (0.001) Stock market capitalization/bank assets -0.001 (0.001) (Table continued on following page.) 398 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 2 Table 5. (continued) Regression results Indicator (1) (2) (3) (4) (S) Stock market capitalization/bank assets interacted with GDP per capita -0.001 (0.001) Number of banks 0.000 (0.000) Market concentration 0.010* (0.007) Total assets (U.S. dollars) 0.000 (0.000) Legal and institutional indicators Contract enforcement dummy -0.022" (0.007) Contract enforcement dummy interacted with GDP per capita 0.001" (0.001) Law and order index -0.000 (0.001) Law and order index interacted with GDP per capita -0.000* (0.000) Corruption -0.002* (0.001) Corruption interacted with GDP per capita -0.000 (0.000) Adjusted R2 0.21 0.27 0.27 0.31 0.35 Number of observations 5,841 5,276 5,212 5,054 4,497 * Significant at the 10 percent level. *• Significant at the 5 percent level. *** Significant at the 1 percent level. Note: The regression is estimated using weighted least squares pooling bank-level data across 80 countries for the 1988-95 time period. The number of banks in each period is used to weight the observations. The regressions also include country and time dummy variables that are not reported. The dependent variable is before-tax profits divided by total assets. Standard errors are given in parentheses. Source: Authors' calculations. in constant thousands dollars for the year 1987). The positive coefficient on the interaction variables in the before-tax profits regression may reflect a higher bank franchise value in wealthier countries. The coefficients on £/TA M and the interac- tion variable together indicate how the ratio of equity to assets affects bank vari- ables in countries with different income levels. For a country with a per capita GDP of $10,000, for instance, the point estimate of the effect of £/TAr_, on before-tax profits divided by total assets is 0.067 (or 0.047 + 10 x 0.002). There is a negative and significant coefficient on non-interest-earning assets as a share of total assets in the net interest margin equation, but there is no signifi- cant relationship for the before-tax profits equation. Note that the sign on this variable interacted with per capita GDP is negative in both the net interest margin Demirguf-Kunt and Huizmga 399 and the before-tax profits specifications. Apparently, the presence of non-interest earning assets depresses net interest income and profitability more in wealthier countries than in poorer countries. By contrast, the sign on loans divided by total assets is positive in the net interest margin equation and negative in the before- tax profits equation. However, the coefficient of this variable interacted with GDP in the profits equation is positive, indicating that at higher income levels bank lending activities tend to be more profitable. On the liability side, customer and short-term funding consists of demand deposits, savings deposits, and time deposits. On average, this type of customer funding may carry a low interest cost, but it is costly in terms of the required branching network. This liability category does not significantly affect the net interest variable, although there is evidence that it lowers bank profitability. Differences in overhead may also capture differences in bank business and product mix, as well as variation in the range and quality of services. The ratio of overhead to total assets has an estimated coefficient of 0.173 in the net interest margin regression, which suggests that about a sixth of a bank's overhead cost is passed on to its depositors and lenders. The interaction of overhead with per capita GDP also enters with a positive coefficient, indicating that a larger share of overhead is passed onto financial customers in wealthier countries. This may reflect more competitive conditions in banking markets in industrial than in de- veloping countries. In the before-tax profits regression the interaction of over- head with per capita GDP enters negatively, indicating that higher overheads eat into bank profits. The foreign ownership variable equals 1 if at least 50 percent of the bank's stock is in foreign hands and equals 0 otherwise. In both tables 4 and 5 this variable has a positive coefficient, while its interaction with per capita GDP has a negative coefficient. These results suggest that foreign banks realize relatively high net interest margins and profitability in relatively poor countries. It may be that foreign banks are frequently exempt from unfavorable domestic banking regulations and apply superior banking techniques. Note, however, that the point estimate of the effect of foreign ownership for a wealthy country with a per capita GDP of $20,000 is negative in the net interest margin equation at -0.016 (that is, 0.004 - 20 x 0.001) and in the profitability equation at -0.015 (that is, 0.005 - 20 x 0.001). Foreign banks' technological and efficiency advantages in countries may be insignificant because, while there, they face informational dis- advantages. This could explain why foreign banks in industrial countries are relatively unprofitable on average. Turning to the macroeconomic indicators, we see, first, that per capita GDP has no significant impact on the realized net interest margin, although it enters the profitability equation with a positive coefficient. Per capita GDP is a general index of economic development, and it thus reflects differences in banking technology, the mix of banking opportunities, and any aspects of banking regulations omit- ted from the regression. Growth, defined as the growth rate of real per capita GDP, is insignificant in both regressions. The percentage change in the GDP defla- 400 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 2 tor, or inflation, is estimated to increase the net interest margin and bank profit- ability. However, the significance of the coefficients in the profitability regres- sions is low, possibly because banks obtain higher earnings from float or because there are delays in crediting customer accounts in an inflationary environment. With inflation, bank costs also tend to rise. A larger number of transactions may lead to higher labor costs and, as shown by Hanson and Rocha (1986:40), result in a higher ratio of bank branches per capita. On net, however, the regression results suggest that the impact of inflation on profitability, although not very significant, is positive throughout. We constructed the real interest rate using the short-term government debt yield and, if that measure was not available, other short-term market rates. The real interest rate enters the net interest margin and before-tax profits regressions positively, although this variable interacted with per capita GDP has a signifi- cantly negative coefficient in the ne t interest margin equation. Thus there is some evidence that increases in the real interest rate do not raise spreads as much in industrial countries, perhaps because their deposit rates are not tied down by deposit rate ceilings. Taxation Variables Banks are subject to direct taxation through the corporate income tax and other taxes, and they are subject to indirect taxation through reserve require- ments. Reserve requirements are an implicit tax on banks if, as is usual, official reserves are remunerated at less-than-market rates. The corporate income tax and the reserve tax differ in important respects. First, the corporate income tax, in principle at least, can be targeted at pure profit. To the extent that it is a profit tax, the corporate income tax is relatively nondistorting. In practice, however, it may not be a pure profit tax if complete expensing of costs is not allowed. The reserve tax, by its very nature, is proportional to the volume of deposit taking and is therefore a distorting tax. From a welfare perspective the corporate income tax thus appears to be superior to the reserve tax. A second important difference is that the severity of the reserve tax depends on the opportunity cost of holding reserves. This may depend on financial market conditions as much as on any tax code. Related to this second condition, reserve requirements are also an instrument of monetary policy. As far as we know, there has been no other empirical research on the effect of the corporate income tax on the banking sector. Several studies have considered the impact of reserve requirements on bank profitability. Some, in particular, have examined how Federal Reserve membership affeaed the profitability of U.S. commercial banks in the 1970s (see Rose and Rose 1979 and Gilbert and Rasche 1980). Federal Reserve membership subjected banks to generally higher reserve requirements. Most of the studies in this area support the notion that nonmember banks were more profitable than member banks (with similar char- acteristics) because nonmember banks held relatively little cash. Competition among member and nonmember banks in the same market appears to have pre- DemirgUf-Kunt and Huizinga 401 vented member banks from passing their higher reserve costs onto their custom- ers. In related work Kolari, Mahajan, and Saunders (1988) study the impact of announced changes in reserve requirements on bank stock prices using an event study methodology. Huizinga (1996) and Eijffinger, Huizinga, and Lemmen (1998) examine how nonresident withholding taxes affect interest rates, while Fabozzi and Thurston (1986) examine how differences in reserve requirements are priced into money-market instruments. Because detailed information on the reserve regulation of all countries in our sample is not available, we use a proxy to capture bank reserves. We construct this variable in the regressions as the product of the banking system's ratio of aggregate reserves to deposits (as in table 3) and the individual bank's ratio of short-term funding to total assets. Customer and short-term funding, consisting of demand deposits, savings deposits, and time deposits, here proxy for reservable deposits. The reserves variable is thus an approximation of actual bank reserves that reflects differences in systemwide reserve requirement rules. In tables 4 and 5 the reserves variable enters the regressions negatively. The coefficients in the net interest margin equations show two effects: less-than- market remuneration and the impact on banks' lending and deposit rates. The first effect is expected to be negative because underremunerated reserves lower a bank's net interest income and profitability. The second effect could be either zero, in which case the bank bears the full cost of higher reserves, or positive, in which case the cost of reserves is passed onto bank customers in the form of higher interest margins. In table 5 we see that the reserves variable negatively affects bank profitability. This suggests that the second, or pass-through, effect is either nonexistent or too small to offset the first, or direct, effect. Abstracting from any pass-through, the coefficient on the reserves variable in either regres- sion can also be interpreted as a bank's opportunity cost of holding reserves. The reserves variable interacted with per capita GDP enters both regressions positively, which may reflect the fact that the opportunity cost of holding reserves is higher in wealthier countries. We capture the explicit taxes that banks pay with the variable tax rate, which is measured by a bank's tax bill divided by its pretax profits. This variable has a significantly positive impact on interest margins and profitability. The tax rate interacted with per capita GDP is negative and significant in both regressions. These results suggest that both the net interest margin and profitability increase with tax rates, but less so in richer countries. Thus the corporate income tax is passed through to bank customers to some degree. To calculate the extent of this pass-through, we use the estimated coefficients on the tax rate variable and its interaction with per capita GDP. Let the pass-through be defined as the increase in before-tax profits following a one-unit increase in the corporate tax bill, or dBTP/dTX. Next, note that (3BTP/