Banking Sector Performance in East Asian Countries: The Effects of Competition, Diversification, and Ownership

Transcription

1 Banking Sector Performance in East Asian Countries: The Effects of Competition, Diversification, and Ownership Luc Laeven* (The World Bank and CEPR) Abstract: This paper takes stock of the bank restructuring process in five East Asian countries Hong Kong (China), Indonesia, the Republic of Korea, Malaysia, the Philippines, Singapore, and Thailand with a particular goal of assessing whether bank performance and stability has improved following the Asian financial crisis of We find that the banking systems in all East Asian countries look markedly different today than during the period before the crisis, both in terms of ownership and market structure. The ongoing process of consolidation of local banking markets and an increase in foreign ownership of banks have improved performance and stability. We conclude with several policy recommendations regarding foreign bank entry, bank consolidation, and bank governance going forward. * This paper was prepared as a background paper for East Asian Finance: the Road to Robust Markets published by the World Bank. The author would like to thank Stijn Claessens and Swati Ghosh for helpful comments and Ying Lin for excellent research assistance. This paper s finding, interpretations, and conclusions are entirely those of the author and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. 0

2 1. Introduction In retrospect, we now know from the recent crisis experience in East Asia that banks were taking excessive risks, largely unknown to small investors and depositors, although bank performance varied markedly across banks depending on such factors as the quality of management and the type of ownership (see Laeven (1999, 2002), among others). Following the onset of the crisis in , the banking systems of many countries in the East Asia region, and especially the crisis-affected countries, have undergone major restructuring efforts, often with major government involvement. Some banks in the respective countries were taken over by the State, while others received government support (Klingebiel, Kroszner, Laeven, and Van Oijen 2001). By now, many of these nationalized banks have been sold to the private sector, mostly to domestic investors, although foreign interest in local banks has also increased (both from outside the region and from within the region, e.g. foreign equity investments by the Development Bank of Singapore). As a result, the banking systems of most countries in the region look markedly different today than before the crisis, both in terms of ownership and market structure. This raises a number of policy-relevant questions: Should further consolidation be encouraged? Should foreign banks be allowed to enter the market? Commercial banks in many East Asian countries have traditionally been linked through ownership to other financial institutions, such as merchant banks and finance companies. In Korea, banks often owned merchant banks; in Thailand, banks often owned finance companies; and in Malaysia, bank holding companies often include commercial banking, investment banking, asset management, and insurance companies. 1

3 While the crisis has led some banks to focus on more traditional banking activities, other banks have continued to expand the range of their activities, with increasing focus on income from fee-based activities. This raises another important policy question: Is diversification a better strategy than focusing on core activities? In this paper, we address these questions by studying the performance and stability of the banking systems in East Asia. We first assess whether performance and stability have improved since the financial crisis of , and then identify the determinants of bank performance and stability today. Based on this analysis, we make several predictions about the impact of the ongoing process of consolidation on the performance and stability of these banking systems. We also derive some policy recommendations regarding bank diversification, foreign bank entry, consolidation of local banking markets, and bank governance more generally. 2. Methodology Measures of bank performance can broadly be broken down in two categories: those based on accounting information and those based on market information. Bongini et al. (2002) show that accounting-based measures are lagging market-based measures, and therefore a market-based approach would be the preferred choice. As a market-based measure of bank-risk we will use the implicit deposit insurance premium measure of risk developed in Laeven (2002b). This measure estimates the cost of insuring all bank deposits in a particular banking system and can be interpreted, as shown by Laeven (2002a), as a measure of bank risk. The riskier the banks in the system, 2

4 the costlier it will be to insure all bank deposits, and the higher the deposit insurance premium. We will calculate this measure for the portfolio of all banks in the system to allow for risk diversification, which can significantly reduce to cost of deposit insurance (Laeven 2000b). The disadvantage of this method is that it can only be applied to listed banks. Most banks in the East Asia region are not listed (although many of the largest banks are) and we will therefore focus on accounting-based measures of bank performance. As accounting-based measures of bank performance we will use the ratio of operating income to total assets. The banking literature has also developed methods to calculate the X-efficiency of cost efficiency of banks using accounting-based information. As argued by Laeven (1999), these methods heavily rely on reliable data on nonperforming loans and measures of bank risk more generally. For most countries in East Asia, such data is not available, and we therefore resort to simple financial ratios as accounting-based measures of bank performance. The basic model will look as follows: Performance ijt = country j + bank i + time t + B ijt + ownership it + market jt + regulation jt + ε ijt where Performance is a measure of bank measure; Country, Bank, and Time capture country, bank, and time-specific effects, respectively; B is a vector of bank-specific variables, such as liquidity ratios, capital adequacy ratios, and other CAMEL-type indicators; Ownership is a bank-specific measure of bank ownership, such as type of controlling owner; Market is a country-specific measure of market contestability, such as market concentration or market share; Regulation is a country-specific measure that 3

5 includes bank regulatory and supervisory variables; and i denotes bank i, j denotes country j, and t denotes year t. As measure of bank performance, we use the ratio of operating income to total assets. This measure has been widely used in the literature (together with pre-tax return on total assets) as a measure of bank profitability. Note that operating income is gross income before operating costs (including personnel expenses) and before taxes. It includes net interest income and income from fees, commissions, and trading income. Under the assumption that banks are profit maximizers, higher profits denote better bank performance. Also, to the extent that high bank profits reflect greater stability of the banking system, thus reducing the likelihood of costly bank runs and bank defaults, they may improve a society s welfare. However, higher bank profits do not necessarily enhance a society s welfare. If banks earn superprofits by extracting excessive rents from consumers, then high profits may be an indication that the banking system is not competitive and that consumer welfare is negatively affected. While it is generally accepted that banks should have a positive franchise value to enhance financial sector stability, very high profits are generally taken as a sign of lack of competition. In our empirical analysis, we focus on within-country (rather than cross-country) variation in bank profitability. This allows us to keep country effects, such as the competitiveness of the banking system, and to analyze what drives differential performance of banks in a given country. Because the performance of banks may differ depending on the diversity of activities they engage in, we construct an activity-adjusted performance measure based on the work by Laeven and Levine (2005). Theory provides conflicting predictions about 4

6 the impact of greater diversity of activities on the performance of financial intermediaries. As suggested by the work of Diamond (1991), Rajan (1992), Saunders and Walter (1994), and Stein (2002), banks acquire information about clients during the process of making loans that may facilitate the efficient provision of other financial services, including the underwriting of securities. Similarly, securities and insurance underwriting, brokerage and mutual fund services, and other activities may produce information that improves loan making. Thus, banks that engage in a variety of activities may enjoy economies of scope that boost performance. Alternatively, diversification of activities within a single financial conglomerate may intensify agency problems between corporate insiders and small shareholders with adverse implications on bank performance (Jensen, 1986; Jensen and Meckling, 1986). Laeven and Levine (2005) find that, on average, diversity of activities by banks destroys value and reduces bank performance. We use the method developed by Laeven and Levine (2005) to control for the possibility that the performance of different financial activities is inherently different. For example, if securities underwriting is more income than loan making, then a bank that does both may have higher operating income than a bank that only makes loans. We abstract from these activity-effects on bank performance to identify the independent impact of diversity by compare the operating income of diversified banks to the estimates of operating income these banks would have if they were decomposed into a bank specialized in loan-making activities and a bank specialized in non-lending activities. Due to data constraints, we differentiate banks by (i) interest income versus noninterest income and by (ii) loans versus other earning assets. Thus, we do not distinguish among securities underwriting, brokerage services, and insurance underwriting. We 5

7 simply differentiate banks by lending versus non-lending activities. First, we construct asset-based and income-based measures of the extent to which banks engage in loan making activities or fee generating activities. One can think of specialized commercial banks as converting deposits into loans, and one can think of specialized investment banks as underwriting securities but not making loans. Second, we construct asset-based and income-based measures of diversity. That is, we measure the degree to which banks specialize in lending or non-lending services, or whether they perform a diversity of activities. Lower values of these diversity indexes imply more specialization, while higher values signify that the bank engages in a mixture of lending and non-lending activities. Clearly there is a link between these diversity measures and the measures of the degree to which banks engage in loan making or nonloan making activities. If a bank only makes loans, it will be classified as having zero diversity. The two measures, however, also capture different traits. The diversity indexes measure diversity per se, while the activity measures gauge where each bank falls along the spectrum from a pure lending bank to a pure fee-generating bank. To measure where along the spectrum each bank falls from pure commercial banking to specialized investment banking, we first construct an asset-based measure that equals loans relative to total earning assets. Total-earning assets include loans, securities, and investments. Very high values signal that the bank specializes in loan making, like the specialized commercial banks mentioned above. Very low values of these ratios signal that the bank is not specialized in loan making and indicates the financial institution specializes in non-loan making activities. 6

8 The second measure of where each bank falls along the continuum from pure lending to pure fee/trading-based activities is an income-based indicator that equals the ratio of net interest income-to-total operating income. Total operating income includes net interest income, net fee income, net trading income, and net commission income. In terms of assessing where along the spectrum each bank falls, a specialized loan-making bank will have a larger ratio of net interest income-to-total operating income, while a specialized investment bank is expected to have a larger share of other operating income. The asset-based measure suffers from fewer measurement problems than the income-based measure, but we include both for robustness. In particular, since loans may yield fee income, the income-based measure may overestimate the degree to which some lending institutions engage in non-lending activities. Also, we would prefer to use gross rather than net income to measure bank activities, but as noted above, we simply do not have gross income for many banks. Next, we construct two measures that focus on diversity per se. Asset diversity is a measure of diversification across different types of assets and is calculated as ( Net loans Other earning assets) 1, where Other earning assets include securities Total earning assets and investments. Total earning assets is the sum of Net loans and Other earning assets, and. denotes the absolute value indicator. Asset diversity takes values between 0 and 1 and is increasing in the degree of diversification. Income diversity is a measure of diversification across different sources of income and is calculated as ( Net interest income Other operating income) 1. Net Total operating income interest income is interest income minus interest expense and Other operating income 7

9 includes net fee income, net commission income, and net trading income. Income diversity takes values between 0 and 1 and is increasing in the degree of diversification. Since different banking activities may generate different income streams, it is important to control for the degree to which banks engage in different activities when comparing their performance. For example, if investment banking generates generally more income than commercial banking, one needs to control for the extent to which the bank is engaged in either activity in order to isolate the relationship between performance and diversity per se. Thus, we compute an excess performance measure following a modified version of the chop-shop approach introduced by LeBaron and Speidell (1987) and Lang and Stulz (1994) and adopted and applied to banks by Laeven and Levine (2005). The idea is to compare the operating income of each bank with the operating income that would exist if the bank were chopped into separate financial shops (pure-activity banks) that each specializes in a financial activity (e.g., lending or fee/income generation). Activity-adjusted π j is our estimate of the ratio of operating income to total assets π that would prevail if bank j were divided into activity-specific financial institutions that each generates income according to the π s associated with each of those activity-specific activities. At a general level, consider bank j that engages in n activities. Letα equal the share of the ith activity in the total activity of bank j, so that α = 1. Let π i equal the ratio of operating income to total assets of financial institutions that specialize in activity i (pure-activity π). Then, Activity adjusted π = j n i= 1 i α π ji n i= 1 ji ji 8

10 More specifically, we primarily consider two banking activities: lending operations versus non-lending operations, including trading, investments, and advisory services. From an asset perspective, we focus on the distinction between investments in loans and investments in securities or other companies. From an income perspective, we focus on the distinction between interest income (mainly from loans) and non-interest income, including fees, commissions, and trading income. For simplicity, we refer in what follows to the first activity as commercial banking and to the second as investment banking. Thus, π 1 is the operating income of an activity-specific bank focused on commercial banking, while π 2 is the operating income of an activity-specific bank focused on investment banking. With two activities, the definition of activityadjusted π for bank j simplifies to the following: Activity adjusted π j = ( α π + α π ) = ( α π + (1 α ) ) (1) j1 j 2 j1 j1 π In what follows, we compute two activity-adjusted π measures. That is, we calculate activity-adjusted π based on both the asset and income measures of the share of bank activity. Thus, α j1 equals either the ratio of net interest income to total operating income or the ratio of net loans to earnings assets for bank j. Excess value equals the difference between a bank s actual π and the activityadjusted π, so that the excess value for bank j is Excess value j = π ( α π + α π ) = q ( α π + (1 α ) ) (2) j1 j 2 j1 j1 π Again, we compute two measures of excess value, one based on weights determined by the asset composition of the bank and the other determined by the income composition of the bank. 9

11 To measure activity-adjusted π s and compute excess value, we construct π 1 and π 2 from banks that specialize in one activity. We follow the literature in defining what constitutes specialization. For asset-based measures, banks where 90% of the assets are associated with one activity are classified as specialized. In this case, π 1 is the average π of banks with a ratio of net loans to earnings assets of more than 0.9. Similarly, for income-based measures, specialized banks receive 90% of their income from one activity, so that π 1 equals the average π of banks with a ratio of net interest income to total operating income of more than 0.9. These pure-activity π s are calculated by averaging across banks from the different East Asian countries in our sample. Most countries do not have a sufficiently large number of pure-activity banks to estimate pureactivity π s at the country-level. In the regression analyses below, we use country fixed effects and year dummy variables to control for differences in π across countries and years. In constructing activity-adjusted π s and excess values, we need to compute α j1 and α j2, which are the shares of pure commercial banking and investment banking in bank j s activities. The weights are based on the relative importance of interest income to total operating income in the case of the income diversity measure. In case of the asset diversity measure, the weights are based on the relative importance of loans to total earning assets. In our empirical work we will also control for the ownership structure of the banks. As shown by Caprio et al. (2004), among others, the type of ownership and the cash flow rights of ultimate controlling shareholders are key determinants of bank performance and valuation. We use hand-collected data on the type of the ultimate owner 10

12 of the bank. We do not have detailed enough information about the ownership structures of all the banks in our sample to calculate the cash flow rights. This would require a detailed study of the often times complex ownership structures of banks in East Asia (socalled pyramidal structures). We consider a bank to be controlled by a shareholder, if the controlling shareholder owns more than 50% of the control rights of the bank. We consider four categories of ultimate ownership: state, foreign state, private domestic, and foreign. We aggregate the stakes of all shareholders by each of these four categories and determine ultimate ownership by attaching the ownership category to the group of shareholders with the largest ownership stake. Since many of the variables under consideration are bound to be endogenous (for example, performance and ownership are expected to be endogenous), efficient estimation of the above relationships will depend on the use of time series data. We will thus construct a dataset that varies over time. This will not only help us to deal with potential endogeneity issues but will also enable us to analyze whether effects have changed over time, for example, whether the effect of foreign bank entry on local bank performance has changed over time. We also develop and estimate different types of bank competition measures. Here we rely on the Panzar-Rosse (1982, 1987) approach developed in Claessens and Laeven (2004, 2005), as well as on more traditional measures of bank concentration, such as the 3-bank concentration ratio and the market shares of individual banks. For most of these measures, it is important to have data on a sufficiently large number of banks in the respective countries, and therefore we are only able to implement this approach by 11

13 pooling country-level data over several years. As a consequence, we can only estimate changes in competition over time across all countries in the region, not for individual countries. The Panzar and Rosse H statistics are calculated from reduced form bank revenue equations and measures the sum of the elasticities of the total revenue of the banks with respect to the bank s input prices. The H statistic is interpreted as follows. H<0 indicates a monopoly; H=1 indicates perfect competition; and 0<H<1 indicates monopolistic competition. Nathan and Neave (1989) point out that this interpretation assumes that the test is undertaken on observations that are in long-run equilibrium. We therefore also test whether the observations are in long-run equilibrium. Following Claessens and Laeven (2004), we estimate reduced form revenue equations on pooled samples for each country (see equation (1) in Claessens and Laeven (2004)). We let the ratio of gross interest revenue to total assets (proxy for output price of loans) be a function of (i) the ratio of interest expenses to total deposits and money market funding (proxy for input price of deposits), (ii) the ratio of personnel expense to total assets (proxy for input price of labor), and (iii) the ratio of other operating and administrative expense to total assets (proxy for input price of equipment/fixed capital). This model is similar to models used previously in the literature to estimate H-statistics for banking industries. We include several control variables at the individual bank level, including the ratio of equity to total assets, the ratio of net loans to total assets, and the logarithm of total assets (to control for potential size effects). We also include year-fixed effects. We take natural logarithms of all variables. We estimate the model both using OLS with time dummies and GLS with fixed bank-specific effects. The H-statistic equals 12

14 the sum of the coefficients on three main explanatory variables: interest expenses to total funding, personnel expense to total assets, and other operating and administrative expense to total assets. 3. Data We collect financial data and ownership data on banks from Bankscope, a commercial data provider of data on over 10,000 publicly listed and private banks around the world. Most of the data come from audited financial statements. We also have data on the type of specialization of the bank (i.e., whether the bank is a commercial bank, an investment bank, a savings bank, a bank holding company, a development bank, etc.). Although nonbank financial institutions are important players in the financial systems of some of the East Asian countries (for example, the finance companies in Thailand and the merchant banks in Korea), we focus on commercial banks. To enhance comparability of banks in our sample, we limit the sample to banks identified by Bankscope as commercial banks, savings banks, and bank holding companies with major commercial banking operations. We collect data for the period (when available) for 7 East Asian countries: Hong Kong (China), Indonesia, the Republic of Korea, Malaysia, the Philippines, Singapore and Thailand. We have data for about 2,157 bank-year observations, although not all variables are available for all banks in all years. The coverage of banks is particularly problematic during the early years of our sample period, because Bankscope does not always keep information for banks that have failed during the sample period. This produces a survivorship bias in the results. We also miss data on 13

15 many banks for the year 2004 because many banks have not yet reported their financial statements for the year 2004 to Bankscope. This explains why the number of banks in our sample drops from 213 in 2003 to only 122 in Table 1 presents a breakdown of the banks in our sample by ownership category. We distinguish between four different ultimate ownership categories: state, foreign state, private, and foreign. Foreign state banks are banks that are owned by a foreign state. An example is the Development bank of Singapore, which is majority owned by the government of Singapore, and has operations in other East Asian countries. Private denotes domestic banks that are majority-owned by domestic citizens. This includes family owned banks as well as banks that are widely held by a large number of private shareholders. Foreign banks are banks that are owned by foreign shareholders (excluding foreign states). The latter group often includes subsidiaries of multinational banks but also includes the Hong Kong and Shanghai Banking Corporation, the largest bank in Hong Kong and one of the largest banks in the world, with stock market listings in several countries and a large shareholder base around the world. The table shows that the majority of banks in the East Asian countries are privately-owned. However, the importance of family ownership and ownership by other private parties has dwindled from almost 80 percent in 1994 to 36 percent in 2000, only recovering somewhat to about 47 percent by the year State-ownership one the other hand has increased over the same period, with the state controlling about 20 percent of the banks in 1994 to about 30 percent in However, the foreign ownership category has recorded the largest increase over this period. While foreigners owned a mere 1.5 percent of banking assets in East Asia in 1994, this number has increased to about 23 14

16 percent by the year The most important reason for these shifts in ownership structure is the East Asian financial crisis and the governments response to the crisis. Many family-owned and other privately-owned banks failed during the crisis, and unless these banks were of systemic importance or had strong links to the political elite, they were unlikely not to be bailed out. This explains the sharp drop in the number of private banks post However, a significant share of these failed private banks did get bailed out by the government, resulting in a temporary increase in state banks, reflected in the significant increase in state banks during the years from 21% to 37%. While some of these banks are still in state hands, others have been successfully privatized to the public, often to foreigners, explaining to a large extent the increasing importance of foreign ownership. Panel B of Table 1 reports the ownership breakdown by country. Although the patterns of changes in ownership are broadly consistent across countries, there are some differences. State ownership, for example, plays much less of an important role in Korea and the Philippines than in the other countries. While ownership of banks by the state in Korea increased after the 1997 financial crisis to about 21% in 2001, it decreased to only 7% by the year State-ownership in the Philippines stood at a level of about 18% by year-end In Indonesia, Singapore, and Thailand, on the other hand, the state still owns more than 50% of the banks (As measured in terms of total assets). The importance of foreign shareholders of local banks also varies significantly across countries. While foreigners are important shareholders of banks in Hong Kong and to a lesser extent in Indonesia and Malaysia, they do not play an important role in the other East Asian countries. 15

17 Next we look at the market structure of the East Asian banking systems. Panel A of Table 2 reports for each country the average size of banks, the 3-bank concentration ratio, and the average market share (all in terms of either total assets or total deposits). Panel B reports values of the same variables for the year We find that Korean banks are much larger on average than their counterparts elsewhere in the region, both in terms of total assets and in terms of deposits, while banks in Indonesia and the Philippines are much smaller. The typical bank in Korea has about US$ 30 billion worth of total assets, while the average bank in Indonesia or the Philippines has about US$ 1.6 billion in total assets. These differences remain large and significant when we control for differences in economic development using per capita GDP (not shown in the Table). We also notice stark differences in the market concentration across the East Asian banking systems with the banking systems of Singapore and Hong Kong being the most concentrated and the banking systems of Korea and Malaysia the least concentrated, although the average bank concentration in the region does not differ significantly from the world average. The 3-bank concentration ratio (in terms of total deposits and for the period 2004) varies from a low of 0.44 in Malaysia to a high of 0.82 in Hong Kong, and the regional average of 0.56 is very similar to the average world-average of about 0.55 (as reported in Claessens and Laeven 2005). Bank concentration is often used as a measure of bank competition, although work by Claessens and Laeven (2004) suggests that concentration ratios are not highly correlated to measures of market contestability and therefore capture other aspects beyond competition. Nevertheless, with this caveat in mind, the figures suggest that competitive pressure may be low in some of the banking markets of East Asia because of high concentration of bank assets and deposits. 16

18 Next we look at differences in bank performance. Panel C of Table 2 reports the country averages of a commonly used measure of bank performance, the ratio of total operating income to total assets. Operating income includes net interest income and income from fees, commissions and other services. This figure is before operating expense (such as labor costs) and before taxes. We find that the average bank in all countries is profitable. The average ratio of operating income to total assets is about 4.8 percent over the period Banks in the Philippines and Indonesia are the most profitable, with operating income to total asset ratios of about 6 percent. In Table 3, we report a measure of regulatory restrictions on banking in the East Asian countries compiled by the Heritage Foundation, as well as similar measures of government intervention and regulation in other aspects of the economy. Two things are striking. First, banking is somewhat more regulated than other aspects of the economy (such as trade and monetary policy) in all countries except Hong Kong and the Philippines. Second, regulatory restrictions on banking have not changed much in any of the East Asian countries over the period , despite the financial crisis. This suggest that despite government intervention in the banking systems of most countries following the crisis, this was not perceived by outside observers such as the Heritage Foundation to negatively affect the freedom of banking in any of these markets. We would have preferred to use more detailed data on specific bank regulatory variables (for example on entry and capital regulation) from the World Bank Database on Bank Regulation and Supervision and Barth et al. (2001) but because such data is only available for two years in our sample we prefer to use the more aggregate measure of banking freedom of the Heritage Foundation instead. 17

19 4. Empirical Results We first assess the risk embedded in East Asian the banking systems. To this end, we apply the option pricing methodology developed by Ronn and Verma (1986) and adapted by Laeven (2002) to the sample of listed banks in each of the East Asian countries. We assume that all bank debt is insured and that there is no regulatory forbearance. In practice, regulatory forbearance can be substantial, especially around times of systemic distress, which is also when the value of the put option of deposit insurance is highest. As a result, we are underestimating the implicit cost of deposit insurance. It is also important to note that most of the East Asian countries explicitly insure deposits; in three cases coverage is even unlimited due to a government blanket guarantee on deposits installed shortly after the onset of the 1997 financial crisis. Blanket guarantees on deposits were enacted in Thailand in 1997 and in Indonesia and Malaysia in The Philippines was the first country to region to adopt explicit deposit insurance, following the example of the United States. The Philippines has had explicit deposit insurance since 1963 with annual premium of 0.2% on deposits. Coverage limit on deposits in the Philippines has been 100,000 Pesos since The Republic of Korea has enacted explicit deposit insurance in 1996 with a coverage limit on deposits of 20 million Won which increased to unlimited coverage in 1997 at the time of the crisis and was subsequentially reduced to a coverage limit of 50 million Won (about US dollars) in 2003 with annual premiums of 0.05%. Hong Kong and Singapore have no explicit deposit insurance (Demirguc-Kunt et al. 2005). 18

20 Our deposit insurance measure estimates the implicit cost of insuring the deposits in a particular banking system. Laeven (2002) shows that we can interpret higher implicit deposit insurance premiums as a measure of banking system risk. By comparing the implicit cost estimates with the actual premiums charged for deposit insurance, we can also infer whether deposit insurance is underpriced. In countries that have not adopted explicit deposit insurance, the estimates give an indication of how much it would cost to insure all deposits in the system if deposit insurance were made explicit. We estimate the implicit cost for the banking system as a whole, thereby allowing for diversification potential by aggregating risks of banks that are not perfectly correlated. As Laeven (2003) shows, the diversification potential can be substantial, particularly in large banking systems with a diverse set of banks, thereby significantly reducing the cost of deposit insurance. Table 4 presents our estimates of deposit insurance for the year 1998 and for the period In all countries, the peak in the implicit cost of deposit insurance was reached in 1998, which not surprisingly corresponds with the height of East Asian financial crisis, except Indonesia where the peak of implicit premiums was reached in 1999 (not shown). In Thailand, the implicit annual deposit insurance premium on deposits would be 0.63% of deposits, which is substantial for a bank with a typical interest rate margin of 2-4%. Again, we note that these estimates are likely to substantially underestimate the actual cost of deposit insurance because we have not allowed for regulatory forbearance. While the implicit cost of deposit insurance reached high levels around the time of the crisis, the cost of insurance is much lower when calculated over the entire period 19

21 The implicit annual premiums on deposits vary from a low of 0.01% in Hong Kong and Singapore (two countries that do not have explicit deposit insurance) to a high of 0.84% in Indonesia. For the Republic of Korea we estimate an implicit cost of deposit insurance amounting to an annual premium of 0.05% per annum, with is identical to the actual premiums that banks are being charged today. For the Philippines, our implicit premium estimates are substantially lower than those actually charged (but again, we should keep in mind that we are potentially underestimating the cost of deposit insurance). Overall, the deposit insurance estimates summarized in Table 4 suggest that while systemic risk increased dramatically in almost all East Asian banking systems during the period , that with the exception of Indonesia, systemic risk today is quite low. Next, we measure the level of competition in each of the banking systems in our sample and investigate whether the financial crisis has affected the level of competition. Table 5 presents estimates of the H-statistics developed in Claessens and Laeven (2004) for our sample of banks. We report both the point estimate of the H-statistic and the standard deviation of the H-statistic. We also report a test of perfect competition (i.e., the H- statistic is equal to one) and monopoly (i.e., the H-statistic equals zero). We do not have enough observations for each country to compute the market competition measure developed in Claessens and Laeven (2004) for each country and year. Panel A of Table 5 presents the estimates of the H-statistic when we estimate the model described in section 2 and in more detail in Claessens and Laeveen (2004) for each year using pooled OLS across all countries and include country fixed-effects. 20

22 On average, we find that the banking systems in our sample are not perfectly competitive but rather display oligopolistic competition. Interestingly, the banking systems were more competitive on average in 1994, prior to the financial crisis, than today. The H-statistic in 1994 was about 0.83 on average, much higher than in 2004 when the H-statistic averaged only We also find that banking systems were least competitive during the height of the financial crisis in 1998, when the H-statistic averaged only Since the crisis in 1998, competition has increased but has yet to reach pre-crisis levels. Next, we study cross-country variation in the level of bank competition. Panel B reports estimates of H-statistics by country based on estimating the model for each country using pooled OLS across years and including year fixed-effects. In panel C, we estimate the model for each country using polled OLS across years and include fixed bank effects as well as fixed year effects. We find that the Korean banking system is most competitive, with an H-statistic of about and not significantly different from one, closely followed by Singapore and Hong Kong. The banking systems of Indonesia, the Philippines and especially Thailand are the least competitive. Malaysia s banking system is somewhere in the middle in terms of competition. Taking the estimates of the implicit deposit insurance measure of risk and the H- statistic measure of competition together, the results suggest that the banking systems of Hong Kong and Singapore are both stable and competitive, while the banking system of Indonesia still embeds a lot of risk and is not very competitive. In the remainder of this section of the paper, we analyze the relationship between bank performance, diversity of bank activities, bank ownership, and regulations. In Table 21

23 6, we report OLS regressions with as dependent variable either the simple ratio of operating income to total assets or the activity-adjusted ratio of operating income to total assets. The difference between the two indicators of bank performance is explained in section 2 of this paper and described in more detail in Laeven and Levine (2005). The first four columns in Table 6 present results where the dependent variable is the simple ratio of operating income to total assets. We include both the diversity measure (Income diversity or Asset diversity) and an activity measure (Net interest income to total operating income or Loans to total earning assets). We include the activity measure to control for the mixture of activities conducted by each bank and to therefore identify the relationship between valuation and diversity per se. We estimates the regressions either for the year 2004 (columns (1) and (3)) or for the period (columns (2) and (4)). As in Laeven and Levine (2005), we find a diversification discount: the coefficient on the income diversity and asset diversity variables enters negatively and significantly (columns (1) to (4) in Table 6). Operating income of banks that engage in multiple activities is much lower than if those banks were broken-up into financial intermediaries that specialize in the individual activities. The results are consistent with the view that diversification intensifies agency problems in financial conglomerates with adverse implications on performance and these costs to diversification outweigh any benefits accruing from economies of scope. Nevertheless, because we do not directly measure agency problems, we cannot unequivocally conclude that intensified agency problems in financial conglomerates drive the results. We can more confidently argue that economies of scope are not sufficiently large to produce a diversification premium. 22

24 Next, we investigate the robustness of the diversification discount in banks to controlling for bank-level and country-level characteristics. As dependent variable, we use excess performance as described in section 2 of the paper. We only present regression results that focus on income diversity (columns (5) to (8) of Table 6) but find similar results when using asset diversity instead. We control for a number of bank-level traits and also use country fixed-effects. The regressions are estimated for the year When we control for numerous bank-level traits in Table 6, we continue to find a negative, significant relationship between measures of the diversity of bank activities and the performance of the bank. First, size is often thought to affect performance through economies of scale. We therefore control for the logarithm of total assets (column (6)). Furthermore, we also include the logarithm of total operating income as an alternative measure of bank size (column (7)). Total operating income may better capture the importance of a bank s off-balance sheet items. While the logarithm of total operating income enters the valuation regressions positively and significantly, we continue to find that diversity is associated with lower valuation. Second, competition in the product market may influence the governance of banks, so that omitting information on the structure of the banking industry may lead to inappropriate inferences regarding the relationship between performance and diversity. Toward this end, we include each bank s market share of deposits as an indicator of the degree of competition facing the bank. Banks with a large market share may exert market power and enjoy correspondingly higher performance. We find no evidence of this. 23

25 Third, we include the ratio of total deposits to total liabilities (Deposits/Liabilities). To the extent that a higher Deposits/Liabilities ratio implies that the bank has access to low cost, subsidized funding (deposits generally being an inexpensive source of funding and deposits generally enjoying government subsidized insurance), then a higher Deposits/Liabilities ratio might signal higher valuations. Fourth, we control for the book value capitalization of the bank (Equity/Assets). A well-capitalized bank may have fewer incentives to engage in excessive risk-taking. If this were the case, we would expect a positive correlation between the ratio of book value of equity to total assets (Equity/Assets) and our excess performance measure. We find that this is the case. Equity/Assets enters with a positive and statistically significant coefficient. Fifth, we control for past performance by including the lag of the growth in total operating income. Past performance is commonly used as a proxy for growth opportunities. We indeed find a strong relationship between current and past performance. When including these variables, however, income and asset diversity still enter negatively and significantly: There is still a significant diversification discount. In Table 7 we control for bank ownership. In columns (1) to (4) we estimate the performance regression for the subset of banks that belong to one of the following ownership categories: state, foreign state, private domestic, and foreign. We find a diversification discount for all four groups of banks. The diversification discount is somewhat larger for domestic banks than for foreign banks. We also find that differences in equity capitalization explain more of the variation in bank performance for state banks and foreign banks than for private domestic banks. 24

26 In columns (5) to (7), we control for the share of private domestic ownership and the share of foreign ownership (both measured in terms of total assets). The default category is state banks. We find that private domestic banks perform slightly better than state-owned banks in East Asia, although the difference is not statistically significant. Foreign-owned banks on the other perform much better and the difference is statistically significant. We find roughly a one-to-one correspondence between increases in foreign ownership and increases in bank performance. So, if foreign ownership were to increase by 10%, then the performance ratio would also increase by about 10%. This is a large effect compared to the average ratio of operating income to total assets in the sample of about 4%. In Table 8 we also include country-level measures of banking sector regulations other government regulations, and a measure of market concentration. All of the countrylevel controls vary over time but in the reported regression we only use data for the year We find similar results when estimating the regressions over the period Specifically, in the first three columns we include a measure of government intervention in banking (including regulatory restrictions on banks and state ownership of banks) from the Heritage Foundation. We find that banking systems with less government interventions (and that are less heavily regulated) perform better. The effect is statistically significant. In columns (4) to (6), we also include other dimensions of the economic freedom index computed by the Heritage Foundation, including a measure of the fiscal burden of government (Fiscal policy), a measure of the effectiveness and independence of monetary policy (Monetary policy), a measure of wage development and price inflation (Price control), and a measure of the protection of property rights 25

27 (Property rights). Like the banking policy variable, all of these indexes are constructed such that higher values denote more economic freedom. When we control for other dimensions of economic freedom, banking sector policy does no longer enter significantly. We find that fiscal and monetary policies are the most highly correlated with bank performance. In columns (7) to (9), we also include the 3-bank concentration ratio (measured in terms of deposits) to control for the market structure of the banking system. We find that banks in more concentrated banking systems generate more income, possibly because they can extract more rents. Of the other country-level traits considered, fiscal policy has the largest effect on bank performance, followed by monetary policy, banking policy, and price controls. Property rights do not appear correlated with bank performance once we control for these other country characteristics. Of course, these results should be interpreted with caution because some of the country level characteristics are highly correlated. 5. Conclusions We study the effect of ownership, diversity of activities, and government policy on the performance of banks in East Asia. We find that foreign banks perform significantly better than domestic banks. Nevertheless, banking systems have been slow to open up to foreigners. The results in this paper suggest that foreign ownership should be encouraged and call for a revision of current policy adopted by countries on this topic. We also find that some of the banking systems the Indonesian banking system in particular are not very competitive and that competition is generally still at lower 26

28 levels than prior to the financial crisis in While our calculations suggest that competition has improved somewhat since the crisis in most countries, much remains to be done in this area. The entry of foreign banks may be one way to put competitive pressure on local banks. Further consolidation of local banks does not seem warranted. Bank concentration ratios are already at par with the world average and banks in the more concentrated markets seem to generate excessive rents. This suggests that existing banks should grow by improving the quality of their services rather than through further consolidation. Finally, we find that improvements in the area of fiscal and monetary policy are equally important and needed to enhance banking sector stability and performance. 27

29 References: Barth, James, Gerard Caprio, and Ross Levine (2001). Bank Regulation and Supervision: A New Database, Policy Research Working Paper, World Bank. Bongini, Paola, Luc Laeven, and Giovanni Majnoni (2002). How Good is the Market at Assessing Bank Fragility? A Horse Race Between Different Indicators, Journal of Banking and Finance 26(5), Calomiris, Charles, Daniela Klingebiel and Luc Laeven (2005), Financial Crisis Policies and Resolution Mechanisms: A Taxonomy from Cross-Country Experience, in: Patrick Honohan and Luc Laeven (eds.), Systemic Financial Distress: Containment and Resolution, Cambridge: Cambridge University Press. Caprio, Gerard, Luc Laeven, and Ross Levine (2004). Governance and Bank Valuation, Policy Research Working Paper No. 3202, World Bank. Claessens, Stijn, Daniela Klingebiel and Luc Laeven (2005), Crisis Resolution, Policies, and Institutions: Empirical Evidence, in: Patrick Honohan and Luc Laeven (eds.), Systemic Financial Distress: Containment and Resolution, Cambridge: Cambridge University Press. Claessens, Stijn and Luc Laeven (2004). What Drives Bank Competition? Some International Evidence, Journal of Money, Credit, and Banking 36(3), Claessens, Stijn and Luc Laeven (2005). Financial Sector Competition, Financial Dependence, and Growth, Journal of the European Economic Association 3(1), Demirgüç-Kunt, Asli, Baybars Karacaovali, and Luc Laeven, (2005). Deposit Insurance around the World: A Comprehensive Database, Policy Research Working Paper 3628, Washington, DC: World Bank. Demirgüç-Kunt, Asli, Luc Laeven, and Ross Levine (2004), Regulations, Market Structure, Institutions, and the Cost of Financial Intermediation, Journal of Money, Credit, and Banking 36(3), Jensen, Michael C., Agency costs of free cash flow, corporate finance, and takeovers. American Economic Review 76, Jensen, Michael C. and William H. Meckling, Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics 3,