Foreign Banking in Developing Countries; Origin Matters

Transcription

1 Foreign Banking in Developing Countries; Origin Matters Neeltje Van Horen * May 2006 Abstract Driven by globalization and increased financial integration, the last decade has seen many foreign banks entering developing countries. Although the majority of these banks are from high-income countries, recently banks from developing countries have followed suit. This paper looks at this phenomenon, by examining the differences and similarities between developing and high-income country foreign banks. Using a large dataset on banking sector FDI in developing countries, we find that 27 percent of all foreign banks in developing countries are owned by a bank from another developing country, while these banks hold 5 percent of the foreign assets. The importance of developing country foreign banks is much larger in low-income countries (both in number of banks and in terms of assets) and this type of foreign banking is strongly regionally concentrated. Although foreign bank entry by both developing country as well as high-income country banks seems to be driven by economic integration, common language and proximity, banks from developing countries are more likely to invest in small developing countries with weak institutions where high-income country banks are reluctant to go. This result seems to suggest that developing country banks have a competitive advantage dealing with countries with a weak institutional climate. Furthermore, our results indicate that developing country foreign banks have a higher interest margin and are less profitable than foreign banks from high-income countries. JEL Classification Codes: F21, F23, G21 Keywords: foreign direct investment, international banking * Van Horen is with the World Bank and the University of Amsterdam. I am grateful to Stijn Claessens for helpful comments and to Allaeddin Twebti, Siret Dinc, Matias Gutierrez, and Haocong Ren for their help with collecting the data. Financial support for this project by United Kingdom's Department for International Development received through DECRG trade and services project is gratefully acknowledged. The views expressed in this paper are those of the author and do not necessarily represent those of the institutions with which affiliated. address: nvanhoren@worldbank.org.

2 1. Introduction A number of factors have recently led to an increase in FDI in the banking sector of developing countries (World Bank, 2006). These include global economic trends such as an increase in trade and FDI flows, with which the internationalization of banking traditionally has been closely tied. In addition, rules governing cross-border lending and the establishment of branches and subsidiaries by foreign banks have been eased and many state-owned banks have been sold, increasing opportunities for investment in developing countries. Furthermore, advances in telecommunications and information technology have enabled banks and other financial institutions to better manage crossborder activities. Moreover, banks have expanded cross-border activities to serve a growing number of expatriates. Bolstered by these developments, many foreign banks have, in the last decade, decided to enter developing countries. The majority of those banks comes from high income, industrialized countries. However, more recently, developing country banks have also started to expand across borders into other developing countries. For example, regional trade agreements and the Central American Free Trade Agreement (CAFTA) have stimulated a number of Central American banks (e.g., Panama s Banistmo, El Salvador s Banco Cuscatlan) to expand into other Central American retail financial markets. South Africa s increased integration with the region, both in terms of trade as well as in terms of investment, has been a driving factor behind South Africa s Standard Bank s increased presence in southern and eastern Africa. Furthermore, Pakistan s Habib Bank has targeted a well-established customer base of expatriates through its branch network in South Asia. (World Bank, 2006). Banks engage in foreign lending for several reasons. For example, it provides a possibility for risk diversification. In addition, entering new markets can increase the bank s client base. Furthermore, by following their international customers in order to provide them with informational financial services banks can exploit informational advantages derived from long-term bank-client relationships (see Petersen and Rajan, 1994 and Rajan, 1998). 1

3 In the literature several factors have been identified that can explain why a bank goes abroad and enters a specific country. First, several studies have found evidence that foreign direct investment in banking is correlated with the degree of bilateral trade and FDI between host and source country, indicating that the internalization of banks is closely tied to the internalization of non-financial firms (Grosse and Goldberg, 1991, Brealey and Kaplanis, 1996, Williams, 1998, and Yamori, 1998). Second, banks, presumably, engage in foreign entry to increase the bank s profitability, within an acceptable risk profile and risk diversification goals. Indeed, host and source countries characteristics related to profitability and risks have been found to be important drivers of a bank s decision to penetrate a foreign market. For example, Focarelli and Pozzolo (2000) find that banks prefer to have subsidiaries in countries were expected profits are larger, owing to higher expected economic growth and the prospect of reducing local banks inefficiency. And Galindo, Micco, and Serra (2003) find that unifying rules and regulations and converging towards international institutional standards have a positive impact on foreign bank penetration, presumably as it lessens the costs of operating in the host country. This literature, however, implicitly assumed that the behavior of foreign banks is independent of the country of residence of the foreign owner. However, anecdotal evidence suggests that developing country banks are attracted to different countries than high-income country banks (World Bank, 2006). This might be purely the result of differences in trade and FDI linkages, however other factors might also play a role. Developing country banks are more familiar with working in challenging investment climates. Greater familiarity means you can take more risks, which suggests that developing country banks might be more likely to venture in countries with a weak investment climate. Besides possible differences in factors driving foreign bank entry, developing country and high-income country foreign banks might also differ in other respects, most notably their efficiency and profitability. Demirguc-Kunt, and Huizinga (1999) found that a bank s efficiency and profitability are affected by a variety of determinants, including bank characteristics like ownership, and macroeconomic conditions. If developing country and high-income country foreign banks differ in their client base, product mix 2

4 and countries they invest in, it is possible that there exist substantial differences between the two types of foreign banks with respect to their efficiency and profitability. As policymakers have an interest in promoting banking sectors that are both stable (i.e. were banking profitability is sufficiently high) and efficient, it is useful to examine whether significant differences exist. This paper attempts to shed a light on the differences between high-income country banks entering a developing country (so-called North-South foreign banking) and a developing country bank entering a developing country (South-South foreign banking), by documenting the characteristics of these two types of foreign banks, by testing whether the determinants of foreign bank entry differ between them, and by examining whether high-income and developing country banks differ in terms of efficiency and profitability. In order to do so we first construct a database covering most banks in all developing countries, including their ownership and the source country of all foreign owners. We find that in terms of number of banks 27 percent of foreign banks in developing countries is owned by developing country banks, in terms of assets developing country foreign ownership amounts to 5 percent. Furthermore, South-South foreign banking is especially significant in low-income countries and is almost entirely intra-regional. Comparing the balance sheets of developing country and high-income country foreign banks we find that the asset mix is very similar, but there are substantial differences in the composition of funding between the two types of foreign banks. Overall we find support for the conclusion in the literature that FDI in foreign banking is strongly related to economic integration, common language and proximity; this holds true for both high-income and developing country banks. More interesting, it appears that developing country banks are more likely to invest in small developing countries with weak institutions, where high-income country banks are reluctant to go. These results indicate that FDI decisions are not so much influenced by the absolute amount of risk faced by firms, but rather by a given firm s ability to bear that risk better than other investors. Looking at the efficiency and profitability of the two types of foreign banks we find evidence that developing country foreign banks have a higher interest margin and are less profitable than high-income country foreign banks. 3

5 The rest of the paper is structured as follows. In Section 2 we document the characteristics of South-South foreign banking relative to North-South foreign banking. In the next section we investigate the similarities and differences in determinants of foreign bank entry for high-income and developing country banks. In Section 4 we examine whether the two types of foreign banks differ in their efficiency and profitability. The last section concludes. 2. Importance of developing country banks as sources of FDI in banking sector Driven by globalization and increased financial integration, the last decade has seen many foreign banks entering developing countries. Foreign banks now account for 35 percent of the banking sector of developing countries in terms of number of banks (based on Bankscope data, see Appendix for full description of the database). Moreover, total foreign assets in developing countries amounted to 944 billion, or 16.2 percent of total bank assets in developing countries. 1 However, foreign bank participation in developing countries differs largely between countries. Table 1 shows the foreign participation in the banking sector of 102 developing countries. While a few countries have remained closed to foreign participation, the vast majority of developing countries has had some foreign bank activity, varying from less then one percent (China) to 100 percent (Madagascar and Mozambique) in terms of assets, with a standard deviation of 31 percent. In the majority of countries (58), the share of foreign owned banks is higher in terms of number of banks than in terms of assets indicating that many of the foreign banks have been relatively small. However, while this is the case for all countries in the Middle East and Northern Africa and also in South Asia, in the majority of countries in Sub-Saharan Africa and in Europe and Central Asia foreign bank participation is higher in number of assets. This indicates that in those two regions many of the largest banks are foreign owned, possibly resulting from historical, including colonial, links as well as proximity to potential investors. This finding is, for Sub-Saharan Africa, confirmed by Claessens and Lee (2002). 1 Assets are based on average. 4

6 Foreign ownership in developing countries, however, does not solely result from high-income country banks investing in a developing country (North-South foreign banking) but also from investment by banks from other developing countries (South- South foreign banking). Of all foreign banks in developing countries, 27 percent is owned by a bank from another developing country. In terms of assets, the magnitude of South- South foreign banking is much smaller. Of total foreign bank assets in developing countries, 5 percent is held by banks from other developing countries. In total there are banks from 48 developing countries that have invested in the banking sector of other developing countries. As expected the vast majority of these countries are (upper) middle income countries (41), which hold 90 percent of the total foreign assets held by banks from developing countries. Low income countries that invest in other developing countries are Benin, India, Kenya, Nicaragua, Nigeria, Pakistan, and Togo. Excluding Panama (an important offshore center), the biggest investors in other developing countries are South-Africa, Malaysia and Hungary. From Table 1, however, it is clear that the importance of developing country banks as investors in the banking sector of other developing countries differs substantially per country. While in some countries (Algeria, Antigua and Barbuda, Burundi, Hungary, Jordan, FYR of Macedonia, Mauritania, Mexico, Mozambique, Pakistan, Poland, Serbia and Montenegro, Seychelles, Thailand, Togo, Trinidad and Tobago, and Turkey) none of the foreign banks are owned by banks from other developing countries, in others (Azerbaijan, Benin, Cambodia, Costa Rica, Kyrgyz Republic, Namibia, and Nicaragua) only developing countries banks have invested in the banking sector. In 74 out of the 102 developing countries in our sample at least one foreign bank is owned by a bank from another developing country. In terms of assets South foreign ownership is smaller than in terms of banks in 76 percent of the countries, indicating that developing country foreign banks tend to be smaller than their highincome country counterparts. In fact, foreign banks in developing countries owned by high-income country banks have median assets of 361 million while foreign banks owned by developing country banks 92 million. While these data indicate that participation by developing country banks is significant, it is important to keep this in perspective. Banks from high-income countries 5

7 accounted for 95 percent of total foreign bank assets in developing countries. Moreover, foreign banks in developing countries account for only 16 percent of total banking sector assets. South-South bank ownership thus accounts for less than one percent of total bank assets in developing countries. Table 2 groups countries according to their income level and region. Foreign ownership is calculated both as the simple average of the foreign ownership of the individual countries (country-based share) and as the total foreign ownership for the group as a share of the total banking assets (total number of banks) in the group (groupbased share). Comparing foreign participation at the different income levels we see that, while upper middle income countries have the highest average share of foreign bank penetration both in terms of assets as well as in terms of number of banks, the differences with low income and lower middle income countries are small. This equality in the share of foreign ownership might be caused by the fact that the different groups can attract FDI for different reasons. Upper middle income countries might be attractive for foreign investors that seek new markets and look for countries with relative well developed institutions but strong growth potential, while a number of low income countries have throughout history received FDI because of their colonial ties. At the same time, low income countries might be attractive for banks from other developing countries that have a harder time competing with banks from industrialized countries in the upper middle income countries. Indeed, when we look at the South share in foreign banking we see that the average cross-border investment by developing country banks is more significant in low-income countries (35 percent in terms of assets and 46 percent in terms of number of banks) than in lower middle income countries (25 and 31 percent) and in upper middle income countries (12 and 24 percent). Foreign ownership also differs substantially per region. The regions with the highest percentage of foreign banks are Europe and Central Asia and Sub-Saharan Africa (47 percent), followed by Latin America and the Caribbean (40 percent), East Asia and Pacific (24 percent) and Middle East and Northern Africa (20 percent). The region with the lowest percentage of foreign ownership is South Asia, were, on average, only 6 percent of the banks is foreign owned. Partly reflecting the negative correlation between the income level in a country and the share of South foreign ownership, the region with 6

8 the highest level of foreign ownership by developing country banks is Sub-Saharan Africa (33 percent in terms of assets and 45 percent in terms of number of banks). The region with the lowest level of South foreign ownership is Middle East and Northern Africa (only 9 percent in terms of assets and 18 percent in terms of number of banks). Note that the relative high share of South foreign ownership in South-Asia almost entirely reflects banks domiciled in Mauritius (an offshore banking center) owned by banks from industrialized countries, that have set up subsidiaries in India. The shares calculated on a group basis are different from the simple averages of the individual countries, especially when looking at the asset shares. For example, the group based share of foreign banks in low income countries is only 6 percent while the average country share is 41 percent. This reflects the fact that the Indian banking sector is very large compared to the bank sector of other low income countries but its share of foreign participation is at the same time really small. In the group of lower middle income countries a similar difference is, for the same reasons, caused by the strong influence of China in the group-based measures. The lower South share that we find in most cases for the group-based measure compared to the country-based measure, indicates that foreign banks owned by banks in other developing countries are more represented in the smaller markets. This suggests that industrialized country banks are more attracted to the larger markets while developing country banks tend to invest in the smaller developing countries. In order to determine whether South-South foreign banking is regionally concentrated or not, we calculated how much of the foreign bank assets in each region are owned by banks located in each of the six regions and by banks from high-income countries (see Table 3). The results show that for all regions, except South-Asia, most if not all South-South foreign banking is intraregional. 2 For example in East Asia and Pacific over 6 percent of the investors in the banking sector are developing country banks and they all come from within the region. After South-Asia, Sub-Saharan Africa is the 2 Note that the data for South-Asia, as mentioned before, reflect banks domiciled in Mauritius (on offshore banking center), most of which are owned by banks from high-income countries that have set up subsidiaries in India. When excluding offshore banking centers from the sample, foreign investment by developing country banks in South-Asia is very small and entirely intra-regional. 7

9 region that receives the highest share of its FDI from developing country banks, with the largest part (85 percent) coming from within the region. The dominance of intraregional foreign banking in part reflects the importance of intraregional trade and FDI flows and the priority being given to regional cooperation and integration in policy agendas. In addition, geographic proximity often implies a common cultural heritage, language and ethnic ties, making it easier for banks to assume more risk. Just as local banks have an advantage over foreign banks due to their greater knowledge of local conditions and their ability to screen and monitor local borrowers (Nini, 2004), foreign banks from within the same geographic region have an advantage over other nonlocal lenders. The fact that both banks from developing countries as well as from high-income countries operate in developing countries raises questions about whether both types of foreign banks behave differently in their lending activities, their forms of income and expenses, and their efficiency and profitability (the last two will be discussed in more detail in Section 4). Table 4 provides a comparison of the assets and the funding of foreign banks from developing and from high-income countries, including the results of t-tests to determine whether differences between the two types of foreign banks are statistically significant. 3 The individual variables (averages of their values between 2000 and 2004) are calculated for each bank in each country and then averaged for the two types of foreign banks separately within a given group. A few developing countries that are offshore centers (Barbados, Lebanon, Mauritius, and Panama) are excluded from our sample. In addition, banks owned by a bank headquartered in an offshore center are also not included. The results show that the composition of assets is not very different between foreign banks owned by banks from developing country and foreign banks owned by banks from high-income countries, except in terms of fixed assets. South foreign banks appear to have a higher share of fixed assets, especially when they have invested in upper-middle income countries. In addition, in the countries in Europe and Central Asia a 3 Note that one has to be careful when comparing the asset and liability composition of banks as data definitions and quality might vary across banks and across countries. Furthermore, the t-test could not be performed for South-Asia as only one developing country bank has invested in this region. 8

10 smaller share of total assets of southern foreign banks is made up by loans, while a larger part of assets reflects other earning assets and fixed assets. The composition of the funding of the two types of foreign banks, however, is substantially different. Foreign banks owned by developing country banks tend to have lower non-interest bearing liabilities and also other funding is lower. At the same time, the loan loss reserve is significantly higher for those foreign banks. Those differences are especially strong in countries in Latin America and the Caribbean. 3. Does bank origin affect determinants of foreign bank entry? The characteristics of South-South foreign banking as opposed to North-South foreign banking described in the previous section, suggest that high-income country banks and developing country banks might differ in their preferred location to invest in. The empirical literature on the determinants of foreign bank entry, however, has not distinguished between foreign ownership by banks from high-income countries and from developing countries (see, for example, Buch and DeLong, 2004, Focarelli and Pozzolo, 2000, and Galindo et al., 2003). In this section we test formally whether the determinants of foreign bank entry differ between the two types of foreign banks, by estimating a model of decisions by foreign banks to enter developing country markets. Empirical methodology Following recent developments in the FDI literature (see, for example, Galindo et al., 2003, and Di Giovanni, 2005) we estimate a gravity model of bilateral FDI in banking. The model in its most simple format explains bilateral flows (i.e. trade or investment) between two countries on the basis of the product of the GDPs of both countries, and the distance between them. The model is typically extended by including dummies that indicate whether the two countries share a common border, common language, past colonial links, possibly time-zone and other geographical information etc. In order to determine whether high-income country banks invest in developing countries for different reasons than developing country banks do, we need to use bilateral data. As described in the Appendix, our data include 102 host countries. We exclude the 9

11 developing countries that are offshore centers from our sample, as decisions to enter those markets are often driven by tax incentives. This leaves us with a sample of 98 host countries. For all foreign banks in those countries, we determine the country of residence of the owner, or the so-called source country. For each host country, we construct country-pairs with all possible source countries in the sample. We restrict the source countries to all high-income and developing countries that have a presence in the banking sector of at least one developing country. This to avoid a bias in the estimation due to the fact that some potential source countries might have capital account restrictions or other economic or institutional factors that make it impossible for their banks to expand to other countries and for which we cannot easily control. In addition source countries that are offshore centers are excluded. This leaves us with a total of 6,382 country-pair observations. As dependent variable we use two measures of overall foreign bank penetration. The first measure of foreign bank penetration is based on banks assets. For each host country, we determine per source country the sum of assets of foreign owned banks and divide that by the total amount of bank assets in the host country. Bank assets come from Bankscope and are based on the consolidated balance sheets. 4 The second measure captures foreign bank penetration related to the number of banks. For each host country, we determine per source country the number of foreign owned banks and divide that by the total number of banks in the host country. The model includes as explanatory variables a number of variables that have been identified in the literature to explain foreign bank entry. To determine whether certain determinants of foreign bank entry affect developing country banks differently we create a dummy that is one if both host and source country are a developing country and zero otherwise and interact the variables with this, so-called, South dummy. We include a number of standard gravity model variables like common border, common language, past colonial links and distance between the host and source countries. These variables have proven to have explanatory power as drivers of foreign bank entry (see Galindo et al., 2003). A number of studies have shown that foreign direct investment in banking is 4 In order to minimize the effects of particular events, data on bank assets are averages of annual values from

12 correlated with economic integration as measured by trade and FDI flows between host and source country (Grosse and Goldberg, 1991, Brealey and Kaplanis, 1996, Williams, 1998, and Yamori, 1998), so we include as an explanatory variable bilateral trade. 5 In addition, the model includes a variable that captures differences in the origins of the legal system in host and source country. Galindo et al. (2003) found that banks are more willing to locate in foreign countries with which they share legal origin, presumably as operating costs and risks in such countries are lower. Economic size (as measured by a country s GDP) is a standard variable included in gravity models. In our case this variable is of particular interest as the statistics in the previous section indicate that banks from developing countries tend to enter the economic small countries while banks from high-income countries tend to target the large countries. Including this variable in the model allows us to test whether this result still holds when other determinants of foreign bank entry are controlled for. The impact of the depth of the financial system on foreign bank entry can potentially go two ways. On the one hand, a well developed financial system suggests a better operating environment, making entry more likely. At the same time, a well developed system may make for fewer opportunities for foreign banks to expand into and make profits. In order to identify whether foreign banks from developing countries and those from industrialized countries are influenced differently by the depth of the financial system, we include this variable. In addition, we include a measure of quality of institutions in the host country. The role of institutions in FDI location has received some attention in recent years (see, for example, Wei, 1997 and 2000, Galindo et al., 2003, and Stein and Daude, 2004). In general, studies find that quality of institutions has a positive effect on FDI. 6 This impact, however, might not hold for multinationals from all potential source countries. FDI decisions are not so much influenced by the absolute amount of risk that a firm is facing but its ability to bear it better than anyone else. Greater familiarity means that you can 5 Another useful indicator would be bilateral FDI flows, but data are not available, and FDI flows are highly correlated with trade flows, which is included in the model. 6 Galindo et al. (2003) do not look at the quality of institutions in the host country as a determinant for foreign direct investment in the banking sector of several industrialized and (mostly middle income) developing countries, but at the differences in institutional quality between host and source country. They find that similarity in institutional quality has a positive effect on foreign entry. 11

13 take on more risks, which would suggest that developing country banks, being more familiar with working in domestic environments where institutional development is low, might be attracted to countries with weak institutions in which high-income country banks are reluctant to invest. Finally we include some additional control variables, which are not interacted with the South dummy. These include a dummy that captures whether foreign bank entry is restricted in the host country. Furthermore, to capture differences in the magnitude of foreign bank entry in different regions of the developing world, we also include regional dummies. We did not use country dummies to capture source country specific effects. Since FDI takes only place in a small portion of our country pairs, adding source country dummies could potentially create a bias in our estimations. To capture some of the source country characteristics that could explain the decision to enter a foreign market, we include, besides the size of the source country economy also its GDP per capita. To limit the possibility that our results are affected by endogeneity, we use for all explanatory variables that are time-varying only their levels in For a complete description of all variables in the model, see the appendix Table 1. Our benchmark model is as follows: FC ij ' = α * South + γ Z + ε (1) ' ' ' ' 0 + α1bij + α 2Bij South + β1 X i + β 2 X i * 1 ij where i refers to the host country and j to the source country. FC ij is our dependent variable which is based on assets or number of banks. B ij is a vector of variables describing the relationship between host and source countries (including colonial linkages, common border, common language, distance between host and source countries, bilateral trade flows, and the difference in legal system). X i is a vector of variables describing host country characteristics that might have a different impact on the decision to penetrate a certain host county for developing country banks as compared to high-income country banks (these include GDP, depth of the financial sector and institutional quality). Z is a vector of additional control variables. We estimate our 12

14 gravity model using Tobit to account for the many zeros in the dependent variable. In addition, the standard errors are corrected for heteroskedasticity. Results Using regression model (1) we test whether foreign bank penetration in developing countries by developing country banks is affected by different factors than entry by banks from high-income countries. The results are presented in Table 5. To aid the economic interpretation we show, instead of parameter estimates, the marginal effects of the unconditional expected value of the dependent variable, E(y), where y=max(a, min(y*,b)) where a is the lower limit for left censoring (0) and b is the upper limit for right censoring (100). The marginal effects are calculated at the mean of the independent variable, except when the independent variable is a dummy in which case the marginal effect is calculated as the change in the dummy variable from 0 to 1. The marginal effects capture the combined effect of the impact of the explanatory variable on the probability of entering the host country as well as on the amount of FDI. The mean of the dependent variable is equal to 0.57 percent when based on assets and 0.54 percent when based on banks. Table 5 reveals some important similarities and differences between the determinants of foreign investment in the banking sector of developing countries by highincome country and developing country banks. Colonial links between source and host country have a strong positive impact on foreign control, but less so for developing countries when foreign control is measured by assets. The impact of colonial links on foreign bank presence is economically very relevant. For high-income country banks, when colonial links exist between the host country and the source country foreign control in terms of assets will increase with 0.76 percent, while for developing country banks it will increase with 0.70 percent. This is a substantial increase considering that the mean of foreign control in all host countries is 0.57 percent. In other words, when colonial ties exist between host and source country, both high-income and developing country foreign banks are willing to more than double their presence in terms of assets. Interestingly, in terms of number of banks the increase is substantially less suggesting that banks that enter developing countries because of colonial links are on average large banks. 13

15 For both developing country as well as high-income country banks, a common language, which reduces the cost of foreign banking, is a significant determinant of foreign bank entry, while distance is negatively related to foreign banking, albeit less for developing country banks. After controlling for distance, a common border is not a significant determinant anymore. FDI by both types of foreign banks is strongly related to bilateral trade flows, indicating that banks tend to follow their customers. At the same time, banks from high-income and developing countries are equally likely to be deterred from entering a developing country with a different legal system, presumably because costs of entering are higher when the legal systems of host and source country are from different origin. The highly significant and opposite signs for GDP in the host country indicate that banks from industrialized countries tend to go to large developing countries, while banks from developing countries tend to enter the smaller developing countries, confirming the results found in the previous section. The results show that a move from the economically smallest country (Seychelles) to the largest (China) will increase foreign control in assets with 0.30 percent and in terms of number of banks with 0.24 percent, or an increase of 53 and 44 percent respectively. 7 For developing country banks, on the other hand, that same move will lead to a drop in foreign control of 12 percent in terms of assets and 10 percent in terms of number of banks. At the same time, the depth of the financial sector is negatively correlated with foreign ownership by high-income country banks, but positively with ownership by developing country banks. Implying that foreign control in assets (banks) decreases with 28 (21) percent for high-income country banks, when moving from the country with the least developed financial sector (El Salvador) to the country with the most developed financial sector (China), while it increases with 5 (7) percent for developing country banks. After controlling for all of the above determinants of FDI, the quality of institutions does not appear to influence the decision by high-income country banks to enter a developing country. However, banks from developing countries are more likely to enter developing countries with weak institutions. For those banks, a move from the host 7 The difference in economic size between the largest and the smallest country is 7.46; this multiplied by the marginal effect implies a change in foreign control of 0.30 and 0.24 percent respectively. 14

16 country with the lowest institutional quality (Democratic Republic of Congo) to the host country with the highest institutional quality (Chile) would increase banks share in foreign markets with 0.20 percent in terms of assets and 0.19 percent in terms of number of banks, or an increase in presence of 35 percent in both cases. This suggests that banks from developing countries are willing to enter countries in which individual country banks are reluctant to invest, possibly because they have a competitive advantage doing business in those countries. Overall we find support for the conclusion in the literature that FDI in foreign banking is strongly related to economic integration, common language and proximity; this holds true for both high-income and developing country banks. More interesting, it appears that developing country banks are more likely to invest in small developing countries with weak institutions, where high-income country banks are reluctant to go. These results suggest that FDI decisions are not so much influenced by the absolute amount of risk faced by firms, but rather by a given firm s ability to bear that risk better than other investors. 4. Impact of origin on the net interest margin and profitability of foreign banks Policymakers have an interest in promoting banking sectors that are both stable and efficient. The banking sector is thought to be economic efficient when banking spreads are not too large. At the same time stability requires sufficient banking profitability. Differences in client base, product mix and country in which is invested can impact a bank s efficiency and profitability. Although information about differences in client base and product mix between developing country and high-income country foreign banks is currently not available on a systematic basis, the previous sections have shown that differences do exist between the two types of foreign banks with respect to their preferred country to expand into. In this section we therefore examine whether the net interest margins and profitability of foreign banks is dependent on the origin of the foreign owner. 15

17 Empirical methodology To examine whether banks net interest margins and profitability differ between developing country and high-income country foreign banks, we follow closely the procedure used by Demirguc-Kunt and Huizinga (1999). Using bank level data for 80 high income and (mostly middle income) developing countries in the years , they provide an empirical analysis on how bank s efficiency (as proxied by changes in net interest margins) and profitability (measured by the bank s before-tax profit divided by total assets) are affected by a large variety of determinants including bank characteristics, macroeconomic conditions, explicit and implicit bank taxation, deposit insurance regulation, overall financial structure and underlying legal and institutional indicators. 8 Our focus of interest is to determine whether origin of the foreign bank affects the efficiency and profitability of the bank. The benchmark model used by Demirguc-Kunt and Huizinga (1999) includes foreign ownership as a possible determinant of the net interest margin and profitability of a bank. So, we can simply adapt their model by introducing a term interacting foreign ownership with the South dummy introduced in the previous section (i.e. a dummy which is one if the owner of the foreign bank is from a developing country and zero otherwise). This gives us the following model: I ict ' + β B = α + α Ownership 1 ict 0 ' + β B 2 1 ict ic ' * GDPcap + γ X + α Ownership 2 1 ct + γ T t t ic + ε * GDPcap ict ct + α Ownership * South 3 (2) where i refers to the individual bank, c to the country in which the bank is located and t to the time period. I ict is the dependent variable (either the net interest margin or the ratio of profit before tax to total assets). Ownership ic is a dummy which is one if more than 50 percent of the bank s shares are held by foreigners. Note that ownership is 8 Tighter interest margins are frequently viewed as representing greater competition and efficiency. However, as noted by Demirguc-Kunt and Huizinga (1999), a reduction in the net interest margin does not necessarily have to indicate an increase in efficiency. Since the net interest margin reflects the ratio of net interest income (which consists of several components) to total assets (which also consists of several components) it depends on what is driving the reduction in the interest margin whether it truly reflects a drop in efficiency. For example, if the reduction is caused by a decrease in bank taxation (one component of net interest income) it may reflect an improvement in the functioning of the banking system. If, on the other hand, the reduction is caused by a higher loan default rate (another component of net interest income) it might reflect the opposite. 16

18 time invariant, reflecting our assumption that in the period , the period for which we estimate the model, the bank s ownership and the source country of the owner are unchanged relative to the bank s status recorded in Ownership, like all the other bank characteristics, is interacted with GDP per capita of the country to control for the possibility than the income level of the country might affect the impact ownership has on the dependent variable. B ict is a vector of bank characteristics, including the ratio of equity to total assets one period lagged, the ratio of loans to total assets, the ratio of noninterest earning assets to total assets, the ratio of customer and short-term funding to total assets, and the ratio of overhead to total assets. X ct is a vector of country characteristics, including GDP per capita, the growth rate, the inflation rate, the real interest rate, and the real interest rate interacted with GDP per capita. T t are time dummies. The model is estimated using a generalized least squares random effects model. 10 description of all variables in the model, see appendix Table 2. For a complete Results Before turning to the regression results, it is useful to have a closer look at how the mean of bank interest spreads and profitability differs between developing country and high income country foreign banks. We do this by breaking down the banks profit before taxes in its several components, satisfying the following accounting identity: Before tax profit/total assets = net interest margin + (non-interest income - overheads - loan loss provisioning)/total assets 9 Our database only provides ownership information based on the information available in At the same time other bank variables (like total assets) are recorded up till This leaves us with two options. On the one hand, we can estimate the model based on cross-section data, leaving out the time dimension. This is problematic for two reasons. First, due to lags in submission of bank information to Bankscope, many banks do not have any entries for the variables of interest in 2004, substantially decreasing the number of observations. Second, focusing only on 2004 data implies that the results could be strongly affected by particular events. On the other hand, we can exploit the time dimension we have available for all bank characteristics in the model with the exception of ownership and keep the ownership (and the source country in case of foreign ownership) constant over time. Since the time period is relative short and we do account for the year the bank became active, we feel this is the preferred option. 10 We did not include country dummies in the model as this creates a bias due to multicollinearity, which, in our view, is more problematic than the bias created by the possibility that the assumption of no correlation between individual effects and the regressors that underlies the random effects estimation might not hold. However, the general conclusions drawn by this study are independent from inclusion of the country dummies. 17

19 The results are shown in Table 6. The values shown in the table are averages of the period Indicators are calculated for each bank in each country and then averaged for high income (northern) foreign banks and developing country (southern) foreign banks separately within a given grouping. Host countries that are offshore centers are excluded from the sample. Furthermore, foreign banks that are owned by a bank headquartered in an offshore center are also excluded from the sample. The table also reports t-test for statistically significant differences between the two types of foreign banks. We should note that not all banks report all income items in all periods. As a result, averages are taken over samples that may differ for each variable, and country and regional averages themselves do not necessarily satisfy the above accounting identity. Notice that a t-test could not be performed for South-Asia as only one developing country bank has invested in this region. We find that on average net interest margins are higher for southern foreign banks, indicating that foreign banks owned by banks in other developing countries are in general less efficient than high-income country foreign banks. However this result is not consistent when looking at the various groups. In East Asia and Pacific, for example, net interest margins are lower for southern foreign banks. Non-interest income as a percentage of total assets indicates how much the bank engages in non-lending activities such as investment banking and brokerage services, to generate income. We find in general no statistically different values for the two types of foreign banks, except in the upper middle income countries where non-interest income is higher for northern foreign banks. Loan loss provision as a percentage of total assets, which measures the actual provisioning for bad debts over total assets, is higher for foreign banks owned by a bank from a developing country. This might reflect a difference in customer base, where northern banks concentrate on the less risky large multinationals, while southern banks lend to more risky smaller clients. The overhead is on average higher for southern foreign banks, and this result is especially apparent for the developing countries in Europe and Central Asia. This difference might indicate that northern foreign banks engage more in wholesale 18

20 transactions (for which overhead is relative low) as compared to their southern counterparts. The lower tax bill of southern foreign banks in low-income countries and in East Asia and Pacific and Sub-Saharan Africa, might indicate that the activity mix of the two types of foreign banks differs. Also, since banks from developing countries tend to invest in the smaller economies with weak institutions, this lower tax bill can reflect the lack of tax enforcement in those countries. Finally, table 6 shows that in low-income countries, northern foreign banks tend to be more profitable than foreign banks from developing countries. However, after taxes are paid the difference between the two is insignificant. The statistics provided in Table 6, give only a very rough indication of the impact of foreign bank s origin on efficiency and profitability of the bank. A more thorough investigation can be done using model (2). The results are presented in Table 7. The first and fourth columns show the basic specification as used by Demirguc-Kunt and Huizinga (1999). The results indicate that foreign banks realize higher interest margins in developing countries. However, the negative and significant coefficient of the interaction between ownership and per capita GDP, suggests that foreign banks realize higher interest margins especially in poor developing countries. When per capita income is above 4,810 US dollars (which holds for over 50 percent of the countries in our sample) foreign banks tend to have lower interest margins than their domestic counterparts. These results confirm the results found by Demirguc-Kunt and Huizinga (1999). With respect to profitability we find that there is no difference between domestic and foreign banks in the poorest countries. However, when per capita income increases domestic banks tend to be more profitable than their foreign counterparts. In the second and fifth column we included a term interacting ownership with the South dummy. The results suggest that especially developing country foreign banks have higher interest margins compared to domestic banks, not high-income country foreign banks. This suggests that foreign banks from high-income countries are more efficient on average. To determine whether the difference between developing country and highincome country foreign banks is different in poorer countries, we included an additional interaction term (see third column of Table 7). We find no differential impact of southern 19

21 versus northern foreign banks in countries with different income levels. Furthermore, including this variable causes the parameter on the variable interacting ownership with the South dummy to become insignificant. However, since foreign ownership by developing country banks is highly concentrated in low-income developing countries, the negative and significant coefficient for the variable interacting ownership with income level confirms the story that developing country foreign banks have relative high interest margins. At the same time we find that profitability of developing country foreign banks is substantially lower compared to high-income country foreign banks, albeit still higher than domestic banks in the poorest countries (see column 5 of Table 7). Note, however, that this last result does not hold up when including the variable that interacts ownership with the South dummy and income level. In addition, differences in profitability between developing country and high-income country foreign banks seem not to be affected by the income level of the host country. The results presented in this section indicate that origin of the foreign bank does have an impact on the efficiency and (some components of) operational performance of the bank. However, in the absence of systematic information about the clients served and the products provided by foreign banks from developing countries, one can only speculate about the causes of these differences. To understand better the factors underlying the dissimilarities and how they might impact the host countries, further investigation is necessary. 5. Conclusion Increased integration of developing countries with each other in terms of trade, FDI and remittances, in combination with substantial growth in many developing countries, financial sector liberalization and technology advances, have provoked many foreign banks to enter developing countries. Although the majority of these banks are from highincome countries, recently banks from developing countries have followed suit. Currently owning 27 percent of all foreign banks in developing countries, these banks have become a substantial investor in developing countries. This study has attempted to document the 20