Foreign banks: Trends, Impact and Financial Stability

Transcription

1 Foreign banks: Trends, Impact and Financial Stability Stijn Claessens and Neeltje van Horen * February 22 Abstract Over the past two decades, foreign banks have become more important in domestic financial intermediation, heightening the need to understand the behavior of foreign banks better. Using a new, comprehensive database on bank ownership for 37 countries over the period , identifying also the home country of banks, we document substantial increases in foreign bank presence in many countries. We also show large heterogeneity in terms of home and host countries, bilateral patterns, and performance. In terms of impact, we find a negative relation between private credit and foreign bank presence, but only in countries with relatively distant foreign banks. Furthermore, leading up to the crisis, private credit grew faster when foreign bank presence was large, but not in countries with relatively distant foreign banks. In addition, foreign banks reduced credit more compared to domestic banks during the global crisis in countries where they have a small role in financial intermediation, but not so when they were dominant or funded through local deposits. Our results show that accounting for heterogeneity is crucial to better understand the implications of foreign bank ownership. JEL Classification Codes: F2, F23, G2 Keywords: foreign banks, financial globalization, financial sector development, financial stability, financial crisis. * Claessens is with the International Monetary Fund, University of Amsterdam, and CEPR, and Van Horen is with De Nederlandsche Bank (DNB). This paper was partly written while Van Horen was visiting the Research Department of the Federal Reserve Bank of Chicago. We would like to thank Thorsten Beck for valuable comments. We are grateful to Tugba Gurcanlar, Matias Gutierrez, Joaquin Mercado, Deimante Morkunaite, Krisztina Orban, Jeanne Verrier and Chen Yeh for their extensive help with collecting the data and providing excellent research assistance. The data collection was started while the authors were at the World Bank and financial support for this project from the World Bank s Research Support Budget and the United Kingdom's Department for International Development (DECRG trade and services project) is gratefully acknowledged. The views expressed in this paper are those of the authors and do not necessarily represent those of the institutions with which they are or have been affiliated. addresses: sclaessens@imf.org, n.van.horen@dnb.nl.

2 . Introduction Although interrupted by the recent financial crisis, the past two decades have seen an unprecedented degree of globalization, especially in financial services. Not only have crossborder bank (and other capital) flows increased dramatically, but also many banks, from both advanced and developing countries, have ventured abroad and established presence in other countries. Although there are exceptions and regional differences, few countries have been left out from this trend of increased financial integration. Given the importance of foreign banks in many countries, understanding the motivations of foreign banks to enter a particular host country, the mode by which they do so, and the impact they have on financial sector development and lending stability has become essential. These questions have become even more prominent as a result of the financial crisis. Although much research has been conducted, many questions remain unanswered, however, partly because data availability has been limited. This paper contributes to the literature on foreign banking in three ways. First, introducing and using a new and comprehensive database on bank ownership, including the home country of foreign banks, covering 37 countries from 995 to 29, it describes salient facts on trends in foreign bank ownership. Second, it re-examines the link between foreign bank presence and private credit. Third, it re-examines the link between foreign bank presence and financial stability, focusing on the recent financial crisis. We document a sharp increase in foreign bank ownership from 995 to 29 affecting a large number of countries. However, much heterogeneity exists with respect to the relative importance of foreign banks in the host country, the home country of the parent bank, the bilateral patterns, and the business models used by and performance of foreign banks. Our empirical analyses show that the link between foreign bank participation and private credit depends importantly on the average distance between parent bank and host country. Only when this distance is relatively large does a negative correlation between the two exist. In addition, in the years leading up to the crisis private credit growth was faster in countries with a larger presence of foreign banks but only when their parents were located in neighboring countries. When studying the link between foreign bank presence and financial stability, we find that during the global financial crisis foreign banks reduced credit more compared to domestic banks in countries where the level of foreign bank participation is relatively low. However, in countries with high levels of foreign bank participation foreign bank behavior 2

3 did not differ from that of domestic banks. Furthermore, foreign banks that relied more on deposit funding remained a more stable source of credit. Many studies have examined the causes and consequences of foreign bank ownership. Before the crisis, the general consensus was that the benefits of foreign banks greatly outweigh costs in many dimensions. Particularly, it was generally thought that foreign banks add to domestic competition, increase access to financial services, enhance financial and economic performance of their borrowers, and bring greater financial stability (Clarke, Cull, Martinez Peria and Sanchez, 23, Claessens, 26, Chopra, 27, and Cull and Martinez Peria, 2). Generally, lower costs of financial intermediation (measured by margins, spreads, overheads) and lower profitability are documented with greater foreign bank presence (Claessens, Demirguc-Kunt, and Huizinga, 2 and related studies, e.g., Mian, 23, Berger, Clarke, Cull, Klapper and Udell, 25). Also, evidence exists of better quality financial intermediation, e.g., lower loan-loss provisioning with more foreign entry (Martinez- Peria and Mody, 24). Likely a number of factors are behind these effects, such as the introduction of new, more diverse products, greater use of up-to-date technologies, and knowhow spillovers (e.g., as people learn new skills from foreign banks, they migrate over time to domestic banks). In addition, foreign banks likely pressured governments to improve regulation and supervision, increase transparency, and more generally catalyze domestic reform (Levine 996, Dobson, 25, and Mishkin, 26). The effects of the entry of foreign banks on development and efficiency appear to depend though on some conditions. Limited general development and barriers can hinder the effectiveness of foreign banks (Garcia-Herrero and Martinez Peria, 25; Demirguc-Kunt, Laeven and Levine, 24). Also, the relative size of foreign banks presence seems to matter. With more limited entry (as a share of the total host banking system), fewer spillovers seem to arise, suggesting some threshold effect (Claessens and Lee, 23). In terms of individual bank characteristics, it seems that larger foreign banks are associated with greater effects on access to financial services for small and medium-sized enterprises, perhaps as they are more committed to the market, while smaller banks are more niche players (Clarke et al. 25). Furthermore, the health of both the home and the local host bank operation seem to matter, with healthier banks showing better credit growth (Dages, Goldberg and Kinney, 2; see also Haber and Musacchio, 25 and De Haas and Van Lelyveld, 26). While the entry of foreign banks is generally thought to have favorable effects on the development of host banking systems, including through increased credit extension, some studies find more ambiguous results. Some show that foreign banks cherry pick borrowers 3

4 (Detragiache, Gupta, and Tressel, 28; Beck and Martinez Peria, 27). This can undermine overall access to financial services, since cherry picking worsens the remaining credit pool, and lower financial development, especially in low-income countries where relationship lending is important. Indeed, Detragiache, Tressel and Gupta (28) show the presence of foreign banks in low-income countries to be associated with less credit being extended. However, Cull and Martinez Peria (27) show that this relationship disappears or even reverses once crisis-induced acquisition of (distressed) banks by foreigners is accounted for. Besides the impact of foreign banks on the development and efficiency of the domestic financial sector, many studies examined the impact of foreign banks on financial stability. Some papers find that global banks support their foreign affiliates during times of financial stress through internal capital markets (De Haas and Van Lelyveld, 26 and 2; and Barba-Navaretti, Calzolari, Levi and Pozzolo, 2). At the same time, however, a number of studies show that (funding) shocks to parent banks can be transmitted to their foreign subsidiaries with negative consequences for their lending (Peek and Rosengren (999, 2); Acharya and Schnabl (2), Chava and Purnandam (2); Cetorelli and Goldberg, 2). Studying the global financial crisis, studies showed that foreign subsidiaries reduced lending more compared to domestic banks (Popov and Udell 2; De Haas, Korniyenko, Loukoianova and Pivovarsk 2; De Haas and Van Lelyveld 2). Ongena, Peydro Alcalde and Van Horen (2), however, show for Eastern European countries that while foreign banks reduced lending more compared to locally funded domestic banks, they did not compared to banks that had financed their lending from international capital markets. Although providing very interesting and valuable insights, these studies analyze in general only a limited number of countries and/or a selected number of (large) banks. While this can allow for a more careful isolation of the mechanisms the authors are after, it can come at the cost of not being able to generalize the results or to address the influence of heterogeneity across foreign banks. By introducing and using a new, comprehensive database covering (close to) the universe of domestic and foreign banks over a lengthy period, this paper extends the literature on foreign banking in regard to two important questions: One, how does foreign bank ownership affect private credit? And two, how does foreign bank ownership affect financial stability? As the database includes the home country of each foreign bank, we also study the impact of bilateral factors, which up till now received only limited attention due to data limitations. This is important given the interplay between home and host countries features in entry decisions (Berger, Buch, DeLong, and DeYoung, 24; 4

5 Galindo, Micco and Serra 23, and Claessens and Van Horen, 2) and between (cultural and institutional) distance and performance (Claessens and Van Horen, 2). The paper itself is structured as follows. Section 2 describes the database. Section 3 provides an overview of the main trends in foreign banking, and highlights the heterogeneity and differences in foreign bank ownership. Section 4 examines the impact of foreign bank ownership on private credit in the host country. Section 5 studies the impact of foreign bank ownership on lending stability during the global financial crisis. Section 6 concludes. 2. Data This paper uses an original and newly collected database on bank ownership. The database covers the period , and includes 5,377 banks active at least one year in 37 countries, thus covering banks in advanced countries like the United Kingdom and the United States, emerging markets like Brazil and China, and developing countries like Cambodia and Zambia. 2 The database includes commercial banks, savings banks, cooperative banks, bank holdings, and holding companies. 3 For each year and for each bank in our sample, we determined the exact moment of entry and exit. If the exact year of establishment could not be determined, but additional information indicated that the bank was in operation prior to 995 (e.g., the presence of financial statements), we coded 5 as the fictive year of establishment. 4 In terms of exit, we took in general the year a bank became inactive in Bankscope as the year of exit. In all cases, we cross-checked this using additional sources and made corrections if necessary. In addition, we carefully accounted for mergers and acquisitions to make sure that only the merged entity or the acquiring bank remained in the sample after a take-over. 5 In building the database, many sources were used, including Bankscope, individual banks annual reports, corporate governance reports, Central Bank publications and web-sites, information from banking regulation agencies, local stock exchanges, US SEC forms F-2, parent company s reports, The Economist Intelligence Unit, Factiva, The Banker, etc. 2 Our sample includes all countries with more than 5 active banks reporting to Bankscope in 28. For the advanced countries in our sample, we restrict our coverage to the largest banks in terms of 28 assets, so smaller (typically regional) banks are not included in the database for these countries. For all these countries, we cover at least 9 percent of the banking system in terms of assets. 3 Including bank holdings and holding companies can potentially lead to double counting as both the holding company and the bank are often included in Bankscope. In all countries where bank holdings were included, we excluded the holding company if the bank itself was represented 4 For 34 banks (2.5 percent of our sample) we were not able to find the exact year of establishment nor could we determine whether the bank was active prior to 995. In these cases the year of establishment is left blank. 5 For example, if two banks, bank A and bank B merged in 2 to create a new entity, bank C, then the two individual banks A and B were each included in the dataset until 2. Then, from 2 on, these two banks were considered inactive and the new bank (bank C) was included in the database. Similarly, if bank A was acquired by bank B in 2, both banks were included in the database until 2, with bank A then becoming inactive after 2 and bank B remaining active after 2. 5

6 For each bank in every year, we identified its shareholders, the nationality of the shareholder(s), and recorded any change in ownership. We classified a bank as foreign-owned using the definition generally applied in the literature, i.e., if 5 percent or more of its shares are owned by foreigners. For all foreign banks we summed the percentages of shares held by foreigners by country of residence, with the country with the highest percentage of shares then considered the home country. 6 Ownership, and country of ownership, is based on direct ownership, i.e., we do not consider indirect ownership. We do, however, take into account that in some cases the direct owner is an entity purely established for tax purposes. In such cases we recorded the country of nationality of the ultimate owner as the source country (these cases typically involve entities registered in Mauritius, Panama and Luxembourg). When determining ownership, we erred on the side of caution and reported ownership as missing when the reliability of the information was in question. 7 Using this procedure we determined complete ownership structure, including the home country of the largest foreign shareholder, for 5,59 of the 5,377 banks in the sample (i.e., 94 percent) for all the years each bank was active. For 92 banks only partial ownership and for 226 banks no ownership could be determined. All in all, the data provide an almost complete picture of bank ownership around the world and changes therein over time. Compared to other databases on bank ownership used in the literature (most notably the one constructed by Micco, Panizza and Yañez 27) our database differs in a number of respects: First, it includes virtually all banks active in a particular country, which allows not only for the comparison of domestic and foreign banks but also for comparisons of types of banks (i.e., small versus large, old versus young, deposit taking versus non-deposit taking). Second, it includes countries at all levels of development, allowing one to test whether foreign banks behave differently (from domestic banks) in certain countries. Third, it captures bank ownership and changes therein over an extensive time period, allowing one to disentangle short-run versus long-run implications of foreign ownership. Fourth, as it includes the home country of foreign banks, it allows one to differentiate between foreign banks from different types of home countries and examine the impact of bilateral factors (like distance, trade linkages, institutional similarity, etc.) on the behavior of foreign banks. 6 In recent years, identifying home countries and tracing ownership information has become more complicated with more banks raising equity through capital markets. To overcome the problem of determining the nationality of anonymous shareholders, we only consider block shareholdings when determining ownership. 7 For specific examples of how the home country was determined see the Working Paper version of this paper available at: 6

7 3. Trends and heterogeneity in foreign banking Over the period 995 to 29, banking systems in many countries experienced important transformations. 8 While the total number of domestic and foreign banks in our sample stayed virtually the same (3,894 in 995 and 3,9 in 29), these aggregate numbers mask two counteracting trends (Table ). 9 The number of domestic banks decreased by about 7 percent, due to consolidation driven by technological changes and deregulation in many advanced and developing markets as well as by some financial crises. At the same time, the number of foreign banks increased by 72 percent. These differing trends mean that the relative importance of foreign banks increased substantially, from a share of 2 percent in 995 to 34 percent in 29. Figure shows this steady increase in the number of foreign banks present, from 774 in 995 to 334 in 29, with the foreign share rising as well. While there was a steady increase, foreign bank entry did fluctuate substantially over the period (Figure 2). Foreign bank investment activity was especially high in the late nineties and early 2s and again in This reflected in part waves of reforms, including the opening up of Eastern Europe and other transition economies, and the liberalization of entry by East Asian countries. It also reflected the sharp increase in financial globalization before the crisis. Entry peaked in 27 but slowed down markedly after the start of the crisis. While 28 still saw entries at levels similar to 25, 29 entries were the lowest of the period, reflecting bank-specific and home- and host-country factors. Many banks suffered capital losses and other balance sheet impairments and, due to market forces or government interventions, were forced to consolidate. Many choose to retreat from cross-border banking in general. At the same time, host countries affected by the crisis became less interesting as investment opportunities. Foreign banks did exit markets also earlier, with 2 standing out with 48 banks exiting, mostly due to the various crises affecting emerging markets and related consolidation trends and reforms. While the recent crisis has affected new entry, until 29 exit remained at levels similar to earlier periods. Even though few countries are left out (in countries did not have any foreign bank present and in 29 only countries), the share of foreign banks and their role as 8 We exclude from all further analyses 8 offshore host countries (Antigua and Barbuda, Bahrain, Barbados, Cyprus, Mauritius, Panama, Seychelles and Singapore) that are included in the database. 9 Note that for some OECD countries in our sample the database does not include all active banks. Very small (typically regional) banks (outside the top banks in terms of assets in 28) are not included. Parent banks, apparently, did not (yet) feel the need to close or sell their foreign affiliates. This likely reflects that most affiliates were located in countries at the time only marginally affected by the crisis and/or with substantial long-term growth opportunities (with fixed costs involved in setting up a foreign affiliate large, exits are in general not driven by short-run fluctuations). 7

8 financial intermediaries differs widely across countries. Figure 3a and 3b show the relative foreign bank presence across all 29 host countries in terms of numbers and assets in 27 (to avoid impact of the financial crisis). It highlights two aspects. First, in terms of numbers, shares vary from % for the top host countries to single digit percentages for the bottom 2 host countries. Similarly, the range of asset shares varies from % to single digits. The drop in asset share though is steeper in the middle range. This reflects that host countries tend to have either many large foreign banks or few small banks. The growing importance of foreign ownership has rendered foreign banks important conduits of financial intermediation. Foreign bank presence is not evenly distributed across income groups, however. Using again 27 data, Figures 4a and 4b, depicting the relation between GDP per capita and shares in numbers and total domestic asset, show the large variations in foreign bank presence. Especially in emerging markets and developing countries, foreign banks play important roles in financial intermediation, with average loan, deposit and profit shares close to 45 percent in emerging markets and close to 5 percent in developing countries. In contrast, and perhaps not surprisingly, in OECD countries the vast majority of financial intermediation is done by domestic banks, with foreign bank loan, deposit and profit shares on average ranging only about 2 percent. Two more two facts are noteworthy. First, the large dispersion in the average size of banks across host countries, with little apparent relation to the number of foreign banks present in the country (Figure 4c). And second, the fact that the share of foreign banks in numbers is less than the share in terms of assets at low levels of presence in numbers, and the reverse at higher level of presence in numbers (Figure 4d). This suggests that foreign banks tend to be small when few in numbers and large when they dominate the market in numbers. Figure 5a and 5b show the home countries foreign bank distribution in terms of share in numbers and assets of total local home banking systems in 27. It highlights again the large variation, similar to the host country distribution, with shares varying between % and zero percent. Still, there is more concentration in the distribution of foreign banks across home countries than across host countries as the shares drop off even faster below the top 2 home countries in terms of numbers, but even more so in terms of assets. This implies that some home countries export relatively many (large) foreign banks. Patterns have also changed over time. While entry largely reflects investments from established countries, there The figures depict the share (in number or assets) of banks in a home country that have foreign bank presence, where each presence is counted separately, relative to all the banks in that home country which the data cover. For example, a % number share means that all banks in that home country have some foreign bank presence. 8

9 have been investors from new countries. While in 995, 47 banks came from emerging markets and 38 from developing countries, in 29 these numbers increased to 279 and 77 respectively. Foreign ownership thus became less concentrated. In 995 the five biggest investors (France, Germany, the Netherlands, the United Kingdom and the United States) owned 45 percent of all foreign banks, compared to 38 percent in 29. Foreign bank ownership is not evenly distributed across home countries. Figures 6a and 6b depict the relation between home country GDP per capita and the country s degree of exporting foreign banks, in shares of numbers and total assets. It shows that more developed countries tend to export more foreign banks and that fewer foreign banks come from low-income countries. Still, while the vast majority of parent banks are from an OECD country, at least 3 percent of foreign banks are owned by a bank from an emerging market or developing country. And there remain large dispersions in the degree of exporting. Two more facts are noteworthy. First, the dominance of small banks, with little apparent relation to the number of banks exported from that home country (Figure 6c). And second, that the share of foreign banks in numbers is almost always higher than the share in assets (Figure 6d), indicating that foreign bank operations tend to be small relative to the size of domestic banks in most home countries. Not only do distributions across home and host country differ, but foreign bank presence also differs bilaterally. Table 2 shows strong regional patterns in foreign bank presence, with the diagonal entries showing much larger presence. For example, of the foreign banks coming from the Americas, 63% are present in the same region and only 37% elsewhere. Figure 7a and b depict this regional pattern in more detail in the form of heat-maps for respectively the number share of foreign banks from a particular home country going to a certain host country, and the assets share of foreign banks in a particular host country coming from a certain home country. Both Figures show the same, strong regional pattern. 2 Finally, foreign banks can differ from domestic banks in their business models, size and profitability, with some of these differences to vary by level of host country development. Figures 8a-8d plot a number of balance sheets and performance measures for domestic and foreign banks against the host countries GDP per capita. The figures show much heterogeneity, but also some differences which vary by level of development. 2 This regional pattern may not surprise since research has shown that foreign banks tend to follow their customers and therefore enter countries with strong trade linkages (e.g., Goldberg and Grosse 994). In addition, studies have found that banks tend to invest in countries that are (geographical and/or institutional) close (e.g., Buch and De Long (24); Galindo, Micco and Serra 23)). Furthermore, Claessens and Van Horen (2) show not only absolute distance to matters but also the distance of competitor countries. 9

10 Loan to deposits ratio (Figure 8a) is a proxy for the degree to which banks are active in traditional forms of financial intermediation, i.e., lending, and the importance of wholesale funding relative to traditional deposits. This ratio tends to be higher for domestic banks compared to foreign banks, consistent with the notion than foreign banks are relatively less active in lending. This is especially so in lower income countries. However, in middle-income countries, foreign banks tend to have higher loan to deposits ratios, suggesting that they are more active in lending and more able to attract non-deposit sources of funding. Foreign banks also tend to have significant more liquid assets (Figure 8b), especially so in high-income countries, but less so in middle-income countries. In terms of solvency, foreign banks tend to have higher ratios of capital to weighted assets, especially in high-income and developing countries, and less so in middle-income countries (Figure 8c). This suggests that foreign banks are more conservative with respect to their capital buffers. In terms of performance, foreign banks tend to underperform domestic banks in middle-income and developing countries (Figure 8d). This may surprise since foreign banks, with greater access to know-how, technology and lower cost of funds, are generally believed to be more profitable in such markets. Some of this probably reflects differences in activities, such as the fact that foreign banks have more conservative portfolios. However, it may also reflect differences of origin of foreign banks and variations in the ease by which they operate in emerging markets and developing countries Foreign banks and domestic credit creation In this section we re-examine the link between foreign ownership and the provision of credit to the private sector. The extent to which foreign banks contribute to financial sector development is possibly one of the most controversial aspects of foreign bank entry. Although some studies have looked at the relationship between private credit and foreign bank ownership (most notably Detragiache, Tressel and Gupta 28 and Cull and Martinez Peria 27), surprisingly is known under what conditions foreign ownership positively relates to private credit and when negatively. Given the substantial differences across foreign banks as 3 Claessens and Van Horen (2) show that the profitability of foreign banks is importantly affected by home, host and institutional factors. They find, for example, that foreign banks perform better when from a high income country and when regulations in the host country are relatively weak. Also foreign banks from home countries with the same language and similar regulation as the host country tend to perform better. These factors may explain some of the differences in the simple averages.

11 highlighted in the previous section, allowing for different types of heterogeneity can potentially be important. Methodology To assess the impact of foreign bank ownership on the provision of credit, we first use cross-country OLS regressions to test whether foreign bank presence is associated with higher or lower levels of private credit and what factors are driving this relationship. Our sample consists of countries that, contrary to other studies, represent all levels of development. Our dependent variable is the ratio of private credit to GDP (from the IMF International Financial Statistics). We take the 3-year average (25-27) to smooth out business cycle fluctuations. Our variable of interest, Foreign presence, is the ratio of foreign bank assets to total bank assets in the country, and is measured in 24 to limit as much as possible joint endogeneity concerns. As control variables, we add a number of variables known to affect the level of private credit in an economy (see, for example, Djankov, McLiesh and Shleifer 27; Detragiache, Tressel and Gupta 28), all measured in 24. The first variable is the overall level of development in the country as measured by GDP per capita. We control for Inflation as it has been shown to correlate with measures of financial depth. In addition, we include two variables capturing the ease of doing business (related to banking): the availability of information to creditors (Creditor info), and the cost of enforcing contracts (Enforcement contracts) (both variables come from the World Bank, Doing Business Indicators). In order to test whether the relationship between foreign bank presence and the provision of private credit is conditional on certain features we introduce next a number of interactions. First, we test whether the relationship between foreign bank presence and the provision of private credit is affected by the level of development of the host country. Detragiache, Tressel and Gupta (28) develop a theoretical model that predicts for poor countries a negative relationship between the two. Studying a sample of poor countries, they find that this is indeed the case. In order to test the robustness of this finding, we interact Foreign presence with a dummy which is one if the country is a developing country. Second, we test whether the average distance of the parent bank to the host country affects the relationship between foreign bank presence and the provision of credit. Both theoretical (e.g., Stein 22; Hausman and Marquez 26) as well as empirical studies (e.g., Berger and DeYoung 2, 26; Mian 26) suggest that foreign banks that are closer to the host country are better equipped to provide loans as they can better access soft information.

12 Therefore, one would expect the provision of credit to be more likely positively correlated to foreign bank presence when banks from neighboring countries have entered. In order to test whether this is indeed the case, we interact Foreign presence with a dummy that indicates whether foreign banks active in a country are relatively close or distant. The variable Distant banks is created as follows: First, we calculate for each host country the average distance between the host country and the home country of all foreign banks active in the host country. Next we take the median of this variable across host countries. The variable Distant banks is one if the average distance for the host country is above this median and zero otherwise. The empirical model we estimate is as follows: Y FP FP X i i i i Z where Y i is the ratio of private credit to GDP in host country i, held by foreign banks, i i FB i is the share of bank assets X i is a matrix of variables that are expected to affect the relationship between foreign presence and the provision of credit, Zi is a matrix of control variables, i is the error term and α, β, γ and δ are parameters to be estimated. Our parameters of interest are β and γ. We estimate our model using OLS regressions with robust standard errors. The obvious drawback of this type of cross-sectional regression is that the market share of foreign banks could be endogenous to the host country s financial development. As also pointed out by Detragiache, Tressel and Gupta (28) the bias could, however, run both ways. On the one hand, foreign banks might be more willing to enter countries where (for other reasons) financial development is particularly low as they expect these markets to grow faster. On the other hand, business prospects might be worse in countries with low levels of development, making foreign banks more reluctant to enter. One way to address this endogeneity problem is to estimate a panel regression. However, due to the short sample period (balance sheet information is only reliably available from 24 onwards) and the credit boom that took place in many countries before the global financial crisis, the private credit to GDP ratio is high persistent (which results in failure of the autocorrelation tests). As an alternative we therefore estimate regressions in which we test if a large foreign bank presence in 24 affects credit growth between 24 and 27. By first differencing the data we directly control for unobserved country-specific effects that might affect the level of private credit. This second set of regressions we estimate is thus as follows: Yi Yi FPi FPi X i Z i t i 2

13 where Yt is the log difference of the ratio of private credit to GDP between 24 and 27. Yi is the level of private credit to GDP in 24, to control for financial development in the initial year. 4 FPi measures foreign bank presence in 24. The other variables are defined as before. 5 The model is estimated using OLS with robust standard errors. Empirical results Table 3 presents the results of our cross-country regression. We find a negative correlation between the presence of foreign banks and private credit to GDP (Table 3, column ). A one standard deviation increase in Foreign presence is associated with a decline in private credit by some 6 percentage points, economically significant, since the average ratio of private credit to GDP in our sample is 5 percent. We also find some evidence of a nonlinear relationship (column 2). This, however, is mostly driven by a few outliers: high-income countries with very large financial sectors compared to the size of their economy that are either dominated by foreign banks (Hong Kong, Ireland and New Zealand) or with hardly any foreign banks (Iceland). When we exclude these countries, we find again a negative (linear) relation that is even stronger (not reported). Pooling all countries, however, masks significant differences across countries. If we split our sample between high-income (OECD and non-oecd) countries and emerging markets and developing countries, we find the relation with foreign bank presence to vary with economic development. For the group of high-income countries (columns 3-4), we find no significant relationship between foreign ownership and credit. However, as column 5 and 6 show, for emerging markets and developing countries there is a strong negative relationship between foreign banks presence and credit. A one standard deviation increase in Foreign presence is associated with a decline in private credit of 5 percentage points (compared to a mean private credit to GDP ratio of 3 percent in this group of countries). In terms of our control variables, we find (as expected) GDP per capita to be generally associated with more financial development and inflation with less financial development, except for high-income countries, where both variables are not significant, in part as differences are limited in these countries. Access to creditor information has a positive impact on financial sector development, but is imprecisely estimated. In emerging markets and developing countries, the longer it takes for contracts to be enforced, the less credit is created. 4 As we control for the initial level of financial development, the coefficient on foreign bank presence should not be biased due to omitted country characteristics or shocks that affect private credit contemporaneously. 5 GDP per capita is not significant and is omitted from the model, but including it does not change results. 3

14 Next, we investigate for the sample of emerging markets and developing countries whether the effect of foreign bank presence differs by type of host country and type of foreign bank. 6 First, we test whether the impact of foreign bank presence differs between developing countries and emerging markets. The result in Table 4 column shows that this is indeed the case. We find that foreign bank presence is only negatively related to private credit to GDP in developing countries. This confirms the results of Detragiache, Tressel and Gupta (28). In the next column we investigate whether distance affects the relationship between foreign bank presence and private credit. Here results indicate that foreign bank presence is only negatively related to private credit when foreign banks are relatively distant. This is in line with theoretical and empirical evidence that suggests that distance increases information asymmetries between bank and borrower (e.g., Hauswald and Marquez 26, Mian 26). We next run a horserace between the impact of economic development of the host country and of the relative distance. Our results (column 3) show the negative relationship between foreign bank presence and private credit to be driven by the type of foreign banks that enter the country, not so much by the level of development of the host country. Although one has to be careful to make any inferences about the direction of causality, these results suggest that in emerging markets and developing countries entry of foreign banks that are more remote can lead to cream-skimming which negatively affects access to credit. A number of studies have found evidence that the behavior of foreign banks might be related to the relative importance of foreign banks in the host country. For example, Claessens and Lee (23) find fewer spillovers with more limited entry. And Cull and Martinez Peria (27) show that in countries where foreign banks hold more than percent of the assets, private credit was significantly higher than in other countries before, during and after crises. Since the relative importance of foreign bank presence varies a lot across the countries in our sample (see Section 3), it is natural to examine whether some kind of threshold effect exists with respect to the relative importance of foreign banks in the host country. Therefore, we split our sample of emerging market and developing countries between countries where foreign banks control less than half of the assets in the country and countries in which they hold more than half. The results in the last two columns show the negative relation between distant foreign banks and private credit to be driven by countries where foreign banks control less than half the assets. This could suggest that foreign banks are more 6 We ran the same regressions on the All country sample and the results were very similar. However, since foreign bank presence only seems to affect financial sector development in emerging markets and developing countries, we only show these results as in our view they more insightful. 4

15 likely to invest in (large) domestic banks that engage in financial intermediation in countries they are more familiar with (i.e., those that are geographically close). In contrast, in more distant countries foreign banks are more likely to enter through a greenfield investment with a focus to remain a niche player, targeting only specific customers. Our next set of regressions studies the impact of foreign bank presence on private credit growth between 24 and 27. The results in Table 5 indicate that in the years leading up to the global financial crisis, private credit growth was higher in countries with more foreign banks. However, interacting Foreign presence with Developing and Distant banks, shows that this higher credit growth is driven primarily by countries in which foreign banks are relatively close to their home country (like Eastern European countries where many Western European banks invested). For those countries with relatively close foreign banks, one standard deviation increase in Foreign presence is associated with 2 percentage points higher credit growth (compared to a mean credit growth for these countries of 28 percent). We cannot reject the null hypothesis of joint significance of Foreign presence and Foreign presence*distant banks, suggesting that in host countries with relative distant foreign banks, the actual importance of these banks in the domestic financial system did not affect private credit growth in the boom years. This holds for emerging markets and developing countries as well as for high-income countries. 5. Foreign banks and financial stability during the global financial crisis The role of foreign banks in lending stability has received renewed attention amidst the current financial turmoil. Concerns have been raised that, when faced with capital or funding shocks, foreign banks withdraw from cross-border banking activities, and reduce credit extension in host markets. While most studies (in large part due to data limitations) focus on specific regions (often Central and Eastern Europe), our database allows us to study how the global financial crisis has affected lending stability of foreign banks in a large number of countries. This allows us to analyze how different levels of development and differences in relative market share of foreign banks affect foreign bank behavior. Furthermore, our database includes virtually all domestic and foreign banks, allowing for a detailed comparison of the stability of lending for both types of banks. Finally, since we know the home country of all foreign banks, we can examine whether foreign banks owned by certain home countries are a more stable source of credit. 5

16 Methodology Our empirical strategy builds on Ongena, Peydro Alcalde and Van Horen (22) who study changes in credit extension by foreign and locally and internationally funded domestic banks in Central and Eastern Europe during the global financial crisis. We start in 25 and end in 29. We include 8 countries which have at least one foreign bank active over this period. We exclude banks that became inactive or entered the market during the sample period, leaving us with a sample of 3,65 banks, of which,98 foreign owned. Our dependent variable is the growth rate of loans, measured as the log difference in total lending (sum of net loans and loan loss reserves) of bank b in country j in year t. Since we want to study whether foreign banks behaved differently from domestic banks, our main variable of interest is a foreign bank dummy (Foreign), interacted with a crisis dummy (Crisis). Similar to Ongena, Peydro Alcalde and Van Horen (2), we distinguish the first crisis year (28) from the second crisis year (29) We add a number of controls. We control for time-invariant differences between banks (like different business models, funding structure, strategies, etc.) using bank fixed effects. We control for differences across countries by including country-year fixed effects. To the extent that there are no systematic differences with respect to the type of firms in the portfolios of foreign versus domestic banks or those differences are not correlated with crisisinduced shocks to credit demand, these fixed effects should control for differential changes in credit demand across countries. We control for a number of bank-specific characteristics that we interact with our two crisis dummies to allow for differential impact of the crisis by type of bank. These characteristics are Size (log assets), Solvency (equity to asset ratio), Liquidity (liquid to total assets) and Deposits (deposits to liabilities). All bank characteristics are measured as of end-27 to limit any endogeneity problems. Our baseline model is as follows: Where Lbt Lit Foreign Crisis X b t b Crisis t Lbt is the yearly growth in loans by bank b in year t (t goes from 25 to 29), Foreign is a dummy that equals one if the bank was foreign owned in 27 and equals zero otherwise, and Crisis stands for two dummies that equal one for t equal to 28 or 29, respectively, and equal zero otherwise. X b is a matrix of bank level control variables that might affect the impact of the crisis on the bank s decision to continue lending, b are bank fixed effects and t are year fixed effects that are allowed to vary per country, i is the error term and α, μ, β and γ are parameters to be estimated. Our parameter of interest is β. b t ij 6

17 To estimate our model we use three procedures. First we estimate the model using fixed effects OLS excluding the lagged dependent variable. Second, we include the lagged dependent variable to assure that our results are not biased due to the existence of serial autocorrelation. Third, to allow for the fact that lagged loan growth might be correlated with the panel effects (Nickell, 98) we estimate the model using difference GMM with the lags of the levels of the explanatory and dependent variables as instruments (Arellano and Bond 99). In all models, standard errors are clustered at the bank level and observations below the st percentile and above the 99 th percentile of loan growth are excluded to reduce the impact of possibly influential outliers. Empirical results Our regression results are presented in Table 6. We first present our three regression specifications based on our full sample of countries (column -3). The results are very similar and show that in 29 foreign banks reduced lending more compared to their domestic counterparts. 7 Based on the results in column, foreign banks reduced their lending in 29 by some 6 percentage points more compared (note the mean credit growth in 29 was 5 percent). We do not find loan growth to differ between domestic and foreign banks in 28, probably as the financial crisis only became truly global in 29 and did not yet affect many foreign banks operations in 28. Our control variables show some interesting relations. Banks that generated a relative large part of their funding from deposits (a relatively stable source of funding during the crisis) continued to lend relatively more in 29. This effect is large: a one standard deviation increase in the deposit to liability ratio means loan growth in 29 was some 4 percentage points higher. We also find that larger banks generally reduced their credit more. And more solvent and liquid banks maintained credit more, with effects larger in 28 than in 29. In the next columns, we split our sample in countries where foreign banks hold less than 5 percent of domestic assets (column 4-6) and countries in which they hold more than 5 percent (column 7-9). The results indicate a very distinct difference between the two groups of countries. We find the loan growth of foreign banks in 29 compared to that of domestic banks to be 8 percent less in countries where they have a low market share (the results in column 4). In countries where they dominate, however, foreign banks actually do 7 The AR (2) tests for the Arellano-Bond regressions consistently show that we cannot reject the null hypothesis of no second-order autocorrelation (since the estimator is in first differences, first-order autocorrelation does not imply inconsistent estimates). Furthermore, the Hansen J test results indicate that the instruments used are valid. 7

18 not show any difference in their loan growth in 29 compared to domestic banks. This result clearly shows how important it is to allow for heterogeneity across foreign banks when examining their (crisis) behavior. Following the base regression, we allow the impact of foreign ownership on lending stability to differ with respect to a number of home and host country, and bank characteristics. The results, presented in Table 7, are based on the fixed effects model without a lagged dependent variable. The results, however, are similar using the other two procedures. In the first two regressions we examine whether this difference in credit growth was specific to particular host countries. First, we examine, using an OECD versus non-oecd country dummy, whether the general economic development of the host country played a role. We find (column ) that in 28 foreign banks in OECD countries already grew their lending less compared to domestic banks (note that on average overall credit growth in 28 was still percent, double that in 29). Since the crisis started in OECD-countries, this suggests that in these countries foreign banks responded faster to the onset of the crisis than domestic banks did. In 29, however, as the crisis spread, foreign banks reduced lending, compared to their domestic counterparts, similarly in OECD and non-oecd countries. 8 Second, we test whether the average distance of the foreign banks active in the country (as defined previously) impacted the lending stability of foreign banks. Even though we found the average distance of foreign banks to have affected the provision of credit, we do not find a differential impact between foreign banks that are close or that are distant. Next, we examine the importance of home country characteristics. First, we allow the impact of foreign bank ownership on lending stability to differ between foreign banks owned by a parent located in an OECD country or in a non-oecd country. Somewhat surprising, we do not find significant differences between the two types of foreign banks (column 3). This may be because not all OECD countries were as severely affected by the financial crisis. We therefore interact foreign ownership with a dummy whether the home country experienced a systemic banking crisis during the period (as classified by Laeven and Valencia, 2). Our results (column 4) indicate that foreign banks owned by a parent located in a country that experienced a crisis reduced their lending more in both 28 and 29 compared to domestic banks; however, the parameter is imprecisely estimated. 8 We also examined the impact of foreign ownership on lending behavior in the four income groups (OECD countries, other high-income countries, emerging markets and developing countries) separately. We found similar magnitude and significance for the foreign ownership times crisis interactions for all regions except other high-income countries, where we found no significant difference. 8