WPS3755 The Mix of International Banks' Foreign Claims: Determinants and Implications Alicia García Herrero and Maria Soledad Martínez Pería* Bank of Spain and World Bank Abstract: This paper analyzes the determinants and implications for financial stability of the mix of international banks' claims countries receive. In particular, we distinguish between local claims, extended by international banks through their affiliates in a host (or claim recipient) country, and cross-border claims, booked from outside the host country, typically from banks' headquarters in their home countries. Using data on U.S., Spanish, and Italian banks' foreign claims across countries, we find that the share of local foreign claims is primarily driven by the degree of "freedom" in the host banking sector and by business opportunities in the local market. Entry requirements, startup and informational costs associated with international banking also play a role, but their influence is less robust. Finally, we find that the mix of international bank claims has implications for financial stability, since foreign claim volatility is lower in countries that receive a larger share of local claims. Keywords: foreign bank financing, financial FDI, cross-border claims JEL: F36, F37, G21 World Bank Policy Research Working Paper 3755, October 2005 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. *We thank Cristina Luna for her valuable assistance in obtaining and interpreting data from Banco of Spain and Bank of Italy on the Spanish and Italian banks' foreign claims, respectively. We also thank both central banks for access to these data. We are grateful to Maria del Pilar Casal and Sarojini Hirshleifer for excellent research assistance and to Harald Anderson for providing us data. We received useful comments and suggestions from Gerard Caprio, Asli Demirgüc-Kunt, Enrica Detragiache, Patrick Honohan, Ross Levine, Jesus Saurina, Adrian Tschoegl, and participants at the Stanford Center for International Development (SCID) conference on Latin American Financial Systems and the Challenges of Economic Growth and those at the World Bank and Journal of Banking and Finance conference on Globalization and Financial Services in Emerging Economies. Contact author: Maria Soledad Martínez Pería, World Bank, 1818 H St. N.W., Washington, DC 20433, email mmartinezperia@worldbank.org. Introduction After a decade of financial repression and stagnant international financial flows, the 1990s saw a resurgence in financial globalization. Countries opened up their economies to capital flows and liberalized their financial sectors. As part of this process of financial integration, multinational banking gained momentum once again and international banks' foreign claims--those extended on residents outside the country in which these banks are headquartered--took off. According to Bank for International Settlement statistics, which monitor foreign claims held by banks from OECD countries vis-à-vis the rest of the world, these claims rose from 1.3 trillion dollars in 1990 to close to 3 trillion dollars in 2002.1 In real terms, foreign claims to all non-BIS reporting countries rose by 120 percent over this period and claims to developing countries increased by 104 percent. These claims consist of financial assets such as loans, debt securities and equities, including equity participations in subsidiaries. International banks may grow their foreign claims portfolio in two ways. First, they may establish affiliates in different countries and extend claims locally through their branches and subsidiaries in these countries. Second, international banks may also extend cross-border claims by financing and booking the claims from outside the recipient or host countries (e.g., originating the claim in their home countries, where their headquarters are located). While the first type of international bank claims involve some form of foreign direct investment in the host country's financial sector, cross-border claims do not. In practice, we observe significant disparities in how banks conduct their business across countries. In some instance, like in the case of Albania, Burundi, Bhutan, Cambodia, and Moldova, foreign banks extend only cross-border claims. 1BIS-reporting countries over this period include: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Luxembourg, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, and U.S. 1 While in other instances, such as in Brazil, Chile, and Hong Kong, foreign banks' exposure is largely local. The purpose of this paper is to empirically analyze the determinants of the mix of foreign bank claims and to study its implications for financial stability. In terms of determinants of the composition of foreign bank financing, we examine the role of entry requirements, startup and informational costs, government intervention in the financial sector (primarily in the form of taxes, and restrictions in bank entry and operations), and business opportunities across host countries. The empirical model we propose is a reduced form model which tries to capture the underlying preferences of foreign banks, as well as those of governments and private agents in the host countries in our sample.2 With respect to the implications of the mix of foreign bank claims, we focus exclusively on the impact of the composition of these claims on the overall volatility of foreign bank financing.3 Because FDI or local claims require paying higher fixed and irreversible costs, it seems reasonable to expect these flows to be more stable and less responsive to bad news than cross-border claims. Also, economic fixed costs aside, banks trying to shrink the size or close down their overseas operations will have to pay the reputational costs of doing so and, therefore, may be less likely to run when conditions deteriorate. At the same time, while in the face of good economic conditions, foreign banks can relatively quickly and, perhaps cheaply, extend cross- border financing, augmenting their local claims might require lumpy investments that are often decided on the basis of long-term rather than short-term profit opportunities. 2Data limitations do not allow us to study the role of bank specific characteristics or home country variables. In other words, because we only have aggregate data for banks from three countries we cannot examine the impact of bank and home country characteristics on the share of local claims. 3There may be other implications from the mix of foreign bank claims, such as the impact on the balance of payments, which we ignore here and leave for future research. 2 Even though the literature on international banking is quite vast, a limitation of existing studies is that the determinants of financial FDI (or local claims) and those driving cross-border claims have been considered largely in isolation.4 Furthermore, though some studies have documented differences in the behavior of international banks' cross-border and local claims across countries (see Haas and Lelyveld, 2002, for Eastern Europe, Peek and Rosengren, 2000, for Latin America, and Palmer, 2000, for U.S. bank claims around the world), the implications of the compositions of foreign bank claims have been largely overlooked in the international banking literature and still require a systematic empirical analysis. The majority of existing papers on international or foreign banking have examined why international banks enter certain markets. An early strand of the literature on FDI by international banks focused on the experience of developed countries (especially the U.S.) with foreign bank entry and bank internationalization during the 1970s and 1980s (e.g., Goldberg and Saunders, 1980, 1981a, b; Ball and Tschoegl, 1982, Nigh et al., 1986; Cho et al., 1987; Hultman and McGee, 1989; Goldberg and Johnson, 1990; Goldberg and Grosse, 1994; Brealey and Kaplanis, 1996; Fisher and Molyneaux, 1996). More recently, several authors have examined the decision of international banks to establish operations overseas during the 1990s, especially in developing countries (e.g., Buch, 2000; Claessens et al., 2000; Focarelli and Pozzolo, 2001, 2006; Guillen and Tschoegl, 2000; Buch and DeLong, 2001; Moshirian, 2001; Galindo, Micco 4Buch and Lipponer (2004) is an exception. This study recognizes that banks have a choice to lend cross-border or to do so locally in host countries via their foreign affiliates. Using data for German banks, the authors test whether these two forms of financing are complements or substitutes by running regressions of one type of foreign funds vis- à-vis the other, while controlling for other determinants of international banking activities. They conclude that in the case of German banks both types of claims are complementary. The problem with their approach is that complementarity could be a result of endogeneity, where both types of foreign claims are driven by omitted factors. Also, the authors do not investigate why different countries get different shares of local claims relative to what they receive in cross-border flows. 3 and Serra, 2003; Buch, 2003, Buch and Lipponer, 2004; Wezel, 2004).5 At the same time, a smaller number of papers have examined the determinants and behavior of cross-border claims (e.g., Dahl and Shrieves, 1999; Buch, 2000; Jeanneau and Micu, 2002; Kawai and Liu, 2002; and Buch and Lipponer, 2004). This paper assembles a database on foreign claims, both cross-border and local, extended by Italian, Spanish, and U.S. banks to over 100 countries around the world during the period 1997-2002, to investigate the determinants of the share of local claims across countries and its implications for financial stability. Banks from Italy, Spain, and the United States are dominant players in the international banking market and they jointly account for approximately 30 percent of all outstanding foreign claims vis-à-vis the countries in our sample. In studying the mix of cross-border and local claims extended by international banks across countries, we rely on the conceptual/theoretical framework used in the trade/multinational firm literature to study the choice between exporting goods (the equivalent to cross-border lending) and producing them abroad for foreign markets (FDI). In particular, studies such as Brainard (1997) and Helpman et al. (2003) discuss the tradeoff faced by multinational firms between paying the higher sunk costs of establishing affiliates overseas vis-à-vis confronting the transportation costs and trade barriers that arise from exporting their goods instead. Following the trade literature, we investigate how the share of local claims to total foreign claims extended by international banks across countries is affected by the entry requirements and startup sunk costs of setting up operations overseas. These costs have been 5A growing literature has examined the implications of foreign bank presence in developing countries. See Clarke et al. (2003) for a review of this literature. Among others, Claessens et al. (2000), Barajas et al. (2000), Denizer (2000), and Martínez Pería and Mody (2004) discuss the implications on competition and efficiency in the banking sector. Dages et al. (2000), Detragiache and Gupta (2004), Peek and Rosengren (2000), de Haas and Levyveld (2002, 2004) compare the lending behavior of foreign and domestic banks during crises. Berger et al. (2001), Mian (2004), and Clarke et al. (2005) address the consequences on access to financing by small businesses. 4 largely neglected in the literature on international banking. We use new survey data on the minimum capital requirements for opening banks across countries, as well as recently available information on the general costs (fees, costs of procedures and forms, fiscal stamps, legal and notary charges, etc.) of starting up a business to study their importance in driving the share of local to total foreign claims. Also, since banking is an informationally intensive industry, we reinterpret the role of variable transportation costs in the manufacturing trade/multinational firm literature as information costs that arise in the international banking context. However, contrary to the case of transportation costs, which have an unambiguously negative effect on the share of exports to FDI, the impact of informational costs on international banking is harder to measure and a priori unclear. On the one hand, there are the informational costs of screening and monitoring foreign clients, which tend to be higher the larger the "distance" (geographic, cultural, legal, etc) between the borrower and the bank (Petersen and Rajan, 1994, 2002; Berger et al., 2002; and Bonaccorsi and Gobbi, 2001). Because extending claims through overseas affiliates allows banks proximity to its foreign clients, local claims are expected to carry lower informational costs of screening and monitoring borrowers than cross-border claims. On the other hand, extending claims through overseas affiliates raises the costs of information and oversight faced by international banks' CEOs and/or shareholders in trying to monitor the actions of managers and loan officers at these affiliates. In other words, extending claims through affiliates might increase information costs by augmenting the "distance" between bank CEOs, managers, and loan officers (Mian 2004). Thus, as a result, the net impact of information costs on the share of local claims is largely an empirical question, which we hope to address. 5 Relative to information costs, the banking policies of the host country are expected to have a clearer impact on the share of local claims. Policies that limit cross-border activities such as capital controls on loans from abroad are anticipated to diminish this form of bank presence, in the same way that trade tariffs discourage trade. On the other hand, limits on "banking freedom" such as restrictions on bank activities and foreign bank entry; controls on foreign currency lending by banks operating in the local market; and high corporate taxes are expected to reduce the attractiveness for international banks of extending claims locally in a foreign country. There are a number of reasons why governments might put in place banking policies that promote one type of foreign bank claims or the other. First, by deterring foreign banks from freely operating locally in a host country, governments might seek to protect the domestic banking sector from foreign competition. Second, governments might promote certain types of foreign claims over others because of their different implications for the balance of payments. In particular, while cross-border financing virtually always implies obtaining foreign currency, in the case of local claims, beyond the initial FDI investments, the local activity of foreign banks involves intermediating local savings. Third, as discussed above, governments might want to influence the type of foreign bank financing to the extent that this has an impact on volatility. Finally, consistent with what other international banking studies on FDI and cross-border claims have found, we allow for measures of economies of scale, profit opportunities, as well as, default, price and expropriation risks to influence the share of local to total foreign claims extended to a country. While some of these variables have a clear expected impact on the share of local claims, the effect of others on the mix of foreign financing is an empirical question. In our view, this makes our empirical investigation all the more interesting. 6 Our results for the share of Italian, Spanish and U.S. banks local claims indicate that this ratio is negatively impacted by restrictions on banks' freedom to operate in the local market. On the other hand, larger economies of scale and business opportunities lead to more local foreign claims. Other factors such as entry bank requirements, startup and informational costs are also significant, but their impact is less consistent across international banks and also varies depending on the specifications. Regarding the implications of the mix of foreign bank claims for the stability of foreign financing, we find that countries with a higher share of local foreign claims observe lower total foreign claims volatility. This result, which is robust to controlling for a number of other factors that might affect foreign claims volatility, helps us confirm in a systematic manner some of the descriptive evidence offered by other studies favoring local foreign claims to cross-border bank financing. The rest of the paper is organized as follows. Section II discusses the data on Italian, Spanish, and U.S. foreign claims. Section III presents the empirical methodology pursued. Section IV discusses the empirical results. Section V concludes. II- The data on foreign claims Perhaps one of the reasons why not much research exists on the mix of foreign claims extended by international banks across countries has to do with the fact that the data are not readily available. The main source of international banking data is the BIS. The BIS' Consolidated Banking Statistics contain information on the foreign claims extended by international banks from more than 20, primarily OECD, countries (referred to as BIS-reporting countries). The data reported by the BIS aggregates the information provided by individual banks in each BIS reporting country, so it is really country level as opposed to bank level data. 7 An important advantage of the BIS Consolidated Banking Statistics is that they net out intrabank claims so, for example, loans between a bank's headquarter and its branches and or subsidiaries overseas are not included. On the other hand, the BIS data does not adequately discriminate between cross-border and local foreign claims. While cross-border claims are mainly denominated in foreign currency, local claims can be denominated either in the local currency of the country where the claimholder resides or in a foreign currency. The latter is particularly true in the case of highly dollarized countries (such as many in Latin America). In general, BIS data on local claims only captures those denominated in local currency. Foreign currency denominated local claims are combined with cross-border claims and reported under what the BIS calls "international claims". Thus, BIS data is largely ill-suited for an analysis of the determinants or the implications of the mix of cross-border versus local international bank claims. Due to the limitations of the existing BIS information, the data used in our analysis had to be specially requested, through the BIS, from the Italian and Spanish central banks. The exception was the U.S., which is the only country that since 1997 reports separate information on cross-border and local claims to the BIS. For all three countries, the data available to us aggregates claims held by all banks vis-à-vis each host country. Thus, like in the case of the BIS data, the information we have is not bank, but rather country level data. For the U.S., we have information for the period 1997-2002. In the case of Spain and Italy the data covers the period 1998-2002. None of these countries collected separate information on local and cross-border 8 claims prior to this period. In the case of all three countries, the information we have nets intra- bank positions.6 Table 1 presents information on Italian banks' foreign claims in 105 countries. Claims are zero in 9 countries (Bhutan, Botswana, Burkina Faso, Cambodia, Honduras, Niger, Papua New Guinea and Rwanda) and are 100 percent cross-border in 60 countries or 60 percent of the sample. Countries where Italian banks extend only cross-border claims include both developed and developing economies across all regions. Italian banks hold local claims in 35 percent of the countries in our sample. These include primarily developed economies or countries in Eastern Europe and Latin America. However, among these countries, the share of local claims is never 100 percent (the largest share of local claims hovering around 90 percent is observed for countries such as Croatia, Poland, and Peru). This indicates that Italian banks do not substitute entirely local claims for cross-border lending. Even in countries where most of the banking business is conducted through the local affiliates, there is still some cross-border financing taking place. Table 2 shows information pertaining to the overseas or foreign claims of Spanish banks. Spanish banks are not active in anyway in 14 countries in the sample --Armenia, Bangladesh, Bhutan, Botswana, Burundi, Cambodia, Jamaica, Nepal, Lesotho, Nepal, Niger, Rwanda, Sri Lanka, and Zambia--, half of which are in Africa. Spanish banks hold exclusively cross-border claims vis-à-vis 48 countries or 47 percent of the sample. Like in the case of Italian banks, the list includes developed (mostly Nordic) and developing countries across all regions. Spanish banks tend to have local operations in developed countries and in emerging economies. The 6This is an important advantage of this dataset since it allows us to measure net exposure to a country. Bank level data from sources such as Bankscope do not disclose information on intragroup positions and on cross-border lending to specific hosts. 9 extent of local presence is particularly high in Latin America (especially Mexico and Bolivia) where in many countries the share of local presence exceeds 75 percent, but never reaches 100. As shown in Table 3, U.S. banks appear to be active in one way or the other in all but 5 countries in our sample (i.e., Armenia, Burkina Faso, Burundi, Madagascar, and Rwanda). Furthermore, contrary to the case of Spanish and Italian banks, U.S. banks have some form of local presence in more than three-quarters of the sample and those countries where U.S. claims are exclusively cross-border are mainly small developing economies (with the exception of Saudi Arabia). Finally, relative to Spain and Italy, it is harder to detect any regional pattern in the share of U.S. bank local foreign claims, since countries with high shares include both developed and developing countries in almost every region. III - Empirical Methodology The empirical analysis conducted in this paper has two objectives: (1) to study the determinants of the mix of foreign claims and (2) to examine the implications of this mix for the stability of foreign bank financing. Below, we discuss the empirical strategies pursued to accomplish these objectives. Because the time span covered by the data is relatively short and since many of the variables of interest do not vary over time, we focus on explaining differences across countries. Therefore, we consider only the average share of local foreign claims across countries and, for each country, the volatility of total foreign claims over the entire sample period. Taking into account the existing empirical international banking literature and borrowing from the theoretical framework used in related trade and multinational firm studies, we model 10 the average share of local to total foreign claims held by Italian, Spanish, and U.S. banks overseas as follows: Share of local foreign claimsi,j= 0 + 1Entry requirements & startup costsj + 2Informational costsi,j + 3Bank Regulationj + 4Taxesj + 5Scale & profit opportunitiesj + 6Risksj+ i,j (1) where i indicates the international banks' country of origin and j refers to the host or claim recipient country. We estimate separate cross-country equations for Italian, Spanish, and U.S. banks' foreign claims.7 In each of these equations, the Share of local foreign claims refers to the ratio of local foreign claims from i country banks to host country j out of the total foreign claims extended from i country banks to host country j. Equation (1) is a reduced form model that tries to capture the underlying preferences of foreign banks and of governments and private agents in the host countries. Unfortunately, because we only have aggregate data for foreign banks from three countries, we cannot examine the role of home country factors and the importance of certain bank specific characteristics. According to the trade/multinational firm literature, one of the main differences between exports and FDI, is that the latter involves paying sunk costs, typically associated with entry requirements and startup costs. Depending on their size, sunk costs can in principle be an important consideration for banks in deciding the type of presence to have in a country. If significant, we expect to find that sunk costs have an adverse effect on the share of local claims held by foreign banks. Obviously, measuring sunk costs is difficult. In the trade literature, these 7A panel estimation would not add much information given the very small number of home countries for which this data is available. In addition, cross-country equations allow us to compare the results obtained for each home country and for each regressor. 11 have often been captured by country fixed effects. However, this cannot be applied to a cross- section analysis such as ours. Furthermore, many other characteristics of a country not necessarily related to sunk costs may be captured by the fixed effects. In our study, we interpret sunk costs as the costs associated with entering the market and starting up a business. We include two separate measures of these costs. Our first measure controls for the minimum capital requirements banks need to comply with to begin operations in the countries in our sample. These data comes from a worldwide survey of bank regulators summarized by Barth, Caprio, and Levine (2001). We use data on the actual monetary amount of capital (expressed in U.S. dollars) banks have to put down and we also take into account the regulations in each country as to what may constitute bank capital. In particular, we build an index of initial capital stringency that can take values from 0 to 3, with higher number indicating greater stringency. This index comes from summing up responses to the questions below in the following way: (1) Can initial and subsequent infusions of regulatory capital include assets other than cash or government securities? 1 if no. (2) Can the initial infusion of capital be based on borrowed funds? 1 if no. (3) Are the sources of funds that count as regulatory capital verified by the regulatory or supervisory authorities? 1 if yes. Finally, we combine the dollar measure with the index of initial capital stringency by obtaining the principal component of the two series. We refer to the first principal component of the two series as Minimum capital requirements.8 8It should be noted that cross-border lending also requires capital for the head office and that it will generally be higher for more vulnerable host countries, for the same loan size. While the two needs (the minimum capital requirement to operate locally and the capital needed for cross-border lending) are not easily comparable, minimum capital requirements should generally involve larger costs when a bank starts to operate locally. 12 Our second measure of entry and startup costs, tries to capture the actual number of procedures, time, and costs to initiate operations in a country. Data on the procedures required to open banks across countries come from Barth, Caprio, and Levine (2001). They construct an index of bank entry requirements that takes values from 0 to 8 depending on whether banks are required to submit none, some, or all of the following to issue a bank license: (1) Draft by-laws. (2) Intended organizational chart. (3) First 3-year financial projections. (4) Financial information on main potential shareholders. (5) Background/experience of future directors. (6) Background/experience of future managers. (7) Sources of funds to be used to capitalize the new bank. (8) Intended differentiation of new bank from other banks. This index is combined using the method of principal components with two other measures of the costs of starting a business that come from the World Bank Doing Business Indicators. In particular, we include survey estimates of the actual monetary costs (in dollars) and, separately, the time (in number of days) involved in opening a business across countries. The first measures all identifiable official expenses in setting up any business (not necessarily banking), namely fees, costs of procedures and forms, fiscal stamps, legal and notary charges, etc.9 The second, the 9 As stated in the Doing Business Report "the text of the Company Law, the Commercial Code, and specific regulations and fee schedules are used to calculate the costs [of starting a business]. If there are conflicting sources and the laws are not clear, the most authoritative source is used. The constitution supersedes the company law, and the law prevails over regulations and decrees. If conflicting sources are of the same rank, the source indicating the most costly procedure is used, since an entrepreneur never second-guesses a government official. In the absence of fee schedules, a governmental officer's estimate is taken as an official source. In the absence of a government officer's estimates, estimates of incorporation lawyers are used. If several incorporation lawyers provide different estimates, the median reported value is applied. In all cases, the cost excludes bribes." A list of all procedures considered in estimating the costs of starting a business can be found in: 13 time variable, captures the number of days to satisfy all procedures that need to be completed before a business license is issued. We refer to the first principal component of the entry bank requirement index, the costs, and time to start a business as Non-capital entry requirements and start up costs. In the export vs. FDI literature, the higher sunk costs involved in the latter are traded off against the higher transport costs involved in the former. In banking, physical transportation costs are not likely to be as important. Instead, the literature has discussed the role of informational costs (see for example Buch, 2003 and Galindo et al., 2003). There are two main types of informational costs in the international banking business. On the one hand, there are the costs of screening and monitoring borrowers. These costs will tend to increase the larger the distance between the bank and its customers. On the other hand, there are the costs to banks' CEO and shareholders of monitoring the affiliates' managers and loan officers on the ground. These costs will also increase the larger the distance between the CEO and the manager/loan officer. Therefore, in extending bank claims overseas, the informational costs of monitoring borrowers decline when banks decide to extend claims locally to foreign countries via their affiliates in these countries. But, at the same time, by growing the share of foreign claims they extend locally, international banks increase the costs of monitoring loan officers and affiliate managers. Thus, the net impact of informational costs on the share of local lending is an empirical question. Testing the importance of informational costs in international banking decisions is complicated further by the fact that these cannot be measured directly. Instead, they are often captured by indicators of geographic, cultural, legal and economic "distance" between countries that are expected to affect information costs. As distance between an international bank and a http://rru.worldbank.org/Documents/DoingBusiness/Methodology/StartingBusiness/StartingBusiness.pdf 14 country increases, both, the costs of monitoring foreign clients and bank managers or loan officers in bank affiliates rise. The discussion above suggests that the distance between the bank and its customers increases the comparative advantage to international banks of extending claims locally, since local presence helps to overcome some of these informational costs. Therefore, on the basis of the costs of monitoring borrowers, distance measures should have a positive impact on the share of local claims. On the other hand, as distance measures increase, the rising costs of monitoring loan officers and affiliate managers suggest that distance might have a negative impact on the share of local claims. Thus, a priori the sign of distance measures on the share of local claims is unclear. We control for the geographic distance between the foreign banks' country of origin and the host country by including the log of the miles between the two countries' capital cities. This information comes primarily from Glick and Rose (2002) and the Central Intelligence Agency (CIA) World Factbook.10 We also take into account that, nowadays, instant methods of communications like phones and internet connections can help bridge physical distances, making project and management oversight and access to information possible from far away places. In our regressions, we use the number of internet hosts per 1000 people in each borrowing/claim recipient country as measures of access to communications, which are like to affect monitoring and informational costs.11 These data come from the World Bank World Development Indicators. We expect that lending cross-border is easier to countries with better communications, allowing 10When data was missing from this source, we used a software provided on the US Department of Agriculture webpage that calculates distances between more than 220 capital cities worldwide (see http://www.wcrl.ars.usda.gov/cec/java/capitals.htm). 11In estimations that we do not report here, we also included measures of the number of phone lines in each claim recipient country. The results are not significantly different. Therefore, to save space we only report results using the number of internet hosts. 15 project monitoring and information gathering to be done "offsite". Thus, we foresee that this variable will have a negative impact on the share of local to total foreign claims. We measure "economic" proximity between the home and the host country by including a measure of bilateral trade between these economies. This information comes from the IMF Direction of Trade Statistics.12 Finally, we measure home-host country "cultural" and "institutional" distance by including, respectively, a dummy that equals one if countries share a common language with the lender country and a dummy that equals one if countries have a common legal origin (which will then result in similar institutions). Data on common language and legal origin come from Glick and Rose (2002) and the CIA World Factbook. Government intervention in a country's financial sector via regulations and/or taxation might affect the share of local to total foreign claims countries receive in different ways, depending on the kind of intervention. For example, controls on cross-border activity will negatively affect the level of cross-border claims and, therefore, will increase the share of local claims. Similarly, restriction on on-shore foreign currency lending, might also lead to a rise in cross-border claims to the degree that banks prefer to extend claims in foreign currencies. On the other hand, limits on "banking freedom", in particular restrictions on bank activities and controls that affect foreign entry and foreign ownership of banks, will have the opposite effect. In our estimations, we control for regulations affecting the volume of cross-border and local claims, respectively. First, we include a dummy that equals one if the country has controls on cross-border financial credits. Second, we control for restrictions on on-shore foreign currency lending by including a dummy for the countries where such restrictions are in place. Data for both variables come from the IMF Annual Report on Exchange Arrangements and 12Ideally, we would have liked to control as well for the share of non-financial FDI from the home to the host country, however, this information was missing for many of the countries in our sample. 16 Restrictions. Third, in order to control for restrictions on bank activities and on financial FDI, we include the Index of Banking Freedom produced by the Heritage Foundation. This index takes into account the extent of restrictions on the ability of foreign banks to open branches and/or subsidiaries, how heavily regulated the financial system is, the presence of state-owned banks, whether the government influences the allocation of credit, and whether banks are free to offer all types of financial services. This index takes values from 1 to 5, where higher values represent less freedom in the banking sector. Bank profits arising from cross-border activities are taxed at the rate prevalent in the banks' country of origin, while taxes on their FDI activities or local claims depend on the hosts' tax rates. Other things equal, we expect to find relatively lower levels of FDI activity or a lower share of local claims in countries with higher corporate taxes. In our study, taxes refer to the top corporate income taxes in each country as reported by the Heritage Foundation. Ideally, we would have also liked to control for taxes on financial intermediation, but such data is not available on a cross-country basis. Studies on both, cross-border claims and financial FDI, document that economies of scale, profit opportunities, and risks might also affect banks' decisions to expand internationally. Thus, the degree to which each of these factors might affect the share of local claims is an empirical question. However, a priori we could speculate that since the returns from cross-border lending are limited to the initial claims extended plus an interest rate, while extending local claims often implies an equity participation in a local affiliate (bringing the potential for unbounded gains and losses13) scale economies, profit opportunities, and risks might be more important in driving FDI or local claims relative to cross-border loans. 13This might be different for branches and subsidiaries due to the financial independence but also differences exist depending on the banks' country of origin. 17 Following existing international banking studies, we capture the potential for scale economies and profit opportunities in two ways. First, we include the log of constant dollar GDP. Second, we include the growth of GDP as a measure of growth opportunities in the country. Economic growth may also be interpreted as a measure of default risk, since in countries that do not grow, borrowers might have a harder time repaying their obligations to banks. Data on both of these variables come from the World Bank World Development Indicators or the IMF International Financial Statistics, depending on the country. In international banking, pricing risks materialize when the value of the claims or return on the claims held by foreign banks declines as a result of exchange rate or other price changes. We measure price risk by controlling for inflation in our estimations.14 Data on inflation comes from the World Development Indicators. While both cross-border and local foreign claims are exposed to price and default risks, local claims are also subject to expropriation risks. We capture the extent of expropriation risks by including a measure of property rights in the host countries. This measure is an index compiled by the Heritage Foundation and reported as part of the Index of Economic Freedom. Higher levels of this index indicate worse property rights in the country. In some estimations, we replace the measure of default, price, and expropriation risks for one composite country risk measure as collected by the International Country Risk Guide (ICRG) in its country risk index.15 We estimate the following equation to investigate the implications for financial stability of the mix (cross-border versus local) of foreign claims: Foreign claims volatilityi,j=0 + 1Share of local claimsi,j + 2Growth volatility,j + 3Price volatilityj + 4Volatility of claims to other countriesi,j + 5Banking crisis indicatorj + µi,j (2) 14Results remain unchanged if we used the change in the exchange rate instead. 15This index is defined by ICRG so that higher values mean less risk, but in our estimations we reverse the sign on this variable to give it the more intuitive interpretation that higher values mean more risk. 18 where, as before, i indicates the lender or home country and j refers to the host or borrowing country. Once again, we estimate separate cross-country equations for Italian, Spanish and US bank claims. In each of these equations, Foreign claims volatility refers to the standard deviation of i country banks' total foreign claims vis-à-vis country j over the sample period divided by the average level of foreign claims to that country. The purpose of estimating equation (2) is to uncover whether, as some have argued (see de Haas and Levyled 2002, Peek and Rosengren 2000, and Palmer 2000), the volatility of foreign claims is lower for countries where foreign claims are predominantly local (i.e., extended through the overseas affiliates of international banks). Thus, we are primarily interested in the coefficient on the share of local claims. However, in order to obtain consistent estimates of this coefficient, it is important to adequately control for other factors that might influence the volatility of foreign claims. Specifically, we control for the volatility of growth and inflation in the host economy, for whether the host country experienced a banking crisis over the period of study, and for the volatility of foreign claims to countries other than j. Growth and inflation volatilities are captured by the standard deviation of GDP growth and inflation in the host countries over the entire period of study. In some estimations, we replace these variables for the composite ICRG country risk index. The crisis indicator is a dummy, which equals one for those host countries that experienced a banking crisis over the period of study (see Caprio and Klingebiel 2003). The variable labeled volatility of claims to other countries is the standard deviation of claims from banks from country i to countries other than host j divided by the average level of foreign claims to countries other than j. The idea behind this variable is that the volatility of claims to a specific host might be influenced by what 19 foreign banks do and experience in other host countries or even in their home countries (which affect all hosts simultaneously). IV- Empirical Results We first discuss our results on the determinants of the mix of foreign bank claims across countries and then proceed to present our findings regarding the volatility of total foreign claims. A detailed list and explanation of all the variables included in our empirical models is presented in Table 4. Because in many cases the share of local claims is zero, we report Tobit estimations to take into account the censoring that occurs at zero. In other words, we assume that the underlying dependent variable is a latent variable, which measures the desired share of local claims that banks would like to hold. As a result, the observed share is above zero only when this latent variable is above a certain threshold.16 Table 5 presents the result from six specifications, two for each country. In particular, the first three columns report results for Italian, Spanish and U.S. banks, controlling for minimum capital requirements. Because this information is missing for a number of countries, we also report estimations excluding this variable, as shown on the last three columns of Table 5. A number of interesting results emerge from the estimations shown in Table 5. First, lack of freedom for banks to operate in a host country has a significant negative impact on the willingness of foreign banks to extend local claims. A standard deviation change in this variable leads to at least a 0.7, 0.5 and 0.35 standard deviation reduction in the share of Italian, Spanish, and U.S. bank local foreign claims, respectively. Other forms of government intervention in the 16We also tried to estimate a Heckman model to take into account the possibility of selection bias, but estimations did not converge because there are two few cases when total claims are 0 (i.e., when the share of local claims is not observed). 20 economy, such as restrictions on foreign credits or the level of taxation, are only significant for Spanish and Italian banks, respectively. Second, scale economies and profit opportunities, as captured primarily by the size of the country, have a consistently positive impact on the share of local claims. Foreign banks are more willing to extend claims locally where the potential for business is larger. In small countries, banks might not be willing to establish local operations, which involve paying fixed cost of entry and operations, given that the potential for business is limited. A one standard deviation change in the log of constant dollar GDP leads to at least a 1.8, 1.2, and 0.7 standard deviation change in the share of Italian, Spanish , and U.S. bank local foreign claims, respectively. Third, controlling for size, minimum capital requirements and other entry costs are also important determinants of the share of local claims, but less consistently so than the other factors mentioned so far. Minimum capital requirements have a negative impact on the share of local claims held by Spanish banks, while non-capital entry startup costs seem to affect U.S. banks' share of local claims. A one standard deviation change in minimum capital requirements leads to a 0.35 standard deviation decline in Spanish banks' share of local claims. A one standard deviation change in entry costs results in a 0.2 standard deviation drop in U.S. banks' share of local claims. Fourth, informational costs matter, since several of the variables included to proxy for these costs are found to be significant. Better means of communications (as proxied by the number of internet hosts) have a negative impact on the share of local claims across all banks, suggesting that access to communications may lessen the need for local operations and increase banks' willingness to conduct more of their business cross-border. A one standard deviation change in the number of internet hosts per 1,000 people, results in at the least a 0.6, 0.2, and a 21 0.35 standard deviation change in the share of local claims from Italian, Spanish, and U.S. banks, respectively. On the other hand, controlling for access to communications, geographic distance has a negative impact on the share of local claims, signifying perhaps that banks feel that their ability to monitor local operations is adversely affected by distance. However, this is true only for Italian and Spanish banks, maybe because U.S. banks' governance structure or greater experience in international banking allows them to view distance as less of an obstacle. Cultural and institutional affinity is especially important for both Spanish and U.S. banks, since we find that having a common language in the case of Spanish banks and sharing the same legal origin both for Spanish and U.S. banks has a positive impact on the share of local claims. The share of Spanish banks' local claims is approximately 63 percentage points higher in Spanish speaking countries, relative to those where Spanish is not the local language. For countries with a common legal origin with the U.S., the share of U.S. banks' local claims is up to 18 percentage points higher than the rest. Fifth, economic and expropriation risks do not seem to affect the mix of foreign bank claims, once we control for other factors. This does not necessarily mean that banks do not pay attention to these factors. It might be that these factors affect both cross-border and local claims in similar ways, making their impact on the ratio statistically indistinguishable from zero. Also, it is worthwhile pointing out that there is a significant correlation between country size and property right protection, which might cause the latter to be insignificant in our sample. To verify the robustness of the results reported so far, Table 6 presents a number of additional estimations. In particular, in the first three columns of Table 6 we substitute the measure of lack of banking freedom for a more direct measure of restrictions on foreign bank entry, as measured by the actual percentage of foreign bank applications denied. The data come 22 from Barth, Caprio, and Levine (2001). The number of countries for which this information is available is smaller than in our previous estimations. This might explain why this variable is significant for Italian and Spanish banks, but not for U.S. banks. In columns (3)-(6) of Table 6, we present estimations where real GDP growth, inflation and property rights, treated so far as measures of default, price and expropriation risks, respectively, are replaced by one composite country risk measure. As with the individual measures, we find that risks do not seem to consistently affect the mix of foreign bank claims. Finally, the last three columns in Table 6 show results controlling for the total level of foreign bank claims each country receives. These estimations try to address the point that perhaps local presence is larger in countries where banks are willing to have a larger exposure. We find no consistent evidence of this. Our second set of results, reported on Tables 7 through 9, examine the implications of the mix of local versus cross-border foreign claims for the stability of foreign bank financing across countries. In each table, we include seven different specifications from least to most general. In the first column of each table, we include only the variable of interest: the share of local to total foreign claims.17 The remaining columns include other controls that might also affect total foreign bank claims volatility. In particular, we control for price and output volatility, the level of financial sector development (as proxied by the ratio of private credit to GDP), and the behavior of foreign claims vis-à-vis other countries. Furthermore, in some estimations, we include a composite measure of country risk. 17We treat as outliers and, therefore, we exclude from the regressions cases where the volatility of the share was very high due to some unusual event or transaction. This meant dropping Croatia, Poland, and the Slovak republic from our sample of countries in the estimates for Italian banks, since these countries experienced an abrupt change in the share of local Italian claims due to mergers and acquisitions. 23 Table 7 presents the results for Italian banks. The share of local claims has a consistently negative impact on the volatility of foreign bank claims, no matter what other variables are included in the regression. On average, a one standard deviation change in the share of Italian banks' local claims would result in a 0.19 standard deviation change in the volatility of Italian banks' total claims. In some estimations, the volatility of Italian banks' foreign claims is also impacted by the volatility of prices, the level of development of the local financial sector, and the composite measure of country risk in the host countries. Results for the volatility of Spanish bank foreign claims are shown on Table 8. Here too, we find that the share of local claims seems to be associated with lower foreign bank claims volatility. A one standard deviation change in the share of local Spanish banks' claims would result in a 0.24 standard deviation change in the volatility of total Spanish banks' claims. In the case of Spain, there is also evidence that output volatility in the host country contributes to foreign bank lending volatility or procyclicality. The negative relationship between the volatility of foreign bank claims and the composition of these claims is further supported by the results for U.S. banks. Once again, we find that the volatility of foreign bank claims is lower in countries where banks' exposure to the country is primarily in the form of local claims. A one standard deviation change in U.S. banks' share of local claims would result in a 0.45 standard deviation change in the volatility of U.S. banks' total foreign claims. As in the case of Spanish banks, U.S. bank volatility seems to be higher for countries with large output volatility and less developed financial sectors. 24 V- Conclusions While a vast literature exists analyzing the cross-border and local activities of multinational banks across countries, the determinants and implications of the mix of international bank claims had not, to our knowledge, been explored before. Using data on the Italian, Spanish, and U.S. bank claims vis-à-vis more than 100 countries, this paper sought to fill this void in the multinational banking literature. First, we estimated a reduced form empirical model of the share of local to total foreign claims across host countries. In particular, we modeled the share of local to total foreign claims as a function of entry requirements and startup costs, informational costs, regulatory barriers to banking activities and foreign bank participation (or what we call lack of banking freedom), taxation, as well as profit opportunities and risks. We found that regulatory barriers to banking and measures of business opportunities and scale economies have the most consistently significant impact on the share of local bank claims. This share is smaller in countries that limit banking freedom. On the other hand, the share of local foreign bank claims tends to be higher in larger countries with better business opportunities. Entry requirements and startup costs, as well as informational costs, also seem to be important determinants of the share of local claims, but their impact is smaller and less robust. Second, we presented evidence that the mix or composition of foreign bank financing affects the stability of foreign bank claims to host countries. Countries where a larger share of foreign claims is extended through the local affiliates of foreign banks, rather than through cross- border loans, tend to enjoy more stable foreign financing. Our findings suggest that in order to reduce the volatility of foreign bank financing, countries should try to increase the share of foreign claims channeled through the local affiliates 25 of foreign banks in their countries. According to our results on the determinants of the share of local presence, host countries can influence this share by lowering regulatory barriers to bank activities and foreign bank participation and by improving business opportunities in their market. While this study is, to our knowledge, the first to examine the determinants and implications of the mix of foreign bank claims, the analysis conducted is not without limitations. In particular, the time series analyzed is short and the number of home countries considered is small. Thus, it would be important that as more data become available, further research is conducted to verify the robustness of our results over more comprehensive datasets. Also, it should be noted that the mix of local versus cross-border foreign financing may have a bearing on questions other than financial stability, which host countries may consider important. For example, host countries may be concerned about the type of financing received, because of their balance of payments needs. Exploring these alternative implications from the mix of foreign bank claims is also left for future research. 26 References Ball, C. A., Adrian Tschoegl (1982). The Decision to Establish a Foreign Branch or Subsidiary: An Application of Binary Classification Procedures. Journal of Financial and Quantitative Analysis 17: 411­24. Bank for International Settlements Consolidated Banking Statistics http://www.bis.org/statistics/consstats.htm Barajas, Adolfo, Roberto Steiner, and Natalia Salazar (2000). The Impact of Liberalization and Foreign Investment in Colombia's Financial Sector. Journal of Development Economics 63(1): 157-96. Barth, James, Gerard Caprio and Ross Levine (2001). The Regulation and Supervision of Banks around the World: A New Database. Policy Research Working Paper No. 2588. World Bank: Washington, D.C. Berger, Allen, Leora Klapper and Gregory Udell (2001). The Ability of Banks to Lend to Informationally Opaque Small Businesses." Journal of Banking and Finance 25 (12): 2127-2167. Berger, Allen, Nathan Miller, Mitchell Peterson, Raghuram Rajan, and Jeremy Stein (2002). Does Function Follow Organizational Form? Evidence from the Lending Practices of Large and Small Banks. National Bureau of Economic Research Working Paper No. 8752. Bonaccorsi di Patti, Emilia and Giorgio Gobbi (2001). The Changing Structure of Local Credit Markets: Are Small Businesses Special? Journal of Banking and Finance 25(12): 2209- 37. 27 Brainard, S. Lael. (1997). An Empirical Assessment of the Proximity-Concentration Trade-off Between Multinational Sales and Trade. American Economic Review 87 (4): 520-544. Brealey, Richard A., and Evi C. Kaplanis (1996). The Determination of Foreign Banking Location. Journal of International Money and Finance 15 (4): 577­97. Buch, Claudia M. (2000). Why Do Banks Go Abroad? Evidence from German Data. Financial Markets, Institutions and Instruments 9 (1), 34-67. Buch, Claudia M. (2003). Information or Regulation: What Drives the International Activities of Commercial Banks? Journal of Money, Credit and Banking 35(6): 851-869. Buch, Claudia M. and Gayle L. DeLong, (2001). Cross-Border Bank Mergers: What Lures the Rare Animal? Kiel Institute of World Economics Working Paper No. 1070. Buch, Claudia M. and Alexander Lipponer (2004). FDI Versus Cross-border Financial Services: The Globalization of German Banks. Deutsche Bundesbank Discussion Paper - Series 1 05/2004. Caprio, Gerard and Daniela Klingebiel (2003). Episodes of Systemic and Borderline Banking Crises. Mimeo. The World Bank. Central Intelligence Agency World Factbook (2004). www.cia.gov/cia/publications/factbook/ Cho, Kang Rae, Suresh Krishnan, and Douglas Nigh (1987). The State of Foreign Banking Presence in the United States. International Journal of Bank Marketing 5 (2): 59­75. 28 Claessens, Stijn, Asli Demirgüç-Kunt, and Harry Huizinga (2000). The Role of Foreign Banks in Domestic Banking Systems. In Stijn Claessens and Marion Jansen, eds., The Internationalization of Financial Services: Issues and Lessons for Developing Countries. Boston: Kluwer Academic. Clarke, George, Robert Cull, Maria S. Martínez Pería, and Susana Sánchez (2003). Foreign Bank Entry: Experience, Implications for Developing Countries, and Agenda for Further Research. World Bank Research Observer, 18 (1): 25-40. Clarke, George, Robert Cull, Maria S. Martínez Pería, and Susana Sánchez (2005). Bank lending to Small Businesses in Latin America: Does Bank Origin Matter? Journal of Money, Credit, and Banking, 37(1): 83-118. Dages, Gerard, Linda Goldberg, and Daniel Kinney (2000). Foreign and Domestic Bank Participation in Emerging Markets: Lessons from Mexico and Argentina. Federal Reserve Bank of New York Economic Policy Review, 17-36. Dahl, Drew and Ronald E. Shrieves (1999). The Extensions of International Credit by U.S. Banks: A Disaggregated Analysis, 1988-1994. Journal of International Money and Finance 18: 153-167. Demirguc-Kunt, Asli, Luc Laeven and Ross Levine (2004). Regulations, Market Structure, Institutions, and the Cost of Financial Intermediation. Journal of Money Credit and Banking 36(3.2): 593-622. Denizer, Cevdet (2000). Foreign entry in Turkey's banking sector, 1980-1997. In S. Claessens and M. Jansen (Eds.), The internationalization of financial services: Issues and lessons for developing countries pp. 389-406. Boston, MA: Kluwer Academic Press. 29 Detragiache, Enrica and Poonam Gupta (2004). Foreign Banks in Emerging Market Crises: Evidence from Malaysia. IMF working paper No. 04/129. Fisher, Adlai, and Philip Molyneux (1996). A Note on the Determinants of Foreign Bank Activity in London between 1980 and 1989. Applied Financial Economics 6 (3): 271­77. Focarelli, Dario and Alberto Pozzolo (2001). The Patterns of Cross-Border Bank Mergers and Shareholdings in OECD Countries. Journal of Banking and Finance, Vol. 25 (12): 2305-2337. Focarelli, Dario and Alberto Pozzolo (2006). Where Do Banks Expand Abroad? An Empirical Analysis. Journal of Business, Vol. 79 (1). Galindo, Arturo, Alejandro Micco, and Cesar Serra (2003). Better the Devil That You Know: Evidence on Entry Costs Faced By Foreign Banks. Inter-American Development Bank Working Paper No. 477. Glick, Reuven and Andrew K. Rose (2002). Does a Currency Union Affect Trade? The Time Series Evidence. European Economic Review 46 (6): 1125-1151. Goldberg, Lawrence, and Robert Grosse (1994). Location Choice of Foreign Banks in the United States. Journal of Economics and Business 46 (5): 367­79. Goldberg, Lawrence, and Denise Johnson (1990). The Determinants of U.S. Banking Activity Abroad. Journal of International Money and Finance 9 (2): 123­37. Goldberg, Lawrence, and Anthony Saunders (1980). The Causes of U.S. Bank Expansion Overseas: The Case of Great Britain. Journal of Money, Credit, and Banking 12 (4): 630­ 43. 30 Goldberg, Lawrence and Anthony Saunders (1981a). The Determinants of Foreign Banking Activity in the United States. Journal of Banking and Finance 5(1): 17-32. Goldberg, Lawrence and Anthony Saunders (1981b). The Growth of Organizational Forms of Foreign Banks in the U.S.: A Note. Journal of Money, Credit, and Banking 13(3): 365- 74. Guillen, Mauro and Adrian Tschoegl (2000). The Internationalization of Retail Banking: The Case of the Spanish Banks in Latin America. Transnational Corporations, Vol. 9(3): 63- 98. Haas, de, Ralph and Iman van Lelyveld (2002). Foreign Bank Penetration and Private Sector Credit in Central and Eastern Europe. De Nederlandsche Bank Staff Reports No. 91. Haas, de, Ralph and Iman van Lelyveld (2004). Foreign Banks and Credit Stability in Central and Eastern Europe: A Panel Data Analysis. De Nederlandsche Bank Staff Reports No. 109. Helpman, Elhanan, Marc J. Melitz and Stephen R. Yeaple. (2003). Exports Versus FDI. National Bureau of Economic Research Working Paper No. 9439. Heritage Foundation. Index of Economic Freedom. Washington, DC: Heritage Foundation Press. Hultman, Charles W., and Randolph McGee (1989). Factors Affecting the Foreign Banking Presence in the United States. Journal of Banking and Finance 13 (3): 383­96. Jeannau, Serge and Marian Micu (2002). Determinants of International Bank Lending to Emerging Market Countries. Bank of International Settlements Working Papers No. 112. Kawai, Masahiro and Li-Gang Liu (2002). Determinants of International Commercial Bank Loans to Developing Countries. Mimeo. 31 Martínez Pería, Maria Soledad and Ashoka Mody (2004), How Foreign Participation and Market Concentration Impact Bank Spreads: Evidence from Latin America, Journal of Money, Credit, and Banking, 36(3): 511-539. Mian, Atif (2004). Distance Constraints: The Limits of Foreign Lending in Poor Economies. University of Chicago Business School. Mimeo. Moshirian, Fariborz (2001). International Investment in Financial Services. Journal of Banking and Finance 25: 317-337. Nigh, Douglas, Kang Rae Cho, and Suresh Krishnan (1986). The Role of Location-Related Factors in U.S. Banking Involvement Abroad: An Empirical Analysis. Journal of International Business Studies 17 (Fall): 59­72. Palmer, David (2000). U.S. Bank Exposure to Emerging-Market Countries during Recent Financial Crises. Federal Reserve Bulletin, (February): 81-96. Peek, Joe and Eric Rosengren (2000). Implications of the Globalization of the Banking Sector: The Latin American Experience. Federal Reserve Bank of Boston New England Economic Review (September­October): 45­62. Peterson, Mitchell A., Raghuram G. Rajan (1994). The Benefits of Firm-Creditor Relationships: Evidence from Small Business Data. Journal of Finance 49: 3-37. Peterson, Mitchell A., Raghuram G. Rajan (2002). Does Distance Still Matter? The Information Revolution in Small Business Lending. Journal of Finance 57: 2533-2570. Wezel, Torsten. (2004). Foreign Bank Entry into Emerging Economies: An Empirical Assessment of the Determinants and Risks Predicated on German FDI Data. Deutsche Bundesbank Discussion Paper - Series 1 01/2004. 32 World Bank Doing Business Indicators. http://rru.worldbank.org/DoingBusiness/ 33 Table 1: Italian banks' foreign claims vis-à-vis other countries No presence Only cross border Share of local claims (% of total foreign claims) 0.1%-25% 25%-50% 50%-75% >75% Bhutan Albania Slovenia China Singapore Japan Botswana Algeria Austria Chile Canada Brazil Burkina Faso Armenia Netherlands UnitedKingdom Colombia Hungary Burundi Australia Turkey Lebanon Argentina Slovak Rep. Cambodia Bangladesh Ukraine France Hong Kong Croatia Honduras Belarus Czech Rep. United Arab Emirates Poland Niger Bolivia Israel Bulgaria Paraguay Papua Congo Greece Spain Peru Rwanda Costa Rica Belgium United States Cote d'Ivoire Ireland Ecuador Switzerland Egypt Panama El Salvador Germany Estonia Romania Finland Ghana Guatemala India Indonesia Jamaica Jordan Kazakhstan Kenya Korea Kuwait Latvia Lesotho Lithuania Madagascar Malawi Malaysia Mali Mexico Moldova Morocco Namibia Nepal New D11Zealand Nicaragua Nigeria Norway Oman Pakistan Philippines Portugal Russia SaudiArabia Senegal SouthAfrica SriLanka Sweden Sweden Taiwan Thailand Togo Tunisia Venezuela Vietnam Zambia Zimbabwe Table 2: Spanish banks' foreign claims vis-à-vis other countries Share of local claims (% of total foreign claims) No presence Only cross border 0.1%-25% 25%-50% 50%-75% >75% Armenia Albania Poland Switzerland Czech Rep. Colombia Bangladesh Algeria Australia Morocco United States Argentina Bhutan Belarus China France Philippines Hong Kong Botswana Bulgaria Turkey Italy India Portugal Burundi Congo Korea Panama Brazil Cambodia Cote d'Ivoire Netherlands Chile Jamaica Croatia Russia Peru Lesotho Egypt Ecuador Paraguay Nepal Estonia Austria Venezuela Niger Finland South Africa Mexico Rwanda Ghana CostaRica Bolivia SriLanka Guatemala Canada Zambia Indonesia Greece Israel Singapore Jordan Honduras Kazakhstan Ireland Kenya Hungary Kuwait Japan Lebanon Belgium Lithuania El Salvador Madagascar Germany Malawi United Kingdom Malaysia Mali Moldova Namibia New Zealand Nicaragua Nigeria Norway Oman Pakistan Papua Romania Saudi Arabia Senegal Slovak Republic Slovenia Sweden Taiwan Thailand Togo Tunisia Ukraine United Arab Emirates Vietnam Zimbabwe Table 3: U.S. banks' foreign claims vis-à-vis other countries Share of local claims (% of total foreign claims) No presence Only cross border 0.1%-25% 25%-50% 50%-75% >75% Armenia Albania Kuwait CostaRica Mexico Singapore Burkina Faso Belarus Slovenia Germany Lebanon Cote d' Ivoire Burundi Bhutan Ghana Morocco Brazil Taiwan Madagascar Botswana Austria China Romania Pakistan Rwanda Cambodia Zimbabwe Oman Nigeria Senegal Congo Finland Indonesia Panama Malaysia Croatia Netherlands Peru Hungary Paraguay Estonia Malawi Colombia Greece Zambia Latvia Norway South Africa Korea Lithuania Sweden Bolivia Argentina Mali Sweden Algeria Philippines Moldova Israel Jamaica Slovak Republic Namibia Ecuador Spain New Zealand Nepal France Kazakhstan Chile Nicaragua Bulgaria Japan Niger Honduras Egypt Saudi Arabia Guatemala Jordan Togo Switzerland Bangladesh Venezuela Canada Italy Kenya Russia SriLanka Ireland Papua Turkey Vietnam Portugal Poland El Salvador Thailand Ukraine United Arab Emirates Belgium Tunisia India Australia Czech Republic Hong Kong Table 4: Definition and sources for the variables used Variable Name Definition Source Dependent variables Banks' share of local to total foreign Banks' local claims vis-à-vis host j divided by banks' total foreign Consolidated Banking Statistics (BIS), claims vis-à-vis host (claim recipient) claims to that host. (Constructed separately for Italian, Spanish, the Italian, and Spanish central banks. country j. and U.S. banks) Banks' foreign claims volatility vis-à- Standard deviation of foreign claims vis-à-vis host j divided by the Consolidated Banking Statistics (BIS), vis host (claim recipient) country j. average foreign claims to that host. (Constructed separately for the Italian, and Spanish central banks. Italian, Spanish, and U.S. banks) Independent variables Minimum capital requirements First principal component of (1) dollar capital requirements to start Regulatory survey summarized in a bank and (2) an index of initial capital stringency which varies Barth, Caprio, and Levine (2001) between 0 and 3 depending on whether the country (a) prohibits capital in forms other than cash or government paper, (b) prohibits the use of borrowed funds, and (c) verifies the sources of capital. Larger values indicate greater stringency. Non-capital entry requirements First principal component of (1) index of entry bank requirements World Bank Doing Business Report and startup costs a la Barth et al. (2001), (2) all identifiable official expenses in setting up a business (fees, costs of procedures and forms, fiscal stamps, legal and notary charges, etc.) from Doing Business Indicators (World Bank) and (3) time it takes to start up a business from the same source. Bilateral distance Log of the distance in miles from a host (claim recipient) country's Glick and Rose (2002) and World capital to the banks' country of origin. Constructed separately for Factbook (CIA). Italy, Spain, and the U.S., respectively. Bilateral trade Share of exports plus imports between a host (claim recipient) Export and import data are from country and the banks' country of origin (home) relative to home Direction of Trade Statistics (IMF). GDP. Constructed separately for Italy, Spain and the U.S. GDP is from the World Development Indicators (World Bank). Common language Dummy equal to 1 if the host (or claim recipient country) and World Factbook (CIA) home country (Italy, Spain, and U.S., respectively) speak the same language. Common legal origin Dummy equal to 1 if the host (or claim recipient country) and World Factbook (CIA) home country (Italy, Spain, and U.S., respectively) share the same legal origin. Internet hosts Internet hosts per 1000 people in host or claim recipient country. World Development Indicators (World Bank) Log of GDP Logarithm of constant dollars GDP. World Development Indicators (World Bank) Corporate tax rate Top corporate tax rate. Index of Economic Freedom (Heritage Foundation) Restriction on foreign credits Dummy variable that equals one if there are any restrictions on non-Annual Report on Exchange resident lending to residents (i.e., cross-border loans). Arrangements and Exchange Restrictions (IMF) Table 4: Definition and sources for the variables used (cont.) Variable Name Definition Source Independent variables Lack of banking freedom An index from 1 to 5 with 1 indicating the fewest controls. Index of Economic Freedom Index measures the following factors: (1) government (Heritage Foundation) ownership of banks (2) restrictions on the ability of foreign banks to open branches and subsidiaries (3) government influence over the allocation of credit (4) government regulations (5) restrictions on bank activities. Proportion of foreign bank applications Percentage of requests by foreign banks to start operations Barth, Caprio, and Levine (2001) denied that were rejected out of total applications received. Restrictions on foreign currency lending Dummy variable that equals 1 if the country restricts foreign Annual Report on Exchange currency lending in the local banking sector. Arrangements and Exchange Restrictions (IMF) Inflation Change in host (claim recipient) country GDP deflator. World Development Indicators (World Bank) GDP growth Host (claim recipient) GDP growth (annual percentage). World Development Indicators (World Bank) Property rights Index of Property Rights from 1 to 5 with 1 being the Index of Economic Freedom highest level of protection. The index considers (1) freedom (Heritage Foundation) from government influence over the judicial system, (2) commerical code defining contracts, (3) sanctioning of foreign arbitration, (4) government expropriation, (5) delays in the judicial system, (6) legally granted and protected property. Country risk An index from 1 to 100 that is the composite of twelve International Country Risk Guide variables measuring political, economic and financial risk. Higher values indicate more risk. Inflation volatility The standard deviation of the change in host (claim World Development Indicators recipient) GDP deflator. (World Bank) Real growth volatility The standard deviation of host (claim recipient) GDP World Development Indicators growth. (World Bank) Private credit to GDP Bank claims on the private sector in current local currency International Financial Statistics over local currency GDP. (IMF) Bank crisis Dummy variable that equals one if the host experienced a Caprio and Klingebiel (2003) bank crisis between 1995 and 2002. Volatility of banks' claims to other Standard deviation of all banks' foreign claims to countries Consolidated Banking Statistics countries other than j divided by the average claims to countries other (BIS), Italian and Spanish central than host j. banks. Table 5: Determinants of the share of local to total foreign claims Tobit regressions for the share of Italian, Spanish, and U.S. banks' local claims, respectively, vis-à-vis a given host (claim recipient country), divided by their total claims to that host. t-statistics are in brackets. *, **, *** denote significance at 10, 5, and 1 percent respectively. Controlling for minimum capital Excluding minimum capital Italian Spanish U.S. Italian Spanish U.S. Variables banks banks banks banks banks banks Minimum capital requirements 10.692 -9.631 -0.706 [1.48] [2.08]** [0.23] Non-capital entry req. & startup costs -4.002 4.448 -5.472 -3.65 2.397 -6.561 [0.62] [0.94] [1.84]* [0.58] [0.50] [2.38]** Bilateral distance -16.357 -15.253 9.129 -20.187 -13.46 4.025 [2.21]** [2.03]** [1.02] [2.76]*** [1.95]* [0.52] Bilateral trade -12.433 -5.143 2.201 -10.304 -3.654 1.736 [1.73]* [1.17] [0.30] [1.38] [0.79] [0.24] Common language 62.74 4.535 76.095 0.964 [3.47]*** [0.67] [4.64]*** [0.14] Common legal origin 0.376 20.901 12.865 0.345 24.258 18.526 [0.03] [1.93]* [1.52] [0.02] [2.15]** [2.38]** Number of internet hosts -0.199 -0.121 -0.064 -0.09 -0.029 -0.072 [2.29]** [1.87]* [2.54]** [1.79]* [0.84] [2.83]*** Log of GDP 28.27 19.046 10.293 29.062 21.217 10.644 [3.94]*** [3.67]*** [4.15]*** [3.75]*** [4.13]*** [4.60]*** Corporate tax rate -1.409 -0.321 -0.57 -2.145 -0.562 -0.56 [1.72]* [0.52] [1.45] [2.59]** [0.87] [1.45] Restrictions on foreign credits 2.584 28.549 4.555 6.833 16.417 -0.186 [0.13] [2.10]** [0.51] [0.33] [1.21] [0.02] Lack of banking freedom -25.159 -20.695 -10.679 -27.706 -17.751 -12.38 [2.47]** [2.76]*** [2.20]** [2.70]*** [2.47]** [2.59]** Restrictions on foreign currency lending -10.624 3.195 6.231 -10.931 11.264 6.109 [0.67] [0.25] [0.80] [0.65] [0.90] [0.78] Inflation -0.003 -0.514 -0.231 0.047 -0.245 -0.193 [0.01] [1.45] [1.78]* [0.18] [0.68] [1.47] Real GDP growth 0.549 -2.615 0.644 0.114 1.142 0.231 [0.13] [0.80] [0.34] [0.03] [0.34] [0.12] Property righs 7.993 -0.902 8.041 17.413 0.676 8.966 [0.70] [0.10] [1.57] [1.56] [0.08] [1.78]* Constant -472.093 -305.304 -280.08 -474.176 -399.861 -239.39 [2.80]*** [2.45]** [3.00]*** [2.70]*** [3.21]*** [2.97]*** Observations 76 73 79 86 79 88 Pseudo R-squared 0.099 0.195 0.069 0.096 0.181 0.066 s mia ** ** *** *** *** * *** . cl nks ng U.S ba 6.807- 2.50][ 5.413 0.71][ 10.392 1.25][ 0.449 0.07][ 16.734 2.17][ 0.08- 3.14][ 12.989 4.94][ 0.523- 1.37][ 1.489- 0.17][ 12.867- 2.73][ 6.558 0.85][ 0.21- 1.60][ 0.276- 0.15][ 8.188 1.65][ 0.001- 1.91][ 300.124- 3.49][ 88 reioflatot h *** ** *** *** * ** isna nks ,)yrtnu gnidu Sp ba 1.728 0.39][ 8.441- 1.23][ 4.182- 0.97][ 55.559 3.02][ 22.36 2.11][ 0.041- 1.20][ 15.44 2.89][ 0.328- 0.54][ 18.698 1.47][ 19.894- 2.84][ 4.45 0.36][ 0.22- 0.68][ 0.293- 0.09][ 0.473- 0.06][ 0.001 1.80][ 282.523- 2.24][ 79 .y cl n *** * *** ** ** ** cot elvi In lia nks enip Ita ba 3.321- 0.53][ 20.004- 2.76][ 11.812- 1.41][ 0.488 0.03][ 0.091- 1.81][ 27.733 3.36][ 2.06- 2.43][ 6.865 0.33][ 26.438- 2.48][ 9.89- 0.58][ 0.041 0.16][ 0.147- 0.04][ 15.979 1.38][ 0.000 0.40][ 444.721- 2.38][ 86 ectp ecir laimc(tsoh esrtnecr ** ** *** *** * ** . sk nks ri U.S ba 5.879- 2.11][ 2.852 0.36][ 0.494 0.07][ 2.31 0.34][ 19.206 2.39][ 0.078- 2.99][ 9.026 4.12][ 0.489- 1.30][ 1.72- 0.20][ 8.424- 1.86][ 7.452 0.93][ 0.194 0.44][ 167.329- 2.34][ 86 pe 1 rytnu h *** ** *** ** *** s mialc nevig dna co 5, a gnidu isna nks Sp ba 3.14 0.70][ 9.774- 1.41][ 3.072- 0.69][ 74.667 4.73][ 31.347 2.61][ 0.032- 0.96][ 18.981 4.53][ 0.45- 0.72][ 23.215 1.62][ 16.692- 2.42][ 11.137 0.90][ 1.212- 1.65][ 470.825- 4.03][ 79 ngier siv 10,ta cl In ** * *** ** ** ** fol à--siv,yle ecnacifi n lia nks Ita ba 1.308- 0.21][ 19.552- 2.59][ 8.91- 1.18][ 2.95 0.20][ 0.09- 1.77][ 23.492 3.59][ 1.697- 2.18][ 14.38 0.73][ 22.145- 2.23][ 7.502- 0.45][ 0.175 0.18][ 319.382- 2.29][ 85 ta to tolacolfo tivcepser, gnis ** ** ** *** ** eton edined de aimslc *** snoitacilppa . nks U.S ba 6.812- 2.03][ 2.457- 0.29][ 1.263- 0.17][ 1.894- 0.23][ 21.952 2.38][ 0.073- 2.54][ 9.284 2.99][ 0.151 0.30][ 1.445- 0.13][ 0.896 0.10][ 0.715 0.07][ 0.309- 1.31][ 0.487- 0.21][ 5.458 0.95][ 189.716- 2.02][ 60 h ** *** *** *** ** *** ** erahs **, alc *, ng isna nks lo'sk .st reiof Sp ba 5.28- 1.17][ 16.953- 2.45][ 3.724- 0.78][ 94.626 5.02][ 38.842 3.49][ 0.008 0.29][ 17.727 3.03][ 0.159- 0.22][ 38.23 2.71][ 56.757- 3.52][ 9.024 0.59][ 0.107- 0.20][ 4.045 1.29][ 0.775 0.09][ 351.975- 2.54][ 51 e thfo anb.S. kecarb gnidu nks ba n * * ** *** *** *** *** cl lia tsnani U ni d era In taI 13.621- 1.76][ 7.991- 0.97][ 9.333- 1.43][ 33.076 1.93][ 0.109- 2.40][ 30.93 3.54][ 3.207- 3.16][ 23.809 1.09][ 119.599- 2.96][ 22.761- 1.12][ 0.608 1.39][ 1.869 0.45][ 1.343 0.10][ 614.906- 3.03][ 57 an,h mr is te de anpS, scitsitat stsoc d nie e de th no aliantIfo -st.tsohtaht tuprats ns nding sts e ot nda tioa le y te arhs s nts pplica ssents ncer e mee th rof sts ditserc m nk urc ba bu Ro:6 e snoisserger mialclatot quirer y e ge iginrol hot ner tear ignerof doeerf rieht ignerof th ims ntrel ncat wo lac yb dea ngua ga x no no la le P ta nking igneroffo gr ns s n n D ns n ns P ghsir ksir ign d leb bla itb deiv pitaac dislar trlar intefor te te Gfo etar bafo n D tyr y reofl nt tioa ta riaa n- ila ila mmoo mmoo T To di V No B B C C Numbe goL poro tiocirtse C R kcaL tioroporP tiocirtse tioal rvesb R nfI Glae R peorP untro C taoT nso C O s aimlc gnie for total * ** xed 20 brackets. in ni banks' skir .0040- [2.45]* 0.012 [1.61] 0.008 [0.07] 656.44- [1.04] .00- [1.29] 1.98 [2.74]* 82 0.17 are yrtnu * ** Italian co g .0040- [2.48]* 0.017 [3.03]* 0.014 [0.11] 877.54- ** [1.30] 2.277 [3.64]* 82 0.16 of statisticst- ind ility clunI * ** ** .0040- [2.56]* 0.017 [3.03]* 0.014 [0.11] 1.759 [3.89]* 82 0.16 volat countries. the * on latilityov s recipient) th .0040- [2.62]* 100.0 0.62][ 130.0- [0.83] 0.068 [0.49] 916.44- ** * 50 [0.99] .00- [3.06]* 1.289 [2.39]* 83 0.15 laimc worg (claim d ** igne an host n atiolf .0050- [2.90]* 100.0 2 1.68]*[ .00- [0.51] 0.141 [1.01] 706.83- * forla [1.05] 0.963 [2.15]* 83 0.08 across .y in s loc elvi g ind ** ** claim clunI respect .0050- [2.98]* 100.0 91 [1.68]* .00- [0.50] 0.139 [1.00] 0.531 [4.81]* 83 0.08 banks' ls foreign lian rotn ** ** percent Ita 1 co banks' of No and .0050- [3.05]* 0.554 [7.70]* 83 0.05 e aliantI 5, shar of 10,ta the of t latilityov pac the ecnacifingis esirtnu s m ai cl corehto ng y m to im of The etoned reiof m s GDP latilityov dexni du to it s 7: calol latilityov th ks sis claimfo ed *** n regressions of worg ri cri oni SL **, rytnuo gnik Cr at der ate latility rve ua bs Table O *, arehS latiofnI Real C Ban Vo ivrP tnatsno C O R-sq s laimc igne for ** total xed 18 10 in .0030- [2.98]* 0.003 [0.55] .010- [0.09] .74- [0.76] .00- [0.67] 0.93 [1.33] 79 0.08 brackets. banks' ni skir are yrtnu ** 72 nish co g .0030- [3.15]* 0.006 [0.97] .0120- [0.10] .26- [1.02] 1.122 1.60][ 79 0.07 Spa ind of statisticst- clunI ** .0030- [3.00]* 0.005 [0.93] .0010- [0.00] 0.785 1.72]*[ 79 0.07 volatility countries. the latilityov on ** s th recipient) aimlc worg d .0030- [2.76]* 000.0 [0.19] 840.0 10 10 1.83]*[ 0.051- [0.47] .53- [0.66] .00- [0.73] 0.515 1.80]*[ 79 0.12 (claim an n igne ** host atiolf forla in .0030- [2.82]* 000.0 [0.08] 150.0 85 1.89]*[ .040- [0.40] .04- [0.76] 0.483 [1.83]* 79 0.12 across g s .y loc elvi ind ** ** claim clunI .0030- [2.71]* 000.0 [0.07] 150.0 [1.90]* .0350- [0.36] 0.29 [5.26]* 79 0.11 banks' respect ls foreign rotn ** ** percent co Spanish banks' 1 h No of .0030- [2.92]* 0.421 [8.16]* 79 0.05 and e 5, Spanis shar of 10,ta the oft latilityov pac the ecnacifingis esirtnu s m ai cl corehto ng y m to im of The etoned reiof m s GDP latilityov dexni du to it s 8: calol latilityov th ks sis claimfo ed *** n regressions of worg ri cri oni SL **, rytnuo gnik Cr at der ate latility rve ua bs Table O *, arehS latiofnI Real C Ban Vo ivrP tnatsno C O R-sq s aimlc gnie xed ** for in skir .0050- [3.93]* 0.004 [0.71] 0.039 [0.61] 355.1 20 [0.46] .00- [1.76]* 0.836 1.51][ 85 0.35 total rackets.b in yrtnu ** are co banks' g .0060- [4.44]* 0.008 [1.41] 0.045 [0.69] 517.2 [0.80] 0.965 [1.76]* 85 0.33 ind U.S. ** ** ** of statisticst- clunI .0050- [4.49]* 0.01 [2.96]* 0.053 [0.81] 1.245 [4.44]* 85 0.32 volatility countries. latilityov ** * * ** 10 20 the th on s recipient) worg .0050- [4.12]* .00- [1.01] 640.0 [2.31]* 0.01 [0.16] 790.2 [0.80] .00- [2.53]* 0.408 [2.95]* 87 0.4 d an aimlc n ** * * ** (claim igne host atiolf in .0060- [4.74]* 000.0 [0.60] 640.0 [2.14]* 0.036 [0.61] 955.4 [2.21]* 0.208 [2.64]* 87 0.36 g forla .y ind ** ** ** across s elvi loc clunI .0050- [4.36]* 000.0 [0.39] 550.0 [2.75]* 0.06 [0.99] 0.364 [5.99]* 87 0.32 claim respect ls banks' foreign rotn percent co ** ** U.S. 1 No of banks' and .0050- [4.36]* 0.527 [8.18]* 87 0.2 e S.. 5, U shar of 10,ta the oft latilityov pac the ecnacifingis esirtnu s m ai cl corehto ng y m to im of The etoned reiof m s GDP latilityov dexni du to it s 9: calol latilityov th ks sis claimfo ed *** n regressions of worg ri cri oni SL **, rytnuo gnik Cr at der ate latility rve ua bs Table O *, arehS atiolfnI Real C Ban Vo ivrP tnatsno C O R-sq