The push-pull determinants of cross-border bank claims on emerging countries: the role of parent banks and their subsidiaries.

Transcription

1 The push-pull determinants of cross-border bank claims on emerging countries: the role of parent banks and their subsidiaries. Sophie Brana & Delphine Lahet * Larefi, Bordeaux University Abstract Studies of the determinants of cross-border bank claims are based on the economic situation of the lending and borrowing countries (the push/pull factors), but fail to take account of the strategies of the banks that are at the origin of these flows. International banks may have a dual strategy for the financing of emerging countries: they may grant cross-border loans and/or set up local subsidiaries. So is the presence of foreign banks a factor of attraction or a means of substitution for outside financing? In the frame of a Push & Pull analysis, we study the determinants of cross-border bank claims on a panel of 28 emerging countries vis-à-vis all sectors, then vis-à-vis the non-bank sector, and finally loans granted to the banking sector alone in order to approach parent bank/subsidiary relations. We explicitly integrate banking determinants. We show that the location of foreign banks is indeed a factor of attraction for cross-border bank claims thanks to inter-bank relations between parent banks and their subsidiaries. It is not, however, a guarantee of stability for any such financing. Classification JEL : F23, F32, F36, G01, G21 Keywords: cross border bank claims; located affiliates, banks health; emerging countries * Larefi, Department of Economics, Bordeaux University, Campus de Pessac, Ave. Léon Duguit, PESSAC Cedex, FRANCE. sophie.brana@u-bordeaux.fr; delphine.lahet@u-bordeaux.fr Corresponding author: Delphine Lahet: delphine.lahet@u-bordeaux.fr

2 The push-pull determinants of cross-border bank claims on emerging countries: the role of parent banks and their subsidiaries. 1. INTRODUCTION Further to the collapse of Lehman Brothers in September 2008, many emerging countries suffered from withdrawals of capital, particularly by banks by way of a fall in cross-border claims held by international banks. On average, between September 2008 and March 2009, these claims fell by 12% across the emerging Europe zone (Slovakia down by 41.9%, the Czech Republic 18.4%, Poland 16.7% and Latvia 13.2%). The drop was 15% in Latin America (Argentina down by 21%, Venezuela 30% and Brazil 20%) and 24% in the emerging Asia zone (China down by 32%, Malaysia 26% and Korea 30%) 1. The liquidity and solvability crisis between 2007 and 2009 that hit the major international banks posed many questions regarding the international spillover effects of financial shocks (Cetorelli & Goldberg [2010]; Takats [2010b]; Kamil & Rai [2010]; Hoggarth et al. [2010]). In particular, it raised the question of the vulnerability of emerging countries to the rationing of crossborder bank financing and that of the potentially destabilizing role, for the host country, of parent banks in a situation of financial stress maintaining or withdrawing investment in their subsidiaries located in emerging countries. In reality, international banks have a dual strategy for financing in emerging countries. Firstly, they may finance them directly via cross-border bank claims. In 2013, for instance, these claims appear to have accounted for 31.23% of the GDP of central and eastern European countries (CEECs), 14.26% of the GDP of emerging countries in Asia and 9.42% of GDP in Latin America, thereby reflecting significant external dependency, particularly for Eastern countries (Table 3, Appendix). Secondly, foreign banks set up subsidiaries in emerging countries. This is particularly so in CEECs, where towards the end of the 1990s massive privatisation programmes, aimed at national banks at a time when few domestic private banks existed and when few local investors could afford to buy them out, left the door wide open to foreign investors. On average, foreign banks own three-quarters of CEEC banking assets (Table 3 and Graph 2, Appendix), meaning in this case a double dependency on foreign banks via local subsidiaries and via cross-border claims. The presence of subsidiaries in Latin America and Asia came about much earlier (in the mid-19 th century) but is generally less marked. It increased considerably after the crises that shook Mexico (1994) and Argentina ( ). Today in Latin America, foreign banks own 34% of banking assets (but 75% in Mexico, 47% in Uruguay, 39% in Paraguay and 50% in Peru). In Asia, despite the restructuring operations that followed the financial 1 Calculations by the authors based on data from cross-border bank claim from the Bank for International Settlements (BIS) (locational banking statistics). 2

3 crises of and despite a speed-up of locations in 2003, the presence of foreign banks is lower 2 than in other zones, with an average market share of 11% (but 32% in Indonesia, 19% in Korea and 18% in Malaysia). The aim of this article is to study the determinants of cross-border bank claims. The financial crisis of 2008 showed the vulnerability of emerging countries to the rationing of international bank financing and the necessity to take much more specific account of international bank strategies, most notably by factoring in relations between parent banks and subsidiaries. In traditional literature covering the determinants of capital inflows (Calvo et al [1993], Fernandez-Arias [1996]), very little space is given to either the study of cross-border bank claims or the accounting for bank-related explicative variables. The objective of this article is to study the determinants of cross-border bank claims on 28 emerging countries 3 between 1995 and 2009, basing our work on traditional Push & Pull determinants. The period of estimation includes the crisis of , when major international banks endured significant financial stress. In this framework, our contribution is threefold: - Firstly, we take explicit account of the existence of subsidiary locations as determinants of crossborder bank claims in order to analyse the dual financing strategy of international banks. Are these two financing strategies substitutable or are they complementary? And is the presence of foreign bank subsidiaries a factor of attraction for foreign bank claims? - Secondly, we take account of the financial situation of international banks and the state of health of local banking systems, an approach seldom seen in traditional literature. - Thirdly, we study more specifically cross-border interbank loans in order to appreciate parent bank/subsidiary strategies, a method, to our knowledge, seldom seen in earlier literature. The existence of bank subsidiaries located abroad can have a two-fold impact on cross-border bank claims. On the one hand, the significant presence of foreign bank subsidiaries in an emerging country can be an attractive factor for funding. It may effectively be seen by foreign investors as a positive signal as to the soundness and development of the domestic banking system and more generally of the country as a whole, and attract cross-border flows given that the country risk is low. The presence of these subsidiaries is, in particular, a factor pointing to greater efficiency for the local banking system owing to the transfer of banking techniques and skills (Claessens et al. [2001]; Maechler et al. [2009]; Mihaljek [2006]). This attractive aspect was highlighted in literature prior to the crisis but without ever having been empirically tested (Bhattacharya [1993], Levine [1996], Caprio & Honohan [1999], Agénor [2003]). The presence of bank subsidiaries may also be a pull factor via the relations between 2 This is explained by the fact that Asian authorities are upholding restrictions on foreign shareholding and wish to keep a relatively significant share of State-owned banks or banks with family-based or regional capital ownership (Singapore, Hong Kong) (Domanski [2005]; Mihaljek [2010]). 3 Argentina, Bulgaria, Bolivia, Brazil, Chile, China, Colombia, Czech Republic, Estonia, Hungary, India, Indonesia, Korea, Latvia, Lithuania, Malaysia, Mexico, Peru, Paraguay, the Philippines, Poland, Romania, Slovakia, Slovenia, Thailand, Uruguay, Venezuela and Vietnam. 3

4 a parent bank and its subsidiaries. Loans from the parent bank to a local subsidiary (cross-border interbank loans) may in particular take the form of financial support in times of crisis in the host country, thereby enabling subsidiaries to maintain their supply of domestic credit (Haselman [2006]; de Haas & Van Lelyveld [2006], [2010]; Arena et al. [2007]). However, specific relations between a parent bank and subsidiaries must take into account the soundness of the parent bank. A parent bank in trouble may restrict its financial support to subsidiaries (McGuire & Tarashev [2008]). It may also arbitrate between its subsidiaries and modify investment depending on the state of the host country and on expected returns (the substitution effect), thereby providing more than just a support effect or playing a role of last resort lender to its most fragile subsidiaries (de Haas & Van Lelyveld [2006] and [2010]; Dinger [2009]). The result of this is that the parent bank-to-subsidiary effect is ambiguous. From another angle, however, the existence of local banking subsidiaries may also have a restrictive effect on foreign financial flows through the so-called substitution effect. The operations of located bank subsidiaries can be analysed in terms of development of the local activities of banks (in the host country) based on local resources which, in this case, may serve as a substitute for crossborder claims. The widespread presence of foreign bank subsidiaries in a country should then be accompanied by a fall in cross-border bank claims. The observation made by Kamil & Rai [2010], whereby the countries the least affected by the rationing imposed by international banks in Latin America further to the crisis of 2007 are those where foreign subsidiaries do high levels of local business in local currency, supports this claim. Thus, to secure a comprehensive vision of the determinants of cross-border bank claims, it is important, in addition to the traditional determinants of capital inflows, to take into account the variables regarding the soundness of both banks in the local banking sector and foreign creditor banks. Moreover, account must be taken of the expansion strategy of international banks. To this end, our contribution lies with the construction of two variables. The first measures the presence of subsidiaries of foreign banks within an economy through the percentage of assets in the local banking system that are in the hands of foreign banks, and also through the percentage of the number of local banks in the hands of foreign banks. The second stems from the creation, for each emerging country, of a health indicator for international banks, one which integrates the weight of external claims on the emerging country per nationality of the foreign creditor banks. Our contribution also lies with the study of total cross-border bank claims firstly vis-à-vis all sectors in an emerging country, then vis-à-vis the nonbank sector, but also of cross-border loans to the bank sector alone in order to study interbank relations and approach the links between parent banks and subsidiaries that may be at the origin of capital inflows. Data comes from banking statistics provided by the Bank for International Settlements (BIS) and more precisely from Locational banking statistics, which are strictly cross-border claims where intra-group positions are not netted out. 4

5 We firstly proceed with a review of literature focusing on the determinants of capital inflows, particularly claims from international banks. We then estimate, from an econometric standpoint, the determinants of cross-border bank claims on a panel of 28 emerging countries over the period between 1995 and 2009 in the framework of a Push & Pull analysis. 2. REVIEW OF LITERATURE Traditionally, the determinants of capital inflows are analysed through Push & Pull factors. Since the pioneering articles from Calvo et al. [1993] and Fernandez-Arias [1996], the determinants of capital inflows 4 into emerging economies have been classified into two categories. The first includes the push factors, or factors outside the emerging country, i.e. conditions ill-suited to investment in industrialised countries, which push capital into the emerging country. The second concerns the pull factors, factors inside or specific to the emerging country, i.e. positive fundamentals which pull or draw capital into the country. Following Calvo et al. [1993], who studied the role of traditional push factors (a fall in interest rate or in economic growth in developed countries) during the debt crisis of 1982, empirical studies have evaluated the respective role of each category of factor and delivered qualified results geared to emerging regions. In particular, Jeanneau & Micu [2002] show the more significant role of pull factors (trade flows, GDP growth, exchange rate and exchange rate regime, the yields of financial markets in borrowing countries) in Asia, and of push factors (real GDP and real 3- month interest rate in lending countries) in Latin America between 1985 and 2000, as determinants of international bank loans. Additionally, the same authors show that, contrary to initial studies, the evolution of the business cycle in lending countries has a pro-cyclical role for capital inflows in emerging countries. After the Asian crisis of 1997, literature focused more on contagion to explain the movements of financial flows: channels of transmission of financial shocks and shifts in foreign investors sentiments (Forbes & Rigobon [2002]; Masson [1999]). Thus, alongside the traditional Push & Pull variables, variables related to risk aversion (such as the VIX or rate spreads) appear to be significant push factors to explain movements of financial flows towards emerging countries (Jeanneau & Micu [2002]; Brana & Lahet [2010]). Most of the recent articles on the determinants of capital inflows into emerging countries take an interest in international bank loans (or claims), which indiscriminately group together cross-border loans (in which loans between parent banks and subsidiaries are offset) and the loans from locallylocated subsidiaries in foreign currency 5. Few articles introduce bank-related variables as determinants for these loans. While some articles test the soundness of foreign banks, they do so indirectly through 4 In initial empirical studies, capital inflows are generally measured by portfolio investments (Fernandez-Arias [1996]; Chuhan et al. [1998]). 5 Data from the BIS, Consolidated banking statistics. 5

6 a stock market index or risk indices, and few articles include the soundness of banks in the host country. McGuire and Tarashev [2008], between the 1 st Q of 1992 and the 2 nd Q of 2007 and with a panel of 19 emerging countries, thus introduce the quality of banks from lending countries, evaluated indirectly by the banking sector stock market index, the expected frequency of bank system default and the volatility of assets 6, alongside traditional macroeconomic factors. They also introduce an indicator for the host country s openness to foreign banks (measured by the share of domestic credit granted by banks with foreign capital), without using any particular frame of analysis or variables reflecting the health of banks in the host country. The health of foreign banks and the degree of openness of host countries explain the growth in credit to emerging countries. Kamil & Rai [2010], with a panel of 13 countries from Latin America over the period between the 4 th Q 1999 and the 4 th Q 2008, show that the health of international banks (measured by their likelihood of default) effectively impacts the supply of total external claims on emerging countries 7. However, the authors conclude that Latin countries were less affected by the drop in foreign bank claims than other emerging countries because they are less dependent on cross-border claims. Many loans are made by subsidiaries in local currency sourced from deposits collected from national customers. In their tests, the authors do not include variables relative to the health of banks in the emerging country s banking sector or the presence of foreign banks. Concerning only cross-border bank claims on all sectors of emerging countries, Hermann & Mihaljek [2010], with a panel of 28 emerging countries from the three zones of Asia, Latin America and transition Europe, show, over the period between the 1 st Q 1993 and the 4 th Q of 2008 and in the frame of a gravitational model, that in addition to distance, to traditional global and specific factors and to indicators of risk aversion, the soundness of banks in the lending country is an explanatory factor for loans to emerging countries, while the health of local banks is an attractive factor. The presence of foreign banks is not tested. The soundness of banks is measured indirectly by the gap between the stock market index for the banking sector and the index for the reference stock market. 8 Again concerning only cross-border bank claims on 40 emerging countries from the 3 rd Q of 2005 to the 2 nd Q of 2012, Avdjiev et al [2012] show that the deterioration of the health of foreign banks (measured by the CDS of these banks and by the volatility of the stock market index in the banking sub-sector) has a negative impact on capital inflows. Neither the presence of foreign banks, nor the health of the banking sector in the emerging country, is included in the study. In all, the presence of subsidiaries operating in the host country is very rarely taken into account. If it is, it is usually measured by the amount of bank credit granted by these subsidiaries. This variable is often used as an explanatory factor for all external bank claims, a point that raises a 6 These latter two variables correspond to the indicators developed by Moody s. 7 Besides cross-border claims and claims in currency from local subsidiaries, total external claims also include claims from local subsidiaries in local currency. 8 The authors are critical of this measurement: a growing index implicitly translates to inflows of foreign capital. 6

7 problem of endogeneity given that external bank claims in the studies mentioned herein include loans made by subsidiaries in foreign currency or in local currency. Furthermore, these studies use claims on all sectors of emerging countries, taken globally, and consequently overlook interbank relations. So, in order to study the strategies of international banks for the external financing of emerging countries through the impact of the presence of foreign bank subsidiaries on the amount of cross-border bank claims, we broaden the range of determinants of capital inflows inside a macroeconomic Push & Pull framework. Alongside traditional factors, we introduce into pull factors variables to qualify the soundness of banks in the host country s banking system. We then integrate variables relative to foreign banks in order to better gauge their strategies, in two ways. Firstly, we use a variable to measure specifically the implantation of foreign banks into the local banking system as a pull factor. We use the percentage of banking system assets in the hands of foreign banks, then the percentage of the number of foreign banks, something that literature to date has failed to do for the determinants of bank financing. Secondly, we include the health of international lending banks as push factors, given that the constraints in terms of liquidity, solvability or profitability weighing on each one of them may lead to a change in their positions in emerging markets. To do this, we built ratios (liquidity, solvability, profitability) that we weight through the share of the largest creditors for each borrowing country. Our econometric study proceeds in two stages. We firstly study the determinants of cross-border bank claims on all sectors in emerging countries, then we make a distinction between claims on the non-bank sector and loans made to the banking sector alone in order to appreciate relations between parent banks and subsidiaries. By using dummy or interaction terms, we are able to approach the role of located foreign banks with greater precision, geared to the state of health of the emerging country and its banking system. 3. AN EMPRICIAL ANALYSIS OF CROSS-BORDER BANK CLAIMS: THE IMPACT OF FOREIGN BANK LOCATIONS An econometric study lets us specify the determinants of cross-border bank flows. The estimated equation takes the form of: 9 CBC it = c + X it α + Y it β + u it (1) with CBC it, the cross-border bank claims on country i at the period t, X a vector of pull (domestic) factors and Y a vector of push (external) factors. Index i relates to the host country (i = 1 to 28), and index t to the period (t = 1 to 17). The period of estimation ranges from 1995 to The countries under study are Argentina, Brazil, 9 The variables, their construction and their sources are explained in Table 4 in Appendix. 7

8 Bolivia, Chile, Colombia, Mexico, Paraguay, Peru, Uruguay and Venezuela for Latin America, China, India, Indonesia, South Korea, Malaysia, the Philippines, Thailand and Vietnam for the Asia zone, and finally Bulgaria, Hungary, the Czech Republic, Estonia, Latvia, Lithuania, Poland, Romania, Slovenia and Slovakia for East European countries. We have tested three dependent variables, all built based on locational international banking statistics from the BIS: 11 gross amounts of total cross-border bank claims vis-à-vis all sectors in an emerging country, gross amounts of cross-border bank claims vis-à-vis the non-bank sector, and crossborder loans to the borrowing country s banking sector alone, an approach which to our knowledge has to date not been studied in literature. Positions between parent banks and their subsidiaries have not been offset, meaning that we are now able to appreciate the relations between the two when studying loans for the banking sector alone. These variables are expressed in growth rates in order to study capital inflows. 12 Amongst the explanatory factors, in line with existing literature, we distinguish between pull factors attracting capital, and push factors that push capital into a country. We also take account of the financial situation of the banking sectors both lenders and borrowers through our ratios of liquidity, profitability and solvability Explanatory variables Traditional pull factors Pull factors are represented by the host country s fundamentals. We have estimated these variables through the sovereign rating built using the numerical values of indices published by Standard & Poor s and Moody s 13 and by calculating the average. This variable has the advantage of being a good summary indicator of a country s fundamentals, monitored by international investors in their investment strategies (Rating Host) 14. We have also taken into account a more cyclical variable, namely the GDP growth rate for each country i (Growth Host). The expected sign for this variable is positive. We have also endeavoured to ascertain whether or not more institutional variables could have an impact on international banks claims. 15 For this, we have used the Index of Economic Freedom 10 Sample size is restricted by the availability of data. In particular, 2009 is the last year for which data concerning located foreign banks is available on a harmonised basis (cf. Claessens and Van Horen [2014]). 11 This data is compatible with national accounts and balance of payments, based on creditor banks reporting claims to the BIS. For further information about these measures, see McGuire P., Wooldridge P. [2005], The BIS Consolidated banking statistics: structure, uses and recent enhancements, Bank for International Settlements Quarterly Review. See also Guidelines to international locational banking statistics, Monetary and Economic Department, BIS, November See, too, Table 4 in Appendix. 12 And to work with stationary variables. 13 Long-term ratings and in foreign currency for the sovereign debt. 14 This variable is expressed in difference for it to be stationary. 15 See Papaioannou [2009] for an analysis of the impact of the institutional environment on the international movements of bank funding. 8

9 from The Heritage Foundation 16 (Score), together with a more specific indicator for the banking sector (Bank Score). We have also tested dummy variables such as the country s exchange rate regime or whether or not it belongs to the OECD group as a factor of attractiveness for cross-border claims, but these variables did not appear to be significant Traditional push factors Capital inflows may also be explained by push factors. These include the economic and financial situation of the country of origin of international lending banks. Given that more than 70% of foreign bank funding invested in emerging countries originates from Europe, North America or Japan, we have introduced a variable representing the weighted growth of these different zones (Growth Home). Weighting is specific to each host country and is a function of the origins of the received cross-border bank claims. If we stay within the framework of a push-pull analysis, the expected sign for the growth variable for home countries is negative. It is because the outlooks for growth (and returns) in (from) investment in the home country are disappointing that international banks turn to cross-border investments with more favourable economic prospects (Calvo et al [1993]). We have also introduced a dummy variable that takes the value of 1 in 2008, the time when the financial crisis, originating in the United States, turned worldwide, most notably further to the collapse of Lehman Brothers (Subprime). We have also taken account of the VIX (CBOE volatility index), as a measure of risk aversion to test the potential presence of speculation and of any crisis-related contagion effect. The expected sign is negative Accounting for the strategies of international banks To test the impact of the presence of foreign banks via their local subsidiaries on capital inflows, we have added a foreign banks variable to the pull factors, a variable which measures the share of the assets in the host country s banking system held by foreign banks (Foreign assets). To build this variable, we based our work most notably on Claessens & Van Horen [2014], Claessens et al [2008], and Jeon et al [2011]. For Korea, we used OECD data. In order to take greater account of international bank strategies, we have also integrated other explanatory factors such as the health of lender-country banks (push factor) and that of banks from the borrower country (pull factor). In doing so, we have used data from the Bankscope data base. The variables used (as in de Haas & Van Lelyveld [2006], Haselman [2006], Derviz et al. [2007]) measure the degree of bank capitalisation (capitalisation variable evaluated via the ratio of capital to assets), the degree of liquidity of these assets (the liquidity variable calculated via the ratio of liquid assets to 16 The index is an arithmetic average of 10 criteria, each given a score of 0 to 100, where 100 represents maximum freedom: freedom of enterprise; freedom of trade; weight of taxes; government expenditures; monetary stability; freedom of investment; financial deregulation; protection of private property; war on corruption; liberalisation of labour. The bank score is more interested in the level of intervention and control by the public authorities vis-à-vis the banking and financial system. 9

10 total assets), profitability (profits to assets or ROA, and interest rate margin (margin variable)) and the quality of the credit portfolio (percentage of non-performing loans (NPL variable)). We have also calculated a Z-score variable to evaluate bank solvability, i.e. the sum of the capital ratio at date t and the average returns on assets over the period (ROA), divided by the standard deviation for asset returns over the period. The Z-score measures the distance to insolvability, a higher Z-score indicating a lower likelihood of the banking system s insolvability. The expected signs for these variables are ambiguous. Of course, banks in a sounder state of health in the host country are an attractive factor for cross-border claims, particularly interbank loans, but at the same time local banks in good health may render the country less dependent on external financing (substitution effects), including interbank financing. Given the construction of the Z-score, this variable will be tested alternately with the capital ratio and ROA, thereby verifying the robustness of results. As cross-border bank claims are the doing of international banks, we have also taken an interest in the health of these banks as an explanatory push factor for capital inflows into emerging countries. As with domestic banks, we have calculated indicators for capitalisation, liquidity, profitability (ROA), impaired loans (NPLs as a % of total gross loans) and a Z-score variable for international banks as well. We have calculated weighted indicators for each emerging country and for each year. Weighting takes account of the nationality of the country s main creditor banks. It represents the amount of claims on an emerging country from banks of the same nationality, divided by the amount of total claims for all banks reporting to the BIS on that country. For each emerging country, we have calculated the weight of each creditor by selecting the most important lender countries (Austria, Belgium, Canada, Denmark, France, Germany, Italy, Japan, the Netherlands, Sweden, Switzerland, the UK, the United States and Spain), i.e. 14 countries from the 17 registered with the BIS 17 (cf. Table 5, Appendix) 18. So for each emerging country, we thus obtain specific indicators of international banks soundness geared to the nationality of creditor banks. Here again, the expected sign with health variables for international banks may be ambiguous. We may think, certainly, that banks in a healthier financial situation will have greater resources and capacities to lend internationally. But likewise, there is also reason to feel that an improvement in this financial situation, thanks to the healthy situation of their home market, might prompt them to stay in their national market and temper any incitement to prospect on the international stage. Finally, we have introduced interaction terms. We wanted to ascertain whether or not the presence of foreign banks had an impact on the sign and the significativity of the bank crisis variable in the emerging country and on the growth rate variable for the host country. We have firstly built an 17 These statistics also include Australia, Ireland and Portugal, which we have omitted because their exposure values were very low, nil or not given. 18 European banks (especially Austrian and German) are thus the most important for CEECs and have been regularly since In Asia, Japanese banks held the upper hand until 2001 but have since been supplanted by British and American banks. In Latin America, American banks are still very heavily exposed but Spanish banks are gaining ground, even overtaking American and British banks from 2001 onwards. 10

11 interaction variable by interacting the growth rate of the emerging country s GDP with the presence of foreign subsidiaries in order to measure the role of foreign locations in a growing emerging country (Growth Host* Foreign Assets variable). We have then built a dummy variable concerning banking crises in emerging countries based on Laeven & Valencia [2010]: it takes the value of 1 during the banking crisis period, otherwise the value is 0. By multiplying this with the foreign subsidiaries variable (Banking Crisis*Foreign assets variable), we are able to test whether or not subsidiaries located in an emerging country in the grip of a banking crisis attract cross-border bank claims or loans. This variable lets us appreciate if parent banks have played a support role for their subsidiaries when the banking situation deteriorates. Estimations are made on stationary variables. 19 Capital inflow variables (total cross-border bank claims, cross-border bank claims on the non-bank sector and cross-border interbank loans) are expressed in growth rate. The variables for the health of local banks and international banks are all in percentage. In order to account for the endogeneity of certain explanatory variables, we use the system of generalized method-of-moments estimators, more efficient than GMM difference estimators (Blundell & Bond, 1998), because it lets us take better account of persistent variables Results Determinants of Cross-Border Bank Claims vis-à-vis all sectors (banks and nonbank sector) in emerging countries Results for total cross-border bank claims (Table 1) show firstly that foreign banks effectively implement a pull strategy. It is the economic conditions of the host country, not the adverse conditions of the home country of international banks (contrary to Calvo et al. [1993]), that determine the investment strategies of international banks. The host country growth rate variable or the change in sovereign rating is a significant factor of attraction for bank financing. The growth rates of home countries also appear to be a motivating factor for cross-border claims (in accordance with Jeanneau & Micu [2002]). These results, confirmed by the fact that the VIX is not significant, support the conclusion drawn by Haselmann [2006], whereby foreign banks seem to be implementing long-term strategies. 19 We run tests of stationarity with first (Maddala, Wu [1999]) and second (Pesaran [2007]) generation panel unit root tests in order to address the problem of cross-sectional dependencies. 11

12 Table 1. Determinants of the growth rate of total cross-border bank claims ( ) Coefficient Coefficient Coefficient (t student) (t student) (t student) Traditional Push & Pull factors CBC t *** (6.70) 0.38*** (6,65) 0.38*** (6.67) Growth Host 0.28*** (3.82) 0.27*** (3.74) Rating Host 5.37*** (3.20) 5.48*** (3.27) 5.55*** (3.44) Bank Score 0.37** (2.38) 0.37** (2.36) 0.36** (2.33) Growth Home 0.88** (2.10) 0.84** (2.01) 0.83** (2.02) Subprime *** (-6.78) -29.4*** (-6.4) -30.9*** (-6.97) Host country bank health (Pull) Domestic bank NPL -0.48* (-1.73) -0.50* (-1.79) Domestic bank capitalisation -2.02*** (-3.45) -2.04*** (-3.50) -1.20** (-2.15) Domestic bank margin 2.8** (2.52) 2.88*** (2.58) 2.8*** (2.65) International bank health (Push) International bank NPL -5.41*** (-2.88) -5.2*** (-2.78) -4.87*** (-2.61) International bank capitalisation 5.65** (2,53) 7.09*** (3.31) International bank Z-score 1.26** (2.24) The role of located affiliates (Pull) Foreign Assets 0.41*** (4.11) 0.40*** (4.03) 0.38*** (3.99) Growth host * Foreign Assets 0.007*** (4.84) Constant -32.9* (-1.94) -32.8* (-1.84) -49.6*** (-3.5) ^ Only significant variables are presented. *Significativity at 10%; ** at 5%; *** at 1%. t-student in parentheses. The different variables are presented in Table 4 in the Appendix. Wald Chi2 (12) = (12) = (11) = N obs = 311 N obs = 311 N obs = 311 Our empirical results then show the importance of the financial situation of international banks, thereby confirming the results produced by Hermann & Mihaljek [2010], McGuire & Tarashev [2008], De Haas & Van Lelyveld [2006 & 2010] and Kamil & Rai [2010]. The deterioration of their 12

13 solvability (increased percentage of impaired loans or fall in capital ratio) leads to a fall in total crossborder bank claims. We obtain the same result using the Z-score as an indicator of solvability 20, a sign of the robustness of results. As a result, multinational banks seem to tend to pass on home market financial shocks. The financial crisis of 2008 was also a powerful factor in the withdrawal of funds. The situation of the host country s banking system also appears significant as a determinant of external bank financing. First of all, the banking system s level of development and openness to competition (dummy variable Bank score) is a significant factor of attraction. Then, the host banking system s financial situation will also play a role, consistent with the works of Hermann & Mihaljek [2010]. In particular, the improvement in the banking situation, marked by a drop in impaired loans, encourages cross-border claims, as does the profitability of the domestic banking system (margin). However, the more banks in emerging countries are capitalised, the less they need external financing and the more capable they are of granting claims on non-financial agents. In this case, cross-border bank claims are reduced. The presence of foreign banks (Foreign Assets) is clearly linked to the increase in cross-border bank claims. This finding shows that the presence of foreign banks operating in the local market is a pull factor for foreign bank financing. We may conclude that the two financing strategies in emerging countries by international banks are complementary. However, the impact of the presence of foreign banks is all the greater in that it is associated with a high growth rate in host country GDP (Growth Host*Foreign Assets). We may therefore see this as a factor of attraction but not as a stabilizing factor for cross-border claims. Globally speaking, cross-border bank claims are sensitive to the local market s openness to foreign banks, a token of economic development and greater maturity of the financial system but dependent above all else on good growth prospects for the host country. Overall, we obtain the same results when using the percentage of foreign banks (number) as a variable measuring the location of foreign banks in lieu of the percentage of foreign bank assets (in relation to total bank assets), another sign of the robustness of our estimations (Table 6 Appendix) Determinants of cross border bank claims vis-à-vis the non-bank sector and of cross border interbank loans We have split up the addressees of capital flows by distinguishing between the non-bank sector and the banking sector. In this way we have studied cross-border bank claims on the non-bank sector and only cross-border loans to the banking sector. We have analysed interbank loans for several reasons. Firstly, we know that 80% of total cross-border bank claims are composed of loans (Hermann & Mihaljek [2010]). Of these, the share of interbank loans is equally significant (Graph 1). In 2011, 20 By contrast, this variable is not significant for the host banking system. 13

14 interbank loans accounted for 46.1% of cross-border bank loans in Latin America, 65.2% in Asia and 57.6% in CEECs. Lastly, this helps us to better appreciate relations between parents banks and subsidiaries and to test whether or not there exists financial support, in the form of loans, from the parent bank to its subsidiary when the banking system or the host country runs into trouble. Graph 1. The share of interbank loans in total cross-border loans. Source: Authors calculation with BIS data When we distinguish between cross-border claims on the non-bank sector and cross-border interbank loans alone (table 2), results are noticeably different. Traditional push factors are no longer significant. Capital flows no longer respond directly to the economic situation of the lender country but are tied more to the state of the host country and the financial situation of both the lending and the borrowing banks. Nevertheless, the presence of foreign subsidiaries does not play the same role when we consider cross-border financings for the non-bank sector or the banking sector alone, a point that allows us to highlight the links between parent bank and subsidiary. 14

15 Table 2. Determinants of the growth rate of cross-border bank claims vis-à-vis the non-bank sector and of cross-border interbank loans ( ) Cross-border bank claims vis-à-vis non-bank sector Cross-border interbank loans Coefficient Coefficient Coefficient Coefficient (t student) (t student) (t student) (t student) Traditional Push & Pull factors CBC t *** (2.56) (1.41) 0.24*** (4.52) 0.21*** (4.08) Growth Host 0.29*** (2.99) 0.31*** (3.25) 0.42*** (2.64) Rating Host 8.41** (2.08) 8.67** (2.22) Subprime *** (-4.88) -29.3*** (-4.43) *** (-4.16) -43.8*** (-4.55) Host country bank health (Pull) Domestic bank NPL -2.26*** (-3.17) Domestic bank ROA 4.95*** (4.84) -4.11*** (-2.56) -2.26*** (-3.21) -4.39*** (-2.75) Domestic bank capitalisation -4.57*** (-5.28) Domestic bank Z-score -2.82*** (-2.89) International bank health (Push) International bank NPL -8.05*** (-3.12) -9.03*** (-3.36) -12.5*** (-3.53) -10.5*** (-2.93) International bank liquidity 1.81* (1.95) 1.69* (1.86) International bank capitalisation 5.25* (1.77) International bank Z-score 1.47* (1.89) The role of located affiliates (Pull) Foreign Assets -0.23* (-1.79) -0.30** (-2.32) 0.49** (2.47) 0.35* (1.74) Growth host * Foreign Assets 0.01*** (3.54) Constant 57.87*** (4.24) 55.56*** (2.97) 6.33 (0.32) (0.54) Wald Chi2 (8) = (7) = (9) = (9) = ^ Only significant variables are presented. *Significativity at 10%;** at 5%; *** at 1%. t-student in parentheses. The different variables are presented in Table 4 in Appendix. N obs = 364 N obs = 364 N obs = 357 N obs =

16 First of all, the analysis of international bank claims on the non-bank sector of emerging countries shows the importance of the financial situation of international banks. Any improvement in terms of balance sheet (a fall in NPL) or in levels of capitalisation encourages their investment operations abroad. We reach the same conclusions with the Z-score variable, a further sign of the robustness of results. A bank will be all the more able to lend abroad to the non-bank sector whenever its risk of insolvability is lower and its financial situation sound. International bank claims on the nonbank sector, however, also depend on growth prospects with the host country (Growth Host) and its banking sector. Higher profitability recorded by domestic banks (ROA) attracts external funding. More generally, banks recording higher returns on assets reflect the country s good health and its outlooks for future growth and yields on investments in the productive sectors of emerging countries. However, when the financial situation of domestic banks improves, reflected by an increase in capital ratio, the financing of non-financial agents by the domestic banking sector tends to replace foreign financing. Better capitalised banks or banks with a lower likelihood of crisis (Z-score) are better able to finance the domestic economy, resulting in lower recourse to cross-border claims. Lastly, foreign bank locations reduce cross-border bank claims on the non-bank sector. This result indicates that the two strategies implemented by international banks are substitutable: either they finance the non-bank sector via cross-border flows, or they finance it via their located subsidiaries. Thus, if we study just financing intended for non-financial agents, the location of foreign bank subsidiaries in emerging countries reduces cross-border financing for these countries. Concerning the analysis of interbank loans, as expected with interbank relations, the liquidity of international banks is a significant push factor, as is the state of their balance sheet (fewer impaired loans - NPLs), which will increase their capacity for cross-border loans. The health of international banks, including parent companies, is therefore a significant explanatory factor for interbank flows. This result is confirmed by the significativity of the dummy variable relating to the subprime crisis. The negative effect of the subprime crisis also plays for the other types of claim under study but it is greater in the case of interbank loans. As a result, the subprime crisis penalised in particular the amount of interbank loans. Regarding demand for financing, banks officiating in a host country will be borrowing all the more in that their profitability will be down (ROA), thereby justifying a growing need for external financing. Interbank loans will meet this need providing that the quality of the loan portfolio of borrowing banks does not deteriorate (a rise in NPLs). The presence of foreign banks located in a country is clearly a factor of attraction for interbank loans. We have shown earlier that the same was true with regards all cross-border claims, even though the effect is negative for the sole claims on non-financial agents. From this we may deduce that this complementarity of strategies cross-border financing/subsidiary locations clearly comes from parent bank/subsidiary relations. Loans from parent banks to their subsidiaries are therefore an 16

17 important explanatory factor for cross-border interbank flows. However, these flows come under a rationale of yield from parent banks, not from a pattern of support for their subsidiaries. Indeed, interbank loans are a response to a slump in the profitability of domestic banks (ROA, implying lower self-financing) but they are decreasing when the situation of banks in the emerging country deteriorates (a rise in NPLs). Additionally, these loans are intended for subsidiaries located in emerging countries that are growing (interaction variable, Growth Host*Foreign Assets). This result is confirmed by the non-significativity of the capital ratio variable and the interaction term of Banking crisis*foreign Assets: bank subsidiaries in trouble, or in a banking system in crisis, do not benefit in any significant way from support from parent banks in the form of loans. Thus, stronger growth from an emerging country works in favour of interbank loans, particularly in countries where foreign banks are located and whose banks are in good health. So relations between parent banks and their subsidiaries effectively fall under an expansion strategy, - winning and moving into new markets-, and not under a parent-subsidiary support strategy. 4. CONCLUSIONS Decisions by international banks to invest in emerging countries are essentially grounded in pull strategies: the macro-economic outlooks of host countries are a decisive factor in their strategies, much more so than the prospects for their home country. This emphasises the importance for an emerging country to introduce sound economic policies. However, our study shows that it is indispensable to equally take account of the situation of international banks. On the one hand, the constraints weighing on these banks (liquidity, solvability) have a significant impact on cross-border claims. On the other, this cross-border financing is written into a more global banking strategy. Are they complementary or can they be a substitute for located subsidiaries? Our results show that when international banks open subsidiaries, there is a clear effect of substitution with regards cross-border claims on the non-bank sector. These claims decrease. However, international banks retain a financing role, firstly indirectly through the activity of loans to the non-bank sector by their subsidiaries, who find their resources locally. 21 They then also maintain their loan business to emerging countries via parent bank/subsidiary relations (interbank loans). So globally, we may conclude that the effect of the presence of international bank subsidiaries in emerging countries remains positive on cross-border capital inflows. The presence of foreign banks is a factor of attraction for cross-border bank claims and reflects the country s economic and financial development. Located subsidiaries benefit from interbank financing from their parent banks, but they 21 This confirms the findings of McGuire & Tarashev [2008], Kamil & Rai [2010]; Takats [2010a] for Latin American countries; or of Moreno & von Kleist [2007] and Lahet [2009] for Asia. 17

18 are not, however, the sole beneficiaries. This also finances national banks and indirectly non-financial agents in the emerging country. Having said this, interbank financing comes under a rationale of finance and profitability. Indeed, when a parent bank comes to support its subsidiary, it is solely to meet its needs for liquidity in the face of lower profitability, providing that the subsidiary s financial situation is sound in other respects (quality of loans) and that the growth prospects for the host country are positive. Our results show that support is not forthcoming if this is not the case. Openness to cross-border bank claims also renders emerging countries more sensitive to the health of international banks (including that of parent banks), but also more vulnerable to international financial crises and to contagion. Parent bank/subsidiary relations are therefore not necessarily a factor for financial stability in the event of a crisis in the host country. 18

19 APPENDIX Table 3. Characteristics of banking systems in emerging countries (2009) Countries Share of foreign banks (% of banking system assets) N of foreign banks (% of total number of banks) Total crossborder bank claims (% GDP) of which for banks (% of GDP) Cross border loans (% of GDP) of which for banks (% of GDP) Argentina Bolivia Brazil Chile Colombia Mexico Paraguay Peru Uruguay Venezuela India Indonesia South Korea Malaysia Philippines Thailand Vietnam Bulgaria Czech Republic Estonia Hungary Latvia Lithuania Poland Romania Slovenia China Source: Columns 1 and 2 based on data from Claessens and van Horen [2014], Claessens et al. [2008], Jeon et al [2011], and the OECD. Columns 3-6: authors own calculation from BIS data, locational banking statistics. 19