The Great Cross-Border Bank Deleveraging: Supply Side Characteristics

Transcription

1 Second Draft December 4, 2013 The Great Cross-Border Bank Deleveraging: Supply Side Characteristics by Eugenio Cerutti and Stijn Claessens IMF Abstract Many international banks have greatly cut their direct foreign lending and local affiliated subsidiaries lending in recent years, notably following the Lehman crisis over the period 2008Q2-2009Q2 and during the height of the Euro crisis the period 2011Q3-2011Q4. They have done so because of shocks to their balance sheets and in the face of reduced borrower demand as well as change in regulations. Controlling for demand and using bilateral data adjusted for breaks and exchange rate movements, we show how cross-border deleveraging largely varied with markets prior assessments of banking systems vulnerabilities, with lender country - borrower country characteristics (e.g., proximity, trade and historical relationships) playing minor roles. Some evidence also suggests barriers to moving resources across borders in that the drivers of reductions in direct loans differed from those for lending by local affiliates. We would like to thank Charles Calomiris, Giovanni Dell Ariccia, Neeltje van Horen, Luc Laeven, Camelia Minoiu, and the participants in IMF brownbag and surveillance meetings for very helpful comments. The views expressed here are those of the authors and should not be attributed to the IMF, or its Executive Directors or Management.

2 2 1. Introduction The recent financial crisis has seen a large retrenchment in cross-border banking, with aggregate gross foreign banking claims as of the end 2012 some 20 percent below their precrisis peak in June 2008 (from about USD 30 trillion to about USD 25 trillion). This retrenchment reflects a multitude of factors, with three notable. One, the deterioration in balance sheets of international active banks, with many facing capital shortfalls and liquidity strains, albeit at different points in times and to differing degrees, and pressures from financial markets to improve their financial positions. This was especially so in 2008/09 and subsequently with the euro crisis, and notably in advanced countries. Two, a weakening of demand among borrowers given worsening economic prospects, and increased default and other risks at both the individual borrower and borrowing country overall level. The developments in the European periphery are a prime but not only the example of the second set of factors. And three, regulatory changes and increased overall uncertainty about the future shape of the international banking system that may have led banks to not only reduce their balance sheets, but also to rebalance their operations away from cross-border banking activities. Teasing out the relative importance of these various factors in determining the volume of cross-border external financing is challenging, with the separation of supply from demand factors being the most difficult task many studies face. The main objective of this paper is to analyze the role of supply and lender-borrower factors in driving changes international banking claims. The secondary objective is to identify the motivations and constraints driving the particular forms of retrenchments, whether in the form of less cross-border foreign loans or reduced local affiliated subsidiaries lending. For both objectives, we are most keen on controlling for demand and other borrower-related factors, as well as for general time-varying factors, such as changes in global financial markets and economic prospects. We do so by exploiting the rich, bilateral Bank of International Settlements (BIS) crossborder banking dataset, enhanced in several ways, using data since the recent crisis, and using an event methodology. We focus on those periods when we can expect to see a large impact of supply factors. Specifically, we study the behavior of cross-border banking flows after or during two financial shocks: that associated with the Lehman failure in 2008; and that with an intense period in the euro crisis in the fall of And we exploit the bilateral nature of our data to control for changes in economic activities and prospects in the destination country. Specifically, since banking systems from various source countries all face the same demand conditions in the destination country, relative differences in bilateral claims represent differences arising from the supply side or the specific bilateral relationships.

3 3 Relative to the existing literature, we innovate in two particular ways. First, we use better data that take into account exchange rate variations and breaks-in-coverage in time series, allowing a more meaningful representation of the evolution of banks foreign claims. Second, we exploit differences between cross-border banking and affiliated lending. The latter was not decline as much during the deleveraging episodes, with aggregate declines of only 5 percent, compared to the 23 percent for cross-border loans during the 2008/9 episode. We find that banking systems deleveraging largely varied with shocks to creditor balance sheets, with banking system-borrower characteristics (e.g., proximity, trade relationships, and historical relationships) playing minor roles. We also find some evidence of barriers to the movement of resources across borders in that the factors explaining the patterns in reductions in direct loans differed from those explaining the reduction in lending by local affiliates. This suggests that international banks capacity to internally reallocate available resources within banking groups was limited during the deleveraging periods analyzed. In terms of outline, the paper proceeds as follows. It first reviews the literature that tries to identify the factors behind cross-border banking flows and develops the specific hypotheses we test. It then describes the data and methodology. The next section reviews the regression results. The last section concludes, also by highlighting possible further research steps. 2. Literature review and hypotheses The work presented in this paper relates to three main strands of research. The closest strand includes those papers that investigate changes in international bank activities using BIS data around periods of financial stress. The key paper here is Cetorelli and Goldberg (2011). They report that banks pulled back their international activities in the fall of 2008 and the first part of 2009, in part in response to a shortage of dollar funding. Also McGuire and von Peter (2009) analyze the importance of dollar funding shortages in explaining the behavior of cross-border banking flows during this period. Other recent studies confirm the general withdrawal over these periods, but note that the behaviors of banking system and banks varied considerably, in part related to the importance of foreign subsidiaries, funding conditions of local subsidiaries, and the distance between source and destination country. Cull and Martinez Peria (2012) show that in Eastern Europe, loan growth by foreign banks fell more than that of domestic private banks during the crisis, but that in Latin America foreign banks did not contract their loans at a faster pace, with the difference driven by the fact that foreign banks in Latin America were mostly funded through domestic deposits, in part due to regulatory requirements. And Claessens and van Horen (2013) show that foreign banks reduced credit more compared to domestic banks in countries where they had a small role, but not so when dominant or funded locally.

4 4 A second set of papers uses very detailed, micro data, notably using information on loans raised through international banking syndications, which has the benefit of being able to control for many individual borrower and bank characteristics (including by using borrower and bank fixed effects). This set of papers studies then the variation therein across source and destination countries, exploring different aspects. Using this data, Giannetti and Laeven (2012a) and De Haas and Van Horen (2013) report evidence of a flight home or flight to core markets, i.e., after the financial crisis banks tended to engage less in syndicated loans that involve a cross-border exposure, and rather lend to borrowers at home. Ongena, Peydro and Van Horen (2012) find that foreign banks in Eastern European countries reduced lending more compared to locally-funded domestic banks, but not compared to domestic banks that had pre-crisis funded themselves more from international capital markets. And De Haas and Van Horen (2012) find that banks, when faced with balance sheets constraints (for example, due to losses on toxic assets or dependence on whole-sale funding), reduced their crossborder syndicated lending, but more likely stayed committed to countries in which they had a subsidiary, especially in countries with weak institutions. The differences between crossborder and local affiliate lending suggest either that local subsidiaries have some specific information about borrowers allowing them to continue to extend loans or that there are some limits to intra-bank capital movements either because of financial frictions within the bank or regulatory barriers erected by the host country. Other papers in this group include Hale, Kapan and Minoiu (2013), which analyzed transformations in the global banking network due to the crisis. A third strand investigates using data for some specific countries on how international banks operating within that jurisdiction altered their operations, due to the financial crisis or in response to regulatory changes. Papers here include Ceterolli and Goldberg (2013) which shows how US banks adjusted their interoffice liquidity and claims during the period of financial turmoil in the fall of Kapan and Minoiu (2013) show that banks reduced their cross-border lending as a function of their pre-crisis exposure to whole-sale funding shocks. For the case of the UK, Aiyar, Calomiris, Hooley, Korniyenko and Wieladek (2013) show that local banks lend less abroad in response to micro-prudential supervisory measures during the 2000s, which required UK banks and UK-based subsidiaries to meet higher capital adequacy requirements. At the same time, Aiyar, Calomiris, and Wieladek (2013) show that foreign bank branches in the UK increased their share of lending in the UK in response to these same measures, a sign of regulatory arbitrage. As such, the net effects of capital shocks or changes in capital adequacy requirements on overall lending can be ambiguous as crossborder lending offsets. Our paper expands on and complements these papers in four ways. First, we study two large events, beside the most-studied global crisis that originated in the US, we analyze developments around the height of the euro crisis, the second half of Second, we are very careful in correcting data for the effects of exchange rate changes and changes in data

5 5 coverage when using BIS data. As noted by Cerutti (2013), such corrections are necessary for a proper interpretation and analysis since they can make for large differences compared with the original series. One notable example is the change in coverage of BIS banking statistics as investment banks in the US became commercial banks in 2009 Q1 which boosted US banking system foreign assets by USD 1.3 trillion. Another notable example is the large effects of the sharp movement in the dollar/euro exchange rate over BIS banking claims are reported in US dollars, so an important source of variation in stocks during the period under study originates from exchange rate movements, and not from bank induced changes in underlying positions. 1 Altogether, there were a total aggregated net amount of adjustments of some USD 1 trillion in each quarter during the period The adjustments for the end-2011 period were smaller (about USD ½ trillion), but still large relative to the changes in underlying positions. Third, by using BIS data we capture fully the on balance-sheet based international banking activities. Syndicated loans data do provide more details in many dimensions (e.g., bank and borrower level information), which allows for controlling better for demand and supply conditions. At the same time, this can come at a cost of: (i) partial coverage (e.g., syndicated loans are only one form of cross-border bank lending); (ii) capturing commitment flows rather than actual disbursements (e.g., information on whether or not credit lines are drawn is not available); and (iii) difficulty in exact identification of the participation shares across the members in each syndication (e.g., there is typical only data available on the participation share of each bank for only about 25 percent of syndicated loans). 2 Our (corrected and adjusted) data does represent the universe of cross-border, on-balance sheet banking claims (coverage is complete from the lender source points of view). And studying changes in the (adjusted) stock of cross-border claims during the two deleveraging episodes allows us to capture the effect of both new flows and effects existing claims maturing. We also cover most borrowing countries (about 150), allowing us to explore differences by both source countries and borrowing-destination combinations. 1 The exchange rate adjustments consist on three corrections. First, the domestic-currency denominated affiliates claims are corrected following the bilateral US dollar domestic currency exchange rate. The domestic-currency denominated affiliates claims are proxied by using its share of total BIS CBS foreign claims at immediate borrower basis. Second, at the same time, this procedure allows the identification of the amount of foreign-currency denominated affiliates claims, which are assumed to be in Euros in Europe and US dollars in the rest of the countries. Finally, bilateral CBS cross-border claims positions are adjusted using, as a proxy, the currency breakdown currency (among US Dollar, Euro, British Pound, Japanese Yen, and Swiss Francs) available from the BIS locational banking statistics. See Cerutti (2013) for more details. 2 For the US, Ivashina and Scharfstein (2009) show that syndicated loans account for at least 26% commercial and industrialized loans on the balance sheets of commercial banks operating in the U.S.

6 6 Lastly, we analyze changes in both cross-border banking claims and local affiliates lending. 3 With increases in local foreign bank presence over the past two decades in many parts of the world, affiliate lending has taken on greater absolute and relative importance (e.g., affiliate lending increased from 40 percent of foreign claims before the crisis to more than 50 percent in 2012). Since many international active commercial banks have a local presence in many borrowing countries (the market share of foreign banks increased from an average of 20 percent in the 1990s to more than 35 percent just before the financial crisis, with shares in some countries having increased to more than 90 percent), they can choose to a degree how to book their loans. While both forms will respond to shocks in source country banking systems, there are likely to be different. Notably, in the presence of intra-group frictions or due regulatory barriers to the movement of capital and liquidity ring-fencing, local affiliates will not be fully affected by shocks to their home banking systems. And to the extent that local affiliates have special information on and relationships with local borrowers, they may not be willing or need to adjust their claims. The main question we try to answer is what factors affect banks decisions to deleverage, i.e., how do home banking systems and other source country characteristics affect cross-border bank flows to specific countries during periods of financial stress? When banks change foreign credit exposures, they obviously many these factors, but their relative importance is unclear. To what extent do banks deleverage in response to market pressures and are marketbased measures therefore useful to predict deleveraging? How do such indicators compare to accounting based measures, such as balance sheet positions? How do banks decide on where to deleverage? Does the size of country exposures matter? Do, controlling for demand, lender-borrower characteristics distance, trade links, common institutions play a role? Second, how do banking systems move capital and liquidity globally and are there changes in intra-group versus direct across borders lending in times of financial turmoil, i.e., does financial turmoil affect banks cross-border and local activities similarly or differently? The role of internal markets may change during periods of financial turmoil as banks headquarters will likely face different incentives and regulations than subsidiaries do, as some papers reviewed indicate. 4 Studying differences between cross-border claims and local 3 Unlike other papers in the literature, our use of BIS consolidated banking statistics at ultimate risk basis provide a clean distinction between banks direct cross-border lending and affiliates lending. Other papers by using BIS data at immediate risk basis cannot exploit a clean breakdown. Affiliates lending would cover lending only in domestic currency but not in foreign currency, since international claims will include direct cross-border lending and affiliates foreign currency lending. 4 Specifically, consider the two extreme modes that could arise. In the international fully integrated model with no intra-group frictions, cross-border and local affiliate claims can be full substitutes as banks can freely move liquidity and capital to the best location headquarter or affiliate. In such a case, shocks to the home banking system will affect the two types of cross-border lending equally. (continued)

7 7 affiliates loans then provides insights into the degree to which there were frictions at times of financial turmoil. 5 While the exact constraints that lead to segmentation remain to be explored in more depth, the general issue is of great policy interest. Data and events Our main data source is BIS consolidated banking statistics (BIS CBS) at ultimate risk basis (i.e., this allocates claims to the country where the ultimate risk lies in a manner consonant with banks' own systems of risk management). This provides a breakdown of foreign claims into: (i) direct cross-border claims, capturing direct lending from banks to a foreign borrower without any presence in the borrower country; and (ii) affiliates claims, including claims by either branches or subsidiaries operating in the borrower countries. Both publically available data (available through BIS website) and restricted data (available through data requests to BIS) were used in the calculations. Following Cerutti (2013), the analysis is performed taking into account coverage break-inseries and exchange rate variations. These corrections are important to achieve a meaningful representation of the evolution of banks claims. The aggregate differences between adjusted and unadjusted series are shown in Figure 1. Total foreign claims were about USD 25 trillion in mid-2012, down from above USD 30 trillion in mid-2008, for the reporting banking systems included in our sample. Local affiliate claims have become relatively more important, with the greater foreign bank presence, and even more so following the financial crises. They represent about 50 percent of total foreign claims at 2012, compared to their 40 percent share before the crisis. This is shown in Figure 1 by the widening gap between foreign claims and cross-border claims, which captures the local affiliates claims. We choose the two events that represent the clearest shocks to international banking systems in the last decade, as also seen from Figure 1. The first one is the bank deleveraging episode which started in mid 2008 and gained momentum after the take-over of Bear Stearns and the failure of Lehman. This period of aggregate bank deleveraging ended in June Although not as acute as the bank deleveraging episode, a second deleveraging episode occurred during the second half of 2011, with two quarters of negative changes in adjusted foreign claims. Of course, these are just aggregate figures and there is great And in the segmented model, capital and liquidity can be trapped and/or ring-fenced, so that cross-border (local affiliate lending) responds to home (local) banks conditions. 5 The literature on multinational banks internal capital markets which includes recent papers such as De Haas and Van Lelyveld (2010) and Cetorelli and Goldberg (2012) has highlighted that global banks were (to some extent) able to reallocate funds and liquidity across locations in response to their relative geographical needs. This literature mostly focuses on the period before the crisis. Only part of the analysis carried by Cetorelli and Goldberg (2012) covers the period They found evidence that US global banks had active internal capital markets in terms of liquidity.

8 8 variation among source, destination and bilateral patterns of the deleveraging over these events. This heterogeneity is clear from Figure 2 and 3, which depict the bilateral lender in the columns, and borrowers in the rows percentage changes in direct cross-border and affiliates claims, where the columns are sorted from left to right by the overall degree of deleveraging of the creditor country and the rows are sorted from top to bottom by the overall degree of deleveraging at the borrowing country. The Figures show that, while there is some relationships, lenders did not uniformly adjust their claims across borrowers. Even lenders that reduced significantly their cross-border positions on aggregate show increases in crossborder claims with respect to some borrowers. Similarly, even borrower countries experiencing the largest aggregate declines in cross-border claims have some heterogeneity at the bilateral level. Methodology We want to control for demand factors, since otherwise any changes in lending patterns can just reflect changes in economic prospects or borrowers risks, rather than supply factors. Controlling for demand is difficult, however, as borrowers economic and financial prospects can as much be driven by the availability of credit as that credit adjust to these prospects. Observing and analyzing actual credit is then not informative on the role of demand or supply conditions. For example, during boom times, both capital adequacy position of banks and demand of borrowers are likely to be higher. Panel regressions during normal times are therefore less likely to provide meaningful results. Controlling for demand is easier to do using a cross-sectional approach during a specific deleveraging period when banks, albeit to different degrees, are known to suffer exceptional and sudden shortage in funding and capital, at the same time that they face increases in risks which vary by borrowers. Specifically, to control for credit demand at the borrower (country) level we use the identification strategy implemented by Khwaja and Mian (2008) and other recent papers (Cetorelli and Goldberg 2011, de Haas and Van Horen 2012, and 2013, Aiyar et al 2013). 6 The approach, which has also been used for within-country level analysis to control for borrower characteristics, is based on the notion that any difference in lending by different lenders to the same borrower must reflect variations in supply conditions among lenders or specific creditor-borrower relationships. 6 Kapan and Miniou (2013) further use industry rather than country level fixed effects.

9 9 We implement this approach by estimating the following baseline cross-sectional specification:, Creditor Borrower where the dependent variable L it is the log difference in the bilateral stock of cross-border claims (or local claims) of creditor banking system i on borrower country j between the beginning and the end of the specific deleveraging episode (adjusted for both coverage break-in-series and exchange rate variations). We use log differences to account for the skewed distribution in the changes in claims (see Figure 4). To control for borrower characteristics, the regressions include the γ j fixed effects for each borrower country, thus accounting for changes in demand factors and other conditions, while exploiting the bilateral nature of our data. The two sets of explanatory variables used in the analysis are variables on the state of the source country banking system, and variables capturing the bilateral relationship between creditor and borrower countries. All these explanatory variables are measured before each of the two events, i.e., at the end of 2007 and as of mid-2011, so as to avoid the deleveraging process from influencing them. The first set of source country variables, BankSystem i, captures the state of the banking system fundamentals, both as perceived by financial markets and as captured in accounting variables and as such, the regressions ask how banking systems respond to a shock depending on their vulnerabilities. Our main variable relates to how the financial markets perceived the riskiness of the banking system prior to the deleveraging period. It is based on the Systemic Risk Contribution (SRISK) measure as developed by Acharya et al (2010). This method uses an option-pricing model, with as inputs the behavior of bank s stock prices and some key balance sheets variables, to derive the perceived riskiness of each bank at each point in time. 7 It is a forward looking measure of the vulnerability of the banking system before each episode, i.e., it is exogenous to the deleveraging process itself. As it provides for a dollar amount of potential capital losses under some adverse scenario, we sum the positive 7 The calculation of SRISK takes three steps. First it estimates the daily drop in equity value of a firm that would be expected if the aggregate market falls more than 2%. This is called Marginal Expected Shortfall or MES. The measure incorporates the volatility of the firm and its correlation with the market, as well as its performance in extremes. These are estimated using asymmetric volatility, correlation and copula methods. In a second step this is extrapolated to a financial crisis which involves a much greater fall over a much greater time period. Finally, equity losses expected in a crisis are combined with current equity market value and outstanding measures of debt to determine how much capital would be needed in such a crisis. A firm is assumed to require at least 8% capital relative to its asset value. The Systemic Risk Contribution, SRISK, is the dollar value of capital shortfall that would be experienced by this firm in the event of a crisis. See for further documentation. Other papers that use SRISK include Idier, Lamé and Mésonnier (2013) and López- Espinosaa, Morenoa, Rubiab, and Valderramac (2012).

10 10 amounts for all domestically-owned banks in each source country to derive an overall measure of banking system capital at risk, which we then scale by the creditor country s GDP. In addition to this market based systemic risk measure, we explore a number of standard accounting based performance, portfolio quality, and solvency variables. Specifically, we include the banking system s prior year return on assets, ratio of non-performing loans to total gross loans, and tier 1 capital adequacy ratio. These measures could also provide an indication of the vulnerabilities of banking systems, but at the same time can suffer from reporting problems and bias. To cover the (subsequent occurrence of) a systemic banking crisis in the source country, we include a dummy based on Laeven and Valencia s (2012) dataset whether the country had a systemic crises as of the last quarter of each deleveraging episode. Since this measure is based in part on the de-facto amount of government support, the estimated coefficient for this dummy is best interpreted as how creditor banking systems deleverage internationally depending on whether or not they received state support ex-post. In terms of bilateral, Creditor Borrower ij, characteristics, we use variables that capture the nature of trade, financial and other linkages between creditor banking system i and borrower country j. Here, we include the traditional distance variables; (i) the log distance between the capitals of the creditor and borrower country; (ii) a dummy of geographical adjacency; (iii) a dummy for common language; (iv) a dummy for legal origin; (v) a dummy for colonial linkages; (vi) bilateral trade as proportion of the creditor country overall trade; and (vii) the direct cross-border concentration exposure of creditor banks to a particular country, measured as the share of the cross-border claims to a particular borrower as percentage of the creditor overall cross-border claims. These variables are proxies for both the severity of informational, financial and other frictions between creditor country banks and the borrower as well as for the presence of (historical) ties. The latter variable (vii) provides an indication whether, once faced with a shock, banks cut back more or less loans depending on how large a borrower concentration is among total cross-border loans. See further Table 1 for exact data definitions. 3. Regression Results Basic statistics Key statistics of the dependent and independent variables are provided in Table 1 and some of the patterns in dependent variables are shown in Figures 2 to 4. Note that, since not all banking systems have affiliates in every borrowing country, the number of observations for affiliates lending is much smaller (some 800) than that for cross-border loans (some 2400).In other words, banking systems are more active through direct cross-border lending than through affiliates. This also means that, in Figure 4, the observations along the vertical lines at zero change in affiliate lending represent those observations of changes in cross-border

11 11 loans for which there were no corresponding bilateral affiliates. In aggregate terms, direct cross-border loans thus dropped more than affiliate lending did. The degree of cross-border deleveraging for lender-borrower-country pairs was large, more so for the first event, but still quite large for the second event, with a median change in direct cross-border and affiliates of -16 percent and 2 percent respectively, for the period 2008Q2-09Q2, and -7 percent and ¼ percent, respectively for the period 2011Q3-11Q4. These median percentage change bilateral figures are very close to the mean of the log differences shown in Table 1. There is a large variation in bilateral patterns though in both events, as Figures 2 to 4 show. Figure 5 provides the distribution of the key independent variables. Our main explanatory variable is the SRISK variable, in the upper two charts. It shows a large variation, with banking systems that are large relative to GDP (such as many of the European systems, like Switzerland) perceived to be quite vulnerable to shocks before each of the two periods. There are also large changes in the vulnerabilities of banking systems over time, with the risk measure on average higher in the second than in the first period. While a number of banking systems became perceived to be much worse, not all deteriorated (e.g., the US became perceived as less vulnerable), making the risk measure less dispersed than in the first period. Banks return on assets (ROA), non-performing loans (NPL) and Tier 1 capital adequacy are depicted in the bottom part of Figure 5. They do not offer the same relative ranking as SRISK, confirming that backward-looking accounting can differ from forward-looking market-based assessments. Still, balance sheet measures markedly deteriorated between end and mid This greater differentiation over time is confirmed in Figure 6 which shows patterns in SRISK for both periods. The figure shows the worsening and greater differentiation over time: while in general, banking systems risk exposures ranked similarly across countries at both points in time, except for the US and Switzerland, all banking systems were perceived to be riskier in mid-2011 than as of end Evidently, most banking systems continued to be hit by large shocks (as is obvious for the banking system in the periphery) and were restructured and recapitalized too little to restore their soundness (e.g., European banks). The US and Swiss banking systems improved though over time, in part due to private sector and regulatory actions, including due to international deleveraging in earlier periods. While before the first period, end 2007, the US system was considered quite at risk, (perceived) risks fell much, and as of June 2011, the US banking system was ranked just above the lowest, the Finish banking system. This drop was presumably due to a combination of market pressure, private recapitalization and various government restructuring measures (notably TARP and the SCAP).

12 12 Regression Results: Supply Side Determinants Table 2 provides the base regression results, with Panel A showing regression results for changes in cross-border claims and Panel B changes in local affiliates claims. Table 2, Panel A shows the importance of supply factors in driving the reduction in cross-border banking claims. Specifically, the SRISK variable is highly statistically significant in the base regression (column 1) and all other specifications (row 1). Differentiating by regions (column 2), we find for the deleveraging episode that creditor banking system in Americas and Asia adjusted their cross-border claims relatively more in a response to perceived capital shortfalls before the period (and European banks, the base case, in contrast, responded less). This could be because the shock originated in the US and other banking systems were less cognizant of the (forthcoming) balance sheets constraints. It could also be that these other banking systems were less inclined to adjust their balance sheets in response as market discipline was less effective (e.g., because of weaker corporate governance or a more extensive public safety net with associated moral hazard). On average, the SRISK coefficients indicate that a one unit increase in SRISK would approximately translate into a 0.05 percent decline in direct cross-border claims; or given that SRISK standard deviation is about 97, a one standard deviation increase in SRISK would translate into a 5 percent decline in direct cross-border claims (as noted, the median bilateral decline was 16 percent). We next explore accounting measures of banking system. We find that ratios of nonperforming loans, return on assets, and Tier 1 capital adequacy generally not to be statistically significant as predictors of subsequent deleveraging actions (column 3). 8 When we combine market-based measures of banking systems vulnerabilities with accounting measures (column 4), we find that market-based measures remain statistically significant and accounting measures insignificant. This suggests that banks international deleveraging was largely driven by market pressures, i.e., it appears shareholders pressured banks to deleverage internationally more when banks were very exposed. When we next the dummy for countries that ran into subsequent systemic crises, we find that these did cut back their cross-border claims more as well (column 6), but the coefficient is not statistically significant.when winsorizing observations using percentiles 5 and 95 percent (column 5 and 7), we find the regression results to be confirmed. When combining all variables and without and with winsorized observations (columns 8 and 9), we find that SRISK remains highly statistically significant, and important again especially for banking 8 We also tried other accounting variables, such as: (i) size (log of assets); (ii) other profitability (e.g. ROE); and (iii) funding structures (the ratio of deposit to loans in the creditor banking system i). These variables were not significant across specifications, or displayed high correlation with the variables presented in the regressions reported.

13 13 system in the Americas. Whether the banking system has higher return on assets, more nonperforming loans, higher tier 1 or a systemic crisis are not significant explanatory factors in deleveraging. We next show the behavior of affiliated claims over this period in Panel B. As not all banking systems have local operations, and not necessarily in the same countries with which they engage in cross-border lending, the sample is smaller, only about 40% of the sample used for analyzing changes in cross-border claims. The regression results (Table 2B, columns 1-9) shows that supply factors are in general not as important in driving the reduction in local affiliate lending as the systemic capital at risk variable is not always statistically significant, and sometimes even positive. Differentiating by regions (column 2) shows that for banking systems from the Americas and Asia, local claims did adjust somewhat upwards in a response to perceived capital shortfalls in banking system home countries. For these banking systems, it seems cross-border and local operations behaved in more segmented ways, where relative to capital constraints at home local claims increased while cross-border claims declined. This suggests that for these banking systems, local affiliate lending was a substitute when hit by a capital shock at home. Regression results for other, that is, European, banking systems contrasts to the pattern in direct claims (which actually decreased in response to shortfalls in the home country banking systems, Table 2A). This differential pattern could be because European banks operated at that time in more integrated banking markets, where shocks originating at home affected both cross-border and affiliated lending similarly. Two of our accounting measures of banking system vulnerabilities, non-performing loans and return on assets, are again not statistically significant, while tier 1 is statistically significant negative (column 3). The latter is perverse as it suggests that better capitalized banking system actually deleveraged more. When combining all variables and without and with winsorized observations (column 4 and 5), we find SRISK again not to be statistically significant and with positive sign for banking systems from the Americas and Asia. Accounting variables remain insignificant, except for Tier 1 which is significantly negative. Home banking systems that ran into subsequent systemic crises cut back more on their affiliated claims (without and with winsorized observations, columns 6-7). Including all variables (without and with winsorized observations, columns 8-9) confirms regression results. Overall, local affiliate lending seems to have acted largely independently of what was happening to the home banking systems, with possibly for banking systems from the Americas and Asia acing as substitute for direct lending. Only the presence of systemic crises in the parent bank country seems to have triggered a decline in affiliates claims. The overall net effect is about a 17 percent decline in local claims at affiliate banks. This suggests that banking systems that received support were under pressure to reduce their international claims, a form of home bias.

14 14 We next study the second deleveraging period, the intense period in the euro crisis, by regressing the change in cross-border and local affiliates claims between the second and fourth quarters of 2011 on the same set of variables, except that all explanatory variables are now measured as mid Regression results are presented in Table 3A and 3B. They are similar in that SRISK relates negatively to the cutbacks in cross-border loans (Table 3A, column 1-9), but it is not statistically significance for all across specifications. But while in the first period local affiliate lending did not adjust to home SRISK, in this later period, it does decline as well in response (Table 3B, column 1). This suggests that over this time period, local affiliates were affected as well by vulnerabilities in home banking systems. There are differences again between the regions of the home banking systems, but the pattern now differs from that in the first period. Now it is not so much the American and Asian banking systems that cut back in lending when more vulnerable to the shock, but rather the European banking systems (Table 3A and 3B, columns 2). There are some differences between cross-border and local affiliates lending though as American banking systems adjust less their cross-border loans and Asian banking systems less their affiliate lending in response to the shock. This pattern likely reflects in part the fact that the crisis was now centered in the euro zone, where the EU single financial market was being questioned, with increased fragmentation of lending conditions. As such, it was the European (and Asian) banking systems this time for which the shock drove them to retrench across borders. We find that accounting measures of banking system vulnerabilities are this time statistically significant in explaining cross-border loans (Table 3A, column 3). For all three accounting variables, however, regressions show negative signs, which could reflect that systems that were expanding faster before the crisis and over-reported their profitability and capital adequacy, yet had more non-performing loans, were more exposed as they had taken on more risks. For affiliate lending, only non-performing loans is statistically significant and has the expected negative sign (Table 3B, column 3). When combining both market-based and accounting variables (Table 3A, without and with winsorized observations, columns 4-6), we find for direct cross-border loans and using all observations, the systemic risk variable to be statistically significant for American banks. All three accounting variables, return on assets, non-performing loans, and capital adequacy, remain significantly negative as well. For affiliate claims (Table 3B, column 4-6), the systemic risk variable remains statistically significant, but again less important for Asian banking system. Of the accounting variables, only the non-performing loans variable is statistically significant and remains negative. When we add the dummy for whether home banking systems that ran into subsequent systemic crises, we find that such system did not cut back significantly more their direct cross-border or affiliated claims (without and with winsorized observations, Table 3A and 3B, columns 6-7). This suggests that government interventions no longer led to a statistically

15 15 meaningful home bias. When using all variables with all observations or a winsorized dataset (columns 8-9), we find that the systemic risk variable not to be statistically significant for direct claims, but in case of affiliate claims, there remains the statistically significant negative effect. The three accounting variables are still significantly negative in case of direct claims, and thus, except for non-performing loans, contradicting their value as indicators of vulnerabilities. For affiliate claims though, non-performing loans remain statistically significant negative. Regression Results: Creditor-Borrower Determinants We next explore the role of bilateral factors in explaining the patterns of deleveraging, while controlling for the systemic capital at risk. Specifically, we first test the role of the exposure of the banking system to the specific country. We then investigate the role of cultural similarity (common language, legal and colonial origin), bilateral distance, geographical contiguity, and institutional environment. These variables are commonly used to explain bilateral patterns in cross-border capital flows (and trade). We explore these factors for the global deleveraging episode. Starting with the role of exposure and consistent with the literature on the differences between the factors driving direct cross-border lending and affiliate lending, we also include the share of direct cross-border lending to a particular borrower out of the total direct crossborder claims of a particular creditor banking system prior to the episode. Unlike de Haas and Van Horen (2013), we find consistent evidence that banks decreased more their direct cross-border lending to countries where they had high pre-episode exposures. (Table 4A, column 1). This rebalancing could reflect a basic adjustment as banks had previously overextended to these markets and they set tighter risks limits during the crisis. It could also be that it was relatively easier to deleverage in larger markets as these were less affected by the financial turmoil and other, local banks were willing to take up the slack. The effect is not there for affiliate loans (Table 4B, column 1). This again hints at that there was no perfect substitution between direct cross-border and affiliates claims when for example banks considered their overall cross-border concentration exposure to a particular borrower. 9 In terms of bilateral relationship, we find less cross-border claims to borrower countries where a recent (after 1945) colonial relationship exists (Table 4A, column 5), or when a common language was present in the case of affiliates claims (Table 4B, column 2). Common language and legal systems and contiguous borders are not associated with less 9 This is also supported by the fact that the interaction of cross-border concentration exposure and having an affiliate in the same country was not statistically significant in direct cross-border regressions.

16 16 reduction in direct cross-border claims, perhaps as with such cultural ties, these claims had actually grown faster before the financial crisis. Common language is statistically significant positive for affiliate claims, however, consistent with the notion that transaction costs with local presence are then less. These regression results are confirmed when including all variables (Table 4A and 4B, column 6). Distance is usually considered linked in the literature a proxy for the degree of transaction costs and information asymmetries (see de Haas and Van Horen, 2013). The greater the distance between the lender and borrower countries, the larger indeed the reduction in banks claims for direct and affiliate claims (Table 4A and 4B, column 7), but distance is never statistically significant. We also include bilateral trade, measured as a share of home banking system s GDP calculated before the crisis episodes. The reduction in cross-border lending could relate to the drop in trade around the crisis periods as banks did cut back in general on trade finance. At the same time, bilateral trade could relate to the familiarity of the creditor banking system with the specific host country and the absence of information asymmetries. As such, higher trade intensity could be associated with less cutback in cross-border lending. We include the bilateral trade variable first alone and then also with the distance variable (Table 4A and 4B, columns 8-9). The regression results show that trade has a negative effect on direct claims and for affiliates claims, which is statistically significant when also distance is included. This suggests that there are these offsetting impacts make for an overall ambiguous effect, but that the trade finance channel is more important as adding the distance variable, a proxy for the absence of information asymmetries, makes trade statistically significant. Importantly, including trade does not change the main results, notably that banking systems with greater vulnerabilities cut back more on their direct cross-border as well as affiliate lending. In general, the results in Table 4 show that some creditor-borrower characteristics (e.g., proximity, trade relationships, and historical relationships) had a role explaining banking systems deleveraging, but not as large as supply side characteristics. The contribution of creditor-borrower characteristics, especially in the case of direct cross-border loans, to the total R-squares were substantially lower than the contribution of supply side factors. We also perform the same analysis during the 2011Q3-11Q4 deleveraging episode. Regression results (not reported) show that mostly deleveraging during this periods was much more idiosyncratic, with just distance and trade as the statistically significant variables for banks direct cross-border and affiliates deleveraging patterns, respectively. This suggests that during this second period banks had mostly rebalanced their portfolios, and largely deleveraged in response to their vulnerabilities across the board.

17 17 Robustness and other tests We conducted some further regressions as robust tests. We already include regressions with winsorized data (in Tables 2 and 3) and these results were similar. We also did single clustering, instead of the double clustering shown, and results do not change. Furthermore, we checked whether some other supply variables made a difference. Besides being confronted with capital shocks, banking systems also suffered from unanticipated liquidity and funding shocks. Especially being unable to fund easily assets in dollars, banks had to adjust their lending dramatically following the events. We use the McGuire and Goetz (2009) source country banking system gross short-term dollar funding need measures (as also used by Cetorelli and Goldberg (2011)).While the data reduce our sample considerably, from about 2000 observations to about 1000 observations, the regression results remain similar as well in terms of coefficient signs but dollar shortfall variables are not statistically significant (especially if double clustering is implemented). Similar results, also with a considerable drop in the sample size, are obtained if the change in the market-to-book ratio of equity of banks of country i as a proxy of funding conditions (similar to Giannetti and Laeven 2012 and De Haas and Van Horen 2012); or the average spread in the overnight swap rate in banking system i during the deleveraging episode (similar to Giannetti and Laeven 2012). We also tested for the importance of local funding conditions for affiliate lending, but this was not statistically significant. We next test whether there are differences in regression results between the sample of banking systems that also have local affiliates in the same destination country and those that do not. One could expect that for those banking system that do have local affiliates as well, that there is greater scope for substitution in forms of lending as well as for spillovers of shocks at home to foreign markets. When using a matching sample, where banks lend to the borrower through both direct cross-border and affiliates activities, we find that the main regression results are confirmed (comparing Table 2A and 5B; note that Table 5B is almost identical to Table 2B as the sample only varies by 14 observations as very few countries have affiliate lending without cross-border lending). Both market-based and accounting measures of vulnerabilities have the same signs and statistically significance. And quantitatively, effects are of similar magnitudes. Interesting though, the role of the systemic crisis dummy becomes more important and is now of the same magnitude for direct loans as for affiliate loans. It suggests that the home bias induced by government interventions in systemic crisis countries affected both direct and affiliates claims equally when both forms of cross-border banking lending were present (in term of other variables, it remains that the two forms of international lending behaved differentially). There are several possible interpretations of this finding. One is that when both forms were present, their magnitudes was also relatively large, making authorities in

18 18 creditor countries call for comparable reductions in both as a quid pro quo for the government support extended. In contrast, for those countries where banks only engage in cross-border loans, amounts may have been smaller, making authorities less interested in calling for comparable reductions. Of course, it could also be that the sample of creditorborrower country pairs differs in such a specific way to generate this finding In addition, we tested the robustness of the creditor-borrower determinants by including credit bank country fixed effects instead of the systemic capital risk variables. The results are very similar with only minor changes in the level of significance of other variables and slightly larger coefficients for some variables (see Table 6 and compared to Table 4). Notably, recent colony remains statistically significant for direct lending and common language, distance and bilateral trade for affiliate lending. 4. Conclusions We analyze the role of supply, borrower and other lender-borrower factors in driving changes in international banking claims, considering both direct foreign loans and local affiliated subsidiaries lending. We study these changes around two shocks: that associated with the Lehman failure in 2008; and that with an intense period in the euro crisis in the Fall of Relative to the existing literature, we innovate in two particular ways. First, we use data that take into account exchange rate variations and break-in-series due to coverage, allowing a more meaningful representation of the evolution of banks claims. Second, we exploit differences between cross-border banking and affiliated lending, with the latter less impacted during the deleveraging episodes (an aggregate decline of only 5 percent, compared to 23 percent for direct cross-border loans). Our findings reconfirm the importance of supply factors for driving international capital flows. Controlling for demand and other borrower related factors, we find that banking systems deleveraging largely varied with market-based measures of vulnerabilities of banking systems to shocks, with accounting variables displaying largely perverse results in that banking system with higher return on assets and capital adequacy deleveraged more. Creditor-borrower characteristics (e.g., concentration of loans, cultural and geographical proximity, trade relationships) played some roles as well. We also find some evidence of barriers to the movement of resources across borders in that the factors explaining the patterns in reductions in direct loans differ from those explaining the reduction in lending by local affiliates. The relevance and importance of our findings, notably those related to supply factors, but also including those relating to the bilateral relationships between the source and destination country, matter for policy. They highlight that accounting measures that are backward-

19 19 looking can be very poor guides to the risk of deleveraging and suggest that market-based that are forward-looking can be much more informative. As such, our findings suggest that financial stability and other assessment should incorporate more such measures. Second, the finding have implications for borrowing countries. It was already established that countries with better macroeconomic and structural policies were less vulnerabilities to deleveraging. Our findings suggest that countries also consider the set of source country of banks in light of possible adverse spillovers. Specifically, besides have a diversified set of lenders to reduce the risks of idiosyncratic shocks affecting their supply of loans, countries may also want to consider the relative importance of their liabilities in lender banking system assets as a large concentration may lead to greater deleveraging during periods of stress.

20 20 References Aiyar, S., C. Calomiris, J. Hooley, Y. Korniyenko, and T. Wieladek, 2013, The International Transmission of Bank Capital Requirements: Evidence from the UK, Aiyar, S. C. Calomiris, and T. Wieladek, 2013, Identifying Channels of Credit Substitution When Bank Capital Requirements Are Varied, Acharya, V., C. Brownlees, R. Engle, F. Farazmand and M. Richardson (2010). Measuring Systemic Risk in Acharya, Viral, Thomas Cooley, and Mathew Richardson (Eds.), Regulating Wall Street: The Dodd-Frank Act and the New Architecture of Global Finance, John Wiley and Sons. Cerutti, E., 2013, Banks Foreign Credit Exposures and Borrowers Rollover Risks: Measurement, Evolution and Determinants, IMF Working Paper 13/9. Cerutti, E., G. Hale, and C. Minoiu, 2013, Major Financial Events and the Composition of Cross-Border Banking Exposures, mimeo. Cetorelli, N., and L. Goldberg, 2011, Global Banks and International Shock Transmission: Evidence from the Crisis. IMF Economic Review, Vol. 59, pp Cetorelli, N., and L. S. Goldberg, 2012a, Follow the money: quantifying domestic effects of foreign bank shocks in the Great Recession, American Economic Review, 102, Cetorelli, N., and L.S. Goldberg, 2012b, Liquidity Management of US Global Banks: Internal Capital Markets in the Great Recession, Journal of International Economics, vol.88, pp Cetorelli, N., and L.S. Goldberg, 2013, Banking Globalization and Monetary Transmission, forthcoming Journal of Finance. Claessens, S., and N. van Horen, 2012, Being a foreigner among domestic banks: asset or liability? Journal of Banking and Finance, 36, Claessens, S., and N. van Horen, 2013a, Foreign banks: trends and impact, Journal of Money, Credit and Banking, forthcoming Claessens, S., and N. van Horen, 2013b, Location decisions of foreign banks and competitor remoteness, Journal of Money, Credit and Banking, forthcoming Cull, R., and M. S. Martinez Peria, 2012, Bank ownership and lending patterns during the financial crisis: evidence from Latin America and Eastern Europe, World Bank Policy Research Working Paper, No De Haas, R., and I. Van Lelyveld, 2006, Foreign banks and credit stability in Central and Eastern Europe: a panel data analysis, Journal of Banking and Finance, 30, De Haas, R., and I. Van Lelyveld, 2010, Internal capital markets and lending by multinational bank subsidiaries, Journal of Financial Intermediation, 19, De Haas, R., and I. Van Lelyveld, 2012, Multinational banks and the global financial crisis: weathering the perfect storm, Journal of Money, Credit and Banking, forthcoming De Haas, R., and N. van Horen, 2012, International shock transmission after the Lehman Brothers collapse: evidence from syndicated lending, American Economic Review Papers & Proceedings, 102, De Haas, R., and N. van Horen, 2013, Running for the exit? International bank lending during a financial crisis, Review of Financial Studies, forthcoming

21 21 Giannetti, M., and L. Laeven, 2012a, The Flight Home Effect: Evidence from Syndicated Loan Market during Financial Crises, Journal of Financial Economics, vol. 104(1), pp Giannetti, M., and L. Laeven, 2012b, Flight Home, Flight Abroad, and International Credit Cycles, American Economic Review: Papers & Proceedings, vol. 102(3), pp Hale, G., T. Kapan, and C. Minoiu, 2013, Crisis Transmission in the Global Banking Network, mimeo. Idier, Lamé and Mésonnier (2013), How Useful is the Marginal Expected Shortfall for the Measurement of Systemic Exposure? A Practical Assessment, ECB Working Paper Ivashina, V., and D. Schrfstein, 2010, Bank Lending during the Financial Crisis, Journal of Financial Economics, 97, no. 3, pp Kapan, Tümer and Camelia Miniou (2013), Balance Sheet Strength and Bank Lending During the Global Financial Crisis, mimeo, IMF. Khwaja, A.; and A. Mian, 2008, Tracing the effect of bank liquidity shocks: evidence from emerging markets, American Economic Review, Vol 98(4). Laeven, L., and F. Valencia, 2012, Systemic Banking Crises Database: An Update, IMF Working Paper 12/163 (Washington: International Monetary Fund). López-Espinosaa, Morenoa, Rubiab, and Valderramac,2012, Short-term Wholesale Funding and Systemic Risk: A Global CoVaR, Journal of Banking and Finance, 36 (12), pp McGuire, P., and G. von Peter, 2009, The US Dollar Shortage in Global Banking, BIS Quarterly Review, March. Ongena, S., J. L. Peydro, and N. van Horen, 2013, Shocks abroad, pain at home? Bank-firm level evidence on financial contagion during the crisis, De Nederlandsche Bank, Working Paper

22 22 Table 1 - Summary Statistics (Period 2008Q2-09Q2) Variable Description Obs Mean Std. Dev. Min Max ur_cb_adjlog09 Adj Direct Cross-border, log difference: 2009Q2 minus 2008Q ur_cb_adjlog09wins ur_l_adjlog09 ur_l_adjlog09wins SRISK_GDP07 SRISK_GDP07* Asia Adj Direct Cross-border, log difference: 2009Q2 minus 2008Q1, winsorised using percentiles 5 and 95% Adj Affil. claims, log difference: 2009Q2 minus 2008Q1 Adj Affil. claims, log difference: 2009Q2 minus 2008Q1, winsorised using percentiles 5 and 95% Creditor country sum of positive SRISK (Weighted by GDP), measured as of Dec 2007; (CAR of 8% was used in all calculations) interaction variable: SRISK_GDP07 * dummy for Asian creditor banking systems (Japan and Australia) SRISK_GDP07* WH interaction variable: qsrisk_gdp07 * dummy for Western Hemisphere creditor banking systems (US and Canada) systemic_crisis systemic banking crisis from Laeven and Valencia (2010) roa07 Return on Assets of domestically-owned banks, as of Dec npl07 Non-performing loans of domestically-owned banks, as of Dec tier07 Tier I capital ratio of domestically-onwed banks, as of Dec comlang_off 1 for common language comlegal 1 for common legal origin contig 1 for geographical contiguity col45 1 for pairs in colonial relationship after bitrade07 Bilateral trade (normalized by homecountry GDP), measured as of Dec distlog log of distance (most populated cities, km) share_cbi07 Share of cross-border claims on borrower j by lender i, wrt total cross-border claims by lender i (in percentage); as of Dec

23 23 Table 1 - Summary Statistics (cont.) (Period 20011Q3-11Q4) Variable Description Obs Mean Std. Dev. Min Max ur_cb_adjlog11 Adj Direct Cross-border, log difference: 2011Q4 minus 2011Q ur_cb_adjlog11wins Adj Direct Cross-border, log difference: 2011Q4 minus 2011Q2, winsorised using percentiles 5 and 95% ur_l_adjlog11 Adj Affil. claims, log difference: 2011Q4 minus 2011Q ur_l_adjlog11wins Adj Affil. claims, log difference: 2011Q4 minus 2011Q2, winsorised using percentiles 5 and 95% SRISK_GDP11 Creditor country sum of positive SRISK (Weighted by GDP), measured as of Jun 2011; (CAR of 8% was used in all calculations) SRISK_GDP11* Asia interaction variable: SRISK_GDP11 * dummy for Asian creditor banking system SRISK_GDP11* WH interaction variable: qsrisk_gdp11 * dummy for European creditor banking system systemic_crisis Laeven and Valencia (2010) roa11 Return on Assets of domestically-owned banks, as of Dec npl11 Non-performing loans of domestically-owned banks, as of Dec tier11 Tier I capital ratio of domestically-onwed banks, as of Dec comlang_off 1 for common language comlegal 1 for common legal origin contig 1 for geographical contiguity col45 1 for pairs in colonial relationship after bitrade11 Bilateral trade (normalized by homecountry GDP), measured as of June distlog log of distance (most populated cities, km) share_cbi11 Share of cross-border claims on borrower j by lender i, wrt total cross-border claims by lender i (in percentage); as of June

24 24 Table 2 - Deleveraging during 2008Q2-09Q2: Supply Side Determinants Panel A - Dependent Variable: Log Changes in Direct Cross-Border Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP ** *** ** ** *** *** ** ** (0.0249) (0.0228) (0.0460) (0.0447) (0.0230) (0.0229) (0.0469) (0.0460) SRISK_GDP07* WH *** *** *** *** *** *** *** (0.165) (0.190) (0.177) (0.182) (0.166) (0.187) (0.172) SRISK_GDP07* Asia *** *** *** *** *** *** *** (0.0877) (0.115) (0.111) (0.0911) (0.0850) (0.128) (0.128) roa (6.741) (10.32) (9.785) (10.53) (9.966) npl (1.416) (1.816) (1.728) (2.054) (1.939) tier (2.034) (2.239) (2.185) (2.321) (2.302) systemic_crisis (6.854) (6.409) (7.761) (7.426) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R-squared Panel B - Dependent Variable: Log Changes in Local Affiliates Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP (0.0288) (0.0244) (0.0563) (0.0495) (0.0262) (0.0252) (0.0561) (0.0488) SRISK_GDP07* WH 0.422*** 0.833*** 0.745** 0.637*** 0.583*** 0.847*** 0.759*** (0.163) (0.323) (0.300) (0.223) (0.203) (0.303) (0.281) SRISK_GDP07* Asia 0.275*** 0.367** 0.363** 0.133*** 0.117** 0.300* 0.294* (0.0845) (0.177) (0.161) (0.0403) (0.0523) (0.179) (0.163) roa (10.32) (14.89) (13.35) (15.56) (13.55) npl (2.586) (3.255) (3.266) (2.879) (2.865) tier * ** *** ** ** (3.094) (3.361) (2.914) (3.316) (2.787) systemic_crisis * ** * (8.949) (8.032) (6.711) (5.914) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R -squared Notes: Cross-sectional regression are estimated. The dependent variable changes were calculated as log differences between the end of the deleveraging episode and its start (coefficients are already reported in percentage). Columns 5, 7 and 9 show regressions with winsorised dependent variable (using percentiles 5 and 95%). All standard errors are double clustered by creditor bank and borrower country. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

25 25 Table 3 - Deleveraging during 2011Q3-11Q4: Supply Side Determinants Panel A - Dependent Variable: Log Changes in Direct Cross-Border Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP (0.0380) (0.0415) (0.0369) (0.0281) (0.0460) (0.0319) (0.0424) (0.0312) SRISK_GDP11* WH 0.583*** 0.530*** *** ** (0.178) (0.186) (0.145) (0.192) (0.146) (0.208) (0.155) SRISK_GDP11* Asia (0.0373) (0.0357) (0.0266) (0.0549) (0.0371) (0.0626) (0.0427) roa *** *** *** *** *** (2.214) (1.915) (1.379) (2.068) (1.461) npl *** *** *** *** *** (1.182) (1.149) (0.673) (1.071) (0.637) tier * * * * * (0.931) (0.874) (0.714) (0.884) (0.698) systemic_crisis (5.531) (3.773) (4.626) (3.171) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R-squared Panel B - Dependent Variable: Log Changes in Local Affiliates Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP ** * ** * * (0.0278) (0.0421) (0.0403) (0.0212) (0.0438) (0.0234) (0.0417) (0.0228) SRISK_GDP11* WH (0.321) (0.346) (0.147) (0.311) (0.130) (0.352) (0.152) SRISK_GDP11* Asia 0.247*** 0.207*** 0.184*** 0.198*** 0.181*** 0.186*** 0.173*** (0.0443) (0.0307) (0.0225) (0.0226) (0.0207) (0.0312) (0.0266) roa (2.346) (2.204) (1.135) (2.249) (1.309) npl *** *** ** *** ** (1.032) (0.816) (0.523) (0.800) (0.522) tier (0.738) (1.098) (0.695) (1.139) (0.699) systemic_crisis (5.297) (2.843) (4.599) (2.667) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R-squared Notes: Cross-sectional regression are estimated. The dependent variable changes were calculated as log differences between the end of the deleveraging episode and its start (coefficients are already reported in percentage). Columns 5, 7 and 9 show regressions with winsorised dependent variable (using percentiles 5 and 95%). All standard errors are double clustered by creditor bank and borrower country. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

26 26 Table 4 - Deleveraging during 2008Q2-09Q2: Creditor-Borrower Determinants Panel A - Dependent Variable: Log Changes in Direct Cross-Border Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP *** *** *** *** *** *** *** *** *** (0.0227) (0.0228) (0.0232) (0.0227) (0.0226) (0.0232) (0.0237) (0.0231) (0.0236) SRISK_GDP07* WH *** *** *** *** *** *** *** *** *** (0.163) (0.150) (0.165) (0.164) (0.167) (0.157) (0.168) (0.154) (0.166) SRISK_GDP07* Asia *** *** *** *** *** *** ** *** ** (0.0880) (0.0912) (0.0921) (0.0891) (0.0881) (0.0933) (0.0992) (0.0935) (0.101) share_cbi ** * * ** ** ** ** * * (0.446) (0.445) (0.439) (0.466) (0.448) (0.454) (0.480) (0.463) (0.485) comlang_off (7.469) (8.429) (8.311) (8.476) (8.407) comlegal (4.873) (5.406) (5.506) (5.470) (5.554) contig * * (8.322) (9.141) (12.49) (12.18) (14.14) col ** 21.70*** 21.72*** 21.44*** 21.51*** (6.513) (7.000) (6.995) (7.051) (7.048) distlog (4.292) (4.357) bitrade * (2.528) (2.457) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R-squared with FE & Supply R -squared Panel B - Dependent Variable: Log Changes in Local Affiliates Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP (0.0241) (0.0236) (0.0233) (0.0241) (0.0243) (0.0233) (0.0239) (0.0248) (0.0256) SRISK_GDP07* WH 0.449*** 0.429*** 0.462*** 0.456*** 0.475*** 0.425*** 0.531*** 0.368*** 0.538*** (0.167) (0.160) (0.169) (0.159) (0.178) (0.156) (0.183) (0.133) (0.186) SRISK_GDP07* Asia 0.278*** 0.327*** 0.287*** 0.288*** 0.275*** 0.318*** 0.378*** 0.270*** 0.377*** (0.0850) (0.0800) (0.0915) (0.0655) (0.0848) (0.0718) (0.110) (0.0758) (0.124) share_cbi (0.636) (0.716) (0.694) (0.629) (0.647) (0.654) (0.729) (0.651) (0.660) comlang_off 14.01* ** 17.83** (8.289) (9.161) (9.423) (8.328) (8.477) comlegal (6.790) (7.322) (7.398) (7.038) (7.174) contig (13.15) (14.30) (15.63) (12.73) (15.73) col (11.84) (11.38) (11.41) (10.75) (10.81) distlog (7.144) (6.860) bitrade * (7.258) (6.779) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R-squared with FE & Supply R -squared Notes: Cross-sectional regression are estimated. The dependent variable changes were calculated as log differences between the end of the deleveraging episode and its start (coefficients are already reported in percentage). All standard errors are double clustered by creditor bank and borrower country. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

27 27 Table 5 - Deleveraging during 2008Q2-09Q2: Supply Side Determinants with Matching Samples Panel A - Dependent Variable: Log Changes in Direct Cross-Border Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP ** *** *** *** ** ** (0.0327) (0.0195) (0.0393) (0.0375) (0.0127) (0.0119) (0.0361) (0.0348) SRISK_GDP07* WH *** *** *** *** *** *** *** (0.153) (0.195) (0.180) (0.184) (0.134) (0.210) (0.190) SRISK_GDP07* Asia *** * (0.0679) (0.106) (0.105) (0.0105) (0.0559) (0.0930) (0.0934) roa (9.121) (11.69) (11.43) (11.41) (11.15) npl (1.948) (2.102) (2.059) (2.280) (2.224) tier *** *** *** * * (1.780) (1.790) (1.745) (1.862) (1.782) systemic_crisis *** *** *** *** (4.079) (3.925) (3.195) (3.126) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R-squared Panel B - Dependent Variable: Log Changes in Local Affiliates Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) SRISK_GDP (0.0293) (0.0249) (0.0567) (0.0500) (0.0263) (0.0253) (0.0562) (0.0490) SRISK_GDP07* WH 0.435*** 0.840*** 0.751** 0.652*** 0.598*** 0.853*** 0.765*** (0.167) (0.325) (0.302) (0.224) (0.205) (0.304) (0.283) SRISK_GDP07* Asia 0.274*** 0.366** 0.361** 0.124*** 0.108** * (0.0845) (0.177) (0.160) (0.0404) (0.0530) (0.178) (0.162) roa (10.39) (14.90) (13.37) (15.63) (13.61) npl (2.602) (3.285) (3.297) (2.877) (2.867) tier * ** *** ** ** (3.106) (3.359) (2.911) (3.289) (2.760) systemic_crisis ** ** * ** (8.942) (8.021) (6.676) (5.892) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with FE only R -squared Notes: Cross-sectional regression are estimated. The dependent variable changes were calculated as log differences between the end of the deleveraging episode and its start (coefficients are already reported in percentage). Columns 5, 7 and 9 show regressions with winsorised dependent variable (using percentiles 5 and 95%). All standard errors are double clustered by creditor bank and borrower country. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

28 28 Table 6 - Deleveraging during 2008Q2-09Q2: Creditor-Borrower Determinants with Creditor and Borrower FE Panel A - Dependent Variable: Log Changes in Direct Cross-Border Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) share_cbi * * * * (0.588) (0.611) (0.595) (0.626) (0.613) (0.636) (0.644) (0.518) (0.551) comlang_off (8.407) (8.448) (8.414) (7.852) (7.654) comlegal (5.104) (4.627) (4.683) (4.512) (4.577) contig 14.12* 15.91* (7.502) (8.681) (11.85) (12.78) (14.26) col *** 28.73*** 28.82*** 29.68*** 30.08*** (7.325) (9.207) (9.293) (9.218) (9.383) distlog (4.593) (4.109) bitrade (2.814) (2.686) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Creditor FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with Borrower FE only R-squared with both FE R -squared Panel B - Dependent Variable: Log Changes in Local Affiliates Claims (in %) (1) (2) (3) (4) (5) (6) (7) (8) (9) share_cbi (0.598) (0.686) (0.676) (0.577) (0.608) (0.623) (0.743) (0.592) (0.626) comlang_off * 13.32* (8.501) (8.442) (8.538) (7.983) (7.984) comlegal (6.598) (7.004) (7.056) (6.871) (6.899) contig (12.61) (14.71) (17.07) (14.33) (17.29) col (11.66) (11.61) (11.75) (11.54) (11.64) distlog * ** (6.361) (5.788) bitrade * (5.938) (5.190) Borrower FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Creditor FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations R-squared with Borrower FE only R-squared with both FEs R -squared Notes: Cross-sectional regression are estimated. The dependent variable changes were calculated as log differences between the end of the deleveraging episode and its start (coefficients are already reported in percentage). All standard errors are double clustered by creditor bank and borrower country. Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

29 29

30 30 Figure 2: Bilateral evolution of banks claims during 2008Q2-09Q2 Total FI GR PT JP AU AT SP IR DE UK LU IT CA SE DN ND FR US CH BE Total SP LU FR JP AU PT SE IT FI IR AT US CA UK DN GR DE CH ND BE E Source: BIS, IFS, and authors estimations. Notes: Each cell depicts the bilateral lender in the columns, and borrowers in the rows percentage changes in cross-border claims (see legend for color scale). The left-hand side panel shows the bilateral evolution of cross-border claims. The righthand side panel displays the evolution of affiliates claims. The first column of each panel displays banks total lending to each borrower country. The columns are sorted from left to right by the overall degree of deleveraging of the creditor country and the rows are sorted from top to bottom by the overall degree of deleveraging at the borrowing country.

31 31 Figure 3: Bilateral evolution of banks claims during 2011Q3-11Q4 Total FI GR AU PT JP CA AT SP DN LU DE FR UK IR SE CH ND IT BE US Total SP BE SE UK IT AU PT FI AT FR US LU GR IR CA DE ND CH JP DN Source: BIS, IFS, and authors estimations. Notes: Each cell depicts the bilateral lender in the columns, and borrowers in the rows percentage changes in cross-border claims (see legend for color scale). The left-hand side panel shows the bilateral evolution of cross-border claims. The righthand side panel displays the evolution of affiliates claims. The first column of each panel displays banks total lending to each borrower country. The columns are sorted from left to right by the overall degree of deleveraging of the creditor country and the rows are sorted from top to bottom by the overall degree of deleveraging at the borrowing country.

32 Second Draft December 4, 2013 Figure 4 - Evolution of Cross-Border and Affiliates' Claims Period 2008Q2-09Q2 Period 2011Q3-11Q4