Policy Research Working Paper 9037 Regulatory Arbitrage and Cross-Border Syndicated Loans Asli Demirgüç-Kunt Bálint L. Horváth Harry Huizinga Europe and Central Asia Region Office of the Chief Economist October 2019 Policy Research Working Paper 9037 Abstract This paper investigates how international regulatory and regulations and institutions. The results indicate that inter- institutional differences affect lending in the cross-bor- national banks have a tendency to switch loan origination der syndicated loan market. Lending provided through a toward countries with less stringent bank regulation and foreign subsidiary is subject to subsidiary-country regula- supervision consistent with regulatory arbitrage, but that tion and institutional arrangements. Multinational banks’ they prefer to originate loans in countries with higher-qual- choices between loan origination through the parent bank ity institutions related to financial market monitoring, or through a foreign subsidiary provide information about creditor rights, and the speed of contract enforcement. these banks’ preferences to operate in countries with varying This paper is a product of the Office of the Chief Economist, Europe and Central Asia Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at ademirguckunt@worldbank.org, H.P.Huizinga@uvt.nl, and balint.horvath@bristol.ac.uk. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Regulatory Arbitrage and Cross-Border Syndicated Loans1 Asli Demirgüç-Kunt World Bank Bálint L. Horváth University of Bristol Harry Huizinga Tilburg University and CEPR Keywords: Regulatory arbitrage, creditor rights JEL classification: G21, G38 1 This paper’s findings, interpretations, and conclusions are entirely those of the authors and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. We thank Iftekhar Hasan and participants at the LAPE-FINEST 2019 Spring Workshop for useful comments. 1. Introduction Banks have become increasingly international through the ownership of foreign subsidiary networks and the provision of cross-border loans. The Bank for International Settlements (BIS) reports that international banks’ foreign claims, which include claims through foreign subsidiaries as well as cross-border loans, stood at 15.9 trillion dollars in 2012.2 While banks have become more international, bank regulation and supervision remain mostly national. This implies that international bank flows to some extent could be driven by international regulatory differences as banks seek to avoid burdensome regulation in their home countries. Using aggregate data on international bank flows from the BIS, Houston, Lin, and Ma (2012) find that banks tend to have more claims on countries with fewer regulations, which they interpret as evidence of regulatory arbitrage by international banks. The provision of loans to countries with less stringent regulations is indeed consistent with regulatory arbitrage, if these loans are originated by subsidiaries located in these countries, as then the more lenient regulatory regime applies. However, the greater claims on countries with fewer regulations could reflect more lending by international banks’ establishments located in other countries with relatively stringent regulation, in which case there is no regulatory arbitrage. The aggregate BIS data do not give details on where international loans have been originated, and hence cannot provide conclusive evidence of whether international banks engage in regulatory arbitrage. In this paper, we examine regulatory arbitrage using micro data on cross-border syndicated loans.3 For each loan, we know where a bank has originated the loan. In 2 Based on BIS Locational Banking Statistics data. 3 Cross-border syndicated loans are an important component of international banking flows with total new commitments of $1.8 trillion in 2012 (BIS Quarterly Review, September 2013, Table 10). 2012 is the last full year for which the BIS reports aggregate cross-border syndicated loan volumes. 2 particular, an international bank can provide a loan through an entity located in its parent country, or alternatively through a foreign subsidiary located either in the borrower’s country or in a third country. In our sample for the years 1995-2016, we find that the shares of the loan volume provided through foreign subsidiaries located in the borrower’s country and in a third country were 8.9% and 3.1%, respectively. For 12.7% of loans provided through borrower-country subsidiaries, the subsidiary country had less stringent capital regulation than the parent country, while this was the case for 30.6% of loans provided through a third- country subsidiary. These data are consistent with some regulatory arbitrage by international banks via loan origination through subsidiaries located in countries with relatively lenient capital regulations. Analogously to Houston, Lin and Ma (2012), we estimate the impact of bank regulation and institutional quality on the aggregate volume of syndicated loans to borrower countries, and also on syndicated loan volumes bilaterally at the borrower country, lender country level. In addition, we examine how regulation and institutions affect the share of loans provided through foreign subsidiaries. We find that loan inflows into borrower countries are positively related to private monitoring incentives related to financial institutions in these countries, consistent with a positive impact of institutional quality on loan inflows. Further, the foreign subsidiary share of loans at the borrower country level is negatively related to borrower country capital regulation stringency consistent with regulatory arbitrage, while it is positively related to the strength of creditor rights and the speed of contract enforcement in borrower countries, reflecting a role for institutional quality to affect the location of loan origination. When considering syndicated loan data at the bilateral level, we find that loan volumes negatively reflect capital regulation stringency in lender countries consistent with regulatory arbitrage, while they positively reflect several aspects of institutional quality in both borrower 3 and lender countries. The subsidiary loan share is positively related to capital stringency and the strength of official supervisory power in lender countries consistent with regulatory arbitrage. Regulatory arbitrage appears to have been more important during the first part of our sample period in the years 1995-2005. The estimation using bilateral loan data also suggests a strong role for the quality of institutions to affect the location of loan origination. In particular, we find that stronger creditor rights and speedier contract enforcement in borrower countries, and weaker monitoring incentives and slower contract enforcement in lender countries, lead to a higher foreign subsidiary share in syndicated loan origination. International regulatory and institutional differences potentially also affect the determination of the lead bank of the loan syndicate among the participating banks. The lead bank performs the main tasks of borrower selection and monitoring on behalf of the entire syndicate, and regulation and institutions could affect a bank’s comparative advantage and incentives to perform these tasks well. We find that stricter capital regulation and greater supervisory power discourage a lead bank role consistent with regulatory arbitrage, while such a role is facilitated by a greater rule of law. Several papers have considered how regulatory arbitrage affects the structure of international banks. Considering international bank M&As, Karolyi and Taboada (2015) find evidence of regulatory arbitrage whereby acquisition flows involve acquirers from countries with stronger regulations than their targets, while abnormal returns are larger when acquirers come from more restrictive banking environments. This suggests that stringent regulations provide banks with a comparative advantage to expand abroad. Consistent with this, Frame, Mihov, and Sanz (2016) show that U.S. Bank Holding Companies (BHCs) are more likely to operate subsidiaries in countries with weak regulation and supervision, and that financial institutions’ decisions to operate in environments with lax environments are associated with 4 an increase in BHC risk and BHCs’ contribution to systemic risk. Carbo-Valverde, Kane, and Rodriguez-Fernandez (2012) find evidence that differences in the size and benefits of safety- net benefits available to banks in individual EU countries help to account for cross-border merger activity.4 Some papers have examined how regulatory arbitrage affects a multinational bank’s operations in foreign banking markets. Examining the riskiness of bank lending, Ongena, Popov, and Udell (2013) find that lower barriers to entry, tighter restrictions on bank activities and to some degree higher minimum capital requirements at home are associated with lower banking standards abroad. Considering international banks operating in the UK, Reinhardt and Sowerbutts (2018) find that a tightening of capital requirements at home reduces UK branches’ interbank lending growth relative to their UK subsidiaries. A few papers address regulatory competition regarding capital standards from a theoretical perspective. Dell’Ariccia and Marquez (2006) analyze a two-country model where higher standards in one country create a positive externality for the other country, as it restricts overall loan supply and banking market competition, thereby rendering the other country’s financial system more stable. In this setting, noncooperative capital standards are too low. Acharya (2003) considers the welfare implications of introducing common capital standards in a world where countries also compete in the area of regulatory forbearance policies. Regulatory forbearance in one country creates a negative spillover for the other country as it increases risk-taking by domestic banks, which reduces the profitability of foreign banks. This leads to noncooperative levels of forbearance that are too high in equilibrium. However, incomplete coordination only in the area of capital standards can be worse than no coordination at all, as it causes countries to compete more fiercely in the area 4 Buchak, Matvos, Piskorvksi and Seru (2018) provide evidence of regulatory arbitrage between banking and shadow banking sectors in the United States. 5 of forbearance policies. Morrison and White (2009) also find that introducing common capital standards can be welfare reducing if there are international differences in the quality of regulation that can induce banks managed by better-skilled managers to relocate towards the better-regulated economy. Several empirical papers address how information asymmetries affect loan syndicate structure. Sufi (2007) finds that the lead bank retains a larger loan share and forms a more concentrated syndicate when it is necessary to undertake more intense monitoring and due diligence. Lin, Ma, Malatesta and Xuan (2012) show that when the control-ownership divergence of a borrower is large, lead arrangers form more concentrated syndicates. Amiran, Beaver, Landsman, and Zhao (2017) find that the introduction of credit default swap (CDS) trading for a borrower’s debt decreases the share of loans retained by loan syndicate arrangers. Ball, Bushman, and Vasvari (2008) document that when a borrower’s accounting information possesses higher debt-contracting value, lead arrangers retain a smaller proportion of new loans. Bosch and Steffen (2011) report that syndicates are smaller if firms are privately held or unrated. Esty and Megginson (2003) find evidence that lenders that cannot rely on legal enforcement mechanisms to protect their claims create larger and more diffuse syndicates as a way to deter strategic default given that larger syndicates make it more costly to restructure loans. Extending the literature on syndicate structure, this paper addresses how international regulatory and institutional differences affect a bank’s propensity to originate syndicated loans through a foreign subsidiary, and how these differences influence which participating bank is the lead bank. Section 2 describes the data. Section 3 presents empirical results on how regulatory and institutional differences affect syndicate loan volume in the aggregate as well as the foreign subsidiary share of these loans. Section 4 presents empirical results on how these differences affect the determination of the lead bank in a loan syndicate. Section 5 concludes. 6 2. Data To study regulatory arbitrage in the cross-border syndicated loan market we obtain data on syndicated loans from Loan Pricing Corporation’s Dealscan database for the period 1995- 2016. This database contains loan level information including the identities and residencies of borrowers and lenders as well as lenders’ contributions to each loan. We use these data to construct aggregate loan volumes that are dependent variables in the regressions. In addition, we investigate credit ratios that reflect international banks’ credit provision through foreign subsidiaries relative to their total cross-border syndicated loan provision. The Dealscan database provides only recent information on lenders and their ultimate parents. In practice, ownership links vary over time due to mergers and acquisitions. To address this issue, we rely on the Dealscan-Compustat link provided by Schwert (2018) that connects the most active lenders in Dealscan to their respective banking groups on a quarterly basis, reflecting mergers and acquisitions over time. We then define a lender as a foreign subsidiary if the banking group that it belongs to according Compustat is headquartered in another country than the lender itself as reported by Dealscan.5 As we are interested in international bank regulatory arbitrage, we drop all loans provided by non-bank lenders and purely domestic loans where the ultimate parent bank and the borrower reside in the same country. Figure 1 plots the development of cross-border syndicated loan provision by banks over the 1995-2016 period, showing a generally upward trend that was temporarily reversed during the financial crisis of 2008-2009. We match the loan data with borrower and lender country variables from various sources as detailed below. After these steps, we obtain a sample of 149,416 individual bank loans to 5  After merging with the Dealscan-Compustat link our database covers about 31% of the entire volume of cross- border syndicated loans in Dealscan, and about 18% of the number of cross-border syndicated loans.   7 borrowers in 119 countries by lenders with ultimate parents in 10 countries.6 This sample is used in the regressions explaining lead bank selection. For about 70% of the individual loan contributions, Dealscan does not report the loan volume, which limits the number of borrower countries in our sample to 82 and the number of lender countries to 10 in the loan volume regressions.7 We consider loan volume regressions where loan volumes are aggregated alternatively at the borrower country-year level and the borrower country-lender country-year level. Panel A of Table 1 provides summary statistics for variables at the borrower-country level. In particular, Volume is the sum of the US dollar value of all loans with a mean of $2.6 billion. To be able to use observations with a zero loan volume, we consider two alternative transformations of the loan volume. First, Log(Volume + 1) is the natural logarithm of 1 plus the sum of the US dollar value of loans with a mean of 11.99. Second, Arsinh(Volume) is the transformed loan volume using the inverse hyperbolic sine function, with a mean of 12.41.8 We construct three credit ratios that inform on potential regulatory arbitrage through the usage of a foreign subsidiary located in the borrower country or in a third country. First, Foreign subsidiary/total volume is computed as the share of loans provided by any foreign subsidiary with a mean of 0.095. Figure 2 plots the development of the share of cross-border loan volume provided by a foreign subsidiary over the period 1995-2015. Foreign subsidiary/total volume is apparently procyclical, increasing from about 7% in 2001 to about 16% in 2008. Furthermore, Borrower-country subsidiary/total volume is the share of loans 6 By a loan we refer to the individual contribution of a lender in a facility. Facilities are credit agreements between a borrower and one or more lenders. 7 For the top 5 lender countries, Table A2 in the Appendix provides information on the top 3 borrower countries and the top 3 subsidiary countries. For lenders in Germany, Japan, and the United Kingdom, the United States is the top borrower as well as subsidiary country. For lenders in the United States, the United Kingdom is the top borrower country as well as the top subsidiary country. For lenders in France the top borrower country is the United States and the Netherlands is the top subsidiary country. 8 The inverse hyperbolic sine transformation transforms loan volumes according to the function arsinh ln √ 1 . 8 provided by a subsidiary located in the borrower country, while Third-country subsidiary/total volume is the share of loans provided by a subsidiary located in a third country. These two variables have means of 0.018 and 0.077, respectively; implying that loan provision through third-country subsidiaries is relatively common. Our main independent variables are indices of the quality of bank regulation taken from the World Bank’s Bank Regulation and Supervision Surveys (Barth et al., 2004 and 2006). The available information is from five consecutive surveys. Following Houston, Lin and Ma (2012), we take values of the regulatory variables for the years 1995 to 1999 from the first survey (measuring regulation in 1999); for the years 2000 to 2003, we use the second survey reflecting the quality of regulation as of the end of 2002; for the years 2004 to 2007, the third survey’s results are used (reflecting regulation at the end of 2005); for the years 2008 to 2012, we take the results of the fourth survey documenting regulation at the end of 2012; finally, for the years 2013 to 2016 we use the results of the fifth survey measuring regulation at the end of 2016.9 Among the regulatory variables, Capital regulation (borrower) is an index of the stringency of capital adequacy standards in the borrower country, with higher values indicating greater stringency. Capital regulation (borrower) ranges between 1 and 10 and has a mean of 6.333. Overall activity restrictions (borrower) is an index of the extent to which banks can engage in securities, insurance and real estate activities in the borrower country, with higher values indicating more restrictions. This variable ranges between 3 and 12 and has a mean of 7.121. Official supervisory power (borrower) is an index of the power of the supervisory authorities in the borrower country to take specific actions to prevent and correct 9 One of the subcomponents of Capital regulation (Overall capital stringency) cannot be calculated using data from the fifth wave of the survey because of missing information. Hence, for this wave we take the values of Capital regulation in the fourth wave and adjust it using changes in the other subcomponent, Initial capital regulation, that is available. 9 problems in banks, with higher values indicating greater power. The range of this variable is 4-16, and it has a mean of 11.00. Following Houston, Lin and Ma (2012), we control for several country institutional variables that may affect international bank flows. Among these, Monitoring (borrower) is an index of the strength of private monitoring of banks through, for instance, certified audits and ratings by international credit rating agencies, with higher values indicating stronger monitoring. This variable ranges between 3 and 11 and is 7.717 on average. Creditor rights (borrower) taken from Djankov et al. (2007) measures the strength of creditors’ rights in case of a bankruptcy in the borrower country, with higher values indicating stronger creditor rights.10 Creditor rights ranges between 0 and 4, and has a sample mean of 1.913 for borrower countries. Information sharing (borrower) measures rules affecting the scope, accessibility, and quality of credit information available through public or private credit registries in the borrower country, with higher values indicating a greater availability of credit information. This variable ranges from 0 to 6 and has a mean of 3.496.11 Time to enforce contracts (borrower) is the time required to resolve a commercial dispute in the borrower country, calculated as the average number of calendar days from the filing of the lawsuit in court until the final determination and, in appropriate cases, payment in a country. Higher values indicate weaker contract enforcement. On average, it takes 629 days to resolve disputes in borrower countries.12 As a final institutional variable, Rule of law (borrower) captures perceptions of the extent to which agents in the borrower country have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property 10 Since the last available data are for 2007, for subsequent years we take the values of this year. 11 This variable is from the World Bank’s Doing Business database. The World Bank changed the methodology of measuring the Information Sharing Index in 2015. Until then, the index ranged between 0 and 6. Since the revised index ranges between 0 and 8, we multiply the post-2014 values of the index by 6/8. The first available year is 2004, and for prior years we take the values for this year. 12 This variable is from the World Bank’s Doing Business database. The first available year is 2004, and for prior years we take the values for this year. 10 rights, the police, and the courts, as well as the likelihood of crime and violence in a country, with higher values indicating higher confidence in the rule of law. This variable is measured in units of a standard normal distribution and ranges approximately from -1.69 to 2.1 with a sample mean of 0.406. We also control for the development and the size of the borrower country, as loan capital may flow to economically less developed countries and larger countries because of economies of scale effects. We include the natural logarithm of GDP per capita measured at constant 2010 US dollar prices, Log real GDP/capita (borrower), and the natural logarithm of the total population, Log population (borrower), in the borrower country. These two variables have means of 9.137 and 16.78, respectively. Panel B provides summary statistics for the data used in bilateral borrower-country, lender country loan volume regressions. In particular, this panel provides summary statistics for the same regulatory and institutional variables as Panel A, but now defined separately for borrower and lender countries.13 Comparing these data for borrower and lender countries, we see that on average capital regulations and monitoring institutions are relatively stronger in lender countries than in borrower countries, while the opposite is true for overall activity restrictions and official supervisory power. In the bilateral regressions, two additional explanatory variables are meant to control for the structure of the banking market in a borrower country, reflecting that higher margins in less competitive markets and markets with less significant government ownership of banks may attract more foreign bank lending. First, Concentration (borrower) is the assets of the five largest banks as a share of total commercial banking assets in the borrower’s country. Second, Government bank ownership (borrower) is the proportion of banking assets in 13 Creditor rights and Information sharing in lender countries are excluded because of a lack of sufficient variation in these variables. 11 government owned banks in the borrower country, where a bank is considered government owned if 50 percent or more of the shares are controlled by the government. These two variables have means of 73.31% and 16.75%, respectively. In the bilateral regressions, we also control for geographic distance and common language between lenders and borrowers. Specifically, Log distance is calculated as the natural logarithm of the physical distance between the capital cities of the respective borrower and lender countries with a mean of 8.417, and Common spoken language is the probability that a pair of people chosen at random from the borrower and lender countries understand one another in some language with a mean of 0.305. The latter variable is taken from Melitz and Toubal (2014). Going beyond loan volumes, we also consider the choice of the lead bank in an international loan syndicate. For this purpose, we define Lead to be a dummy variable indicating a lead arranger role for a bank in a loan syndicate. Following Bharath et al. (2011) and Berg et al. (2016), we set Lead equal to one if 1) the reported lender role in Dealscan is either “Admin agent”, “Agent”, “Arranger”, or “Lead bank”; or 2) the lead arranger credit field equals “Yes”; or 3) if the loan has a sole lender. Lead has a mean of 42.7%. The lead bank regressions include regulatory and institutional variables for the lender country and several additional control variables for the parent lender bank. Log assets is the natural logarithm of the bank’s total assets lagged by one year, and Log syndicated lending is the log of 1 plus the sum of the dollar value of all loans provided by a lender in a given year minus the pertinent loan amount with means of 13.63 and 22.52, respectively.14 Larger banks by assets may be in a better position to assume the lead bank role in a loan syndicate, as this tends to require a bank to retain a relatively large share of the overall facility on its own books (Sufi, 2007). Loans/deposits is the ratio of all loans to deposits lagged by one year with a mean of 0.902. Banks that retain more loans on their balance sheets may be more 14 Bank accounting data are taken from Compustat. 12 traditional, and less inclined to assume lead roles in loan syndicates. Equity/assets is the bank’s total common equity to total assets ratio lagged by one year with a mean of 0.05. Less capitalized banks may be less aggressively entrepreneurial, and hence less inclined to take on lead roles in loan syndicates. Finally, Past relationship is a dummy variable that indicates whether a bank has already provided at least one syndicated loan to the borrower in a previous year with a mean of 43.6%. A past relationship with a borrower makes it less necessary to provide a loan through a foreign subsidiary to facilitate monitoring, and it makes it easier for a bank to take on a lead role in new loan syndicate as the bank has an information advantage regarding the borrower. A multinational bank with foreign subsidiaries can engage in regulatory arbitrage in the area of capital regulation by originating syndicated loans through foreign subsidiaries located in countries with less stringent capital regulation. Figure 3 provides information on the extent to which loans have been ‘arbitraged’ over the 1995-2016 period in the sense that they have been originated through foreign subsidiaries located in countries with less stringent capital regulation than in the parent country. The share of ‘arbitraged’ loans reached a peak of about 8% in 2000, and generally was relatively high in the period 2000-2007 preceding the financial crisis, but has been more modest at less than 1% since then. The much lower share of arbitraged loans in recent years could conceivably reflect international convergence in capital regulatory standards, which would reduce the scope for such arbitrage. To see whether such convergence has occurred, Figure 4 displays the mean value and 10th percentile and 90th percentile values of the capital regulatory variable across countries for each of the World Bank regulation and supervision survey waves. The figure fails to provide evidence of international capital regulatory convergence over time.15 15 Alternatively, the use of foreign subsidiaries as a means to bring about lower capital regulatory standards may have declined over time as home-country regulations regarding banks’ international operations have tightened according to a confidential IMF survey covering the period 2006-2014 (see IMF, 2015). In the empirical work, a time trend in such regulations is controlled for by the inclusion of time fixed effects. 13 Arbitraged loans can be provided by subsidiaries located in either borrower countries or in third countries. As seen in Figure 5, the share of loans provided through subsidiaries in borrower countries that was arbitraged stood at a very high level of around 80% during 2000- 2004, while it has declined to much lower levels in recent years. In contrast, the fraction of loans provided through third countries that was arbitraged has remained relatively high in recent years and stood at 26.7% in 2016. Alternatively, Figure 6 plots the share of loans provided to borrowers located in countries with less stringent capital regulation than the country where the lender’s parent bank is headquartered. As discussed, Houston et al. (2012) interpret the provision of loans to borrowers located in countries with less stringent regulation than the parent bank’s country as potential evidence of regulatory arbitrage. The share of loans provided to borrowers in countries with relatively lax capital regulation reached a peak of 53.8% in 2000, and it stood at 11.6% in 2016. Generally, the share of ‘arbitraged’ loans in Figure 3 and the share of loans to borrowers in countries with relatively lax capital regulation in Figure 6 are seen to move similarly over time.16 To conclude this section, we compare the mean values of the regulatory variables for loans originated through foreign subsidiaries and through parent banks to see how on average loan origination through foreign subsidiaries has affected the stringency of the regulatory regime facing international banks that are active in the cross-border syndicated loan market. To start, Panel A of Table 2 provides the mean values of Capital regulation for loans provided though foreign subsidiaries in borrower countries (column 2), foreign subsidiaries in third countries (column 4), and establishments in parent countries (column 6). Separate mean values are provided for the overall sample period of 1996-2016, and for the subperiods 1996- 16 The similar shapes of the two figures could reflect that other things equal a relaxation of capital regulation in the borrower’s country makes it more likely that this capital regulation drops below the capital regulation of a foreign subsidiary country if applicable and also of the lender’s parent bank country. 14 2005 and 2006-2016. For the overall period, we see that the mean Capital regulation for loans provided through subsidiaries in borrower countries exceeded the mean value for parent- country loans, while it was lower (higher) in the first (second) subperiod (column 8). These results suggest that foreign subsidiaries located in borrower countries were used relatively heavily to avoid burdensome parent-country capital regulation in the first half of the sample period, in line with figure 5.17 On average, subsidiaries located in third countries were subject to more stringent capital regulation than banking establishments in parent countries throughout the sample period (column 9), which suggests that on average subsidiaries in third countries were not used to evade parent-country capital regulations. However, subsidiaries located in third countries were subject to less stringent capital regulation than subsidiaries located in borrower countries during the overall sample period, and also during 2006-2016 (column 10). Panel B provides analogous mean values of the activities restrictions variable for loans provided through foreign subsidiaries and parent banks. Subsidiaries located in borrower countries have been subject to more stringent activities restrictions than banks located in parent countries throughout (column 8), while subsidiaries located in third countries had fewer restricted activities relative to establishments in parent countries (column 9) and subsidiary banks in borrower countries (column 10). From Panel C, we see that on average subsidiaries in both borrower and third countries were subject to greater official supervisory power than banks in parent countries (columns 8 and 9), but that subsidiaries in third countries were subject to less powerful official supervision compared to subsidiaries in borrower countries for the overall period and during 1996-2005 (column 10). Finally, Panel D shows that subsidiaries in both borrower and third countries were subject to greater 17 For the loans provided through a subsidiary in the borrower country in Figure 5, the difference between the mean capital regulation variable in the borrower country and mean capital regulation in the pertinent parent country is significantly positive for the overall sample. It is significantly negative during 1996-2005 and significantly positive during 2006-2016. 15 financial market monitoring incentives than establishments in parent countries during the overall sample period (columns 8 and 9), pointing at the possibility that international banks prefer to originate loans through subsidiaries located in countries with stronger monitoring incentives. 3. Regulatory arbitrage and loan volumes In this section, we consider how bank regulation and other institutional variables affect cross-border syndicated loan flows in the aggregate and the shares of loans originated in subsidiary countries. In turn, we consider how bank regulation in borrower countries affects inflows of loans into these countries, and how bank regulations in borrower and lender countries jointly determine bilateral loan flows.18 3.1 Inflows of loans into borrower countries We estimate the following relation between the inflow of syndicated loans into a borrower country and bank regulatory and other variables: I (borrower)i,t = α + β0 Regulation (borrower)i,t + β1 X (borrower)i,t + ηi + t + εi,t (1) in which I (borrower)i,t is a variable characterizing the inflow of loans into borrower country i at time t. Regulation (borrower)i,t is a set of regulatory variables in country i at time t (Capital regulation, Overall activity restrictions, Official supervisory power), with higher values indicating tighter regulation. X (borrower)i,t is a set of institutional and control variables for country i at time t (Monitoring, Creditor rights, Information sharing, Time to enforce contracts, Log real GDP/capita and Log population), and ηi and t are sets of borrower country and year fixed effects, respectively. We report standard errors that allow for 18 We have too few lender countries in our data set to analyze loan volume variation at the lender country level. 16 clustering at the borrower country level. Similarly to Houston, Lin and Ma (2012, Table III), we expect to find that loan inflows into borrower countries, as measured by Log(Volume + 1) and Arsinh(Volume), are negatively related to the stringency of bank regulation in these countries, consistent with β0 < 0. In addition, we hypothesize that the usage of borrower- country subsidiaries, as reflected in the Foreign subsidiary/total volume and Borrower country subsidiary/total volume variables, is negatively related to borrower-country regulatory stringency consistent with regulatory arbitrage, giving rise to β0 < 0 in the pertinent regressions. Columns 1 and 2 of Table 4 provide the results of regressions of Log(Volume + 1) and Arsinh(Volume) along the lines of (1). Furthermore, column 3 reports a Tobit regression of Arsinh (Volume) as an alternative way to deal with truncation of the sample in case of zero credit flows.19 Capital regulation (borrower), Overall activity restrictions (borrower) and Official supervisory power (borrower) obtain insignificant coefficients in the three regressions. Monitoring (borrower) is estimated with positive and significant coefficients, suggesting that better private monitoring incentives encourage syndicated loan inflow into borrower countries. Information sharing (borrower) enters the three regressions with negative and significant coefficients, perhaps because information sharing on borrowers reduces the profitability of credit relationships to international banks. Credit inflows into borrower countries are further positively and significantly related to the rule of law, the log of GDP per capita and the log of the population in these countries. Next, regressions of Foreign subsidiary/total volume, Borrower country subsidiary/total volume and Third country subsidiary/total volume are provided in columns 4- 6. The foreign subsidiary share variable is negatively and significantly related to Official 19 Santos and Tenreyro (2006) propose a pseudo-maximum likelihood (PML) estimation technique to deal with zero observations in an international trade application. Application of this technique failed to reach convergence in our case. 17 supervisory power (borrower) in column 4, indicating that international banks avoid foreign subsidiary usage in case of greater official supervisory power in borrower countries. The foreign subsidiary and borrower country subsidiary shares vary positively with Monitoring (borrower) in regressions 4 and 5, as better financial market monitoring institutions in borrower countries appear to make borrower country subsidiary usage more attractive. In addition, the foreign subsidiary share is positively and significantly related to Creditor rights (borrower), and negatively and significantly to Time to enforce contracts (borrower). Overall, regression 4 provides evidence that foreign subsidiary usage is positively related to the quality of institutions in borrower countries as related to financial market monitoring, creditor rights, and the time to enforce contracts, while foreign subsidiary usage is negatively related to borrower country supervisory, power consistent with regulatory arbitrage. Bank regulation potentially is endogenous to absolute and relative loan volumes. Borrowing countries experiencing large syndicated loan provision by local subsidiaries of international banks, could, for instance, increase the quality of regulation to discourage such credit provision. To mitigate potential endogeneity, we next re-estimate regressions 4-6 of Table 4 using instrumental variables (IVs), taking Capital regulation (borrower), Overall activity restrictions (borrower), Official supervisory power (borrower), and Monitoring (borrower) to be potentially endogenous. Following Houston et al. (2012, p. 1879), we employ instrumental variables that have been advanced in the literature as possible determinants of regulation. First, we use the time-varying means of the regulatory variables (excluding the pertinent country) to reflect possible ‘regulatory contagion’ (see Demirguc- Kunt and Detragiache, 2002).20 Second, we use a dummy variable indicating that the central bank supervises banks for prudential purposes (from the World Bank regulation 20 Specifically, in case of a borrower-country regulatory variable, we instrument it by the mean of this variable for all borrower countries excluding the pertinent borrower country. In the regressions with bilateral data, in analogous fashion we construct the instruments for lender-country regulatory variables. 18 and supervision survey) to reflect that central bankers are more likely to choose bank regulation that promotes systemic stability (see Goodhart, 2002). A final instrument is the five-year moving average of the Gini index measuring income inequality (from the WDI), as regulation is in part shaped by its distributional consequences (see Beck, Levine, and Levkov, 2010).21 The resulting IV regressions are reported in columns 7-9 of Table 4. In the foreign subsidiary share and third country subsidiary share regressions 7 and 9, the instrumented capital regulation variable is estimated with negative significant coefficients, suggesting less foreign subsidiary usage in case of more stringent borrower-country capital regulation. In the IV regression 7, foreign subsidiary usage is positively related to Creditor rights (borrower) and negatively to Time to enforce contracts (borrower). Thus, in the IV regressions the picture remains one of subsidiary usage being negatively related to borrower country regulatory stringency, but positively to borrower-country institutional quality. As a specification test, for regressions 7-9 we conducted an overidentification test based on Hansen’s J statistic with as the null hypothesis that the instruments are valid, i.e. uncorrelated with the error term and correctly excluded from the estimating equation. As indicated in the table, this null hypothesis is not rejected for the three regressions. In addition, we conducted an under-identification test based on the Kleibergen-Paap rk Wald statistic with as the null hypothesis that the model is not identified, as the excluded instruments are not sufficiently correlated with the endogenous regressors. As seen in the table, in this instance the null hypothesis is rejected for the three regressions. Thus, the IV regressions 7-9 appear to be correctly specified. 3.2 Bilateral loan flows between borrower and lender countries 21 Other instrumental variables used by Houston et al. (2012) are subsumed by included fixed effects in our setting. 19 In this subsection we consider how syndicated loan flows aggregated at the bilateral borrower country, lender country level are related to bank regulatory stringency in both borrower and lender countries. We estimate equations as follows: I (bilateral)i,j,t = α + β0 Regulation (borrower)i,t + β1 X (borrower)i,t + γ0 Regulation (lender)j,t + γ1 X (lender)j,t + θ1 Log distancei,j + θ2 Common languagei,j + ηi + φj + t + εi,j,t (2) in which I (bilateral)i,j,t is a variable characterizing the flow of loans to borrowers in country i from banks headquartered in country j at time t, and ηi, φj and t are fixed effects for borrower and lender countries and time, respectively. Specification (2) includes regulatory variables and other institutional variables for both borrower and lender countries. We report standard errors that allow for clustering at the borrower country level.22 The potential effects of regulatory variables in borrower and lender countries on absolute and relative credit variables are analogous to the discussion in Section 3.1. Columns 1-2 of Table 4 report OLS regressions of Log(Volume + 1) and Arsinh(Volume), and a Tobit regression of Arsinh(Volume) with bilateral syndicated loan data, respectively. Capital regulation (Lender) enters the three regression with negative significant coefficients, suggesting that the total loan volume at the bilateral level declines with lender-country capital stringency consistent with regulatory arbitrage. Overall activity restrictions in the borrower country impact negatively and significantly on bilateral loan volumes in regressions 1-2, while overall activity restrictions in the lender country are positively and significantly related to loan volumes in regressions 1-3. This could reflect that 22 Two-way clustering at the borrower and lender country levels yields qualitatively similar results to the reported results. 20 restricted non-banking activities are complements to syndicated loan provision for borrower- country banks but substitutes for lender-country banks. The strength of monitoring incentives in both borrower and lender countries vary positively and significantly with bilateral loan volumes in regressions 1-3, as stronger monitoring incentives may facilitate funding for borrower-country and lender-country banks alike. In the Tobit regression 3, bilateral loan volume positively reflects borrower-country creditor rights and banking market concentration, and negatively the time to enforce contracts and government bank ownership in borrower countries. Bilateral loan volume is positively related to the rule of law in borrower countries in regressions 1-3, but unexpectedly it varies negatively with the rule of law in lender countries in regressions 1-2. As expected, bilateral loan volume varies negatively with bilateral distance, and positively with a common spoken language of borrower and lender countries. Columns 4-6 report the results of regressions of the overall foreign subsidiary loan ratio, and the borrower-country and third-country subsidiary loan ratios. Capital regulation (lender) enters regressions 4 and 6 with positive significant coefficients, which suggests that foreign subsidiaries generally and third-country subsidiaries specifically avoid burdensome parent-country capital regulation consistent with regulatory arbitrage. Official supervisory power in the lender country impacts positively and significantly on the overall foreign subsidiary ratio in regression 4, consistent with regulatory arbitrage so as to avoid stringent parent-country supervision. Furthermore, the supervisory power variable is estimated with a negative (positive) coefficient in the borrower-country (third-country) subsidiary loan ratio regression 5 (6). These results suggest that more powerful supervisors in parent-countries cause banks to provide fewer syndicated loans through borrower-country subsidiaries, but instead to channel more syndicated loans through subsidiaries in third countries. The negative and significant coefficients for Monitoring (lender) in regressions 4 and 6 indicate that 21 stronger monitoring incentives in lender countries reduce foreign subsidiary usage. Stronger creditor rights in borrower countries provide for greater relative use of foreign subsidiaries especially in third countries (regressions 4 and 6). More time to enforce contracts in borrower (lender) countries discourages (encourages) foreign subsidiary usage (regressions 4 and 6). The estimated coefficients in regression 4 can be used to assess the economic significance of the impact of regulatory variables on the foreign subsidiary share. The coefficient of 0.0281 for Capital Regulation (lender) implies that an increment in this variable by one standard deviation (1.387) increases the foreign subsidiary share by 0.039 (=0.0281*1.387), which corresponds to 16.31% of the standard deviation of the foreign subsidiary share and 40.81% of its mean. A one standard deviation rise in Official supervisory power (lender) increases the foreign subsidiary share by 0.027(=0.0115*2.332), corresponding to 11.22% of its standard deviation and 28.08% of its mean. A one standard deviation rise in Monitoring (lender) decreases the foreign subsidiary share by 0.031(=0.0218*1.421), corresponding to 12.96% of its standard deviation and 32.44% of its mean. These estimated effects of regulation on the foreign subsidiary share are economically meaningful. The results of IV estimation applied to the loan ratio regressions 4-6 are reported in columns 7-9. A positive and significant coefficient for Capital regulation (lender) in the IV regressions 7 and 9 points at greater foreign subsidiary usage by international banks in case of more stringent capital regulation in parent-bank countries, consistent with regulatory arbitrage. This regulatory arbitrage apparently takes the form of greater usage of especially third-country subsidiaries. Similarly, the positive significant coefficient for Official supervisory power (lender) in regression 9 is evidence that international banks increase their usage of third-country subsidiaries if subject to greater supervisory power in their home countries. Better monitoring institutions in lender countries, in contrast, reduce third-country 22 subsidiary usage (regression 9). In the IV regressions 7-9, estimated coefficients for creditor rights in borrower countries, and the time to enforce contracts in borrower and lender countries are very similar to regressions 4-6. The IV regressions appear to be well specified according to reported overidentification and under-identification tests. Overall, the results of Table 4 indicate that international banks are more likely to originate loan through foreign subsidiaries located in especially third countries if subject to stricter capital regulation and greater supervisory power in their home countries, consistent with regulatory arbitrage. In contrast, better institutions in lender countries relative to borrower countries lead to a lower subsidiary usage in the provision of cross-border syndicated loans. In particular, the foreign subsidiary share is negatively related to the quality of monitoring institutions and the speed of contract enforcement in lender countries, while it is positively related to creditor rights and the speed of contract enforcement in borrower countries. As discussed, Figure 3 suggests that arbitrage with respect to capital regulations was relatively prevalent in the first half of our sample period. To test this formally, we re-estimate the subsidiary ratio regressions 4-9 from Table 4 separately with data for the years 1995-2005 and 2007-2016. The results using data for the years 1995-2005 are displayed as columns 1-6 in Table 5. These results provide consistent evidence of higher subsidiary usage in case of more stringency in lender countries as related to capital regulation (columns 1, 3, 4 and 6), overall activity restrictions (columns 4 and 6), and overall supervisory power (columns 1 and 3). There is some evidence that stronger monitoring incentives in lender countries reduce subsidiary usage (column 3). Estimation results for the years 2006-2016 are shown in columns 7-12. Overall, these results suggest that regulatory arbitrage was more limited in the later period. In particular, the capital regulatory index for the lender country is insignificant in all of these regressions. The 23 overall activity restrictions variable for the lender country is negative and significant regressions 7, 9, 10 and 12, which suggests that these restrictions reduced subsidiary usage in the later period, inconsistent with regulatory arbitrage so as to avoid burdensome restrictions. This could reflect that syndicated lending functioned as a substitute rather than a complement to restricted non-bank activities in the later period. Regulatory arbitrage so as to avoid burdensome lender-county supervisory power appears to have continued in the later period (columns 7, 9, 10 and 12). The impact of monitoring institutions in lender countries on subsidiary usage in the later period is ambiguous given a positive impact on borrower country subsidiary usage in column 8, and a negative impact on third-country subsidiary usage in column 12. The IV regressions 4-6 and 1-12 appear to be well specified as they pass the reported overidentification and under-identification tests. 4. Determination of the lead bank A lead bank performs key selection and monitoring functions within a loan syndicate. Specifically, the lead bank initially selects the potential borrower and negotiates key elements of a prospective loan agreement. Subsequently it recruits other participating banks to provide their share of the loan funding. After a loan agreement has entered into force, the lead bank monitors the borrower in the interests of the entire loan syndicate, and it administers the loan. In this section, we examine the impact of bank regulation on the choice of a lead bank among the banks that participate in a loan syndicate. We estimate a relation as follows: Leadi,j,k,t = α + β0 Regulation (lender)i,t + β1 X (lender)i,t + θ1 Bi,t + Θ2 Past relationshipi,j,t + ηk,t + φi,,t + εi,j,k,t (3) 24 in which Leadi,j,k,t is a dummy variable signalling that bank i is a lead bank in a loan to borrower j as part of loan facility k at time t. Bi,t is a set of bank variables (Log assets, Log syndicated lending, Loans/deposits, and Equity/deposits).23 Past relationshipi,j,t is a dummy variable indicating whether the lender has provided a loan to the borrower before the pertinent loan. The basic regressions include facility fixed effects ηk,t, and lender country fixed effects φi,,t. We report standard errors that are clustered at the banking group level. A negatively estimated β0 suggests that stricter regulation makes it more difficult for a bank to assume a lead bank role. Stricter capital regulation could, for instance, discourage such a role, as the lead bank generally retains a relatively large share of the overall syndicated loan on the bank’s own books Table 6 reports the results of estimating (3). Regression 1 includes facility and lender country fixed effects. In regressions 2-4, we replace the lender country fixed effects by bank fixed effects, bank * borrower country fixed effects and bank * borrower company fixed effects, respectively. Columns 5-8 report the results of applying IV estimation to regressions 1-4. Capital regulation in the lender country enters negatively and significantly in the OLS regressions 2-3 and in the IV regressions 6-8. Stricter capital regulation thus discourages a lead bank role, consistent with regulatory arbitrage. Overall activity restrictions in the lender country are estimated with positive significant coefficients in the OLS regressions 1-3 and in the IV regressions 5 and 7, perhaps as a lead bank role can serve as a substitute for restricted non-bank activities. Greater supervisory power is estimated to make a lead bank role less likely given the negative significant coefficients in the OLS regression 3 and the corresponding IV regression 7, consistent with regulatory arbitrage. The rule of law variable obtains positive significant coefficients, as greater rule of law apparently facilitates a lead bank role. Among the bank level variables, larger size as measured by total assets or total 23 The creditor rights variable is not included because of insufficient variation in this variable. 25 syndicated lending, and lower loans-to-deposits and equity-to-assets ratios tend to be associated with a greater propensity to become the lead bank. The IV regression 5 fails the overidentification test, but IV regressions 6-8 pass the reported specification tests. Overall, Table 6 suggests that stricter capital regulation and greater supervisory power discourage a lead bank role consistent with regulatory arbitrage, while such a role is facilitated by a greater rule of law. 5. Conclusions In this paper, we investigate how international regulatory and institutional differences affect lending in the cross-border syndicated loan market. The syndicated loan data enable us to see whether a multinational bank provides a cross-border loan directly from the parent bank or indirectly through a foreign subsidiary. Lending provided through a foreign subsidiary is subject to subsidiary-country regulation and institutional arrangements. International banks thus can engage in regulatory arbitrage by originating loans through foreign subsidiaries located in countries with relatively lax regulation rather than through their parent banks. We find evidence that international banks’ usage of foreign subsidiaries is in part driven by international regulatory differences. In the case of loans aggregated at the borrower- country level, we find that stricter capital regulation in the borrower country negatively affects the share of loans provided through foreign subsidiaries. When considering syndicated loan data at the bilateral level, we find that the subsidiary loan share is positively related to capital regulatory stringency and the strength of official supervisory power in lender countries. In addition, we find that stricter capital regulation and greater supervisory power discourage a lead bank role in an international loan syndicate. These results suggest that international banks prefer to operate in locations with relatively lax bank regulation and 26 supervision, consistent with regulatory arbitrage. In the case of loan provision, regulatory arbitrage is shown to have been more important during the first part of our sample period in the years 1995-2005. Countries stand to benefit from additional banking activity by way of increased local banking employment and profitability. For this reason, regulatory arbitrage by banks could trigger international regulatory competition by countries, tending to a ‘race-to-the-bottom’ in regulatory standards that could possibly threaten financial stability. In addition to regulation, we find a strong role for institutional quality to affect the location of bank activity. The foreign subsidiary share of loans at the borrower country level is positively related to the strength of creditor rights and the speed of contract enforcement in borrower countries. Using bilateral data, we find that that stronger creditor rights and speedier contract enforcement in borrower countries, and weaker monitoring incentives and slower contract enforcement in lender countries, lead to a higher foreign subsidiary share in syndicated loan origination. Further, a bank’s lead bank role in a loan syndicate is encouraged by a greater rule of law. These results suggest that banks prefer to operate in countries with greater institutional quality and that countries that improve institutional quality could be rewarded by being able to attract additional international bank activity. 27 References Acharya, V., 2003, Is the international convergence of capital adequacy regulation desirable?, Journal of Finance 58, 2745-2781. Amiram, D., W. Beaver, W. Landsman, and J. Zhao, 2017, The effects of credit default swap trading on information asymmetry in syndicated loans, Journal of Financial Economics 126, 364-382. Ball, R., R. Bushman, and F. Vasvasi, 2008, The debt-contracting value of accounting information and loan syndicate structure, Journal of Accounting Research 46, 247-287. Barth, J., G. Caprio, and R. Levine, 2004, Bank regulation and supervision: what works best?, Journal of Financial Intermediation 13, 2005-248. Barth, J., G. Caprio, and R. Levine, 2006, Rethinking bank supervision and regulation: till angels govern, Cambridge, UK: Cambridge University Press. Beck, T., R. Levine, and A. Levkov, 2010, Big bad banks? The winners and losers from bank deregulation in the United States, Journal of Finance 65, 1637-1667. Berg, T., A. Saunders, and S. Steffen, 2016, The total cost of corporate borrowing in the loan market: Don't ignore the fees, Journal of Finance 71, 1357-1392. Bharath, S., S. Dahiya, A. Saunders, and A. Srinivasan, 2011, Lending relationships and loan contract terms, Review of Financial Studies 24, 1141-1203. Bosch, O., and S. Steffen, 2011, On syndicate composition, corporate structure and the certification effect of credit ratings, Journal of Banking and Finance 35, 290-299. Buchak, G., G. Matvos, T. Piskorski, and A. Seru, 2018, Fintech, regulatory arbitrage, and the rise of shadow banks, Journal of Financial Economics 130, 453-483. Carbo-Valverde, S., E. Kane, and F. Rodriquez-Fernandez, 2012, Regulatory arbitrage in cross-border banking mergers within the EU, Journal of Money, Credit and Banking 44, 1609-1629. Dell’Ariccia, G., and R. Marquez, 2006, Competition among regulators and credit market integration, Journal of Financial Economics 79, 401-430. Danisewicz, P., D. Reinhardt, and R. Sowerbutts, 2017, On a tight leash: does bank organizational structure matter for macroprudential spillover, Journal of International Economics 109, 174-194. Demirguc-Kunt, A., and E. Detragiache, 2002, Does deposit insurance increase banking system stability? An empirical investigation, Journal of Monetary Economics 49, 1373-1406. Djankov, S., C. McLiesh , and A. Shleifer, 2007, Private credit in 129 countries, Journal of Financial Economics 84, 299–329. 28 Esty, B., and W. Megginson, 2003, Creditor rights, enforcement, and debt ownership structure: evidence from the global syndicated loan market, Journal of Financial and Quantitative Analysis 38, 37-59. Frame, W., A. Mihov, and L. Sanz, 2016, Foreign investment, regulatory arbitrage and the risk of U.S. financial institutions, Federal Reserve Bank of Atlanta. Goodhart, C., 2002, The organizational structure of banking supervision, Economic Notes 31, 1-31. Houston, J., C. Lin, and Y. Ma, 2012, Regulatory arbitrage and international bank flows, Journal of Finance 67, 1845-1895. International Monetary Fund, 2015, Global financial stability report on navigating monetary policy challenges and managing risks, Washington, D.C. Karolyi, G., and A. Taboada, 2015, Regulatory arbitrage and cross-border acquisitions, Journal of Finance 70, 2395-2450. Laeven, L., and R. Levine, 93, Bank governance, regulation and risk taking, Journal of Financial Economics 93, 259-275. Lin, C., Y. Ma, P. Malatesta, and Y. Xuan, 2012, Corporate ownership structure and bank loan syndicate structure, Journal of Financial Economics 104, 1-22. Melitz, J., and F. Toubal, 2014, Native language, spoken language, translation and trade, Journal of International Economics 93, 351-363. Morrison, A., and L. White, 2009, Level playing fields in international financial regulation, Journal of Finance 64, 1099-1142. Ongena, S., A. Popov, and G. Udell, 2013, “When the cat’s away the mice will play”: does regulation at home affect bank risk-taking abroad?, Journal of Financial Economics 108, 727-750. Santos Silva, J., and S. Tenreyro, 2006, The log of gravity, Review of Economics and Statistics 88, 641-658. Schwert, M., 2018, Bank capital and lending Relationships, Journal of Finance 73, 787-830. Sufi, A., 2007, Information asymmetry and financing arrangements: evidence from syndicated loans, Journal of Finance 62, 629-668. 29 Figure 1: Total volume of cross-border syndicated loans provided by banks This graph shows the total US dollar value of syndicated loans of which the borrower and the parent bank of the lender are located in different countries in billions of US dollars. The graph excludes loans for which the exact loan allocation between lenders is not available. 30 Figure 2: Fraction of cross-border syndicated loan volume provided by foreign bank subsidiaries This graph shows the ratio of cross-border loan volume provided by foreign subsidiaries relative to the total volume of cross-border syndicated loans. 31 Figure 3: Fraction of “arbitraged” cross-border syndicated loan volume This graph shows the volume of cross-border loans provided by foreign subsidiaries located in countries with less stringent capital regulation than in the parent bank’s country divided by the total volume of cross-border syndicated loans. 32 Figure 4: Mean, 10th and 90th percentile of the capital regulation index This graph shows the mean value, and 10th and 90th percentiles of the capital regulation index for all countries in the World Bank Regulation and Supervision Survey. 33 Figure 5: Fraction of “arbitraged” cross-border syndicated loan volume by subsidiaries located in borrower countries and third countries This graph shows the volume of cross-border loans provided by foreign subsidiaries located in borrower (third) countries with less stringent capital regulation than in the parent bank’s country divided by the total volume of cross-border syndicated loans provided by borrower-country (third-country) subsidiaries. 34 Figure 6: Fraction of cross-border syndicated loan volume to borrower countries with capital regulation less stringent than in the country of the lender’s parent bank This graph shows the volume of cross-border loans provided to borrowers located in countries with less stringent capital regulation than in the parent bank’s country divided by the total volume of cross-border syndicated loans. 35 Table 1: Descriptive statistics Volume is the value of all cross-border syndicated loans in billions of US dollars. Log(Volume +1) and Arsinh(Volume) are transformations of the loan volume using the natural logarithm and the inverse hyperbolic sine functions, respectively. Foreign subsidiary/total volume is the ratio of the US dollar value of syndicated loans provided by foreign subsidiaries relative to the US dollar value of all syndicated loans. Borrower-country subsidiary/total volume is the ratio of the US dollar value of syndicated loans provided by foreign subsidiaries located in the borrower’s country relative to the US dollar value of all syndicated loans. Third-country subsidiary/total volume is the ratio of the US dollar value of syndicated loans provided by foreign subsidiaries located in a third country relative to the US dollar value of all syndicated loans. Loan volumes are aggregated at the borrower country level in Panel A and at the borrower country-lender country level in Panel B. In Panel A the sample and variables correspond to the regressions in Table 3. In Panel B the sample and variables correspond to the regressions in Table 4. Lead is a dummy variable indicating lead arranger role in a syndicated loan. In Panel C the sample and variables correspond to the regressions in Table 6. See Table A1 in the Appendix for variable definitions of the regulatory, institutional and control variables that are included in the various regression tables. Panel A Observations Mean SD Min Max Volume 1440 2,604 13,512 0 192,890 Log(Volume + 1) 1440 11.99 9.926 0 25.99 Arsinh(Volume) 1440 12.41 10.26 0 26.68 Foreign subsidiary/total volume 866 0.0952 0.173 0 1 Borrower-country subsidiary/total volume 866 0.0179 0.0798 0 1 Third-country subsidiary/total volume 866 0.0774 0.159 0 1 Capital regulation (borrower) 1440 6.333 1.938 1 10 Overall activity restrictions (borrower) 1440 7.121 2.001 3 12 Official supervisory power (borrower) 1440 11.00 2.324 4 16 Monitoring (borrower) 1440 7.717 1.627 3 11 Creditor rights (borrower) 1440 1.913 1.102 0 4 Information sharing (borrower) 1440 3.496 2.234 0 6 Time to enforce contracts (borrower) 1440 628.8 304.0 120 1580 Rule of law (borrower) 1440 0.406 0.957 -1.690 2.100 Log real GDP/capita (borrower) 1440 9.137 1.350 5.386 11.43 Log population (borrower) 1440 16.78 1.422 14.31 21.04 36 Panel B Observations Mean SD Min Max Volume 4991 352 1,811 0 32,183 Log(Volume + 1) 4991 7.902 9.266 0 24.19 Arsinh(Volume) 4991 8.198 9.605 0 24.89 Foreign subsidiary/total volume 2127 0.0955 0.239 0 1 Borrower-country subsidiary/total volume 2127 0.0208 0.103 0 1 Third-country subsidiary/total volume 2127 0.0747 0.221 0 1 Capital regulation (borrower) 4991 6.096 1.860 2 10 Capital regulation (lender) 4991 6.538 1.387 3 9 Overall activity restrictions (borrower) 4991 7.158 1.939 3 12 Overall activity restrictions (lender) 4991 6.028 1.952 3 10 Official supervisory power (borrower) 4991 11.00 2.473 4 16 Official supervisory power (lender) 4991 10.13 2.332 5.385 14.50 Monitoring (borrower) 4991 8.200 1.390 4 11 Monitoring (lender) 4991 8.732 1.421 6 11 Creditor rights (borrower) 4991 1.937 1.038 0 4 Information sharing (borrower) 4991 3.857 2.026 0 6 Time to enforce contracts (borrower) 4991 612.0 307.8 120 1510 Time to enforce contracts (lender) 4991 421.2 79.25 120 570 Concentration (borrower) 4991 76.31 17.69 23.18 100 Government bank ownership (borrower) 4991 16.75 20.71 0 95.78 Rule of law (borrower) 4991 0.615 0.962 -1.676 2.014 Rule of law (lender) 4991 1.603 0.192 1.065 1.983 Log real GDP/capita (borrower) 4991 9.478 1.204 5.683 11.43 Log real GDP/capita (lender) 4991 10.68 0.207 10.25 11.43 Log population (borrower) 4991 16.99 1.456 14.42 21.00 Log population (lender) 4991 17.78 1.132 15.23 19.56 Log distance 4991 8.417 1.037 5.162 9.851 Common spoken language 4991 0.305 0.293 0 1.000 Panel C Observations Mean SD Min Max Lead 149416 0.427 0.495 0 1 Capital regulation (lender) 149416 6.545 1.596 3 10 Overall activity restrictions (lender) 149416 5.835 1.819 3 10 Official supervisory power (lender) 149416 10.42 2.094 5.385 14.50 Monitoring (lender) 149416 8.073 1.774 5 11 Information sharing (lender) 149416 5.322 0.827 3 6 Time to enforce contracts (lender) 149416 442.8 88.99 120 830 Rule of law (lender) 149416 1.605 0.195 0.902 1.923 Log real GDP/capita (lender) 149416 10.66 0.124 10.25 10.93 Log population (lender) 149416 18.06 0.844 15.23 19.59 Log assets 149416 13.63 0.834 7.685 14.90 Log syndicated lending 149416 22.52 2.193 0 24.75 Loans/deposits 149416 0.902 0.741 0.0613 12.10 Equity/assets 149416 0.0497 0.0212 0.0179 0.149 Past relationship 149416 0.436 0.496 0 1 37 Table 2: Means of regulatory variables for loans originated by foreign subsidiaries and parent banks Panels A to D show the means of Capital regulation, Activity restrictions, Official supervisory power and Monitoring, respectively, for loans provided by foreign subsidiaries in borrower countries, foreign subsidiaries in third countries, and parent banks. See Table A1 in the Appendix for variable definitions. In all panels column 1 shows the sample period for which means are calculated. Columns 2, 4, and 6 (3, 5, 7) show the sample means of the pertinent regulatory variables for (number of) loans provided by foreign subsidiaries in borrower countries, foreign subsidiaries in third countries, and parent banks. Columns 8, 9 and 10 show differences between means shown in columns 2 and 6, 4 and 6; and 4 and 2, respectively. *, **, and *** denote significance at 10%, 5%, and 1%. Panel A (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Difference of means Period Mean of Number of Mean of Number Mean of Number of Borrower- Third- Third-country Capital borrower- Capital of third- Capital lender- country and country and and borrower- regulation in country regulation in country regulation in country lender-country lender- country loans borrower loans third loans lender loans loans country (4) - (2) countries countries countries (2) - (6) loans (4) - (6) All years 7.803 13650 6.74 4952 6.526 130495 1.277*** 0.214*** -1.063*** 1996-2005 5.9476 4733 6.04193 2181 6.036 59148 -0.0884*** 0.00593 0.0943*** 2006-2016 8.789 8917 7.291 2771 6.933 71347 1.856*** 0.358*** -1.499*** Panel B (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Difference of means Period Mean of Activity Number of Mean of Activity Number of Mean of Activity Number of Borrower-country Third-country and Third-country and restrictions in borrower- restrictions in third third-country restrictions in lender lender-country and lender-country lender-country borrower-country borrower countries country loans countries loans countries loans loans loans loans (2) - (6) (4) - (6) (4) - (2) All years 7.582 13757 5.521 5163 5.888 130495 1.694*** -0.367*** -2.061*** 1996-2005 7.627 4840 5.517 2392 5.668 59148 1.959*** -0.151*** -2.110*** 2006-2016 7.559 8917 5.525 2771 6.071 71347 1.488*** -0.546*** -2.034*** 38 Panel C (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Difference of means Period Mean of Number of Mean of Number of Mean of Number of Borrower- Third-country Third-country Official borrower- Official third- Official lender- country and and lender- and borrower- supervisory country supervisory country supervisory country lender-country country loans country loans power in loans power in loans power in loans loans (4) - (6) (4) - (2) borrower third lender (2) - (6) countries countries countries All years 12.389 13748 11.13 5106 10.45 130495 1.939*** 0.680*** -1.259*** 1996-2005 12.894 4840 11.145 2392 9.949 59148 2.945*** 1.196*** -1.748*** 2006-2016 12.118 8908 11.121 2714 10.87 71347 1.248*** 0.251*** -0.997*** Panel D (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Difference of means Period Mean of Number of Mean of Number of Mean of Number of Borrower- Third-country Third-country Monitoring borrower- Monitoring third- Monitoring lender- country and and lender- and borrower- in borrower country in third country in lender country lender-country country loans country loans countries loans countries loans countries loans loans (4) - (6) (4) - (2) (2) - (6) All years 8.278 13650 8.805 4952 8.063 130495 0.215*** 0.742*** 0.527*** 1996-2005 9.753 4733 9.664 2181 8.61 59148 1.143*** 1.054*** -0.0895*** 2006-2016 7.495 8917 8.129 2771 7.609 71347 -0.114*** 0.520*** 0.634*** 39 Table 3: Syndicated loan volume at the borrower country level In regression 1 the dependent variable is the natural logarithm of the dollar value of loans plus 1. In regressions 2 and 3 the dependent variable is the dollar value of loans transformed using the inverse hyperbolic sine function. Regressions 1 and 2 are OLS regressions, and regression 3 is a Tobit regression. In regressions 4 and 7 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries to the dollar value of all loans. In regressions 5 and 8 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries located in the borrower’s country to the dollar value of all loans. In regressions 6 and 9 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries located in a country different from the borrower’s and the parent bank’s country to the dollar value of all loans. Regressions 4-6 are OLS regressions. In regressions 7 to 9 regulatory variables are instrumented by their sample means excluding the pertinent country, a dummy variable indicating that the central bank supervises banks for prudential purposes, and the 5-year moving average of the Gini coefficient measuring income inequality in the borrower country. Detailed variable definitions can be found in Table A1 in the Appendix. Variables are for borrowing countries. The sample period is 1995-2016. Borrower country and time fixed effects are included. In all regressions except for regression 3 standard errors are clustered at the borrower country level. *, **, and *** denote significance at 10%, 5%, and 1%. OLS Tobit OLS IV (1) (2) (3) (4) (5) (6) (7) (8) (9) Borrower- Third- Borrower- Third- Foreign Foreign country country country country Log(Volume + 1) Arsinh(Volume) Arsinh(Volume) subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ total volume total volume total volume total volume total volume total volume Capital regulation (borrower) -0.0365 -0.0383 -0.0968 -0.00302 0.00137 -0.00439 -0.00686* -0.000260 -0.00660* (-0.22) (-0.22) (-0.52) (-0.88) (0.74) (-1.30) (-1.82) (-0.20) (-1.87) Overall activity restrictions (borrower) -0.112 -0.114 0.00505 0.00658 0.00623 0.000353 0.00281 0.000621 0.00219 (-0.72) (-0.71) (0.02) (1.01) (1.25) (0.06) (0.53) (0.44) (0.42) Official supervisory power (borrower) -0.0210 -0.0224 -0.0353 -0.00674* -0.00200 -0.00474 -0.00592 -0.00232 -0.00360 (-0.17) (-0.17) (-0.21) (-1.69) (-0.86) (-1.43) (-1.52) (-1.12) (-1.12) Monitoring (borrower) 0.330* 0.340* 0.468* 0.0108* 0.00936* 0.00141 0.0104 0.00878 0.00159 (1.74) (1.74) (1.75) (1.70) (1.81) (0.27) (1.47) (1.62) (0.27) Creditor rights (borrower) 0.679 0.711 -0.182 0.305** 0.177 0.128 0.272*** 0.00717 0.265*** (0.54) (0.55) (-0.25) (2.48) (1.22) (1.66) (3.55) (0.44) (3.45) 40 Information sharing (borrower) -0.657** -0.678** -0.698** 0.000646 0.00240 -0.00176 0.000368 0.00231 -0.00194 (-2.58) (-2.57) (-2.53) (0.06) (0.88) (-0.18) (0.04) (0.80) (-0.20) Time to enforce contracts (borrower) -0.00122 -0.00125 -0.00316 -0.000103* -0.0000306 -0.0000724 -0.000116* -0.0000226 -0.0000935 (-0.31) (-0.31) (-1.22) (-1.68) (-1.06) (-1.19) (-1.95) (-0.96) (-1.50) Rule of law (borrower) 3.477** 3.607** 4.233*** 0.0394 -0.0538 0.0932** 0.0510 -0.0316 0.0826* (2.25) (2.25) (3.59) (0.78) (-1.59) (2.08) (1.14) (-1.16) (2.00) Log real GDP/capita (borrower) 8.071*** 8.351*** 5.463*** -0.159** 0.0186 -0.177*** -0.201*** 0.00334 -0.204*** (3.09) (3.08) (5.84) (-2.21) (0.66) (-2.87) (-2.89) (0.16) (-3.23) Log population (borrower) 9.320** 9.692** 6.238*** -0.427*** -0.0610 -0.366** -0.482*** -0.112 -0.370** (2.36) (2.36) (10.11) (-2.69) (-0.62) (-2.55) (-3.24) (-1.60) (-2.63) Observations 1440 1440 1440 866 866 866 790 790 790 Adjusted R-squared 0.629 0.626 0.257 0.404 0.206 0.170 0.374 0.148 Overid. test (p value) - - - - - - 0.953 0.221 0.568 Underid. test (p value) - - - - - - 0.000 0.000 0.000 41 Table 4: Bilateral syndicated loan volumes In regression 1 the dependent variable is the natural logarithm of the dollar value of loans plus 1. In regressions 2 and 3 the dependent variable is the dollar value of loans transformed using the inverse hyperbolic sine function. Regressions 1 to 2 are OLS regressions, and regression 3 is a Tobit regression. In regressions 4 and 7 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries to the dollar value of all loans. In regressions 5 and 8 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries located in the borrower’s country to the dollar value of all loans. In regressions 6 and 9 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries located in a country different from the borrower’s and the parent bank’s country to the dollar value of all loans. Regressions 4-6 are OLS regressions. In regressions 7 to 9 regulatory variables are instrumented by their sample means excluding the pertinent country, a dummy variable indicating that the central bank supervises banks for prudential purposes, and the 5-year moving average of the Gini coefficient measuring income inequality in the borrower country. Log real GDP/capita (borrower, lender) and Log population (borrower, lender) are included but not reported. Detailed variable definitions can be found in Table A1 in the Appendix. The sample period is 1995-2016. Borrower country, lender country and time fixed effects are included. In all regressions except for regression 3 standard errors are clustered at the borrower country level. *, **, and *** denote significance at 10%, 5%, and 1%. OLS Tobit OLS IV (1) (2) (3) (4) (5) (6) (7) (8) (9) Borrower- Third- Borrower- Third- Foreign Foreign country country country country Log(Volume + 1) Arsinh(Volume) Arsinh(Volume) subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ total volume total volume total volume total volume total volume total volume Capital regulation (borrower) 0.114 0.118 0.0844 -0.000974 0.00296 -0.00394 -0.00148 0.00222 -0.00369 (0.83) (0.83) (0.48) (-0.23) (1.48) (-0.90) (-0.28) (1.34) (-0.74) Capital regulation (lender) -1.008*** -1.043*** -1.398*** 0.0281*** 0.00369 0.0244*** 0.0294*** 0.00345 0.0260*** (-5.88) (-5.88) (-4.51) (3.00) (1.03) (2.72) (2.93) (0.92) (2.70) Overall activity restrictions (borrower) -0.340** -0.351** -0.269 0.00499 0.00624 -0.00125 0.00391 0.00210 0.00181 (-2.07) (-2.06) (-1.29) (0.73) (1.12) (-0.22) (0.51) (0.51) (0.27) Overall activity restrictions (lender) 1.382*** 1.433*** 1.410*** 0.00835 -0.00179 0.0101 0.0103 -0.00186 0.0122 (6.53) (6.55) (5.14) (0.68) (-0.25) (1.06) (0.63) (-0.22) (0.86) Official supervisory power (borrower) -0.0644 -0.0678 0.0739 -0.00190 -0.00302 0.00112 0.0000700 -0.00273 0.00280 (-0.64) (-0.65) (0.50) (-0.55) (-1.43) (0.31) (0.02) (-1.42) (0.65) Official supervisory power (lender) -0.136 -0.141 -0.166 0.0115* -0.00498* 0.0165*** 0.00971 -0.00467 0.0144*** (-1.39) (-1.39) (-0.85) (1.88) (-1.79) (3.42) (1.50) (-1.46) (2.81) 42 Monitoring (borrower) 0.509** 0.526** 1.187*** 0.00912 0.00875 0.000372 0.00825 0.00864 -0.000384 (2.08) (2.08) (4.73) (1.16) (1.33) (0.08) (0.93) (1.17) (-0.08) Monitoring (lender) 0.373** 0.387** 0.959*** -0.0218** 0.00215 -0.0240*** -0.0198 0.00463 -0.0244*** (2.06) (2.06) (3.20) (-2.06) (0.28) (-3.50) (-1.64) (0.54) (-3.11) Creditor rights (borrower) 1.113 1.141 0.990** 0.185** 0.102 0.0824** 0.153*** -0.00545 0.159*** (0.54) (0.53) (2.10) (2.54) (1.23) (2.11) (3.59) (-0.31) (3.68) Information sharing (borrower) -0.470 -0.489 0.363 0.00985 0.00806 0.00179 0.0135 0.00569 0.00776 (-1.21) (-1.22) (1.32) (0.65) (1.31) (0.12) (0.76) (1.32) (0.45) - Time to enforce contracts (borrower) 0.000317 0.000298 -0.00419** -0.000140 -0.000404** -0.000563** -0.0000600 -0.000503** 0.000544*** (0.06) (0.05) (-2.33) (-3.04) (-1.35) (-2.15) (-2.65) (-0.58) (-2.38) Time to enforce contracts (lender) 0.00575 0.00561 0.00996 0.00766*** 0.00244 0.00521*** 0.0118** 0.000846 0.0110*** (0.19) (0.18) (1.54) (4.17) (1.26) (4.24) (2.36) (0.19) (3.36) Concentration (borrower) 0.0321 0.0336 0.0711*** -0.000455 0.000367 -0.000822 -0.000640 0.000180 -0.000820 (1.13) (1.14) (2.94) (-0.58) (0.67) (-1.20) (-0.79) (0.37) (-1.16) Government bank ownership (borrower) -0.0114 -0.0120 -0.0869*** 0.000541 0.0000307 0.000510 0.000103 0.000349 -0.000245 (-0.41) (-0.42) (-3.31) (0.32) (0.06) (0.31) (0.06) (0.91) (-0.15) Rule of law (borrower) 2.473* 2.573* 3.809*** 0.0226 -0.0345 0.0571 0.0260 -0.0111 0.0370 (1.67) (1.68) (4.16) (0.50) (-1.45) (1.29) (0.50) (-0.45) (0.76) Rule of law (lender) -18.61*** -19.32*** -1.243 -0.0587 -0.00835 -0.0503 -0.0640 -0.00238 -0.0616 (-9.13) (-9.13) (-0.38) (-0.51) (-0.36) (-0.46) (-0.53) (-0.09) (-0.52) Log distance -1.262*** -1.303*** -1.709*** -0.00668 -0.00143 -0.00524 -0.00360 -0.000824 -0.00278 (-5.06) (-5.05) (-3.11) (-0.80) (-0.33) (-0.67) (-0.40) (-0.18) (-0.34) Common spoken language 3.365*** 3.486*** 4.383** -0.0261 0.0329 -0.0591 -0.0259 0.0330 -0.0589 (3.16) (3.17) (1.96) (-0.49) (1.02) (-1.65) (-0.42) (0.88) (-1.41) 43 Observations 4991 4991 4994 2122 2122 2122 1907 1907 1907 Adjusted R-squared 0.399 0.398 0.111 0.234 0.111 0.098 0.226 0.105 Overid. test (p value) - - - - - - 0.835 0.241 0.568 Underid. test (p value) - - - - - - 0.000 0.000 0.000 44 Table 5: Bilateral syndicated loan volumes in 1995-2005 and 2006-2016 In regressions 1, 4, 7 and 10 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries to the dollar value of all loans. In regressions 2, 5, 8 and 11 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries located in the borrower’s country to the dollar value of all loans. In regressions 3, 6, 9 and 12 the dependent variable is the ratio of the dollar value of loans provided by subsidiaries located in a country different from the borrower’s and the parent bank’s country to the dollar value of all loans. Regressions 1 to 3 and 7 to 9 are OLS regressions. In regressions 4 to 6 and 10 to 12 regulatory variables are instrumented by their sample means excluding the pertinent country, a dummy variable indicating that the central bank supervises banks for prudential purposes, and the 5-year moving average of the Gini coefficient measuring income inequality in the borrower country. Other variables are included as in Table 4 but not reported. Detailed variable definitions can be found in Table A1 in the Appendix. In regressions 1 to 6 the sample period is 1995-2005. In regressions 7 to 12 the sample period is 2006-2016. Borrower country, lender country and time fixed effects are included. In all regressions standard errors are clustered at the borrower country level. *, **, and *** denote significance at 10%, 5%, and 1%. Sample period: 1995 - 2005 Sample period: 2006 - 2016 OLS IV OLS IV (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Borrower- Third- Borrower- Third- Borrower- Borrower- Foreign Foreign Third- Third- country country country country Foreign country Foreign country subsidiary/ subsidiary/ country country subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ subsidiary/ total total subsidiary/ subsidiary/ total total total total total volume total total volume total volume volume total volume total volume volume volume volume volume volume volume Capital regulation 0.00308 0.00439 -0.00131 -0.000440 0.00585 -0.00629 0.00331 -0.00105 0.00435 0.0145 -0.000812 0.0153 (borrower) (0.50) (1.06) (-0.24) (-0.06) (1.60) (-0.92) (0.42) (-0.38) (0.53) (1.37) (-0.27) (1.37) Capital regulation 0.0493*** 0.00952 0.0397*** 0.0329** 0.00768 0.0252** -0.0568 0.00120 -0.0580 -0.0552 -0.00200 -0.0532 (lender) (4.50) (1.67) (3.95) (2.65) (1.53) (2.16) (-1.59) (0.22) (-1.65) (-1.52) (-0.30) (-1.49) Overall activity restrictions 0.000492 0.0140 -0.0135* -0.0125 -0.00174 -0.0108 -0.00174 0.00560 -0.00734 -0.00211 0.00793 -0.0100 (borrower) (0.04) (1.22) (-1.80) (-1.17) (-0.24) (-1.03) (-0.19) (1.33) (-0.79) (-0.19) (1.26) (-0.87) Overall activity restrictions 0.0122 -0.00453 0.0168 0.0368* -0.00360 0.0404** -0.114*** -0.00417 -0.110*** -0.106** 0.00648 -0.112** (lender) (0.80) (-0.49) (1.46) (1.76) (-0.40) (2.19) (-2.99) (-0.63) (-3.00) (-2.29) (0.71) (-2.60) Official supervisory power 0.00200 0.000328 0.00167 0.00646 -0.000115 0.00657 -0.00961 -0.00671 -0.00290 -0.0153* -0.00807 -0.00727 (borrower) (0.34) (0.12) (0.29) (0.77) (-0.04) (0.84) (-1.19) (-1.51) (-0.38) (-1.70) (-1.58) (-0.81) Official supervisory power 0.0164** -0.00137 0.0177*** 0.00639 -0.00333 0.00971 0.0307** -0.00125 0.0319** 0.0339** 0.000419 0.0335** (lender) (2.35) (-0.46) (3.06) (0.67) (-0.78) (1.23) (2.55) (-0.60) (2.65) (2.67) (0.18) (2.64) Monitoring 0.00594 0.00588 0.0000608 -0.00118 0.00143 -0.00261 0.00160 0.00107 0.000526 0.0115 0.00200 0.00946 45 (borrower) (0.71) (0.66) (0.01) (-0.13) (0.16) (-0.45) (0.14) (0.47) (0.04) (0.88) (0.62) (0.69) Monitoring -0.0239* -0.00141 -0.0225** -0.000459 0.00493 -0.00539 -0.0217 0.00537* -0.0271 -0.0299 0.00210 -0.0320* (lender) (-1.92) (-0.16) (-2.63) (-0.03) (0.45) (-0.42) (-1.31) (1.75) (-1.67) (-1.62) (0.83) (-1.79) Observations 1311 1311 1311 1163 1163 1163 809 809 809 739 739 739 Adjusted R-squared 0.150 0.301 0.136 0.145 0.295 0.135 0.096 0.166 0.104 0.090 0.178 0.101 Overid. test (p value) - - - 0.218 0.150 0.374 - - - 0.332 0.363 0.166 Underid. test (p value) - - - 0.000 0.000 0.000 - - - 0.001 0.001 0.001 46 Table 6: Determinants of the lead bank role The dependent variable is a dummy variable indicating a lead bank role. Regressions 1 to 4 are OLS regressions. In regressions 5 to 8 regulatory variables are instrumented by their sample means excluding the pertinent country, a dummy variable indicating that the central bank supervises banks for prudential purposes, and the 5-year moving average of the Gini coefficient measuring income inequality in the lender country. See Table A1 in the Appendix for the definitions of the other variables. The sample period is 1995-2016. Regressions 1 and 5 include facility and lender country fixed effects. Regressions 2 and 6 include facility and bank fixed effects. Regressions 3 and 7 include facility and bank * borrower country fixed effects. Regressions 4 and 8 include facility and bank borrower company fixed effects. Standard errors are clustered at the banking group level. *, **, and *** denote significance at 10%, 5%, and 1%. OLS IV (1) (2) (3) (4) (5) (6) (7) (8) Capital regulation (lender) -0.00202 -0.00502* -0.00775*** -0.00444 -0.00154 -0.00495* -0.00772*** -0.00562* (-0.55) (-1.70) (-2.87) (-1.43) (-0.41) (-1.70) (-2.83) (-1.79) Overall activity restrictions (lender) 0.00606* 0.00429* 0.00476** -0.000269 0.00780* 0.00474 0.00525* -0.000104 (1.76) (1.88) (2.31) (-0.08) (1.87) (1.65) (1.93) (-0.03) Official supervisory power (lender) 0.000738 -0.00209 -0.00372* 0.00115 0.00127 -0.00206 -0.00420* 0.000181 (0.25) (-0.91) (-1.69) (0.43) (0.44) (-0.84) (-1.75) (0.06) Monitoring (lender) -0.00402 0.00178 0.00339 -0.000707 -0.00466 0.000776 0.00306 -0.00104 (-0.99) (0.47) (0.88) (-0.15) (-1.07) (0.19) (0.68) (-0.23) Information sharing (lender) -0.00982 -0.0236 -0.00649 0.0180 -0.00409 -0.0228 -0.0103 0.0225 (-0.48) (-1.33) (-0.37) (0.78) (-0.19) (-1.29) (-0.55) (1.09) Time to enforce contracts (lender) -0.0000755 0.0000719 0.0000722 0.0000919 -0.0000817 0.0000745 0.0000730 0.0000926 (-0.82) (0.82) (0.85) (0.84) (-0.91) (0.84) (0.83) (0.86) Rule of law (lender) 0.140*** 0.107*** 0.0959** 0.117*** 0.123** 0.103** 0.0969** 0.121*** (2.92) (2.83) (2.56) (2.74) (2.38) (2.52) (2.52) (2.85) Log real GDP/capita (lender) -0.188 0.0969 0.0626 -0.0750 -0.252 0.0836 0.0663 -0.102 (-1.22) (0.66) (0.46) (-0.37) (-1.17) (0.48) (0.39) (-0.45) 47 Log population (lender) -0.600*** -0.306** -0.194 0.361 -0.554*** -0.288** -0.218 0.375* (-3.23) (-2.45) (-1.42) (1.63) (-2.98) (-2.23) (-1.62) (1.69) Log assets 0.113*** 0.0321** 0.0371*** 0.0316*** 0.113*** 0.0325** 0.0374*** 0.0326*** (8.66) (2.07) (2.73) (2.76) (8.46) (2.04) (2.72) (2.94) Log syndicated lending 0.00363** 0.00310 0.00302 0.00183 0.00354** 0.00299 0.00288 0.00185 (2.09) (1.61) (1.57) (1.34) (2.00) (1.53) (1.47) (1.33) Loans/deposits -0.00357 -0.00422** -0.00450*** -0.0102*** -0.00357 -0.00408** -0.00431*** -0.00985*** (-1.03) (-2.43) (-2.89) (-8.87) (-1.06) (-2.30) (-2.71) (-8.23) Equity/assets -1.342*** -0.129 0.0951 -0.588 -1.353*** -0.103 0.121 -0.517 (-3.02) (-0.24) (0.23) (-1.67) (-3.02) (-0.19) (0.28) (-1.40) Past relationship 0.104*** 0.102*** 0.0914*** 0.0182*** 0.105*** 0.104*** 0.0926*** 0.0181*** (15.76) (16.16) (14.89) (4.63) (16.84) (17.41) (16.00) (4.56) Observations 149416 149412 148988 128281 146043 146039 145607 125285 Adjusted R-squared 0.528 0.533 0.564 0.794 -0.365 -0.403 -0.406 -0.420 Facility FE Yes Yes Yes Yes Yes Yes Yes Yes Lender country FE Yes - - - Yes - - - Bank FE No Yes - - No Yes - - Bank * borrower country FE No No Yes - No No Yes - Bank * borrower company FE No No No Yes No No No Yes Overid. test (p value) - - - - 0.022 0.161 0.460 0.908 Underid. test (p value) - - - - 0.005 0.005 0.006 0.007 48 Appendix Table A1: Variable definitions and data sources Variable Definition Source Volume The value of all syndicated loans in billions of US dollars Dealscan aggregated either at the level of the borrower country, or the borrower country-lender country pair. Log(Volume + 1) Natural logarithm of 1 plus the US dollar value of all syndicated Dealscan loans aggregated either at the level of the borrower country, or the borrower country-lender country pair. Arsinh(Volume) Transformed value of the US dollar value of all syndicated loans Dealscan aggregated either at the level of the borrower country, or the borrower country-lender country pair, using the inverse hyperbolic sine function for the transformation defined as: arsinh ln 1 Foreign The ratio of the US dollar value of syndicated loans provided by Dealscan subsidiary/total foreign subsidiaries relative to the US dollar value of all volume syndicated loans aggregated either at the level of the borrower country, the lender country, or the borrower country-lender country pair. Borrower-country The ratio of the US dollar value of syndicated loans provided by Dealscan subsidiary/ total foreign subsidiaries located in the borrower’s country relative to volume the US dollar value of all syndicated loans aggregated either at the level of the borrower country, the lender country, or the borrower country-lender country pair. Third-country The ratio of the US dollar value of syndicated loans provided by Dealscan subsidiary/ total foreign subsidiaries located neither in the borrower’s nor the volume parent bank’s country relative to the US dollar value of all syndicated loans aggregated either at the level of the borrower country, the lender country, or the borrower country-lender country pair. Lead Dummy variable indicating lead arranger role in a syndicated Dealscan loan. Following Bharath et al. (2011) and Berg et al. (2016) we set it equal to one if 1) the reported lender role in Dealscan is either “Admin agent”, “Agent”, “Arranger”, or “Lead bank”; or 2) the lead arranger credit field equals “Yes”; or 3) if the loan has a sole lender. Capital regulation Index measuring the stringency in determining minimum capital World Bank adequacy and initial capital stringency in borrower or lender Regulation and country, with higher values indicating greater stringency. Supervision Survey (Barth et al., 2006) Overall activity Index of the extent to which banks can engage in securities, World Bank restrictions insurance and real estate activities in borrower or lender country, Regulation and with higher values indicating more restrictions. Supervision Survey (Barth et al., 2006) Official Index of the power of the supervisory authorities to take specific World Bank supervisory power actions to prevent and correct problems in banks in borrower or Regulation and lender country, with higher values indicating greater power. Supervision Survey (Barth et al., 2006) 49 Monitoring An index measuring the strength of private monitoring incentives World Bank in borrower or lender country, with higher values indicating more Regulation and private supervision. The index measures whether certified audit is Supervision Survey required; the top ten banks are all rated by international credit (Barth et al., 2006) rating agencies; no explicit deposit insurance scheme exists in the country; the income statement includes accrued or unpaid interest or principal on nonperforming loans and whether banks are required to produce consolidated financial statements; off-balance sheet items are disclosed to the public, banks must disclose risk management procedures to the public, and subordinated debt is allowable (required) as a part of regulatory capital. Creditor rights Index measuring the strength of creditor rights in borrower or Djankov et al. (2007) lender country, with higher values indicating stronger creditor rights. Information Index measuring rules affecting the scope, accessibility, and Doing Business sharing quality of credit information available through public or private Database credit registries in borrower or lender country. The index ranges from 0 to 8, with higher values indicating the availability of more credit information, from either a public registry or a private bureau, to facilitate lending decisions. Time to enforce Index measuring the time required to resolve a commercial Doing Business contracts dispute, calculated as the average number of calendar days from Database the filing of the lawsuit in court until the final determination and, in appropriate cases, payment, in borrower or lender country. Higher values indicate easier contract enforcement. Rule of law Index capturing perceptions of the extent to which agents have World Governance confidence in and abide by the rules of society, and in particular Indicators the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence in borrower or lender country, with higher values indicating a stronger rule of law. Concentration Assets of the five largest banks as a share of total commercial Global Financial banking assets in borrower or lender country. Development Report Government bank The proportion of banking assets in government owned banks World Bank ownership (where a bank is considered government owned if 50 percent or Regulation and more of the shares are controlled by the government) in borrower Supervision Survey or lender country. (Barth et al., 2006) Log real Natural logarithm of GDP per capita measured at constant 2010 World Development GDP/capita US dollar prices in borrower or lender country. Indicators Log population Natural logarithm of the total population in borrower or lender World Development country. Indicators Log distance Natural logarithm of 1 plus the geographic distance between the http://techslides.com/ capital cities of the borrower and lender countries measured in list-of-countries-and- kilometres. capitals (downloaded on June 27, 2016) Common spoken The probability that a pair of people at random from the borrower Melitz and Toubal language and lender countries understand one another in some language. (2014) Log assets Natural logarithm of a bank’s total assets ratio lagged by one Compustat year. Log syndicated Natural logarithm of 1 plus the sum of the US dollar value of all Dealscan lending loans provided by the lender in a given year, minus the pertinent loan amount. Missing loan shares are replaced by zeros. 50 Loans/deposits A bank’s total loans (net of total allowance for loan losses) to Compustat total deposits ratio lagged by one year. Equity/assets A bank’s total common equity to total assets ratio lagged by one Compustat year. Past relationship Dummy variable indicating that the lender bank has provided a Dealscan loan to the borrower before the pertinent loan. 51 Table A2: Total syndicated loan amounts by largest lender and borrower countries This table provides information on the largest lender countries, their most significant borrower countries, and the countries where their most active foreign subsidiaries are located. Columns 1 and 2 show the main lender countries and the total amounts of cross-border syndicated loans provided by banking groups headquartered in these countries during the 1995-2016 period. Columns 3 and 4 show the total amount of syndicated loans provided to a given borrower country during the 1995-2016 period. Columns 5 and 6 show the total amount of syndicated loans provided by subsidiaries located in a given country during the 1995-2016 period. All loan amounts are in constant 2016 US dollars reflecting the US GDP deflator. Total lending Total lending to Lender Total lending in Country of foreign through country of Borrower country borrower country country billions of USD subsidiary foreign subsidiary in billions of USD in billions of USD (1) (2) (3) (4) (5) (6) United States 674 United States 118 United Hong Kong SAR, 1,290 France 76 9 Kingdom China Spain 66 France 6 United States 408 Netherlands 53 France 955 Spain 78 United States 42 United Kingdom 65 Switzerland 13 United States 455 United States 100 Japan 891 Australia 63 China 5 United Kingdom 41 Singapore 3 United States 363 United States 39 Germany 665 United Kingdom 81 Luxembourg 6 France 40 Austria 4 United Kingdom 78 United Kingdom 16 Hong Kong SAR, United States 480 France 39 12 China Netherlands 35 Australia 6 52