Draft. The Role of Foreign Banks in Trade. Stijn Claessens, Omar Hassib, and Neeltje van Horen * December Abstract

Transcription

1 Draft The Role of Foreign Banks in Trade by Stijn Claessens, Omar Hassib, and Neeltje van Horen * December 2014 Abstract Financially developed countries tend to export relatively more in financially vulnerable sectors, suggesting access to finance to be important for promoting trade. Combining detailed data on bilateral, sectoral trade with bilateral data on foreign bank presence for 95 exporting countries, we show that exports are even higher with greater foreign bank presence. Results are robust to including various control variables and controlling for possible reverse causality. Studying the role of foreign banks by exporters level of financial development and contracting and information environments suggests that foreign banks, especially those from the importing country, besides providing additional external financing, help overcome information problems. JEL Classification Numbers: F10, F14, F21, F23, G21 Keywords: Foreign banks, international trade, credit constraints, heterogeneous firms * Stijn Claessens is at the IMF, University of Amsterdam, and CEPR; Omar Hassib is at Maastricht University; and Neeltje van Horen is at De Nederlandsche Bank. We are grateful to JaeBin Ahn, Cagatay Bircan, Frederic Lambert, Luc Laeven, and seminar participants at the IMF/WB/WTO Trade Workshop, De Nederlandsche Bank and Maastricht University for useful comments, and to Yangfan Sun for help with the trade and UNIDO data. The views expressed in this paper are those of the authors and not necessarily of the institutions they are affiliated with. Author s addresses: sclaessens@imf.org; o.hassib@maastrichtuniversity.nl; n.van.horen@dnb.nl.

2 2 I. INTRODUCTION Banks can facilitate economic activity by providing financing, guarantees and other services to entrepreneurs (Levine, 2005). While access to external funds is important for domestic production, it is especially important for exporting firms (Foley and Manova, 2014). 1 Indeed, theoretical papers show that better developed financial systems mean a comparative advantage in trade for industries that rely more on external finance. And substantial empirical evidence, both at the country- and firm-level, indicates that greater financial development facilitates trade. 2 These studies, however, do not consider banking system structures, in which countries can differ greatly, especially regarding bank ownership. While in some countries most financial intermediation is done by domestic banks, in others it is primarily done by foreign owned banks. Since, as Figure 1 shows, there exists little relationship between domestic financial development and the share of foreign banks, i.e., a country can be financially highly developed or underdeveloped with few or many foreign banks present, 3 the question arises whether the presence of foreign banks affects trade over and above the impact of general financial development. By combining detailed data on bilateral, sectoral trade with bilateral data on foreign bank presence for 95 exporting countries, this paper sheds light on this issue. It adds to the literature on how financial development helps with trade and to the literature on how foreign banks can have a positive impact on domestic financial sector and real activity. Using the database s large exporting country heterogeneity with respect to financial development, foreign bank presence and information and contracting environments, it explores the following three arguments why foreign banks might play a special role in enhancing trade. 1 First, there exist substantial upfront sunk and fixed costs that are specific to exports, like learning about profitable export opportunities, setting up and maintaining foreign distribution networks, and costs related to shipping and duties. Furthermore, exporters need for working capital are higher since international transactions take on average between 30 and 90 days longer than domestic transactions. Finally, the added risk of selling products overseas makes insurance and other financial guarantees more necessary. While some of these factors also apply to (the financing of) imports, the literature has found overall effects to be stronger for exports as they require greater upfront investments. 2 Theory papers include Kletzer and Bardhan (1987), Beck (2002), Matsuyama (2005) and Wynne (2005). Country level empirical studies include Beck (2002, 2003), Svaleryd and Vlachos (2005), Hur, Raj and Riyanto (2006), Manova (2013), Becker, Chen and Greenberg (2013). Firm-level studies include Greenaway, Guariglia and Kneller (2007), Muuls (2008), Manova, Wei and Zhang (forthcoming), Berman and Hericourt (2010), Minetti and Zhu (2011). Some papers (Braun and Raddatz 2008, Do and Levchenko 2007), however, point out the possibility of reverse causality: higher export demand could translate into higher observed levels of private credit. 3 Claessens and van Horen (2014) show that foreign bank presence can be explained by various factors, only one of which is a country s general (and financial) development.

3 3 First, foreign banks have in general been found to lower the overall costs and increase the quality of financial intermediation, increase access to financial services, enhance the financial and economic performance of their borrowers and increase real economic growth (Claessens, Demirguc-Kunt and Huizinga, 2001; Clarke, Cull, Martinez Peria and Sanchez, 2003, Martinez- Peria and Mody, 2004; Bruno and Hauswald, 2013). These effects are thought to result from increases in banking competition, the introduction and spill-over of new, more sophisticated technologies and enhanced regulatory reforms. These effects likely extend to trade. Given the size, global focus and reach of many of their parents, foreign banks can for example be expected to provide better than most domestic banks trade-related financial products, like letters of credit. 4 The magnitude of the effects of foreign banks though are found to depend on the economic and institutional development of the host country and the foreign banks market share (Demirguc- Kunt, Laeven and Levine, 2004; Beck and Martinez Peria, 2008; Detragiache, Gupta and Tressel, 2008; Claessens and Van Horen, 2014). These latter findings suggest that foreign banks might increase access to finance of exporting firms especially in countries which financial systems are still underdeveloped and where assessment of credit risk is more difficult, and when foreign banks are more invested in the local economy. Second, foreign banks might be better able than domestic banks to reduce export risk by mitigating enforcement difficulties. Shipping goods internationally is risky and takes time while the enforcement of international contracts and payments tends to be more difficult. Considering these risks, specific forms of trade financing exporter finance (open account), importer finance (cash in advance) and bank finance (letter of credit) are utilized that allocate risk and financing cost differentially between trading partners. The choice of contract is found to importantly depend on financial market characteristics and contract environments in the exporting and importing country (Antras and Foley, forthcoming; Schmidt-Eisenlohr, 2013). Foreign banks, especially those active in both countries, may provide an additional means to overcome such contracting problems and provide a substitute commitment technology if needed (e.g., by providing (additional) insurance through guarantees). Indeed Olsen (2013) shows that bank linkages can help enforce or guarantee payments when reputation mechanisms are not strong enough to overcome weak international contract enforcement. And Niepmann and Schmidt- Eisenlohr (2014) show that when in response to a shock US banks reduce their supply of letters of credit, US exports decline more to smaller and less developed countries where fewer US banks are active. Third, foreign banks might be better able to mitigate export risk that arises from information asymmetries. When assessing the loan application of an exporting firm, a bank needs to assess also the risk of the borrower s trading partners. A foreign bank, especially one from the 4 See further Niepmann and Schmidt-Eisenlohr (2013) and Del Prete and Federico (2014) for a description of the market structure for trade finance in the US and Italy.

4 4 importing country, might, through its international networks, be better equipped to assess and collect information on the economic conditions on the importing side of the transaction and use this information to better evaluate the risks involved. 5 Some recent studies highlight banks role in information collection in trade facilitation. Studying trade patterns across state-pairs in the US, Michalski and Ors (2012) find that, because of better information transmission, trade tends to be higher between two states that allowed entry by each other s financial institutions. Hale, Candelaria, Caballero and Borisov (2013) find that new connections between banks in a given country-pair, established through participation in syndicated lending, lead to an increase in trade flows in the following year. And Bronzini and D Ignazio (2013) find that Italian firms are more likely to export to a county if its bank has an affiliate in that country. If the abovementioned factors, and possibly others, contribute to reducing the cost and enhancing the availability of external finance for exporting firms, then the presence of foreign banks should have a positive impact on export over and above general financial sector development. We test this empirically building on the work of Manova (2013) and others. Our identification strategy relies on exploiting variation in foreign bank presence across 95 exporting countries and variation in financial vulnerability across 28 manufacturing sectors while controlling for financial development using data over the period The use of sectoral exports allows us to control for all country factors that might simultaneously influence foreign bank presence (or financial development) and the overall growth of exports. And it allows us to exploit the notion that sectors, for technological reasons innate to the manufacturing process and unlikely driven by either the level of financial development or the share of foreign banks, vary in their vulnerabilities to financial frictions. If foreign banks (or financial development) play a role in boosting exports through financing, then this should affect firms in vulnerable sectors more. By including interactions of foreign presence and financial development in the exporting country with sectoral indicators of financial vulnerability, measured by external finance dependence and asset tangibility, we can identify the channels as well as importantly reduce statistical concerns, including with respect to omitted variables, endogeneity and reverse causality. 6 Our results show that, controlling for domestic financial development, foreign bank presence is associated with greater exports in sectors more intensive in external capital and with limited 5 Ahn (2011) argues that, as a letter of credit allows the bank of the exporter to transfer non-payment risk from the importer to the importer s bank, the bank replaces its inferior screening technology over the importer with superior technology of the importer s bank, thus reducing the effects of information asymmetry. A foreign bank from the importing country can effectively do this similarly. 6 While higher export demand for sectors intensive in external finance might still endogenously result in higher private credit or a larger presence of foreign banks, this concern does not apply to sectors asset tangibility.

5 5 tangible assets. An increase in general foreign bank presence in terms of assets by one standard deviation means exports in the sectors at the 75 th percentile of the distribution of external financing dependency are 7.2% percentage points higher than in the sectors at the 25 th percentile. Exports in sectors at the 25 th percentile of the distribution of asset tangibility are 6.4% percentage points higher compared to sectors at the 75 th percentile. These results are robust to adding various control variables, including differences in factor endowments, institutional differences, general economic activity and other types of financial integration. As foreign banks might establish presence in countries in which they expect more (future) trade, especially in sectors with greater financing needs, there could be some endogeneity. We exploit differences in foreign bank presence over time and use instrumental variables techniques in order to address such remaining concerns and find results to be again robust. To shed some light on the mechanisms through which foreign banks facilitate trade, we differentiate between the impacts of foreign banks in general and of banks coming from the importing country, as the latter may be especially well equipped to reduce the impact of contracting and information problems. We find that bilateral foreign bank presence further facilitates bilateral exports, over and above general domestic financial development and foreign bank presence. Also indirect foreign bank links between the importing and exporting country, i.e., when subsidiaries of a third country banking system are present in both countries, importantly add to exports. The latter finding suggests an information mechanism playing an important role (with results not affected by any reverse causality of foreign banks establishing themselves in anticipation of growth in bilateral trade or following their multinationals abroad). We lastly exploit the heterogeneity among our exporting countries in terms of economic and financial development, quality of contracting and informational environments, and share of foreign banks present. Differentiating along these dimensions, we find the impact of foreign bank presence on exports not to differ in the sample of financially underdeveloped countries and to be less beneficial in weak contracting and information environment countries, confirming previous results that foreign banks have greater difficulty operating in such environments. Foreign banks appear to have more beneficial impact on trade when they represent a larger share of the local system, suggesting it matters whether they are more invested in the local economy. The impact of bilateral foreign bank presence on exports is especially important when the exporting country is economically or financially less developed and has a relative weak information environment. Since we find fewer differences regarding contracting environment, this suggests that bilateral foreign bank presence mostly helps overcome information asymmetries. Overall, our evidence is consistent with the arguments that foreign banks facilitate trade over and beyond what domestic banks can offer. In addition, our findings suggest that foreign banks facilitate trade not solely through providing funds to (potential) exporters but also by helping overcome agency, information and other problems between (potential) exporters and importers.

6 6 The remainder of the paper is structured as follows. The next section describes the different data sources we use and combine. Section 3 presents the methodology and discusses our empirical findings. Section 4 concludes. II. DATA We want to examine to what extent the presence of foreign banks facilitates firms exports and what mechanisms may play a role. To this end, we need to combine sectoral and bilateral data on exports with detailed and bilateral data on foreign bank presence. We also need sectoral data on external financial dependence and asset tangibility. We describe these data in detail here, leaving the description of the control variables we use to the next section. We obtain data on bilateral trade flows for 134 countries at the 3-digit ISIC industry level for 28 manufacturing sectors from the UN COMTRADE database for the period We purposely end our sample period in 2007 as to avoid our results being influenced by the global financial crisis. To account for the skewed distribution in exports and deal with zero observations, our dependent variable is the log of the value of exports from country i to country j in 3-digit ISIC sector s and year t. The value of exports and number of trade partners differ greatly across countries and sectors. Appendix Table 1 reports for each exporting country total export in the 28 manufacturing sectors, the number of different sectors a country exports in and the number of trading partners (all measured in 2007). To measure foreign bank presence we use the bank ownership database constructed by Claessens and Van Horen (2014). The database contains ownership information of 5,324 banks that were active for at least one year between 1995 and 2009 and that reported financial statements to Bankscope. It covers 137 countries and coverage is very comprehensive, with banks included accounting for 90 percent or more of banking system assets. A bank is considered foreign owned if 50 percent or more of its shares is owned by foreigners, with residence of its main owner determined as the country for which the total shares held by foreigners is the highest. 7 To determine the importance of foreign banks in financial intermediation, we match ownership data with balance sheets data provided by Bankscope. Our main variable of interest is foreign bank presence (FB it ), which we define as the share of the assets of all foreign banks active in exporting country i in total bank assets in exporting country i at time t. The asset share is our preferred measure as it captures the importance of foreign banks in financial intermediation in an exporting country. Unfortunately, asset information is only reliable available from This implies that a foreign bank may be considered French owned, even though French investors only hold 20 percent while German and UK shareholders each hold 15 percent. In most cases, however, a foreign bank is majority owned by one parent bank. For further details see Claessens and Van Horen (2014).

7 7 onwards, reducing our sample period to include only Therefore, we also use two alternative ways to define FB it. The first is a dummy which is one if at least one foreign bank is present in exporting country i at time t. The second is the share of all foreign banks active out of total banks active in exporting country i at time t. Both these variables are measured over the full sample period, An important feature of the database is that it provides for each foreign bank the country in which its parent is headquartered, allowing us to study the bilateral aspects of foreign bank presence. We define bilateral foreign bank presence as the share of foreign banks assets from importing country j active in exporting country i out of total bank assets in exporting country i (BFB ijt ). As an additional measure of foreign bank presence, we study the impact of indirect bank links, which we define as a dummy which is one when countries i and j have no direct foreign bank links, but share a link through a common third banking system (IFB ijt ). Using the indirect links allows us to study the indirect effects of foreign bank presence and avoids some possible endogeneity issues related to (bilateral) foreign bank presence. We exclude a number of countries from our sample: offshore centers, as very specific factors may drive a bank s decision to enter those; 8 and exporting countries for which the share of banks with asset information available from Bankscope is less than 60 percent in at least one year between 2005 and This leaves us with a final sample of 95 exporting and 122 importing countries. Appendix Table 2 provides a list of all exporting countries in our sample the share of foreign banks (in assets and numbers), the number of foreign banks present, and in how many different countries the parent banks are headquartered (all as of 2007). In 2007, 1,043 foreign banks headquartered in 77 different home countries were active in our sample of exporting countries. The importance of foreign banks varies greatly by exporting country and ranges from zero (e.g., Ethiopia) up to 100 percent share, as for some other African countries. On average, 11 foreign banks from six different home countries are present in an exporting country. In most countries where foreign banks are present, banks from several different home countries are active and only in very few countries (only 11, or just 10 percent of the countries in our sample) are only foreign banks from one country present. In 78 percent of the 11,590 possible exporting-importing combinations in our sample is at least one foreign bank present, yet in only six percent of these pairs is a bank headquartered in the importing country present in the exporting country. Furthermore, 53 percent of the exporting-importing pairs share a link through a common, third banking system. These variations allow us to examine both the 8 We define the following countries as offshore centers: Andorra, Antigua and Barbuda, Bahamas, Bahrain, Barbados, Bermuda, British Virgin Islands, Cayman Islands, Cyprus, Liechtenstein, Mauritius, Netherlands Antilles, Panama, Seychelles and Singapore. 9 Including these countries does not affect our main results.

8 8 impact of general as well as specific, bilateral or indirect, foreign bank presence on bilateral exports. Our empirical strategy importantly relies on exploiting industry differences with respect to dependency on external finance and the availability of tangible assets, as done in much prior literature, also that studying the role of finance in exports (e.g., Manova, 2013). For technological reasons innate to the manufacturing process, producers in certain industries incur higher up-front investment that cannot be generated internally, thus typically requiring more external finance (Rajan and Zingales, 1998). Sectors also differ in firms endowments of tangible assets, such as plant, property and equipment, that can serve as collateral for raising outside finance, with firms with less tangible assets likely having more difficulty to attract external finance, especially in less financial developed and institutionally weaker countries (Braun, 2003; Claessens and Laeven, 2003). These two industry characteristics are widely viewed as sectorspecific, technologically-determined characteristics innate to the manufacturing process and unlikely to be determined by either the level of financial development or the share of (bilateral) foreign banks. At the same time, if foreign banks (or financial development) play a role in boosting exports through providing external financing, then this should affect firms in these financially vulnerable sectors more. We follow the literature and define external financial dependence as the fraction of total capital expenditure not financed by internal cash flows from operations, and asset tangibility as the share of net plant, property and equipment in the total book-value assets. Even though these sector characteristics could differ across countries, the measures are typically constructed using US data. 10 We use the values as provided by Braun (2003) who uses data for all publicly-listed US-based companies available in Compustat averaged over Appendix Table 3 lists for all sectors in our sample the ratios used for external finance dependency and asset tangibility. Table 1 provides some summary statistics and Appendix Table 4 provides the detailed description and the sources of the variables used in our analysis. 10 This is for three reasons. First, as the US has one of the most advanced financial systems, the behavior and choices of firms likely reflect optimal choices of external financing and asset structure, and not financing constraints. Second, detailed firm-level data needed to construct the variables are not available for many countries. Finally, for our empirical strategy only a relative ranking of sectors across the two dimensions is needed; therefore using US data is not a problem if industries relative ranking is the same across countries even if the exact magnitudes may vary.

9 9 III. EMPIRICAL RESULTS To examine the role of foreign banks in facilitating trade, we closely follow the specification of Manova (2013), who uses a traditional gravity trade model extended with the level of financial development, captured by private credit to GDP (as is standard in the literature). Since our goal is to examine the role of foreign banks in facilitating trade over and above the impact of financial development, our baseline model extends this specification by including also the presence of foreign banks in the exporting country. Similarly to testing the mechanisms by which financial development affects export activity, we allow the impact of foreign bank presence to vary by sectors that differ with respect to the dependency on external finance and asset tangibility. Our baseline model is as follows: where subscripts i and j denote exporting and importing country respectively, and s and t denote industry and year respectively; stands for the (log of) exports from country i to country j in sector s in year t, captures financial development in the exporting country i at time t, and and measure the external financing dependency and asset tangibility of the sector s; is either a dummy if at least one foreign bank is present in the exporting country at time t, or the share of foreign banks in the exporting country i at time t in terms of numbers or assets; is a coefficient vector and is a matrix of control variables which, in the base line specification, includes the (log of) the distance between the exporting and importing country;, and are vectors of exporter-year, importer-year and industry-fixed effect coefficients, respectively; and is the error term. The inclusion of exporter-year and importer-year fixed effects allows us to control for all (timevarying) exporting country factors that might simultaneously influence foreign bank presence (or financial development) and the level of exports and (time-varying) changes in demand at the importer side. Real GDP in both exporting and importing country, variables standard used in gravity models, are thus subsumed in these fixed effects. Industry fixed effects allow us to control for (time-invariant) sector characteristics that might affect trade patterns. Altogether, the inclusion of this extensive set of fixed effects should importantly reduce concerns that our results are driven by omitted variables or reversed causality. In a number of robustness tests, however, we address possible remaining concerns more formally. In an extension to the baseline model we study the impact of specific types of foreign bank presence, bilateral or indirect, over and above the impact of general foreign bank presence. This model is as follows:

10 10 where either represents the share of foreign banks from importing country j active in exporting country i at time t ( ) or a dummy which is one when subsidiaries of a thirdcountry banking system are present in both the importing and exporting country ( ) and includes, besides distance, and/or un-interacted. All regressions are estimated using OLS and standard errors are clustered by exporting-importing pairs. A. Baseline results In Table 2 we provide the results based on our baseline regression model. Using the full, period, we find (column 1 and 2) that countries with foreign bank presence, either when we use a dummy capturing the presence of at least one foreign bank or the share of foreign banks in terms of numbers, export relatively more in sectors more dependent on external finance and with less tangible assets to pledge as collateral. These effects are clearly in addition to the effects of general financial sector development, which itself also increases exports for those sectors more financially dependent and with less tangible assets (with the coefficients for those variables not differing much from those obtained in by Manova, 2013). We subsequently run the same regression using the share of foreign bank presence in terms of numbers and assets for the period (column 3 and 4). As can be seen, these regression results confirm the role and mechanism of foreign banks in facilitating trade. This suggests that general foreign bank presence provides for additional financing to financially vulnerable firms, which allows them to access export markets and/or increase their exports. Effects are economically significant: an increase in foreign bank presence in terms of assets by one standard deviation means exports in the sector at the 75 th percentile of the distribution of external financing dependency are 7.2 percentage points higher than in the sector at the 25 th percentile of the distribution. Exports in sectors at the 25 th percentile of the distribution of asset tangibility are 6.4 percentage points higher compared to sectors at the 75 th percentile of the distribution (based on the coefficients in column 4). While economically a smaller effect compared to general financial development (28.9 and 17.9 percentage points respectively), these increases are by no means marginal. B. Robustness tests We next conduct a number of robustness tests in which we include additional fixed effects as well as various other country variables that may affect exports through interactions with sectoral financial vulnerabilities. Results are reported in Table 3, where column 1 repeats the base regression results of Table 2, column 4. In general, the results show that adding these fixed

11 11 effects or control variables does not change much the statistical significance or size of the coefficients on foreign bank presence interacted with the two sectoral characteristics. In the first robustness regression (column 2) we include, besides the exporter-year fixed effects, importer-year-industry fixed effects, i.e., a full matrix of all 134 importers times 28 sectors times the three years. This way we control for any demand and price effects that may vary by importer and by sector. The statistical significance and size of the coefficients of all variables including foreign bank presence remain largely unchanged and remain supportive of the positive role that foreign bank presence plays in facilitating trade. Countries differ, besides in financial sector development and the presence of foreign banks, in various other ways that can affect their export performance. While the fixed effects we use already control for any time-invariant and time-varying country characteristics, there could still be country characteristics that interact with sectoral characteristics to affect exports. Therefore, we include, in line with the general law and finance literature, the exporter s country s GDP per capita and an index of the prevalence of the rule of law in the country as key (institutional) development indicators and interact these with the two sectoral characteristics. We find (column 3) that the better export performance comes about in part through easier access to external financing associated with higher general economic development and better rule of law. Tangibility, however, is less important for higher income countries. Importantly, the results on the role of foreign bank presence are not affected. While the coefficient for the interaction of foreign bank presence with financial dependence is somewhat smaller than in the base regression, as the other two country variables assume some of its effects, the coefficient on the interaction with the tangibility sectoral measure is of the same magnitude. Countries might also differ in their natural endowments which can affect their export patterns. For example, a country rich in human capital may have a comparative advantage to export more from a sector that relies naturally more on human capital. Similar to Manova (2013), we use the following three country factor endowments: human capital intensity, physical capital intensity, and natural resource intensity. We interact these with the corresponding sectoral intensities, where the benchmarks are, similar to the financial vulnerability measures, obtained from US corporate data. Regression results (column 4) show indeed that more human and physical capital and natural resource intensive sectors export more in countries that are more endowed with capital and natural resources. Importantly, even with these extensive controls and interactions, the effects of foreign bank presence on export performance are reconfirmed, with the interaction coefficients equally statistically significant and only marginally smaller than in the base regression. Since the effects of foreign bank presence on export performance can come about through overcoming information asymmetries and contracting problems in part related to how far the

12 12 exporting and importing countries are from each other, we next directly explore the role of bilateral country differences by including three additional distance measures: common border; past colonial links; and common language (note that we always already include physical distance, as is common in gravity models). We find these three distance measures to be all statistically significant positive (column 5), i.e., closer countries have more bilateral trade, consistent with fewer information asymmetries and contracting problems. Only the coefficient on physical distance is somewhat smaller, logical as the three other distance measures absorb some of its effects. Importantly, adding these variables does not affect the magnitude of the impact that foreign bank presence has on trade through external financing. Next we address the concern that our main results might be driven by other financial linkages. To this end, we include the stock of FDI in and cross-border loans to the exporting country (both as a share of GDP) and interact them with our sectoral measures of external financial dependency and tangibility. Regression results (column 6) show that countries and sectors with greater FDI indeed export more in financially vulnerable sectors. The effects of foreign bank presence on export performance and its channels through external financing, however, are reconfirmed (albeit parameters become somewhat smaller, a sign that the share of foreign banks and the stock of FDI are somewhat correlated, as also shown by Poelhekke, 2014). 11 Interestingly, the higher stock of cross-border loans does not have a positive impact on exports in sectors with higher financial dependency and a negative impact on exports in sectors with limited tangible assets. This shows that, over and above just merely providing bank loans, foreign banks play an important additional role through their local presence. Finally, we explicitly control for the possibility that foreign bank presence might impact the domestic economic activity of firms, and not directly provide financing for exports (for those firms in sectors that are more dependent on external finance or have less tangible assets). It could for example be that the entry of a foreign bank reduces credit constraints for all firms without scaling up exports. We therefore include the (log) output of all firms by exporting sector, country, and year. The results (in column 7) show that the size and significance of the interactions with financial dependency and asset tangibility become only slightly lower, indicating that foreign bank presence especially benefits exporting firms and that we are not just capturing an increase of cross-border sales similar to an increase in domestic sales. 11 As shown by Poelhekke (2014), the presence of a foreign bank tends to attract non-financial MNEs from the same home country (i.e., firms follow their banks) and this effect is especially strong in countries where investing is more hazardous. If these MNEs subsequently export back to their home country, this can be another channel through which foreign banks can facilitate trade.

13 13 C. Endogeneity It could be that foreign banks choose their investment overseas in those countries and sectors where financing needs are high because of high exports, but local banks are not able to provide financing. In addition, foreign banks might more likely invest in a country in which multinationals from their country are present. This could create a relationship between export growth and foreign bank presence that is based on a reverse causality. While possible, reverse causality is not so likely to affect our results as it requires a systematic correlation of foreign bank entry with exports and the sector-specific characteristics on external financing and asset tangibility. While higher export demand for sectors intensive in external finance might endogenously result in a larger presence of foreign banks, this concern does not apply to sectors asset tangibility. Furthermore, the argument that banks might just follow their customers only holds if multinationals specifically invest in financially vulnerable sectors and if they export their product either to their home country or a third country. To nevertheless address the concern that the decision for foreign banks to enter might be endogenous and therefore biases the estimates for the interactions, we conduct a number of additional robustness tests. We start of by exploiting differences in foreign bank presence over time. Foreign bank presence grew substantially in many countries from the late 90s until the start of the global financial crisis. However, prior to these boom years many banks already had established a presence abroad. While their investment decisions might have been driven by the presence of multinationals from their home country or by current or expected growth in trade in financially vulnerable sectors at the time of entry, they were unlikely influenced by trade opportunities materializing in financially vulnerably sectors years ahead. We therefore conduct the same regression as in the base model except we lag the foreign bank presence variable by 10 years. As we do not have asset information prior to 2005 we take the share in numbers. And as an additional test, we use the asset of only those banks that were present in 1995 to calculate the current foreign bank asset share. As can be seen in Table 4 columns 1 and 2, in both case, we find that all interaction coefficients remain similar and equally statistically significant. We next address endogeneity concerns through an instrumental variable estimation by two-stage least-squares. We follow an identification strategy similar to Jayaratne and Stahan (1996), Giannetti and Ongena (2009) and Bruno and Hauswald (2013) and use improvements in conditions for entry as instruments for the presence of foreign (or outside) banks. Specifically we instrument foreign bank presence with the Heritage Foundation financial freedom index. This index measures, among other things, stringency of banking regulation, ease of entry, and restrictions on foreign banks opening branches or subsidiaries, with 0 being a repressive system and 100 having negligible government interference. Since regulation on foreign bank entry alone is unlikely to have a direct effect on trade, the index should represent an appropriate instrument. We interact the index with financial dependence and tangibility and use both as instruments for

14 14 the interaction of share foreign assets with financial dependence and tangibility. Both instruments are positive and significant at the 1 percent level in the first-stage regression. Column 3 reports the results along with relevant test statistics. Using the Heritage instrument, we find results to be robust, with both interaction terms having very similar economic and statistical significances. The Kleibergen-Paap Wald F-statistic exceeds in all specifications the critical values of Stock and Yogo (2004), i.e., the null hypothesis of weak instruments is rejected. The Kleibergen-Paap LM statistic rejects under-identification again in all specifications (pvalue<0.001). Overall the instrumental variable regression thus confirms our findings that foreign bank presence boosts exports in financially vulnerable sectors. D. Bilateral and indirect foreign bank presence Until now our results show that foreign bank presence is associated with greater trade in financially vulnerable sectors over and above the impact of financial development. We next explore what mechanisms can explain this using information regarding the home country of the parent of the foreign banks to differentiate between different types of foreign ownership. First, we examine to what extent the presence of a foreign bank from the importing country has an additional impact over and above that of foreign banks from other countries. Whereas all foreign banks could facilitate exports to all countries in several ways, foreign banks from the importing country are likely most useful in helping overcome contracting and information problems. Furthermore, as shown by Poelhekke (2014), the presence of a foreign bank tends to attract non-financial MNEs from the same home country (i.e., firms follow their banks), with this effect especially strong in countries where investing is more hazardous. If these MNEs subsequently export back to their home country, this can be another channel through which foreign banks facilitate bilateral trade. 12 The results (Table 5, column 1) show that bilateral exports tend to be higher when a foreign bank from the importing country is present in the exporting country. Importantly, this impact is especially strong for sectors highly dependent on external finance or with low levels of tangible assets. This effect is in addition to that of general foreign bank presence and the coefficients for the other key (interaction) variables hardly change in significance or size. In economic terms, an increase in bilateral foreign bank presence by one standard deviation means exports in sectors at the 75 th percentile of the distribution of external financing dependency are 8.6 percentage points 12 This is also suggested by Bilir, Chor and Manova (2014) who show that financial development attracts more entry by multinational affiliates as it reduces credit constraints, but also reduces affiliates incentives to sell locally versus to their parent- and third-country destinations, i.e., it increases trade. Given the findings of Poelhekke, this channel could be stronger when financial development is driven by the entry of a foreign bank from the home country of the MNE.

15 15 higher than in sectors at the 25 th percentile. Exports in sectors at the 25 th percentile of asset tangibility are 7.4 percentage points higher compared to sectors at the 75 th percentile. Note that, while one could argue that the presence of a foreign bank would especially attract non-financial MNEs in financial dependent industries, it is unlikely that this applies to sectors with fewer tangible assets and that this would thus drive the higher levels of exports. Overall, these findings are therefore consistent with the idea that the presence of a bilateral foreign bank also facilitates trade by overcoming information problems. 13 Next, we consider the roles of indirect banking links between the importing and exporting country. Specifically, we want to examine whether bilateral exports in financially vulnerable sectors are higher when the importing country does not have a direct presence in the exporting country, but both countries share the presence of subsidiaries from a third country. Including this variable in the regression allows us to investigate whether foreign banks can also help with exports without being head-quartered in the importing country directly. Second, the presence of indirect links does not likely suffer from concerns regarding endogeneity as the decision of a third-country foreign bank to establish a presence in both country i and j is unlikely driven by trade (in financially vulnerable sectors) between these two countries. The results (column 2) show that indirect links are important for trade, both in general and especially in the financially vulnerable sectors. In terms of economic importance, the presence of an indirect link boosts exports of the sector at the 75 th percentile of the distribution of external financing dependency compared to the sector at the 25 th percentile by 9.6 percentage points. For a sector at the 25 th percentile of asset tangibility exports are 4.0 percentage points higher compared to a sector at the 75 th percentile when an indirect link exists. This means that the presence of an indirect foreign bank link compares in importance to a one standard deviation increase in bilateral foreign bank presence, thus a large effect. When we put the two effects together (column 3) we find that the results stay virtually the same, except for the interaction between bilateral foreign bank presence with asset tangibility which is now only significant at the 12 percent level (but only a slightly lower coefficient). E. Exporting country heterogeneity It is likely that (bilateral) foreign bank presence is more important in countries economically and financially less developed and that have greater institutional weaknesses. In those countries firms, especially those in more financially vulnerable sectors, are likely to find it more difficult 13 We also estimated the same regression for the longer times series, , using the number shares, and find similar results. When replacing the contemporaneous (bilateral) foreign bank presence with the presence in 1995 results remain again the same. We omitted these results for the sake of brevity, but they are available upon request.

16 16 to raise external financing or get trade finance and thus to export. Foreign bank presence may then be especially important. Also the relative stake of foreign bank presence may matter, as has been found in the general literature on foreign banking. To examine how these country differences may matter and drive the general result on the benefits of foreign bank presence, we split our sample of exporting countries across a number of dimensions. First, in order to capture differences in economic development we split the sample into two exporting groups based on income level: advanced countries vs. emerging markets and developing countries. We next split the countries based on their financial development and their share of foreign banks. In order to capture differences in contracting and information environment, we finally split the countries based on the strength of their contract enforcement, measured by the cost of enforcing contracts and the availability of credit information, as captured by the credit information index, with both variables coming from the World Bank Doing Business Indicators. For the last four country characteristics, we split our sample in groups of exporting countries below and above the median. Note that the correlations between these four variables ranges from -0.6 (between share of foreign banks and credit information) and 0.50 (between financial development and credit information), indicating that the splits capture different groups of countries. The results in Table 6 show that there are important differences among countries regarding the role of foreign banks in trade. We find (columns 1 and 2) that the impact of foreign bank presence in emerging markets and developing countries is somewhat higher in sectors with high external financial dependency (but lower for sectors with limited tangible assets). However, the results are especially striking in terms of bilateral foreign banks. They show that bilateral foreign banks do not play an additional role in advanced countries, but in emerging markets and developing countries exports in general and in financially vulnerable sectors in particular are substantially higher when a foreign bank from the importing country is present. In economic terms, an increase in bilateral foreign bank presence by one standard deviation means exports in sectors at the 75 th percentile of the distribution of external financing dependency are 13.6 percentage points higher than in sectors at the 25 th percentile. Exports in sectors at the 25 th percentile of asset tangibility are 6.8 percentage points higher compared to sectors at the 75 th percentile. We next find that the impact of financial development on trade is larger when countries are financially underdeveloped (column 3 and 4), probably as this helps to overcome other institutional weaknesses in such countries. The additional effect of general foreign bank presence on exports, on the other hand, is, albeit slightly higher, statistically no different for this group of countries. However, in terms of bilateral foreign bank presence, however, effects are much stronger for financial vulnerable sectors in financially less developed countries, with effects somewhat larger compared to those in the emerging market and developing country group as a whole. Indeed, in unreported regressions we split the sample of emerging and developing

17 17 countries between those with low and high financial development and find that the impact of bilateral foreign banks is especially high in the former group. When we split the countries in a group with a large share of foreign banks (more than 23 percent of the assets are held by foreigners) and those with a small share of foreign banks, we find very interesting results (column 5 and 6). Foreign banks, general and bilateral, only seem to have a beneficial impact on trade in financially vulnerable sectors when they represent a larger share of the local banking system. This suggests that they are better able to identify firms with profitable trade opportunities and provide trade-related financing to new and existing firms when they are more invested in the local economy. Finally, while financial development seems to (somewhat) overcome weak contracting and information environments, general foreign bank presence actually matters less for trade in these environments (columns 7-10). This is in line with earlier findings in the literature that show that foreign banks in general have difficulties operating in these environments. However, in these countries exports in sectors with greater external dependence are importantly higher when bilateral foreign bank presence is higher. And for sector with low tangibility, exports are higher in low information environments when bilateral foreign bank presence is higher. This is consistent with the notion that foreign banks from the importing country can facilitate trade by overcoming information and, to a lesser extent, enforcement problems. Overall, our findings suggest that foreign banks facilitate trade through overcoming information and other problems more prevalent in emerging markets and developing countries. But in order to do so, they either have to be strongly engrained in the local economy or they need to come from the importing country. IV. CONCLUSION This paper investigates empirically whether the benefits of foreign bank presence also extend to trade. It explores a number of arguments why foreign banks may play a special role in enhancing trade using a unique dataset of bilateral foreign bank presence combined with data on bilateral sector exports for 95 exporting countries. Our results confirm the established fact that countries with more developed financial sectors tend to export relatively more in sectors that have greater natural external financial dependence and less tangible assets, indicating that access to finance is an important channel. Controlling for this effect, we find that sectors with greater external financial dependence and less tangible assets tend to export more when a larger share of the banking sector is foreign owned. In addition to the general foreign bank presence, foreign banks from the importing country have an important role in facilitating trade, especially in less developed economies and where the information environment is weak. Furthermore, the existence of an indirect link between the importing and exporting country through foreign bank ownership importantly benefits trade. We interpret our findings as suggesting that foreign banks,