U.S. Monetary Policy and Global Credit Cycles

Transcription

1 U.S. Monetary Policy and Global Credit Cycles Falk Bräuning Federal Reserve Bank of Boston Victoria Ivashina Harvard University and NBER First Draft: March 15, 2016 Abstract Using twenty-five years of global syndicated loan issuance, we show that U.S. monetary policy easing is associated with a general increase in cross-border loan volumes by global banks from developed countries. The effect is much larger for borrowers from emerging market economies (EMEs) and holds after controlling for demand factors. Over a typical monetary easing cycle this differential effect amounts to a 32 percent loan volume increase for EMEs, with a similarly large effect upon reversal of U.S. monetary stance. We show that local lenders do not offset this effect. Furthermore, the results hold not only for corporate credit but also for EME infrastructure projects which are particularly dependent on cross-border credit. The effects of monetary policy are not confined to conventional interest rate policy, but also unconventional monetary policy easing has strong expansionary effects on the global flow of capital into EMEs. Key words: Global business cycle, monetary policy, emerging markets, reaching for yield We thank Kovid Puria for excellent research assistance. The views expressed in this paper are those of the authors and do not necessarily reflect those of the Federal Reserve Bank of Boston or the Federal Reserve System.

2 I. INTRODUCTION The mandates of central banks are typically focused on domestic economic conditions and do not account for potential international spillovers. While there are some isolated examples of collaboration among monetary authorities of major currency areas, emerging market economies (EMEs) remain outside of these coordination efforts. Following the 2008 financial crisis, this issue has resurfaced in the public debate in the context of large capital inflows into EMEs associated with unconventional monetary policy in major currency areas. 1 But while there is no denial of the significance of cross-border capital flows, especially for EMEs, substantial skepticism about monetary policy spillovers remains. This is potentially due to the lack of convincing evidence supporting the link between global capital flows to and from EMEs in response to monetary policy in the U.S. and other major economies. 2 Providing such evidence is the main goal of this paper. Our focus is on capital flows channeled by global banks; specifically, on emerging markets lending by banks headquartered in developed economies. Cross-border loans are by far the most important category of capital flows into EMEs. As of 2015, International Monetary Fund (IMF) data indicate that loans represent about half of all external liabilities of emerging market countries, while foreign bond and equity portfolio investments combined represent only about twenty percent. Much of this foreign capital comes from banks from developed economies: Bank for International Settlements (BIS) data show that roughly a third of all external liabilities 1 Rajan (2014) is a notable example of a call for action on this issue. 2 It is true that interest rate differentials associated with differences in national monetary policies can promote cross-border capital flows as investors seek higher returns. But my reading of recent research makes me skeptical that these policy differences are the dominant force behind capital flows to emerging economies; differences in growth prospects across countries and swings in investor risk sentiment seem to have played a larger role. (Bernanke, 2013.)

3 of the emerging markets is held by U.S., European, and Japanese banks. Moreover, the volume of these claims nearly doubled since the onset of the global financial crisis reaching about $7 trillion in We examine the effect of U.S. monetary policy, as the vast majority of the capital flows into EMEs is denominated in U.S. dollars. Consistent with the general dominance of the dollar in international trade and finance (e.g., Shin, 2012 and Gopinath, 2016), a currency breakdown of syndicated loan data shows that about 87 percent of the cross-border loans to EMEs are denominated in U.S. dollars. In turn, according to Gadanecz and Von Kleist (2002), syndicated loans the core data for our study represent a large share of corporate cross-border credit in emerging markets. Figure I illustrates the basic correlation between cross-border loans to EMEs and U.S. monetary policy (as measured by the U.S. federal funds rate). The data are compiled from the IMF International Investment Position and cover the period from 1980 to The significant negative correlation tightening U.S. monetary policy associated with the contraction in credit holds in levels and in changes. As we will show, this relationship can also be detected in the aggregate BIS data on country-by-country cross-border bank claims; see also McCauley et al. (2015). [FIGURE I] Aggregate results, however, could merely pick up relative changes in investment opportunities around the world, or compositional shifts in the investor base in a given country. We therefore use DealScan syndicated loan data which provides detailed information on loan issuance to individual firms from a wide range of EMEs from 1990 through Because in the syndicate loan data we observe the borrower identity, we can control for time-invariant borrower characteristics including any heterogeneity in time-invariant demand for credit by including 2

4 borrower fixed effects. Moreover, since we also observe the lender identity, we can then focus on comparing loans originated by the same lender in the same quarter to borrowers in different markets. In particular, we focus on comparing loans by the same lender to foreign borrowers in emerging markets and in developed markets. Our hypothesis is that flows driven by capital in search of yield are more likely to pursue a riskier, high-yield market like EMEs as opposed to a developed market. Although we are looking at foreign markets, to account for potential heterogeneity in trade linkages with the U.S. (which could affect investment opportunities of foreign companies, introducing a time-varying component to demand for credit), we look at a geographical breakdown of the emerging market regions. 3 We show that U.S. monetary policy easing is associated with a general increase in crossborder loan volumes by global banks from developed countries. However, we document a significant differential effect across markets: when U.S. monetary policy eases, a given bank provides larger cross-border loan amounts to borrowers from EMEs as compared to borrowers from developed markets, in the same quarter. This differential effect amounts to about eight percentage points per one-percentage point decrease in the U.S. fed funds rate and holds after controlling for demand factors using borrower fixed effects. During a typical monetary easing cycle in our sample period, the Fed cut its target rate by about 4 percentage points. Given our estimates, this would amount to an additional increase in loan volumes to emerging market borrowers by a sizable 32 percent. On the other hand, a monetary policy tightening would pull 3 We should emphasize that monetary policy is endogenous to economic conditions (Romer and Romer 2004). In this study, we are using monetary policy rates as a measure of domestic nominal interest rate levels, which are, in particular at the short end, strongly influenced by monetary policy actions, but are intrinsically connected to the economic fundamentals. Our results are informative to consequences of monetary policy actions, but we are not relying on the exogeneity of monetary policy in our analysis. 3

5 out bank flows from emerging markets and lead to a strong contraction of foreign credit in emerging markets. The effect of U.S. monetary policy on emerging market lending holds qualitatively for borrowers from all geographic regions (however, with some heterogeneity: the effects are strongest for borrowers from Africa and weakest for borrowers from Latin America). Moreover, the effects are not constraint to credit provision by U.S. banks; also, non-u.s. banks capital flows significantly more into EMEs when the Fed eases policy. Additionally, we also show that the spillovers of U.S. monetary policy to cross-border bank credit in emerging markets actually generate larger credit fluctuations in countries with more financially open economies. We confirm that the contraction of credit by banks from developed market economies (DMEs) in a U.S. tightening cycle is not offset by an increase in credit by the local banks. Instead, the reduction of foreign bank credit leads to a lower probability of refinancing for borrowers of global banks and a contraction in credit amounts of granted loans. We also stress the relatively short maturity of corporate loans and large amount of loans coming due in any given year. Finally, we show that also EME infrastructure financing is sensitive to U.S. monetary policy. Importantly, it is well known that infrastructure financing is heavily dependent on international capital and, specifically, on syndicated bank credit. Our results are robust to alternative measures of U.S. monetary policy. In particular, we also study the effect of various measures of unconventional U.S. monetary policy that may have particular strong effects on emerging market capital flows. For example, we use the Wu-Xia (2016) shadow rate, which accounts for unconventional monetary policy easing when the fed funds rate was at the zero lower bound. We also use the size of the Fed s balance sheet as a proxy of quantitative easing. Similarly, we look at the 10-year U.S. treasury yield and the term 4

6 spread as measures of monetary policy. In all cases, we find that easier monetary policy significantly boosts the lending to borrowers from EMEs and tightening monetary policy leads to its reversal. At a high level, our paper contributes to the large economic literature on international spillovers via capital flows, monetary policy transmission, and the role of global financial intermediaries. Most directly, our work expands empirical evidence of a global financial cycle that is linked to economic conditions in the center country of the world economy (Rey, 2013). In particular, using a VAR approach, Miranda-Agrippino and Rey (2015) show the importance of transmission of U.S. monetary policy across border via financial intermediaries. McCauley, McGuire, and Sushko (2015) use aggregate data to study the effect of U.S. monetary policy on global dollar credit. We complement this work by providing firm-level evidence of the transmission of the U.S. monetary policy through the balance sheet of global banks. This is in line with the contemporaneous work by Demirgüç-Kunt, Horvath, and Huzinga (2017) who focus on the role of global banks charter in foreign markets for the transmission of foreign monetary policy. Morais, Peydró, and Ruiz (2015) and Altunok, Gumus, Kapan, and Ongena (2016) examine spillovers of monetary policy through global bank activities in Mexico and Turkey, respectively. We show that this result holds in a cross-country setting and is especially pronounced for emerging markets more broadly. Bräuning and Ivashina (2016) discuss firm-level evidence on monetary policy spillovers in major developed economies but the emphasis there is on the interaction with the currency market and part of the mechanism is specific to capital flows across major currency areas. 5

7 Our focus in this paper is specifically on the differential sensitivity of emerging market economies to the transmission of the U.S. monetary policy. In that respect, this paper complements work by Forbes and Warnock (2012), Fratzscher (2012), and Ahmed and Zlate (2013) which have highlighted the importance of U.S. economic conditions, in particular U.S. monetary policy for capital flows into emerging markets from a bond and equity flow perspective. More broadly, our work contributes to the literature on credit cycles in emerging market economies and their implications for financial stability and economic development. For example, Acharya et al. (2015) and Shin (2016) highlight the risks for financial stability when emerging markets borrowers sharply increase dollar leverage during periods of strong capital inflow. The rest of the paper is organized as following: In Section II, we establish our main result on the effect of U.S. monetary policy on lending to EMEs. In Section III, we discuss the effects of foreign bank funding dependence by EMEs. Section IV concludes. II. U.S. MONETARY POLICY AND CORPORATE LENDING According to data from the IMF International Investment Position, loans are by far the largest investment class by foreigners in EMEs, representing about 50% of all EMEs external liabilities. 4 By comparison, as shown in Figure II (a), in 2015, portfolio (vs. direct) bond investment in EMEs were only about 15% of external liabilities, and portfolio equity investments were about 5% (these numbers refer to the median values of EMEs). That said, bank and broader 4 Data on total external liabilities are collected from the IMF statistics and include all claims of foreigners on a given country, including all forms of equity and debt instruments. 6

8 fund flows display a strong correlation: from 1990 onward, the correlation between loan flows, and equity and bond flows to EMEs is about (This relation is weaker for the 1980s.) Combining the IMF data with the BIS Consolidated Banking Statistics shows that global banks play a key role in channeling capital flows into developing markets. About a third of all external liabilities of the emerging markets is held by foreign banks from major economies, twice as large as claims of foreign banks on developed markets (Figure II b). 5 Moreover, the relative importance of cross-border bank claims for developed countries has decreased from 2005 (beginning of the detailed BIS data) to 2015, but the trend has been the opposite for the emerging markets. Consistent with Giannetti and Laeven (2012), the BIS data also show that the total volume of claims on developed market countries decreased since 2008 from $25 trillion to $16 trillion in In contrast, claims on emerging market countries increased by more than threefold throughout the entire sample from about $2 trillion in 2005 to about $7 trillion in [FIGURE II] To establish the connection between monetary policy and credit flow into emerging markets, we rely on syndicated loan market data from DealScan. A syndicated loan is originated and managed by a small group of banks, but it is funded by a wider group of creditors. DealScan tracks information on loan originations, including borrower name and country, the amount of the loan, currency denomination, and the identity of the key lenders. Comprehensive data are available from 1990:Q1 through 2006:Q3. We focus on loans to borrowers domiciled in emerging markets by banks from developed countries. While DealScan only has accurate 5 BIS Consolidated Banking Statistics contains cross-border claims as reported by banks from 21 developed countries. (Claims by banks from the only three emerging market countries in the sample are small.) The reported cross-border claims included all types of bank loans, but also other debt instruments and equity claims. 7

9 coverage for syndicated loans (and not other form of bank credit), these represent a significant part of international bank claims. According to Gadanecz and Von Kleist (2002), the estimated outstanding stock of syndicated loans amount to about 50 percent of all outstanding BIS bank claims on Latin America and developing Europe, and to around 100 percent of all outstanding bank claims on Asia and the Africa-Middle East region. In our analysis, we focus on loan commitments by banks from developed markets to foreign borrowers that are domiciled in a different country than the lending bank. Table I provides summary statistics of the geographical breakdown of these cross-border loans as well as their currency denomination. About 20 percent of all cross-border loans in DealScan are issued to borrowers from the emerging market countries. Similarly, about 20 percent of all borrowers in the sample are from the emerging markets. Overall, the sample covers loan issuance to borrowers from 109 different emerging market countries. [TABLE I] Table II presents the benchmark results. The dependent variable is the logarithm of the total amount of lending by a given bank to a given firm in a given quarter. 6 Our key independent variable is the (effective) federal funds rate (in percent) that measures the stance of U.S. monetary policy. Columns (1) to (6) restrict the analysis to loans denominated in U.S. dollars which, as mentioned earlier, account for the vast majority of loans to emerging market firms. In column (1) we look at the basic relationship between the federal funds rate and DME banks lending abroad. Given the emphasis on foreign capital flow to EMEs, the key coefficient of interest in Table II is the interaction term between the federal funds rate and a dummy indicating whether the borrower is domiciled in an emerging market economy as classified by the BIS. This 6 About 2 percent of all borrowers have more than one loan in a given quarter, in which case we aggregate the amounts. 8

10 coefficient picks up an average differential effect for capital flow to EMEs when the U.S. federal funds rate changes. Because we know the borrower identity and its key lenders, in specification (3) we control for firm, bank, and quarter fixed effects. This helps us to deal with a demanddriven explanation for changes in credit behavior. In particular, this set of controls accounts for time-invariant demand factors such as firms size and industry. The results in Table II consistently indicate that the cross-border loan volumes to both emerging and developed market firms are negatively related to the fed funds rate. An easing of U.S. monetary policy pushes bank flows into foreign markets, while a tightening of U.S. monetary policy reduces banks investment into foreign countries (column 1). Importantly, as columns (2) through (5) indicate, the estimated effect is larger for emerging markets. In economic terms, a 1 percentage point easing in the U.S. federal funds rate increases the volume of cross-border loans to borrowers from emerging markets by an additional 7.7 percent as compared to borrowers from developed market. This result does not change if we include lenderquarter fixed effects (column 4), thereby controlling for time-varying bank heterogeneity, such as individual bank health or differences in business models. The identification of the effect is driven by the differential loan volumes to emerging and developed countries of a given bank in the same quarter after netting out any borrower specific time-invariant characteristics. [TABLE II] A closer look at the geographical breakdown of emerging market countries in column (5) shows that the effect of U.S. monetary policy is present throughout the different geographical regions. 7 The result is economically and statistically weaker for Latin America. This is in line with Takats (2010) who points out that, unlike other EMEs, in Latin America the expansion of 7 We use the BIS classification for the four emerging market regions and exclude offshore centers from the sample. 9

11 international banks mainly took the form of increased domestic currency lending by local affiliates, making cross-border bank lending relatively less important for these regions. The effects of U.S. monetary policy are strongest for Asian and African (including the Middle East) emerging market countries. In column (6) we exclude U.S. banks from the sample. The result indicates that U.S. monetary policy spillovers to emerging markets are also present in the portfolio allocation of other major non-u.s. banks located, e.g., in Europe and Japan. Presumably because we consider cross-border loans denominated in U.S. dollars, and the U.S. monetary policy primarily affects the relative return on dollar assets. In column (7) through (9), the sample includes cross-border lending by global banks denominated in currencies other than U.S. dollars, primarily in euro and yen, but also in local currencies of the borrowers. This shows that the effect of U.S. monetary policy on the differential loan provision to borrowers in emerging vs. developed markets is indeed both economically and statistically weaker for this subsample of loans. Table III shows that the results in Table II are robust to the use of alternative monetary policy measures. In particular, we evaluate the sensitivity of the coefficient in Table II, specification (4) on the interaction term between the federal funds rate and the EME dummy to the use of alternative measures of U.S. monetary policy, keeping the underlying sample the same. Table III shows that the results are quantitatively similar when we use the Wu-Xia (2016) fed funds shadow rate that takes into account unconventional monetary policy measures during the post zero-lower bound period, such as quantitative easing programs that affected long-term yields. In another approach to measure unconventional monetary policy, we find that the size of the balance sheet of the Federal Reserve (relative to U.S. GDP) that was driven by large-scale asset purchases that compressed longer-term yields is positively related to emerging market 10

12 lending. We also use the 10-year U.S. Treasury yield that directly measures the stance of longterm interest rates. The resulting estimate is about twice as large when compared to the baseline regression that uses the fed funds rate suggesting that low long-term yields lead to strong loan origination in EMEs. Quantitatively, a decrease in the 10-year yield increases the volume of EMEs loans by an additional 17 percent as compared to the general increase in cross-border lending. Moreover, our main result remains robust if we use the term spread, defined as the difference between the 10-year and the 2-year Treasury yield, as an alternative monetary policy measure. [TABLE III] There might be a concern that the dummy variable for EME is too coarse. In Table IV, instead of dividing the borrower countries into emerging and developed markets based on the BIS classification, we look in Table IV at underlying country characteristics and how these characteristics interact with U.S. monetary policy. In column (1), we find that borrowers from countries with strong economic growth (as measured by lagged GDP growth in percent) in general receive larger loan volumes. However, borrowers in these countries are also more sensitive to capital flow reversals when U.S. monetary policy tightens. This result holds after controlling for borrower fixed effects and the coefficient is therefore identified from variation in GDP growth of a given country. Column (2) shows that the same holds for borrowers from countries which experience a strong increase in the national stock market index. (Note that the underlying sample changes due to data availability.) In column (3), we find that in general borrowers from high-risk countries (as measured by the lagged rating for long-term sovereign debt from Fitch Ratings) receive smaller loan amounts. However, the negative coefficient on the interaction term shows that loan volumes to high-risk countries increase stronger than loan 11

13 volumes to low-risk countries if U.S. monetary policy eases. Overall, the evidence is in line with a reaching-for-yield behavior, where DME banks increase syndicated loan origination in riskier and high-growth markets when U.S. interest rates are low (e.g., Bruno and Shin, 2015). [TABLE IV] By looking at the syndicated loan data we can analyze the loan flow (vs. stock), focus on specific currencies, and, most importantly, include a tight set of borrower and lender controls to address alternative explanations of our findings. While this cannot be done with the BIS data, it is informative to examine whether a similar relationship between U.S. monetary policy and bank capital flows from developed to emerging markets holds in this sample given that BIS data are not constrained to syndicated credit. In Table V, we look at the logarithm of the cross-border claims on nonbank private firms as reported by banks from developed markets in the BIS consolidated banking statistics. Hence, the data are at the banking-sector, country, quarter level, covering the period from 2005:Q1 through 2016:Q1. 8 For example, one of the observations in the sample corresponds to all claims on nonbank firms in Kenya held by all British banks in a given quarter. Table V, column (1) shows that overall cross-border claims increase when there is a monetary policy easing in the United States. Similarly to results in Table II, when we look more closely in which countries banks invest when there is a U.S. monetary policy easing, we see that the effect is largely driven by investments into emerging markets. In column (3), we add time fixed effects and banking-sector-country pair effects, which absorb any common time trends and time-invariant country-pair-specific effects such as distance between the two countries or similarity in legal origins. After adding these controls, we estimate a significant differential effect in the sensitivity to U.S. monetary policy between claims to emerging and developed 8 Another advantage of the DealScan data is that it covers a longer timer period than the aggregate BIS data. 12

14 market countries. Economically, emerging market claims increase by an additional 8.5 percent as compared with domestic market claims when the U.S. eases monetary policy by 1 percentage point. [TABLE V] In column (4), we add banking-sector-quarter fixed effects to the specification. Thereby, we mimic the strong identification used in the loan-level analysis and control for any time-varying heterogeneity at the banking-sector level, such as macroeconomic conditions in the banks home country or the health of the banking sector, and isolate the differential effect between emerging and developed market claims depending on U.S. monetary policy. For example, this would also accommodate a general contraction in credit abroad of all Japanese banks due to problems in the Japanese banking sector. After this strict identification, the effect decreases slightly but is still sizable, indicating an additional increase of investment in emerging markets of 6.5 percent per 1- percentage-point cut in the fed funds rate. As before, the response of EME bank flows to monetary policy is not confined to certain regions (column 5). Consistent with earlier findings, the effect is smallest for Latin America. Column (6) shows that the sensitivity of cross-border bank flows from developed to emerging market countries is not driven by U.S. banks only, as we find a similar estimate when we exclude claims by the U.S. banking system from the sample. Overall, the results using a broader measure of bank credit confirm the effects that we have identified using syndicated loan-level data. In Table VI we evaluate the sensitivity of bank capital flows to U.S. monetary policy depending on the financial openness of a country. As Rajan (2014) points out: [c]ountries that undertake textbook policies of financial sector liberalization are not immune to the inflows indeed, their deeper markets may draw more flows in, and these liquid markets may be where 13

15 selling takes place when conditions in advanced economies turn. We measure the financial openness of a country using the financial openness index of Chinn and Ito (2006), which measures the degree of a country s capital account openness based on various restrictions on cross-border financial transactions. The index ranges from 0 (no financial openness) to 1 (full financial openness). Column (1) of Table VI shows that not surprisingly foreign banks hold more claims on private nonbank entities from countries which are more financially open. Note that the regression includes banking-sector*host-country fixed effects. Therefore, the coefficient is identified from changes in financial openness of a given country over time. In column (2), we test whether bank capital flows to financially open countries are more sensitive to U.S. monetary policy. Indeed, the negative estimate indicates that financially open countries have a larger increase in bank claims when U.S. monetary policy eases and a stronger retrenchment when monetary policy tightens. The coefficient estimate in column (3) is robust to a tight set of fixed effects and indicates that if the financial openness index of a country increases by one-standard deviation, the sensitivity of loan volumes to U.S. monetary policy increases by an additional 2 percentage points. The result is robust to controlling for country risk in column (5) and the interaction with country risk and U.S. monetary policy (6) and therefore is not driven by the possibility that financially open countries may have a different risk profile. (Note the change in sample form column 3 to 4 due to limited data availability of country ratings.) [TABLE VI] III. FOREIGN CREDIT DEPENDENCE Results in the previous section indicate a substantial dependence of global bank credit to EME borrowers on the stance of U.S monetary policy. We next analyze the extent to which the 14

16 large capital inflow during U.S. easing periods and the subsequent retrenchment of foreign capital during a U.S. monetary policy contraction affects credit conditions of EME borrowers at the firm level. After all, at the individual firm level, inflow and outflow of foreign capital may just lead to a substitution toward domestic lenders, leaving overall firm-level funding conditions unchanged. First, Figure III highlights a slightly positive correlation between lending by global/dme banks and local EME lenders. That is, local creditors do not offset foreign withdrawal of bank credit. (This basic relation is robust to a regression setting.) While this might not be surprising in the banking context, this result is in sharp contrast to findings for securities markets that local investors (at least partly) offset a decline in foreign holdings, see Forbes and Warnock (2012). The lack of substitution, in turn, magnifies the overall effect that foreign capital withdrawal might have on the economy (Caballero and Simsek, 2016). To study the effects of foreign bank capital flows driven by U.S. monetary policy at the firm level, we analyze the extensive and intensive margin of credit for a given firm during periods of U.S. monetary easing and tightening. For the extensive margin, for each firm-quarter, we construct a dummy variable that equals one if the firm obtains a (syndicated) loan and zero otherwise. For the intensive margin, we construct the percentage change (first difference of the logarithm) of the amount of the loan in this quarter and the amount of the last loan in the same sample. We then estimate the probability of a loan to a firm and the change in the loan amount in periods of U.S. monetary policy tightening and monetary easing. We define periods of U.S. monetary policy tightening as quarters when the federal funds rate increases and all subsequent quarters where the rate is not decreasing until a reversal of policy occurs. The remaining periods are defined as periods of monetary policy easing. As a key explanatory variable of interest, we 15

17 use the share of foreign global banks from DMEs participating in the last syndicated loan of the firm. We focus on the period after 1995:Q1 to have a sufficient number of firms in our sample so that we can compute the variable Past Foreign Bank Reliance. Column (1) of Table VII shows that, during an easing of U.S. monetary policy, borrowers who had a larger share of DME banks participating in their last syndicate have a higher probability of obtaining a loan. The result holds after controlling for firm and quarter fixed effects, and we are hence only identifying the cross-sectional differential effect depending on past foreign bank reliance. The estimate indicates that for a borrower that had a one-standard deviation larger share of foreign banks in the past syndicate, the probability of obtaining a loan during an easing cycle increases by 5 basis points. Relative to the mean probability of getting a loan, this corresponds to a sizable increase of about 28 percent. We do not find a differential effect on the loan probability during a U.S. tightening cycle (column 2). That is, given the positive correlation between local and global banks lending behavior, the lack of a differential effect indicates a contraction in credit. When we look at the intensive margin of credit, we find in column (3) that during an U.S. easing cycle, borrowers who relied heavily on foreign bank credit in the last loan receive larger loan volumes. The estimated coefficient indicates that a borrower with a one-standard-deviation larger share of foreign banks in the last syndicates obtains a 6 percent larger loan volume. However, when U.S. monetary policy tightens, these borrowers also experience a strong reduction in the loan volume. The coefficient suggests a additional contraction in loan volumes by 13 percent for a borrower with a one-standarddeviation larger share of foreign banks in the last syndicate. Thus, the bank capital pushed in and out of emerging market by U.S. monetary policy has actual consequences on EME borrowers funding conditions. 16

18 [TABLE VII] To understand the importance of the availability of refinancing, we should emphasize the relatively short maturity of corporate loans. As Figure IV illustrates, the median maturity of a corporate loan in our sample is five years. As of the end of 2016, roughly half of the loans outstanding matured within two years, which points to a substantial macroeconomic risk. Finally, we look at the financing of large infrastructure projects that is projects related to the provision of essential services that are relevant for the broader economic development and growth of an economy due to their dependence on the availability of foreign bank financing (e.g., Ehlers, 2014, World Bank, 2016). Infrastructure projects are in the vast majority of cases greenfield projects with a long maturity. Moreover, infrastructure projects typically generate no cash flow until the completion of the project. 9 Thus, bank loans for infrastructure projects are often syndicated as a way of diversifying risk exposure. The alternative to privately syndicated credit is loans from multinational institutions, which tend to follow a very different and intense compliance process. 10 DME banks have increased investment into infrastructure projects in EMEs in the last 25 years rising to a total volume of $25.8 billion in 2014 which equals 13.5 percent of all new loan volumes committed to EMEs borrowers. Whereas some doubts might remain as to how hard it is to find alternative financing for corporate investing, we consider the sample of infrastructure credit to be an example of high impact projects for which it is very 9 World Bank (2016) reports that 86% of all infrastructure projects in EMEs were greenfield projects. Syndicated loans are the predominant source of funding for infrastructure investments as compared to infrastructure bonds, which only account for 10-20% of infrastructure funding depending on the region and mostly in the operational phase of the project (Ehlers 2014). 10 As an example, failure to close a private syndication due to the unravelling of the 2008 financial crisis led to roughly a two years delay in raising debt funding from multinationals for construction of Egyptian Refinery Corporation. 17

19 difficult to find a substitute to funding by global banks. (On the positive side, as illustrated in Figure V, the maturity of this type of credit tends to be very long and is less likely to be dependent on refinancing.) DealScan data has information on a large set of syndicated project finance loans, the major structure of infrastructure finance. In Table VIII, we look at whether the capital flow triggered by U.S. monetary policy affects the financing of infrastructure projects in EMEs by developed market banks. To identify infrastructure investments, we look at various dimensions related to infrastructure investments. This information includes the type of project (e.g., sea ports, toll roads, or electricity grids). We therefore look at the effects of monetary policy on project financing. One of the key differences between corporate loans and project financing is the maturity of the loan, with corporate loans having an average maturity of 4.3 years, and project finance loans having an average maturity of 10.7 years. Given that infrastructure investments typically have long duration, we also examine syndicated corporate loans (that are not project finance) with maturity larger than five years as a further proxy for infrastructure-related investments. Moreover, we use information on corporate loans to borrowers from infrastructurerelevant sector based on the classifications by the World Bank (2016) and the BIS (Ehlers, 2014). [TABLE VIII] In columns (1) to (4) we look at the effects of monetary policy on project financing. All specifications include bank fixed effects, but, given the structure of the investment where the assets of the sponsor are ring-fenced and repeated projects by the same sponsor are rare, we do not include borrower fixed effects. Unlike the full set of loans, we do not observe multiple project finance loans in the same quarter by the same bank, which is why we do not include fixed 18

20 effects for bank interacted with quarter. The results in columns (1) and (2) indicate that an easing of U.S. monetary policy is associated with an increase in the volume of cross-border project financing, with a significant stronger increase in EMEs that we can identify after netting out common time effects. This result also holds in columns (3) to (4), where we restrict the sample to infrastructure-related projects as classified in World Bank (2016). 11 Economically, we estimate an additional increase in project finance loan volumes to EMEs by 6 percent. In columns (5) to (10), we look at syndicated corporate loans that are not classified as project finance. Hence, the assets of the borrower are not ring-fenced, and the borrower identity matters, why we include borrower fixed effects in all columns. In columns (5) to (8), we restrict the full sample of corporate loans to firms from infrastructure-related sectors, both based on the World Bank and BIS classifications, and in columns (9) to (10), we look at loans with a longer maturity (in excess of five years). Also for these subsets of loans, we find that an easing of U.S. monetary policy increases loan origination in EMEs significantly stronger that in developed markets. V. CONCLUSIONS The years following the global financial crisis had been characterized by highly accommodative monetary policy throughout major currency areas, in particular in the United States. These efforts by central banks to stimulate domestic economies had been argued to incent a reaching for yield phenomenon. This per se is not so alarming. As Yellen (2011) points out, the shift toward riskier assets is a normal channel through which monetary policy supports economic activity. However, in the context of global financial markets, this channel spills to foreign countries, with large amounts of capital flowing from the U.S. and other developed economies 11 The World Bank sectors include airports, electricity, natural gas, railroads, roads, seaports, telecom, and water and sewerage. 19

21 into emerging markets in search for yield. As we have shown, the volume of global banks claims on emerging markets by far, the largest category of foreign capital channeled through financial intermediaries into EMEs nearly doubled since the onset of the global financial crisis reaching about $7 trillion in The problem is that the accommodative periods of monetary policy in large countries unavoidably will come to an end. And the reversal of monetary policy toward a contractionary stance leads to capital withdrawal from high-yield assets, which in the global context means capital withdrawal from the emerging markets. Such externally driven credit cycles are prone to build up of leverage and imbalances making economies vulnerable to global capital outflows. Rajan (2014) postulates that emerging countries wish for stable global capital inflows instead of flows being pushed in by foreign monetary policy given that local policy measures of receiving countries are unlikely to be effective. But substantial skepticism remains among economists and monetary authorities on whether global macro-prudential approach to monetary policy is necessary. In part, this is due to little empirical evidence on the subject. This is the research that we address in this paper. Using loan-level data for syndicated loans around the world, we show that there is a strong connection between U.S. monetary policy and cross-border loan issuance. What is remarkable is that there is significant differential across markets with U.S. monetary policy disproportionately affecting EMEs. We estimate that, during a typical monetary easing cycle over which the Fed cuts its target rate by about 4 percentage points, the increase in loan volumes to emerging market borrowers exceeds the flow into developed markets by 32 percent. On the flip side, a monetary policy tightening would pull out bank flows from emerging markets and lead to a strong contraction of foreign credit in emerging markets. The granularity of the data allows us to control for borrower time-invariant characteristics as well as for bank-quarter level 20

22 effects in lending. Our results are robust to alternative measures of U.S. monetary policy, including unconventional measures. We are able to explore geographical differences in capital flow and get insight into the type of lending that is most exposed to the externally driven credit expansion. In particular, we show that one of the most affected investments in the economy is infrastructure. This type of financing is also heavily dependent on availability of foreign bank funding. 21

23 References Acharya, V., S. G. Cecchetti, J. De Gregorio, Ş. Kalemli-Özcan, P. R. Lane, and U. Panizza (2015), Corporate Debt in Emerging Economies: A Threat to Financial Stability?, The Brookings Institution and the Centre for International Governance Innovation. Ahmed, S. and A. Zlate (2014), Capital Flows to Emerging Market Economies: A Brave New World?," Journal of International Money and Finance 48(PB), Altunok, F., I. Gumus, T. Kapan, and S. Ongena (2016), The Effect of US Unconventional Monetary Policies on Bank Lending in Emerging Markets: Evidence from Turkey, mimeo. Bernanke, B. (2013), Monetary Policy and the Global Economy, Speech at the Suntory and Toyota International Center for Economics and Related Disciplines, London, 2013, March 25. Bräuning, F. and V. Ivashina (2016), Monetary Policy and Global Banking, mimeo. Bruno, V. and H. S. Shin (2015), Capital Flows and the Risk-Taking Channel of Monetary Policy," Journal of Monetary Economics 71(C), Ehlers, T. (2014), Understanding the Challenges for Infrastructure Finance, BIS Working Paper No Chinn, M. D. and H. Ito (2006), What Matters for Financial Development? Capital Controls, Institutions, and Interactions, Journal of Development Economics 81(1), Demirgüç-Kunt, A., B. Horvath, and H. Huzinga (2017), Foreign Banks and International Transmission of Monetary Policy, World Bank Policy Research Working Paper Forbes, K. J. and F. E. Warnock (2012), Capital Flow Waves: Surges, Stops, Flight, and Retrenchment, Journal of International Economics 88(2), Fratzscher, M. (2012), Capital Flows: Push versus Pull Factors and the Global Financial Crisis, Journal of International Economics 88(2), Gadanecz, B., and K. Von Kleist (2002), Do Syndicated Credits Anticipate BIS Consolidated Banking Data?, BIS Quarterly Review, March, 2002, Giannetti, M, and L. Laeven (2012), The Flight Home Effect: Evidence from the Syndicated Loan Market during Financial Crises, Journal of Financial Economics 104(1), Gopinath, G. (2016), The International Price System, Jackson Hole Symposium Proceedings, Kapur, M., and R. Mohan (2014), Monetary Policy Coordination and the Role of Central Banks, IMF Working Paper No McCauley, R., P. McGuire, and V. Sushko (2015), Global Dollar Credit: Links to US Monetary Policy and Leverage, BIS Working Paper No Miranda-Agrippino, S. and H. Rey (2015), "World Asset Markets and the Global Financial Cycle," NBER Working Papers

24 Morais, B., J. L. Peydró, and C. Ruiz (2015), "The International Bank Lending Channel of Monetary Policy Rates and QE: Credit Supply, Reach-for-Yield, and Real Effects," International Finance Discussion Papers Rajan, R. (2014), Competitive Monetary Easing: Is It Yesterday Once More?, Speech at the Brookings Institution, Washington DC, 2014, April 10. Rey, H. (2013), Dilemma Not Trilemma: the Global Cycle and Monetary Policy Independence, Proceedings - Economic Policy Symposium - Jackson Hole, Federal Reserve Bank of Kansas City, pages 1 2. Romer, C., and Romer, D. (2004), A New Measure of Monetary Shocks: Derivation and Implications, American Economic Review 94 (4), Shin, H. S. (2016), Global Liquidity and Procyclicality, Presentation at the World Bank Conference, The State of Economics, the State of the World, Washington DC, 2016, June 8. Shin, H. S. (2012), Global Banking Glut and Loan Risk Premium, IMF Economic Review 60(2), Stein, J. (2013), Overheating in Credit Markets: Origins, Measurement, and Policy Reponses, Speech at the Federal Reserve Bank of St. Louis, St. Louis, Missouri, 2013, February 7. Takats, E. (2010), Cross-Border Bank Lending to Emerging Market Economies, BIS Working Papers No. 54. World Bank (2016), PPI Investments in IDA Countries, 2011 to Wu. J and F. Xia (2016), Measuring the Macroeconomic Impact of Monetary Policy at the Zero Lower Bound, Journal of Money, Credit and Banking, 48(2-3), Yellen, J. (2011), Assessing Potential Financial Imbalances in an Era of Accommodative Monetary Policy, Speech at the Bank of Japan Conference on Real and Financial Linkage and Monetary Policy, 2011, June 2. 23

25 FIGURE I CROSS-BORDER LOANS TO EMERGING MARKETS AND U.S. MONETARY POLICY (a) Relationship in Levels Correlation= Cross-Border Loans to DMEs (% of GDP) Federal Funds Rate (%) (b) Relationship in Changes Correlation= Change in Cross-Border Loans to EMEs (pp) Change in Federal Funds Rate (pp) Note: This figure shows the relationship between cross-border loans to emerging market economies (EMEs) and U.S. monetary policy. Panel a) plots the annual cross-border loans (as a percent of GDP) against the annual U.S. federal funds rate (in percent). Panel b) plots the annual change in cross-border loans (normalized with lagged GDP) and the annual change in the U.S. federal funds rate (in pp). Both figures show the median values of 43 EMEs. Data on cross-border loans are compiled from the IMF International Investment Positions and cover the period from 1980 to

26 FIGURE II CROSS-BORDER CLAIMS ON EMERGING AND DEVELOPED MARKET COUNTRIES (a) Breakdown by Instrument Percent of All External Liabilities Emerging Markets Loans Portfolio Bond Investment Portfolio Equity Investment Developed Markets (b) Importance of Foreign Banks Cross-border Bank Claims (% of all Ext. Liab.) Emerging Markets Developed Markets Note: This figure shows the composition of cross-border claims on countries in emerging markets (EM) and developed markets (DM). Panel a) shows the amount of loans, portfolio bond and equity investment held by foreigners relative to the total amount of all external liabilities in Panel b) shows the share of external liabilities held by foreign banks. Both figures show the median values within each country group. Data on external liabilities are compiled from the IMF International Investment Position. Data on cross-border bank claims are compiled from the BIS Consolidated Banking Statistics, which has cross-border claims held by 24 banking sectors. The sample in both figures contains the same set of 29 DM countries and 43 EM countries. 25

27 FIGURE III LENDING TO EMES BY FOREIGN, DME BANKS VS. LOCAL BANKS Change in (Log) Total Loan Volumes by DMEs Banks R-squared=0.031; Slope= Change in (Log) Total Loan Volumes by Local Banks Note: Changes in the (logarithm of) syndicated loan volumes to EMEs borrowers by local EME lenders (horizontal scale) and foreign DME lenders (vertical scale). Lending volumes by foreign and local volumes are at the quarterly level, covering the period from 1990:Q1 to 2016:Q3. 26

28 FIGURE IV MATURITY STRUCTURE OF CORPORATE LOANS AND PROJECT FINANCING Corporate Loans Project Finance Density Mean = 4.34 Median = 4.68 Mean = Median = Maturity in Years Note: The figure is based on the sample of syndicated cross-border loans from 1990:Q1 through 2016:Q3. Project Finance loans follow the classification in DealScan. 27

29 TABLE I SUMMARY STATISTICS FOR CROSS-BORDER LOANS Geographic Region # Borrowers # Loans Currency Breakdown of Loans USD EUR Other Emerging Africa (incl. Middle East) 944 1,902 92% 5% 3% Emerging Asia and Pacific 3,955 7,618 87% 1% 12% Emerging Europe 1,259 3,379 76% 20% 4% Emerging Americas 1,431 2,661 97% 0% 3% Developed Countries 26,118 59,887 61% 24% 14% Note: The statistics are based on syndicated cross-border loans from 1990:Q1 through 2016:Q3. Country groups are based on the BIS classification. Offshore centers are excluded from the sample. 28