How Do US Global Banks Benefit the Real Economy?

Edith X. Liu* Cornell University

Jonathan Pogach FDIC 22th August 2014 (Preliminary) The views and opinions expressed herein are those of the authors and do not necessarily reflect the official positions of the Federal Deposit Insurance Corporation.

Abstract We examine the potential benefits of increased globalization in the US banking sector on the domestic economy. Using the regulatory filing of US banks’ foreign exposure, we evaluate the association between global banking activities on domestic business and consumer lending, as well as shareholder value. Surprisingly, we find that shareholders of US global banks do not necessarily realize higher Tobin’s Q than shareholders of purely domestic banks. Moreover, we find that while growth in US C&I lending is larger for global banks than non-global banks, the degree of global exposure among US global banks is not correlated with growth in domestic lending. We believe that these results are important to evaluate the channel through which potential gains from financial sector globalization are passed through to the real economy. Keywords: Global banking, Shareholder value, Domestic lending JEL classifications: *Dyson School of Applied Economics and Management, 310E Warren Hall, Ithaca, NY 14850 U.S.A., (607) 354-8308, [email protected]. We thank the participants of the FDIC seminar, Cornell Brown Bag,

and especially helpful conversations with Ricardo Correa,

Sally Davies, Linda Goldberg, and Jack Reidhill. All errors and omissions are our own.

1. Introduction The financial sector is more inter-connected than it has ever been in the past. This was clear during the recent financial crisis, both in the US and in Europe. The severity of the recent financial crisis has given rise to an extensive literature on financial sector stability and macro-prudential policies. While this line of research has highlighted the extensive nature of the risk involved in inter-connected banking systems, surprisingly little work has been done to understand how complex banking institutions may benefit the real economy. We attempt to address this gap in the literature by examining the benefits to the domestic real economy from internationally connected and complex banks, often termed as global banks. This paper explores a unique and confidential data set of US banks with international exposure. We take a US centric perspective and ask: How do US global banks provide benefits to the domestic real economy? To address this question, we explore two specific channels in which US global banks may provide value, to domestic shareholders and domestic commercial loan customers1. We posit that US banks may realize benefits from their international exposure in a variety of ways, both on the asset side and the liabilities side. On the liabilities side, globalized banking activities offer lower cost of capital and increased funding liquidity. As shown by Ceterelli and Goldberg (2012), the funding liquidity of global banks allows banks to be less sensitive to 1

Given the extensive documentation of US equity home bias, we assume here that US bank shareholders are

exclusively domestic agents.

2

domestic monetary shocks as well as lower their cost of capital. For the asset side, banks have greater opportunity to diversify their loan portfolio by lending to customers in foreign countries2. Moreover, these benefits may be time-varying and particularly valuable to the domestic economy when it is most needed. Given the various sources of the benefits from globalization, this paper aims to document stylized facts about how these gains are channeled into the real economy. In particular, we quantify how potential gains from global banking activities can be realized through increased shareholder value (Tobin’s Q) and increased lending offered to domestic customers (US C&I loan growth). We argue that how bank globalization benefits the real economy matters. For example, if the gains from globalization directly fund growth in domestic commercial lending, then global banks add value to the domestic economy through increasing real investments and growth. This effect can be particularly important when the credit availability of domestic banks is low. By contrast, when the benefits of global banking are realized by shareholders3, the impact on the real economy will be realized through shareholder wealth and consumption behavior of these agents. The latter is likely to translate into smaller real side benefits than the former.

2

Of course, we can also think of global banks providing additional product lines, which may provide benefits

of focusing rather than diversification. 3

Similarly, we can think of executive compensation, and other benefits that may be centralized in a small group

of individuals. For limited participation of US consumers in stock market see Vissing-Jørgensen and Attanasio 2003

3

The results of the paper can be summarized as follows. First, we find that after controlling for individual bank balance sheet differences, global banks realize significantly lower Tobin’s Q than domestic banks. Moreover, using the sample of only global banks, we find no significant relationship between Tobin’s Q and the degree of global exposure. Second, when comparing global banks with and without exposure to capital outflow control events, our results show that global banks affected by the exogenous negative shock to their global operations realized a sharper and larger decline in Tobin’s Q. In contrast, negative shocks to global exposure as represented by capital inflow restrictions do not carry the same impact on Tobin’s Q. Third, for domestic lending, we find that global banks provide significantly larger domestic C&I loan growth, as compared to non-global banks. Surprisingly, within the global bank sample, more international exposure is associated with significantly lower domestic C&I loan growth to total assets. Our paper contributes to a growing literature on global banks and the effects of international banks on financial stability, risk, and the aggregate economy. Papers related to this area of research are Ceterelli and Goldberg (2012), Correa, Goldberg, and Rice (2014), and Berger, El Ghoul, Guedhami, and Roman (2013). These papers highlight the potential differences between global banks and non-global banks, particularly related to risk taking and sensitivity to liquidity shocks. Our paper complements this strand of literature by asking how these differences across global and domestic banks are translated into shareholder wealth or

4

more domestic commercial lending. In a broader context, this paper is related to a small but growing literature on regulatory arbitrage and global financial regulation. Recent works of Houston, Lin, and Ma (2012) find that there is some evidence that bank flows and cross-border M&A activity tend to funnel funds from higher regulated environments to less stringent countries. Karolyi and Toboado (2014) examine the attributes of the target and acquirer in cross-border M&A deals to assess deals motivated by either escape from cumbersome regulations or harmful race to the bottom. Moreover, Ongena, Popov, and Udell (2012) find that stricter banking regulations and higher capital requirements tend to push banks to make riskier loans abroad. This paper extends this literature and evaluates the potential consequences of regulatory arbitrage on shareholder value and domestic lending. The paper is organized as follows. Section 2 presents the empirical specification, testable hypotheses. Section 3 provides a brief description of the data. Section 4 presents the main empirical results, and Section 5 concludes.

2. Empirical design To understand the effect of bank globalization and shareholder value4, we begin with a simple panel regression specification:

4

There is potentially a secondary or intermediate question of how globalization affects the risk taking behavior

of banks using metrics such as z-scores, volatility of ROA, earnings volatility, equity return volatility, etc. See Laeven and Levine (JFE 2009)

5


௜,௧ =  + ௜,௧ିଵ +    + ௜,௧

(1)

Where
௜,௧ is the firm value as measured by Tobin’s Q, ௜,௧ିଵ is a measure of global exposure, and Bank Controls include (1) revenue growth, (2) log ratio of bank’s liquid assets to total assets, (3) log asset size, and (4) ratio of non-performing loans to total assets5. We lag all the Bank Control variables by one quarter to avoid issues with simultaneity. We define global exposure in two ways: Dummy variable for global bank6 and Total Absolute Net Due7. We also include year dummies. Clearly, the simultaneity issue arises because the firm’s decision on its level of global exposure likely depends on such a decision’s anticipated effect on Tobin’s Q. We attempt to address this issue in later analysis using capital control events as exogenous negative shocks to a bank’s global exposure. Alternatively, we may be able to run an instrumental variables 2SLS model or Heckman Selection model, however, an instrument highly correlated with the decision to go global, but uncorrelated with Tobin’s Q, is needed for the first stage8. Given that very few banks are observed going from being purely domestic to being global within our sample, it is difficult to verify the validity of potential instruments.

5

Similar bank level controls are used in Ceterolli and Goldberg (JF 2012) and Laeven and Levin (JFE 2009).

6

Global bank is defined by any US bank that has ever filed a FFIEC 009. Identifying global bank as a US bank

filing the FFIEC 009 in that year yields similar results. 7

Net Due is a variable item on the FFIEC 009 that represents aggregate due from the foreign office of the

country to the remaining bank, which may include the US parent office or other foreign offices. 8

An alternative may be run a Heckman Selection model on the degree of global exposure. However, in this

case, the dependent variable becomes a continuous variable and is no longer a binary choice variable.

6

For changes in the lending due to global exposure, we: ௜,௧ = ∑ସ௝ୀଵ ௧௝ ௜,௧ି௝ + ௜,௧ିଵ +    + ௜,௧

(2)

Where ௜,௧ is the change in domestic C&I lending over lagged total assets, ௜,௧ିଵ is a measure of global exposure, and Controls include (1) log ratio of bank’s liquid assets to total assets, (2) banks capitalization ratio, (3) log asset size, and (4) value of non-performing loans. Alternatively, we can evaluate change in the ratio of domestic C&I to Total Assets, which reflects a decision to rebalance the bank’s loan portfolio either more toward domestic C&I as compared to alternative forms of lending, which includes foreign C&I lending. While we believe both measures are interesting to evaluate, the majority of our results will focus on the effects on domestic lending, and relegate the results on the portfolio decision to the Appendix. While our dataset contains an unbalanced panel of firm-quarter observations from 1986 – 2012, some of our variables are highly persistent at the quarterly frequency. Therefore, all empirical above specifications are run at the annual frequency with year-end data.

2.1 Testable hypotheses While globalization in principal should create shareholder value for the bank, there are also competing interests. In particular, having a more defuse global infrastructure of banks may create additional corporate governance opacity and make it easier for managers and insiders to expropriate funds at the expense of minority shareholders. Similar to the arguments presented

7

in Bailey and Liu (WP 2014), the internationalizing of banking activities may benefit shareholders or benefit insiders and managers, at the expense of lower shareholder value. Therefore, we begin with a simple null: H0(A): Global banking activities is irrelevant, after controlling for firm characteristics. There is no valuation difference between banks that have global exposure versus banks with only domestic activities. H1(A): Global banking activities provide shareholders with higher firm value, after controlling for firm characteristics. For domestic lending, if banks find cheaper funding abroad and lend in the US, then we would expect that increased international exposure would have a positive effect on domestic C&I loan growth. On the other hand, if US global banks use international exposure to increase lending opportunities, then we would expect a lower growth rate in domestic C&I lending as funds are invested abroad. Given that some global banks may use their global operations for funding and others for lending, we expect on average no difference in the growth rate of domestic C&I lending between global banks and non-global banks. H0(B): Global banking activities are irrelevant, after controlling for firm characteristics. There should be no difference in the growth rate of US C&I lending between banks that have global exposure versus banks with only domestic activities.

8

H1(B): International exposure of global banks provide higher growth rate of C&I lending, after controlling for firm characteristics. 3. Data We explore a confidential regulatory filing FFIEC 009 (CEX) that requires all US banks with $30 million or more in foreign claims on foreign residents to report their claim and liability exposure by country. While the survey has had several iterations of revisions and new items since its inception in 1977, we begin our study using the data from 1986. For the purposes of the regulatory filing, the bank may elect to report at the bank holding company (BHC) level or at the bank level. The majority of the US banks in the CEX report at the bank holding company level, however, a few bank holding companies have multiple bank subsidiaries reporting in the CEX. Since the foreign exposures do not easily aggregate to the bank holding company level, we leave the foreign exposures as separate observations but use financials of the same bank holding company. While these few observations are treated separately, we cluster our regression standard error at the bank holding company level. For individual bank financial information, we use publicly available and standard data sources including Y9C, Compustat, and CRSP9. We use the Y9C to determine bank specific balance sheet variables such as total assets, liquid assets, bank capitalization ratio, and

9

The majority of the financial information is from Y9C. However, we use Compustat/CRSP is to obtain the

data variables needed to compute Tobin’s Q. Therefore, Total Assets used in Tobin’s Q is from Compustat/CRSP, while Log Total Assets used to as a bank specific control for size is from Y9C.

9

non-performing loans. To complement bank balance sheet information, we use Compustat and CRSP to gain financial variables such as market value of equity. Any bank in the CEX Country Exposure data that does not have financial information in Compustat/CRSP is automatically dropped. Similarly, we exclude a significant number of banks under foreign bank holding companies that do not file Y9C10.

4. Empirical results and discussion 4.1 An overview of the data Table #1 provides a descriptive summary of the financial data for our sample of domestic and global US banks. We report separately the summary statistics for global banks versus non-global bank11. Row #1 of Table #1 show that our sample contains 29,217 non-missing bank-quarter Tobin’s Q observations for domestic banks (Non-CEX), and 5,895 bank-quarter observations for global banks (CEX). For the global banks in column 4 under CEX, the average Tobin’s Q is 1.0536, while for non-global banks in column 3, the average Tobin’s Q is 1.0560. While the average Tobin’s Q appears to be very close for the two bank samples, the maximum Tobin’s Q for non-global banks is much larger in magnitude than for the non-global banks.

10

Our sample is significantly smaller than global banks identified from Call Report data. In particular, our

sample includes only banks that meet the minimum requirement threshold for global exposure and is part of a US bank holding company. 11

The global bank dummy flags any BHC that has ever had exposure in the FFIEC 009. Alternatively, global

banks can be flagged for each year. Both definitions yield very similar empirical results.

10

These large Tobin’s Q partly reflects the merger premiums on target banks during the merger wave of the 1990’s (Calomiris and Kareceski 2000). Table #1 also summarizes the balance sheet variables of global versus non-global banks that will be used as bank specific controls in the later analysis. First, perhaps the most noticeable difference between global and non-global banks is the size of their asset positions. Global banks are much larger in asset size than non-global banks in our sample, with the average asset size of the global banks at $102 billion versus $2.47 billion for domestic banks. Moreover, the maximum asset size in our sample for global banks is $2.44 Trillion, while the maximum asset position for our non-global banks is $200 billion. Given the large size difference across the global and non-global sample, we trim the data and exclude the largest and smallest 5% of banks in our sample12. For the other balance sheet variables, our sample statistics indicate that domestic banks tend to have a higher non-performing loan to total asset ratio at 0.371%, as compared to 0.108% for global banks. The last row of Table #1 reports the average ratio of US domestic C&I lending over total asset for global and non-global banks. The average ratio of US C&I lending to total assets is 11.0% for domestic only banks and 15.3% for US global banks, which implies that on average global banks dedicate a larger portion of their assets to US C&I loans than do non-global banks.

12

Results using 10% or 15% cutoffs for the data trimming generally yield very similar results.

11

4.2 Global banks and Tobin’s q One potential channel in which global banks can bring additional value is to increase value for its shareholders. Given the significant home bias documented for the US equity markets, it can be argued that increases in shareholder value will primarily benefit the wealth of domestic US investors. In this section, we evaluate whether global banks provide higher firm value than non-global banks. We use Tobin’s Q as a measure of firm value, which is defined as (Market Value of Equity + Total Assets – Shareholder’s Equity)/Total Assets. Beyond the difference across US global banks and domestic banks, we use the sample of only global banks to also examine the effect on Tobin’s Q of differing degrees of global exposure. Table #2 reports the results for the pooled regression in Equation #1 of Tobin’s Q on a global bank dummy, while controlling for bank balance sheet differences. While the data is observed at the quarterly frequency, we find that quarterly Tobin’s Q for banks is highly persistent at the quarterly level. Therefore, we use annual year-end data on each bank and we cluster standard error by bank. Column 2 in Table #2 reports the unconditional difference between Tobin’s Q of global banks versus non-global banks with year fixed effects. We find that without any bank specific controls the Tobin’s Q for global banks is significantly higher than non-global banks. However, in subsequent specifications with bank balance sheet controls, this result is overturned.

12

Column 3 through column 6 of Table #2 details the pooled regression results for Tobin’s Q and the global bank dummy, with controls for bank balance sheet differences. After controlling for bank specific performance and financials, we find that the Tobin’s Q for global bank is lower than non-global banks, with marginal statistical significance. In particular, in specification #6 of Table #2, we find that global banks have a Tobin’s Q that is -0.0237 lower than the domestic banks, which is significant at the 10% level. To evaluate the economic magnitude of these figure. The average Tobin’s Q is around 1.05, which means if we take the average coefficient across specifications (3) – (6), difference between global banks and non-global banks is -0.024, which represents a 2.5% reduction in shareholder value. For the average market capitalization in our sample is $2.178 billion13, a 2.4% loss in firm value would be approximately $52.3 million. Table #2 also reveals that balance sheet and firm performance metrics, such as asset size and the ratio of non-performing assets to total assets, are very significant to the Tobin’s Q of a bank. Since current non-performing assets are likely to become future defaulting loans, it is perhaps not surprising that the ratio of non-performing assets to total asset is negatively related with Tobin’s Q. Similarly, log asset size is significantly positive for Tobin’s Q. Again, this result is intuitive, as large banks can often enjoy significant economies of scale within and across different banking products and business lines. 13

The median bank in our sample is substantially smaller at $151 million. Using the median bank, the loss of

2.4% to shareholder value would be approximately $3.6 million.

13

Table #3 shows the relationship between Tobin’s Q and a bank’s ratio of total absolute net due to total asset. Using only the global bank sample, we evaluate the effect of differing levels of foreign exposure on Tobin’s Q. We compute Absolute Total Net Due as the sum of every country’s net due, but since positions can be negative or positive, we first take the absolute value before totaling14. The specifications for Absolute Total Net Due mimic the ordering in Table #2. Table #3 reports that, conditional on being a global firm, a higher global exposure is related to higher Tobin’s Q, but the relationship is not statistically significant. Interestingly, for the sample of only global firms, regression specifications (3) to (6) in Table #2 shows a statistically negative coefficient on log total assets. The negative relationship between size and Tobin’s Q suggests that larger global banks have significantly lower Tobin’s Q than smaller global banks. This stands in contrast to the above pooled regression results with both global and non-global banks in Table #2, which revealed a positive and significant relationship between bank size and Tobin’s Q. Since domestic banks outnumber global banks within our sample, the positive and statistically significant relationship in Table #2 is driven by the economies of scale enjoyed by increasing size in domestic banks. By contrast, the negative and significant result in Table #3, for global banks, suggests that global operations may not

14

Recall that Net Due for each country is the amount due to the foreign office from the rest of the bank, or all

other offices including the US. By taking the absolute amount, we are double counting positions where any internal borrowing between country locations. However, a larger international internal capital market may be viewed as a higher degree of internationalization.

14

enjoy the same economies of scale, and in fact, shareholders may value substantial international expansions negatively. Overall, this section shows that after controlling for bank specific characteristics, domestic banks surprisingly offer higher shareholder value than global banks. Moreover, for the global bank sample, increased international exposure has an insignificant effect on Tobin’s Q. Therefore, our pooled regression results do not appear to support the hypothesis that benefits from globalization in the banking sector is passed onto US investors by increasing firm value. 4.3 Selection Bias, Tobin’s q and Capital Control Events The pooled regressions above suggest significantly lower Tobin’s Q for global banks, as compared with non-global banks, and no significant difference between global banks with varying degrees of international exposure. However, these specifications do not account for the potential endogenous bank decision to become global or change their international exposure. To find an instrument that creates exogenous variation in the decision to switch from a domestic bank to a global bank, and yet uncorrelated with shareholder value, is difficult given the few number of banks in our sample that changed from purely domestic to global. Therefore, this section explores the sample of global banks and exploits capital control events as an exogenous shock to the bank’s degree of international exposure. Since capital control decisions at the country level, and not by the banks, the limiting of capital flows represents an exogenous

15

negative shock to the bank’s global exposure to both the affected country and in general. Given that some global banks are exposed to the capital control country, and others are not, this difference provides a way to measure the impact of an unexpected negative change in global exposure on Tobin’s Q. We separately analyze the effect of inflow versus outflow capital control events. We use capital control events, identified in Magud, Reinheart, Rogoff (2011), as exogenous shocks to the bank’s global exposure. For each bank, we use their one year prior to the event country exposure in the CEX data to classify them as either exposed or not exposed to the capital control event. For each bank, we measure the pre and post event Tobin’s Q. We then compute the average change in Tobin’s Q for the group of global banks with exposure to the event country and without exposure15. Finally, we aggregate the difference in Tobin’s Q for exposed versus non-exposed banks across all capital control events with weights determined by the total number of global banks observed for each event period. Figure #1(a) plots the average Tobin’s Q for exposed versus non-exposed banks, in the three quarters before and after the capital outflow control event, while Figure #1 (b) depicts the difference in Tobin’s Q between the two groups. Zero on the date x-axis marks the quarter in which the event occurred. Figures #1 (a) and #1 (b) clearly show that those global banks with exposure to the capital outflow control shock had lower Tobin’s Q than those global banks 15

This specification equally weights each bank for the capital control event. Alternatively, we evaluate

size-weighted (by assets) change in Tobin’s Q, which yield similar results.

16

without ex-ante exposure. This result supports the intuition that a shock that reduces the global exposure of a bank has an adverse effect on a bank’s Tobin’s Q. This suggests that Tobin’s Q, potentially through bank profitability, is negatively impacted when capital outflow control shocks limit the bank’s ability to reap benefits of being international. Interestingly, the deviation in Tobin’s Q between global banks with and without exposure to the capital outflow event begins three quarters in advance of the actual event. This phenomenon may represent either the ability to anticipate the event or the coarseness of the dating the exact quarter of the event. In addition, we find that the effect on Tobin’s Q persists even after six quarters. These figures suggest that changes in global exposure can have direct persistent effect on Tobin’s Q. In stark contrast to the figures for capital outflow events, Figures #2 (a) and (b) suggests that capital inflow control events do not have the same impact on Tobin’s Q. In particular, the Tobin’s Q for both exposed and non-exposed banks rose throughout the event window when capital inflow controls were imposed. While the slight declining slope in Figure #2 (b) shows that Tobin’s Q of exposed banks rose less than non-exposed banks during the event window, the effect on Tobin’s Q arising from the capital inflow event appears small in comparison to the capital outflow events in Figure #1(b).

17

Overall, we find that a negative shock to global exposure from capital outflow event appears generate the predicted negative to impact Tobin’s Q, while capital inflow events have less of an effect.

4.4 Time varying effect of global banking on Tobin’s q The capital control events analysis above suggests that there are potential important differences in the way capital outflow versus inflow events might affect the firm value of the bank. In addition, there is also a difference in the time period in which these events occur. Moreover, capital outflow, or inflow, events tend to cluster around a group of countries and within a few years. To better understand the potential time variation in the effect of global exposure on global banks and non-global banks, we run regression specification (1) for each quarter from Jan 1990 to June 2008. We modify our sample end date to exclude the recent financial crisis. Figure #3 (a) shows the time variation in the coefficient on the global bank dummy for specification #1, including all individual bank controls. The graph reveals that, after controlling for bank specific balance sheet differences, Tobin’s Q post 2001 has been consistently lower for global banks than non-global banks. Only within the period of Jan 1996 and Dec 1998, and the four quarters of 2000, was Tobin’s Q higher for global banks, as compared to non-global banks. Moreover, large increases in the difference between the

18

Tobin’s Q of global banks and non-global bank’s occurred in the periods of 1995Q4, 1996Q1, and 1999Q4, while sharp declines occurred in 1999Q1 and 2001Q1. Similarly, Figure #3(b) plots the coefficient on the degree of global exposure, as measured by absolute value of the firm’s total net due across all countries. We use only global banks in this figure to understand how global exposure within the set of global banks might be time varying in the way it affect Tobin’s Q. Figure #3 (b) again reveals substantial time variation in the sensitivity of Tobin’s Q to degrees of global exposure. For the sample of only global firms, we find that the positive relationship between Tobin’s Q and the measure of global exposure is confined mostly in the period of 1999Q4 to 2005Q1. The quarters, before and after this period, show a slightly negative relationship between Tobin’s Q and global exposure.

4.5 Domestic US Lending and Global Banks Beyond shareholder value, global banks may channel the benefits of internationalization to the US economy through increased domestic lending, which may be particularly valuable when credit availability is low. This section evaluates the extent to which global banks are able to provide domestic commercial and industrial loans (US C&I) to finance US investment and growth. While all forms of domestic lending are important for the economy16, we focus on the domestic C&I lending to understand the lending channel that directly affects business 16

Alternative sources of domestic lending may include commercial and residential real estate, consumer credit,

and other loans.

19

investments and economic growth. Moreover, rather than evaluating the level of C&I lending, which may reflect bank specific business model that targets a different loan customer base, we focus on the sensitivity of domestic C&I growth, scaled by total assets, to the degree of international exposure. Table #4 reports the pooled regression results for the average growth in domestic C&I lending on the global bank dummy and individual bank characteristics17. We find that on average, global banks provide a significantly larger annual growth in domestic C&I lending, as a percent of total assets, than non-global banks. To evaluate the economic magnitude of this difference in growth rates, the average growth rate of C&I over total assets in our sample is 0.007918. From the pooled regression results, the average difference in domestic C&I growth, as a percent of total assets, between global banks and non-global banks is approximately 0.0073 across the four specifications that include individual bank balance sheet controls. Therefore, the average growth rate of domestic C&I loans to total assets of a global bank, is double that of a non-global bank. Beyond the difference between global versus non-global banks, we also evaluate how the degree of global exposure may affect domestic C&I growth within the global bank sample. Table #5 reports the sensitivity of average domestic C&I growth, relative to total assets, to the 17

In addition to the bank specific controls included in the Tobin Q regressions, bank capitalization ratio

(equity/total assets) is also included. 18

Again, the median bank looks somewhat different than the average statistic, and has an average growth in

C&I over total asset of 0.0059.

20

degree of international exposure for only global banks, as proxied by Absolute Total Net Due. Surprisingly, for global banks, we find that increased international exposure actually imply lower growth in US C&I lending. For the economic magnitude of this effect, we average the coefficients on Absolute Net Due for specifications (2) to (6) and get an average effect of -73.02. The average absolute net due to total assets in our sample of .000150219, implying that the overall effect on the ratio of the growth rate of domestic C&I lending to total assets is -0.0109. Given that the average ratio of domestic C&I to total assets is 0.0079, more globally exposed banks appears to have economically meaningful lower growth rates in domestic lending. This section shows that although global banks appear to have significantly larger domestic C&I loan growth than non-global banks, within the sample of global banks more international exposure may not flow funds to additional domestic C&I lending. One potential hypothesis could be that global banks benefit from international funding channels to supply domestic lending, relative to non-global banks. However, additional international exposure allows global banks to capture foreign lending opportunities, which diverts funds away from domestic lending.

5. Summary and conclusions

19

The median global bank’s Abs Net Due is .0000145

21

This paper evaluates the potential benefits of US global banks to the domestic economy by focusing on two distinct channels: higher firm value and larger domestic C&I loan growth. We find that for firm value, as measured by Tobin’s Q, global banks deliver significantly lower value to shareholders than do domestic banks. Moreover, within the global bank sample, increased global exposure is statistically insignificant for Tobin’s Q. We also find an interesting relationship between size and Tobin’s Q. For the full sample of banks, with a majority of domestic banks, size is positively related to Tobin’s Q, and highly statistically significant. By contrast, for the sample of global banks, we find that size is negatively associated with Tobin’s Q, and also highly statistically significant. The evidence appears to suggest that while domestic banking activities create value through economies of scale, shareholders do not view and value global expansion in the same way. In addition, we use foreign capital control events as a way to evaluate the impact of an exogenous negative shock to a bank’s international exposure on Tobin’s Q. We find that global banks with exposure to the capital outflow control event have a larger negative decline in Tobin’s Q than global banks without exposure. Surprisingly, for capital inflow control events, we do not find the same adverse impact to Tobin’s Q. Finally for domestic C&I lending, our results suggest that global banks provide statistically larger and positive loan growth relative to non-global banks. However, within the

22

sample of global banks, we find that larger foreign exposure is negatively related to loan growth. These results provide some preliminary evidence to guide further research. First, globalization for banks may not provide as much benefit to shareholders and investors as for industrials or other multi-national corporations. Second, while international exposure may allow banks to achieve lower cost fund for domestic lending, the highly international banks are using their foreign exposure to lend abroad at the expense of slower domestic C&I growth.

23

References Berger, A. N., Saunders, A., Scalise, J. M., & Udell, G. F. (1998). The effects of bank mergers and acquisitions on small business lending. Journal of financial Economics, 50(2), 187-229. Berger, A. N., El Ghoul, S., Guedhami, O., & Roman R. (2014). Bank Internationalization and Risk Taking. Working Paper. Calomiris, C., & Karceski, J. (2000). Is the bank merger wave of the 1990s efficient? Lessons from nine case studies. In Mergers and productivity (pp. 93-178). University of Chicago Press. Cetorelli, N. and Goldberg, L., (2013), Bank Globalization and Monetary Transmission, Journal of Finance. Cetorelli, N. and Goldberg, L., (2013), Liquidity Management of U.S. Global Banks: Internal Capital Markets in the Great Recession, Journal of International Economics. Correa, R., Goldberg, L., & Rice, T. (2014). Liquidity Risk and US Bank Lending at Home and Abroad. Manuscript Federal Reserve Bank of New York and Board of Governors of the Federal Reserve System. Karolyi, G. Andrew and Toboado, A., (2014), Regulatory Arbitrage and Cross-Border bank Acquisition, Working Paper. Magud, N. E., Reinhart, C. M., & Rogoff, K. S. (2011). Capital controls: Myth and reality-A portfolio balance approach (No. w16805). National Bureau of Economic Research. Ongena, S., Popov, A., & Udell, G. F. (2013). “When the cat's away the mice will play”: Does regulation

at

home

affect

bank

risk-taking

abroad?. Journal

of

Financial

Economics, 108(3), 727-750. Sapienza, Paola. (2002), The Effects of Banking Mergers on Loan Contracts. Journal of Finance Vol 57, Issue 1, pg 329 – 367

24

Vissing-Jørgensen, A., & Attanasio, O. P. (2003). Stock-market participation, intertemporal substitution, and risk-aversion. American Economic Review, 383-391.

25

Table #1

TobinQ Total Assets Bank Capitalization (Tier 1) Liquid Assets to Tot Assets Non Perform to Tot Assets SaleGrQoQ US C&I to Tot Assets

Num of Obs Non-CEX CEX 29217 5895 34081 6001 24255 2925 34058 6000 30392 4403 25328 5481 34077 6001

Average Minimum Maximum Non-CEX CEX Non-CEX CEX Non-CEX CEX 1.0560 1.0536 0.6645 0.8765 8.4725 2.2691 2477037 102000000 31612 202807 200000000 2440000000 9.4326 7.9328 -8.5400 0.6500 121.8300 20.4600 0.4107 0.3853 0.0125 0.0546 1.8629 1.0294 0.0088 0.0085 0.0000 0.0000 0.3710 0.1084 0.0963 0.1046 -57.4370 -3.5567 94.0620 8.2663 0.1100 0.1533 0.0000 0.0000 0.5761 0.5786

Table #2 TobinQ flagCEX2 t-stat

(1)

(2)

(3)

(4)

0.0071

0.0197

-0.0255

-0.0244

0.87

2.47

-1.94

-1.85

-1.66

-1.76

0.0136

0.0137

0.0134

0.0140

4.53

4.24

3.94

4.08

0.0541

0.0518

0.89

0.91

Log Total Assets t-stat Liquid to Total Assets t-stat NonPerforming to Total Assets

-1.2679

t-stat

-5.05

(5)

(6)

-0.0221 -0.0237

-1.1151 -1.1042 -5.26

Sales Growth (YoY)

-4.87 0.0013

t-stat Constant

0.9 1.06027

0.9815

0.7630

0.8541

0.8314

0.8039

184.87

83.48

15.45

19.51

20

18.4

Obs (Firm-Year)

7740

7740

7740

7574

7574

6703

R-Squared

0.000

0.066

0.075

0.079

0.083

0.089

No

Yes

Yes

Yes

Yes

Yes

1990

1990

1990

1990

1990

1992

t-stat

Year Dummies Base Year

The data is trimmed by eliminating the largest and smallest 5% of banks by asset size for each year Regressions also include Year Dummies and standard errors are clustered by bank

26

Table #3 TobinQ Absolute Net Due to Total Asset t-stat

(1)

(2)

(3)

(4)

(5)

6.135

23.547

69.764

23.517

23.692

19.251

2.47

0.24

0.73

0.27

0.26

0.20

-0.0109

-0.0133

-0.0134

-0.0123

-2.16

-2.75

-2.62

-2.31

-0.0006

0.0033

-0.01

0.08

-2.4153

-2.4195

-2.2915

-4.00

-3.67

-3.54

Log Total Assets t-stat Liquid to Total Assets t-stat NonPerforming to Total Assets t-stat Sales Growth (YoY)

0.0083

t-stat Constant t-stat Obs (Firm-Year) R-Squared Year Dummies Base Year

(6)

0.80 1.0649

0.9891

1.1589

1.2683

1.2691

1.2612

129.75

63.44

14.69

14.02

12.24

11.20

529

529

529

488

488

469

0.000

0.463

0.484

0.507

0.507

0.513

No

Yes

Yes

Yes

Yes

Yes

1990

1990

1990

1990

1990

1990

The data is trimmed by eliminating the largest and smallest 5% of banks by asset size for each year Regressions standard errors are clustered by bank

27

Figure #1 (a)

Tobin's Q: Capital Outflow Control Events 1.13 1.12

Tobin's Q

1.11 1.1 1.09 1.08 1.07 1.06 -3

-2

-1

0

BanksNoExposureToEvent

1

2

3

BanksWithExposureToEvent

Figure #1 (b)

Diff Tobin's Q: Capital Control Outflow Events 0.005

0 -3

-2

-1

0

-0.005

-0.01

-0.015

-0.02

-0.025

28

1

2

3

Figure #2 (a)

Tobin's Q: Capital Inflow Control Events 1.07 1.06 1.05

Tobin's Q

1.04 1.03 1.02 1.01 1 0.99 0.98 -3

-2

-1

0

NoExposureBanks

1

2

3

ExposedBanks

Figure #2 (b)

Diff Tobin's Q: Capital Control Inflow Events 0 -3

-2

-1

0

-0.005 -0.01 -0.015 -0.02 -0.025 -0.03

29

1

2

3

-500

-1000

30

1000

500

0 Dec-04

Jan-04

Feb-03

Mar-02

Apr-01

May-00

Jun-99

Jul-98

Aug-97

Sep-96

Oct-95

Nov-94

Dec-93

Jan-93

Sep-07

Tobin's Q: Time Varying Sensitivity to Abs Net Due (no Crisis) Oct-06

Table #3 (b)

Sep-07

-0.08

Oct-06

-0.06 Nov-05

-0.04

Nov-05

Dec-04

Jan-04

Feb-03

Mar-02

Apr-01

May-00

Jun-99

Jul-98

Aug-97

Sep-96

Oct-95

Nov-94

Dec-93

Jan-93

1500 Feb-92

0.06

Feb-92

Mar-91

-0.02

Mar-91

Figure #3 (a)

Tobin's Q: Time Varying Coefficient on FlagCEX2 (no Crisis)

0.04

0.02

0

Table #5 (Δ Annual US C&I)/Lag Total Asset

(1)

(2)

(3)

(4)

(5)

flagCEX2

0.0035

0.0077

0.0074

0.0073

0.0078

1.82

2.71

2.59

2.56

2.76

-0.0011

-0.0011

-0.0011

-0.0012

-2.11

-1.94

-1.94

-2.16

-0.0122

-0.0123

-0.0137

-4.89

-4.94

-5.33

-0.6030

-0.6088

-0.6261

-7.88

-7.90

-7.44

-0.0116

-0.0076

-1.25

-0.93

t-stat Lag Log Total Assets t-stat Lag Liquid to Total Assets t-stat Lag NonPerforming to Total Assets t-stat Lag Equity Cap Ratio t-stat Sales Growth (YoY)

0.0012

t-stat Constant

1.49 0.0126

0.0296

0.0402

0.0416

0.0357

0.040

3.25

4.78

4.93

4.51

Obs (Firm-Year)

7686

7289

6970

6970

6578

R-Squared

0.077

0.080

0.094

0.094

0.095

Base Year

1990

1991

1991

1991

1991

t-stat

The data is trimmed by eliminating the largest and smallest 5% of banks for each year Regressions also include Year Dummies and standard errors are clustered by bank

31

Table #6 (Δ Annual US C&I)/Lag Total Asset

(1)

(2)

(3)

(4)

(5)

Absolute Net Due to Total Asset

-69.42

-72.40

-77.14

-72.06

-70.48

-3.49

-3.47

-2.89

-2.81

-2.77

-0.0001

-0.0015

-0.0008

-0.0015

-0.03

-0.90

-0.46

-1.03

-0.0235

-0.0149

-0.0161

-2.64

-1.53

-1.76

-0.8215

-0.8234

-0.7357

-4.76

-4.2

-3.96

0.3059

0.2971

2.35

2.34

t-stat Lag Log Total Assets t-stat Lag Liquid to Total Assets t-stat Lag NonPerforming to Total Assets t-stat Lag Equity Cap Ratio t-stat Sales Growth (YoY)

0.0433

t-stat Constant

2.65 0.0390

-0.0189

0.0839

0.0406

0.0582

2.37

-0.59

2.72

1.30

2.18

524

469

438

438

438

R-Squared

0.2427

0.2553

0.2653

0.2773

0.3423

Base Year

1990

1991

1991

1991

1991

t-stat Obs (Firm-Year)

The data is trimmed by eliminating the largest and smallest 5% by assets of banks for each year Regressions also include Year Dummies and standard errors are clustered by bank

32