Cross hedging in Bank Holding Companies

Cross hedging in Bank Holding Companies Congyu Liu 1 This draft: January 2017 First draft: January 2017 Abstract This paper studies interest rate risk management within banking holding companies, and finds cross hedging within the groups: subsidiaries that are capable in risk management not only manage interest rate risk for themselves, but also for other banks in the same parent company. Given everything equal, I find big subsidiaries in bank holding companies hold larger net interest rate swap position for hedging than standalone banks per maturity gap; standalone banks hold more than small subsidiaries do. Moreover, hedging positions of large subsidiaries significantly responds to the aggregate maturity gap of their banking holding companies. Cross hedging level depends on parents companies net worth, capital structure, risk management governance framework and access to external capital. Banking holding companies with higher level of cross hedging have lower volatilities in stock prices and operating income. The paper extends internal capital market literature to risk management redistribution in the internal market, as well as provide a new explanation for the variation in the degree of bank hedging. Keywords: Bank holding company; Internal capital markets; Risk management; Derivatives trading JEL classification: G21 1 Congyu Liu is a PhD student in the Finance Department, Drexel University. Email: cl874@drexel.edu. 1

I. Introduction Derivatives are very important tools for banks to hedge risk. Over 90% of relatively large commercial banks (total assets over $3 billion) 2 trade derivatives for hedging purposes (Table 1). Despite of having exposure to interest rate risk, interestingly, not every bank trades derivatives for hedging purposes. Even in the same parent company, some subsidiaries trade while others do not; the differences in the level of trading per interest rate risk exposed between subsidiaries are also large. Literature about banks interest rate risk management is limited (Vuillemey 2016 WP a). Rampini et al. (2016 WP) documents that net worth plays a crucial role in determine banks hedging level. Since both financing with new debt and risk management requires collateral, financially constraint banks hedge less, to save equity as collateral to retain debt capacity. Vuillemey (2016 WP b) models the capital structure of a bank, and find that differences are consistent with optimal hedging under financial friction. This paper contributes a new explanation for the differences in hedging level among banks. I study the variations within banking holding companies, and explain the difference between subsidiary banks within same group. In particular, I investigate whether bank holding companies efficiently manage interest rate risk of all their subsidiaries as an integral group. More specifically, whether risk management for the entire group rests predominantly on a few subsidiaries that are relatively capable in interest rate management, especially in hedging with derivatives. Do these subsidiaries also cross hedge for other subsidiaries under the same parent banking holding company who are weak in risk management? I define cross hedging efficiency as the difference in the hedging level of subsidiaries in the same group. Large difference indicates higher cross hedging efficiency. What 2 Following the size classification of Berger and Bouwman (2009) 2

are the factors that drive the variation of the efficiency among banking holding companies, such as formation of a risk committee in the banking holding company level, capital structure, net worth (capital) sufficiency and frictions between internal and external market? Lastly, I study the benefits of cross hedging for parent companies and subsidiaries, for instance, lower volatility in operating income and stock prices. This paper also provides a new perspective to examine efficiency in the internal market----- risk management. From the previous internal capital market literature, we learned that parent banking holding companies allocate capital among their various subsidiaries for new investment opportunities through winner picking (Compello 2002), or provide a source of strength to small or weak subsidiaries through channels such as equity injection or deposit smoothing (Ashcraft 2007, Cremers et al. 2011, and many others). All of these paper have shown that parent companies work as central planners to maximize value of all their subsidiaries as a whole through capital reallocation to finance new investments. However, we do not know much about whether parent companies shift risk and risk management tasks among subsidiaries according to their own risk management abilities. This paper extends prior analysis of the function of internal capital market in the banking holding company setting by focusing on a risk management perspective. Lastly, it deepens our understanding about the source of strength that parent companies provide to subsidiaries. The source of strength is not just provided right before the bankruptcy, but also during normal times to decrease the likelihood of bankruptcy. Banking holding companies are similar to investment companies that hold banks in their investment portfolio, operation businesses are done independently at the subsidiary level, instead of at the parent level. Subsidiaries within same parent companies may operate in different regions, the volume of deposits and loans are subjected to local economy. Hence risk can be 3

managed at commercial banks level too, according to their own ability and needs. However, since parent company may oversees all the subsidiaries and risk management resources, as well as skills are unevenly distributed, some subsidiaries may help hedge other subsidiaries in the same group, who are weak in risk management. I use interest rate derivatives trading for hedging purpose as a proxy for interest rate risk management of commercial banks, and study whether some subsidiaries hedge interest rate risk for other subsidiary banks in the same group. I find that, given their own risk exposure, some subsidiary banks, who are more adept in risk management, manage risks for peer banks in addition to their own risk exposure (cross hedging); meanwhile some other subsidiary banks rely on other peer banks under same parent to manage interest rate risk. Given everything else equal, I observe that large subsidiaries have 5% higher net hedge ratio per maturity gap than standalone banks, and 8% higher than small subsidiaries. In addition, only hedging position of large subsidiaries are statistically significantly relate to aggregate interest rate exposure of affiliated BHC, and increase after new (relatively small) subsidiaries are acquired into the group. Thus, these results show that large subsidiaries hedge interest rate risk for other subsidiaries in the same group. The results are robust to propensity score matching between standalone banks, small and large subsidiaries. Some banks may have greater ability in cross hedging than other banks. I also study the determinants of cross hedging efficiency: net worth, access to external market and risk governance structure in the parent company. Net worth level of parent company and the difference in the level between subsidiaries may explain the cross hedging efficiency. Sufficient net worth (capital) work as collateral is a prerequisite for risk management, especially hedging through derivatives trading. Rampini and 4

Viswanathan (2010, 2013) present the dynamic trade-off between financing investment and risk management in their model: financially constrained firms need to allocate limited capital between the two. Rampini et al. (2014) and Rampini et al. (2016 WP) further support their arguments with empirical evidence in the financial stressed airline companies and commercial banks, that these firms hedge less when approaching distress, due to lack of collateral. I conjecture that parent companies with higher net worth will, in general, have more capital as collateral to support subsidiaries in risk management, especially to enable large banks to trade more derivatives for hedging purposes. Furthermore, large net worth differences between large and small subsidiaries within same group may also increase the cross hedging efficiency. Subsidiaries which have sufficient net worth are able to hedge for the ones are short of net worth to manage risk through derivatives trading. As a result, given similar net worth of small subsidiaries and standalone banks, small subsidiaries may have lower level in hedging with derivatives. (To be done) Frictions between internal market and external market may also contribute to the cross hedging efficiency within a conglomerate. In the BHC structure, parent company issue debt and equity and inject to subsidiaries. Cash and other cash equivalents are the most common collateral for derivatives trading. 3 If cross hedging exists in the banking holding company, subsidiaries that hedge for others may require higher level cash holding than the time when they only hedging for themselves. If the friction between internal market and external market is low, parent companies can issue short-term debt, such as commercial paper, or borrow from interbank market to provide cash as collateral for subsidiaries to cross hedging. In addition, subsidiary commercial banks may also raise short term funds through interbank market. Therefore low friction between 3 JPMorgan cash holding drop as a result of increased collateral need of derivatives trading. 5

internal and external capital market may facilitate cross hedging in banking conglomerates. (To be done) Banks with strong corporate governance in risk management may have higher cross hedging efficiency. Cross hedging requires overseeing ability from the parent companies. The efficiency is related to the information transparency from subsidiaries to parent, as well as the effective capital mobility among subsidiaries to support risk management activities, especially derivatives trading for hedging purpose. Information transparency serves as a prerequisite that the top managers are able to accurately understand subsidiaries risk exposures and advice the hedging subsidiaries take the appropriate positions for both themselves and other subsidiaries. Ellul and Yerramilli (2013) construct a risk management index (RMI) based on the risk committee governance features, such as role and the importance of Chief Risk Officer. They find BHCs with higher RMI have lower tail risk, and better operating and stock return from during financial crisis. Additionally, The BHC s Dodd-Frank enhanced risk management program required parent company to integrate subsidiary bank s risk governance frame work. Hence, I expect to find corporate governance structure, especially whether the presence risk management committee, will facilitate cross hedging in a banking holding company. (To be done) Regarding the benefits of cross hedging, banking holding companies that have high cross hedging efficiency may have stable operating income and lower stock price volatilities. As it is required by FDIC, parent companies serve as source of strength to support subsidiaries that are close to bankruptcy. Ashcraft (2004) documents equity injection from parent company to distressed subsidiaries. Subsidizing weak subsidiaries is costly to equity of parents companies. Cross hedging between subsidiaries may mitigate the costs of parent companies by decreasing the probability that weak subsidiaries go bankrupt as result of undermanaged interest rate risk. In addition, Ashcraft (2005) finds 6

that acting as source of strength may lead to failure of healthy banks. Hence, cross hedging may reduce the threat posed by distressed subsidiaries to the rest of banks in the parent companies. Therefore, public parent bank holding companies of high cross hedging efficiency may have lower stock volatilities. The paper proceeds as follows. Section 2 develops my main hypotheses. Section 3 describes data and construction of key variables measurements. Section 4 explains the research designs and empirical testing strategies. Section 5 shows main regression analysis. Section 6 presents empirical evidence of cross hedging and Section 7 discusses about the endogeneity. Lastly, section 8 concludes. II. Hypotheses Hypothesis 1: Parent company manages interest rate risk as a central planer for all the subsidiaries. More specifically, within a parent company, large subsidiaries, which is likely to be more capable in managing interest rate risk through derivatives trading, hedge interest rate risk for smaller banks. In the previous literature in the capital market, parent banks may subsidize subsidiaries through winner picking, where most profitable subsidiaries are funded, even they face same financial constraints with other subsidiaries. In the meantime, Scharfstein and Stein (2000) find that internal market may also allow Socialist, which is that stronger banks may cross subsidize weaker banks in the parent company. I propose cross subsidizing may also exist in the risk management within a parent company, but work in a lightly different way from capital transfer. Capital is limited and consumable resource, and can only be spent on certain subject exclusively. However, when a derivatives platform and a risk management team are built in certain banks, 7

derivatives can be traded not only for those banks themselves, but also for other subsidiaries. The initial equity required to trade derivatives may be allocated to these trading capable banks through the winner picking channel where capital is provided to subsidiaries who is better, or cost efficient, in trading derivatives for risk hedging. Therefore, to manage interest rate risk of all the subsidiaries within a parent company as a integrate group, internal market may combine both winner picking and Socialist channels in supporting derivatives trading for hedging purposes. However, not every banking holding company is able to do cross hedging. Among banking holding companies that are able to do so, some have higher cross hedging efficiency that others. Hence, I also study the factors that have effects on the abilities to cross hedge. Hypothesis 2: Bank holding companies that are better capitalized or access to external capital market have cross hedging efficiencies among their subsidiaries. As Rampini and Viswanathan (2010, 2013) have shown, collaterals are required for derivatives trading, regardless whether the trading is for hedging or not. Therefore, cross hedging by large subsidiaries to smaller ones requires higher net worth of those large subsidiaries. These large subsidiaries can obtain extra capital for cross hedging either from parent company or from external capital market, such as interbank lending market. If the parent company or the larger subsidiaries in the parent company are rich in capital, the larger banks will have enough capital as collateral to hedge for other subsidiaries in a more sufficient level; otherwise, the larger banks can only partially hedge for other subsidiaries according to their maturity gaps. In the same way, if banks have easy access to external capital market, they can also obtain funding from the external market to support their hedging activities. Therefore, if the large banks are rich in capital, regardless through their own net worth accumulation, from parent company or external capital market, I expect to see higher cross hedging efficiency for these subsidiary banks. 8

Hypotheses 3: Banking holding companies with a risk committee have higher cross hedging efficiency. Ellul and Yerramilli (2013) emphasizes the importance of risk committee in a banking holding company. They find that banks with higher risk management index have lower tail risk and better operating and stock return during crisis time. This paper extends the finding of Ellul and Yerramilli (2013), and document other role of risk committee at bank level cross hedging. According to OCC, every bank should develop its own risk governance framework. Unless the subsidiary bank has substantially similar risk profile as the parent company, the subsidiary bank cannot directly adopt the risk governance framework from the parent. Hence, risk management strategies are individualized and specific across the subsidiaries, for interest rate exposure varies among subsidiaries. Without a central planner at the parent company level, it will be hard for subsidiary bank A to precisely understand the risk exposures of other subsidiaries in the same parent and hedge accordingly, especially if other subsidiaries operate in different regions and have different focus of business. Therefore, with presence of a risk committee on the parent level, the risk committee can effectively collect and analyze the information from all subsidiaries, and advise subsidiary banks who are capable in derivatives trading for hedging trade for other subsidiaries. I find (to be found) banking holding companies with a risk committee have higher cross hedging efficiency. Moreover, the net income of banking holding companies with higher cross hedging efficiency less correlate with interest rate volatilities. 9

Hypothesis 4: Parent companies that have higher cross hedging efficiency experience lower volatilities in the stock returns and operation income. Benefits of cross hedging According to FDIC requirement, bank holding companies have the responsibility of cross guarantee, where when a subsidiary is close to bankrupt, the parent company is obligated to eject equity to save the subsidiary. Aschraft (2007) documents that health banks are forced to go bankrupt when they practice cross guarantee to help out with other subsidiaries. If cross hedging is allowed, weak subsidiaries can be rescued at an early stage, lowering the likelihood the bankrupts of multiple subsidiaries found in Aschraft (2002). Therefore, bank holding companies with higher level of cross hedging may have lower volatility in stock prices. III. Data: All my data come from call reports and Y-9c forms, which are provided on the website of Federal Reserve Bank of Chicago. Sample period ranges from 2001Q1 to 2013Q4. FAS 133, Accounting for Derivative Instruments and Hedging Activities, became effective for all quarters of all fiscal years beginning after June 15, 2000. The new accounting rules requires banks to accurately report derivatives trading amounts for both trading and non-trading purposes. Therefore, banks report regarding derivatives trading amount is more consistent after 2001. I include bank-quarter observations that are (1) belong to a bank holding company. Total interest rate derivatives trading for hedging purposes of the BHC is larger than zero; (2) standalone banks that have positive derivatives for hedging purposes (3) to rule out of the 10

possibilities that variation in the level of derivatives hedging results from different business lines, only commercial banks are included. In addition, I exclude observations that report (1) nonpositive total assets; (2) banks have less than 8 consecutive quarters in the sample 4 ; (3) zero total maturity gap. Till this step, my sample is left with 56,694 observations. Following Rampini et al. (2016 WP) and construct my key dependent variable net hedge ratio, the proxy for the level that a bank hedges its interest rate risk. The ratio is calculated as the difference between interest rate swap that banks agree to pay at a fixed rate (RCFD 1589) and a floating rate for hedging purposes, divided by total assets of the bank. Interest rate swap hedging as floating payer is calculated as the remaining of reported total interest rate swap trading for hedging purposes (RCFD 8725) after subtracting the trading portion as fixed payer (RCFD 1589). It is clear to see that net hedge ratio heavily relies on the value of interest rate swap traded as fixed payer for non-trading purposes. However, not all the banks report the value. Some report zero, even those banks who report a positive number of interest rate swap for hedging purposes. h = _ _ (1) = h (2) Concerning about the accuracy, I drop all the observations that had missing value in interest rate swap hedging purposes as fixed payer, or report zero in the fixed payer part while non-zero in interest rate swap for hedging purposes as total. The sample eventually has 9,807 observations. 84% of the observations are dropped out from the sample (Table 2 Panel A), however, the main 4 I follow Berger and Bowman (2009) to calculate the volatility of ROA with information from past 8 quarters. 11

results are also robust to including these observations in the sample. 5 In this paper, the ratio not only indicates the hedging positions of the commercial banks, as it is used in Rampini et al. (2016 WP) and Vuillemey (2016 WP b), but also reveals the level of hedging. That is, given unit interest rate risk exposure measured by the maturity gap, how much net interest rate swap banks hold to hedge the risks. Table 2 presents the sample comparison of groups who report zero and non-zero value for interest rate swap trading as fixed payer for non-trading purposes. Banks that report non zero values in the interest rate swap hedging as fixed payer are likely to be larger, on average remain one year longer during my sample periods. On balance sheet, these banks have higher percentage in Commercial and Industrial (C&I) loans, lower percentage in real estate loans, higher in consumer loan and lower in construction loans. Regarding liabilities, they rely less on core deposit, and relatively more on wholesale funding. In terms of equity, these banks have lower Tier 1 ratio, lower ROA and lower volatilities of ROA, except standalone banks. Both groups have similar percentage in their income sources, roughly 60% from interest income, and 40% from non-interest income. There are big variations of net hedge ratio among banks. In general, large subsidiaries, such as subsidiaries belong to a banking holding company is negative, and smaller in value than small subsidiaries. (Table 2 Panel F). The major control variable in the analysis is maturity gap. Following the literature, maturity gap is calculated as the difference between assets and liabilities that are going to mature in one year divided by total assets. =! "#$%! & ' (!#% )*+,#-.! "#$%! & ' (!#% )* '/#.! )* (3) 5 Among observations in the Mult_BHC, there are fewer observations that are in Bottom and medium subsidiaries compared to top subsidiaries. The reason is that many small subsidiaries do not report interest rate swap trading paying at fixed rate for hedging purpose. The main results are robust to including the full sample. 12

Banks also report assets and liabilities mature in 1-3 years, 3-5 years, 5-15 years and 15 years above. Table 2 Panel E shows the maturity gap distribution of all groups. Maturity gap within a year takes about 20% of total assets. The percentage decreases as the maturing years increase. Main analyses in this paper use maturity gap in one year. For robustness check, I also include maturity gap in other years. In general, the maturity gap is negative related to net hedge ratio. Positive maturity gap indicates that the bank holds more assets than liability that matures in the near term. That is, when short term interest rate increase in the near term, the bank will experience increase in the net interest income. Negative net hedge ratio means that the bank is a net pay-float interest rate swap trader for hedging purposes, and will profit from the position from the time when short term interest rate goes down. Therefore, a positive maturity gap will pair up with a negative net hedge ratio for interest rate hedging purposes; vice versa. Hence, large absolute value of correlation between the two suggested high sensitivity of net hedge ratio to maturity gap. In another words, if the relation between net hedge ratio and maturity gap is negative, the smaller the value, the more interest rate swap that a bank trades for hedging per maturity gap. My first hypothesis predicts that, because of cross hedging, large subsidiaries in a parent company have larger net hedge-maturity gap ratio sensitivity than standalone banks; small subsidiaries have the lowest sensitivity among the three groups. Table 3 shows the by group univariate analysis of relation between net hedge ratio and maturity gap (Standalone -0.16; Top -0.797; Bottom -0.094). The results are consistent with these predictions. IV. Research Design: Cross hedging does not have to only exist in the banking holding companies, but they provide perfect setting in studying the cross hedging problem. First of all, in a banking holding company, the parent works as an investment company that holds all the subsidiaries in its portfolio. From the previous literature in the internal market, we know parent companies play an 13

essential role in reallocating valuable resource, such as capital, among subsidiaries to seek for investment opportunities or subsidizing weak subsidiaries. Therefore, this structure also enables parent company to oversee risks from all types of subsidiaries, understand risk managing capabilities among subsidiaries, and reallocate risk management tasks to according to subsidiaries capabilities. Secondly, the nature of commercial banks business place banks exposed to interest rate risk. As bank taking liabilities through accepting deposits and adding assets via issue loans, banks are exposed to interest rate risk through the maturity differences between assets and liabilities. The larger difference in the dollar amount between assets and liabilities maturing in the same period, the larger interest rate risk the bank is exposed to. Thirdly, banks actively manage interest rate risk through derivatives trading for hedging purposes. Lastly, there are enough variations in derivatives trading for hedging cross parent banking companies and between commercial banks within banking holding companies. Many reasons determines the level of hedging, such as net worth and risk governance framework. Thus there will be enough variations across banking holding companies. As it is mentioned earlier too, not all the subsidiaries within same parent banking holding companies trade derivatives for hedging. Variations across subsidiaries will also be sufficient. Hence, the three features make banking holding companies a good setting to study cross hedging. The basic strategy in this paper to identify cross hedging is to compare interest rate swap trading between relative large and small subsidiaries, using standalone banks as reference. There are many reasons that enable large banks to trade more derivatives than smaller banks in general, regardless whether the bank belongs to a group. The level of hedging, net hedge sensitivity, is measured as the quotient of net hedge ratio over maturity gap, indicating dollar amount of net interest rate swap that per maturity gap that a bank has. The larger the quotient, the higher level 14

that a bank trade interest rate swap for hedging. Using standalone banks a reference level, I assume standalone banks trade derivatives for hedging at the right level based on their maturity gap. If cross hedging does exist, I shall see higher net hedge sensitivity of large subsidiaries than standalone banks, given everything else equal. For in addition to hedge for themselves, large subsidiaries also need to hedge for other smaller subsidiaries in the same group. For same logic, small subsidiaries will hedge less than standalone banks, as large subsidiaries will hedge for them.. V. Regression analysis: Below are the main regressions of my analysis in documenting cross hedging. Baseline: 0 1,3, = 4,3, +6 7,3, +8,3, 9,3, +:,3, (4) With interaction terms: 0 1,3, = 4,3, +6 7,3,,3, +6 ; <809,3, +6 = 809,3, +6 > 809,3, +6 @ 809 8A,3, +8,3, 9,3, +:,3, (5) In the baseline regression, we focus on 6 7, which measures the net hedge sensitivity. It shows the net interest rate swap position for hedging purposes that the a bank holds per maturity gap. As it is explained in the Data section, 6 7 is expected to be negative, if banks are really hedging their fixed income. The smaller (the more negative) the value, or the higher the absolute value, the higher net hedge sensitivity. If cross hedging really exists, according to my basic 15

strategy, I expect to see 6 7 from equation (4) should be smallest for relative large subsidiaries, medium for standalone banks, largest for relative small subsidiaries in a parent banking holding company, when regressions are ran separately by group. The above by group regressions can be integrated into one regression with interaction terms (equation 5). OBHC is a dummy variable that equals to 1 if the commercial bank belongs to a banking holding company that holds only one commercial bank. MBHC is a dummy variable that equals to 1 if the commercial bank belongs to a banking holding companies that hold multiple commercial banks. TOP is a dummy variable that equals to 1 if the commercial bank is one of the largest three banks in a banking holding company. Bottom equals to1 if the commercial bank belongs to the smallest one third subsidiaries in the group. The omitted category in the regression is the standalone banks. Thus 6 7 in the equation (5) represents the net hedge sensitivity for standalone banks. Similar with prediction in the equation (4), 6 7 in the equation (5) shall be negative, which is the basic relation between net hedge ratio and maturity gap for hedging position. 6 7 +6 ; in the equation (5) shows the net hedge sensitivities of subsidiaries belongs to One-banking holding companies. Since the hypothesis of cross hedging relies on the setting of multiple subsidiaries in a group, and subsidiaries in an One-BHC do not have other peer bank to help hedge with, I don t expect One-BHC subsidiaries to hedge differently from standalone banks. Hence, I predict 6 ; will not significantly different from zero. On average, I expect subsidiaries belong to a multi-banking holding company have higher net hedging position than standalone banks. Rampini et al. (2011 and 2013) show that net worth plays a critical role in determine the level of hedging, as derivatives hedging requires sufficient capital as collateral. Banking holding companies are likely richer in capital than standalone 16

banks, which enables them to support higher level of hedging. In equation (5), among the dummies for multi-bhc, subsidiaries of medium size are omitted. Hence 6 7 +6 = represents the net hedge sensitivities for the group. If subsidiaries in the banking holding companies hedge more than standalone banks as I expected, I shall see 6 = < 0, Hypothesis 1 predicts that large subsidiaries hedge more than standalone banks, which in turn hedge more than small subsidiaries. Therefore, we expect the net hedge sensitivity of large subsidiaries (6 7 +6 = +6 > ) to be smaller (more negative) than standalone banks; 6 > shall be statistically negative. In a similar way, 6 7 +6 = +6 @ represents the net hedge ratio of small subsidiaries. If as hypothesis 1 predicts, small subsidiaries that belong to multi-banking holding companies receive help from large subsidiaries in hedging, hence have lower net hedging position given same maturity gap than standalone banks. Then 6 @ shall be positive, so that 6 7 < 6 7 +6 = +6 @ < 0. Large banks tend to have higher level of derivatives trading. I use log form of total assets log (Total Assets) to control for size. Rampini et al. (2011, 2013) stress the importance of net worth in determine the level of hedging activities. I follow Rampini et al. (2016 WP) and use net income and tier 1 capital ratio as a proxies for the net worth in the control variables. Owing to the focus of the analysis is managing interest rate risk, in the alternative regression, I limit income source to net interest income, and split it into interest income and interest expenses. High income volatilities also create incentive for banks to manage interest. Following Berger and Bowman (2009), I include volatilities of ROA, calculated from the past 8 quarters, in the analysis. In addition to the capital and risk management resources required, motivations for trading derivatives for hedging purposes may also relate to the banks asset structure. Trading interest rate derivatives for hedging is meant to mitigate the overall interest rate risk a bank exposed to in 17

the underlying lending and taking deposits business. Banks who have larger interest rate risk may trade more derivatives for hedging. Interest rate risk results from maturity gap. Various maturities of assets and liabilities contribute to the maturity gap. Therefore, I control for the asset structure--the percentage of all types of loans (C&I loans, real estate loans, consumer loans and construction loans) and deposits (demand deposits and time deposits) in the equation 5. Equation 5 is intended to document difference in net hedge sensitivities between standalone banks and subsidiaries. However, the differences may not be brought by the status of belonging to a banking holding company, instead, are driven by variations of other factors, such as size, interest rate exposure and asset structure, etc. In order to tease out the possibility, I use method of propensity score matching to match standalone banks (control group) to subsidiaries in a parent company (treated group), controlling for maturity gap, size, capital ratio, income volatility, asset structure and liability structure. I divided the treated group into group of subsidiaries that are one the largest three in a BHC (TOP), and group of subsidiaries that are one of smallest one third of subsidiaries in a BHC (BOTTOM). If belonging to a group contributes to the subsidiaries net hedging level, then I shall see that given similar propensity score, TOP subsidiaries hedge more than standalone banks; standalone banks hedge more than BOTTOM subsidiaries (Table 5). Similarly, I also match TOP subsidiaries to BOTTOM subsidiaries from different parent companies on the same set of variables. I expect to see the ranking of size in a group also helps determine subsidiaries hedging level: TOP subsidiaries hedge more than BOTTOM subsidiaries because of the responsibility of cross hedging. In addition, I use commercial banks M&A as identification strategy (Table 6). When a bank who do not trade, or trade little derivatives is acquired into a new group, maturity gap of the entire group is likely to increases. 6 Put in another way, if the large subsidiaries within the group carry the responsibility to hedge for other subsidiaries, when new banks join in, the large subsidiary will trade more interest rate 6 By the nature of banking business of maturity transformation, banks are usually have positive maturity gap, where they hold more assets than liability by the same maturity. 18

swap derivatives for hedging purposes, even though its own maturity gap didn t change much. Hence, the net hedge ratio-maturity gap sensitivity of those large subsidiaries will increase as a result of the acquisition in its group. VI. Empirical Results: This session displays the main results. Table 3 shows the results of baseline regression by groups. The analysis divides the whole sample into two major groups, standalone banks and banks belong to a banking holding company. Among subsidiaries, banks are further categorized into TOP, Medium and Bottom subsidiaries according to their size ranking within the group. In general, TOP subsidiaries are the largest three subsidiaries in the parent company; Bottom subsidiaries are the smallest one third subsidiaries in the group; medium banks are the rest. More specific categorization are shown in the Appendix 2. The coefficient of Maturity Gap shows how much net interest rate swap position that banks hold per Maturity Gap that the banks are exposed to. As data session explains, if banks are really hedging, negative relation is expected between net hedge ratio and maturity gap. The deeper in the negative value, the more the bank hold hedging positions against their interest rate exposure. Table 3 shows, among all categories, TOP subsidiaries have the lowest value, -79.7%, which means that 79.7% of the interest rate exposure is hedged. Medium banks and standalone banks followed the TOP, 27.6%, and 16.6% of their interest rate swap were hedged. The Bottom subsidiaries have the lowest hedging ratio. I also include other variables balance sheet information variables. The preliminary results are consistent with our main hypothesis that because of cross hedging, with standalone banks as reference, TOP subsidiaries hedge more than standalone banks, and standalone banks more than Bottom subsidiaries in a group. Table 4 integrates all the regression analyses in Table 3 into single regressions (Equation 5), using category dummy variables and interaction terms to differentiate groups. Regression analysis in session V explains the interpretations of the coefficients. The coefficient of Maturity 19

Gap represents the net hedge sensitivity of standalone banks. 16.9% of their interest rate exposure is hedged. The coefficient of OBHC shows the incremental interest rate swap hedged compared to standalone banks. In column (1) where robust standard errors are applied, subsidiaries in OBHC traded 2.8% more than standalone banks. When standard errors are clustered, the statistical significance is gone. Medium banks in banking holding company are not different from standalone banks. The key variables of interest in this paper is the coefficients of interaction terms between MBHC and TOP, and MBHC and Bottom. Coefficient of MBHC*TOP indicates that TOP subsidiaries in a banking group hedge 5.4% more than the medium banks in the group. As the Medium banks do not hedge significantly differently from the standalone banks, the TOP subsidiaries also hedge 5.4% more than the standalone banks. By the same analysis, Bottom banks hedge 3.8% less than the standalone banks. 7 These results so far are consistent with my hypothesis if cross hedging is present, that if large subsidiaries hedge for small subsidiaries in the same group, then given the same maturity gap, that large subsidiaries should have higher net interest rate swap per maturity gap than standalone and bottom subsidiaries. However, the level differences among subsidiaries are not sufficient to document cross hedging. Table 5 further supports cross hedging through documenting the strong negative relation between interest net hedge ratios of large subsidiaries with the total maturity gap of the affiliated parent company. The same results are not found in Bottom and Medium subsidiaries. Column (1) and (2) show the relation for large subsidiaries. 8 The flipped sign of coefficients of 7 The coefficient for MBHC * Bottom is significant at 11% level when standard errors are robust and clustered at bank level. 8 Maturity Gap of BHC is calculated as the total maturity gap of all subsidiaries divided by total assets of all the subsidiaries. Smaller subsidiaries have lower maturity gap ratio, thus the maturity gap of total parent company is smaller. Therefore, the absolute value of the coefficients are larger than table 3 &4. 20

Maturity Gap of large subsidiaries themselves is consistent with model prediction in Vuillemey (2016 WP b). He argues and finds that banks that face more persistent lending opportunities are more likely to hedge decreases in interest rate with derivatives, which leads to a positive net hedge ratio; while financial constraint banks are likely to hedge increase in interest rate, which results in a negative net hedge ratio. Compared to smaller subsidiaries, large subsidiaries usually geographically cover larger regions and have more persistent investment opportunities, and less stressed in financial constraint. In addition, on average, banks have positive maturity gap. Hence, large subsidiaries may have positive coefficients to its own maturity gap exposure, to secure its interest income from lending; while negative coefficients to the aggregate bank holding company exposure, to hedge against increasing financing costs. VII. Endogeneity Many factors can contribute to variations in net hedge ratio. I use the propensity score matching to control for all the possible factors that impact hedging level, including balance sheet information and net worth (all the control variables in the main regression analyses). I conduct several pairs of comparison. Table 6 shows the results. First, I compare TOP subsidiaries with standalone banks. Given similar size and capital structure, I find that TOP subsidiaries have 7.5% higher net hedge ratio per maturity gap exposure compared to standalone banks. Thus I conclude the role as large subsidiaries in a group make these banks have higher net hedge ratio. Second, I compare small subsidiaries to standalone banks, controlling for the same variables. As expected, given everything equal, relatively small subsidiaries in a group have a lower (3%) hedging ratio than standalone banks. Lastly, I compare large subsidiaries of one banking holding companies to subsidiaries with similar size and capital structure of other banking holding companies, I 21

continue to find large subsidiaries have higher net hedge ratio. Therefore, I argue that the relative size within a group contribute to the hedging level of subsidiary banks. Propensity score matching does not address the endogeneity issue. Hence, I use mergers & acquisition events to document the causality, that being a relative large subsidiary that makes a bank hedge more than it needs, and a relative small subsidiary hedge less. I predict that when cross hedging is present, if new (small or non-derivatives trading) banks are acquired into a parent bank holding company, the large subsidiaries are expected to have higher net hedge ratio. Because there are more interest rate risk exposure (from newly acquired banks) to be covered by large subsidiaries within the parent group. The results are consistent with the prediction (Table 7). I find that the relative large subsidiaries in a parent have 3.7% (4.1%) higher net higher ratio one year (two years) after new banks (at least one) are acquired into the parent. VIII. Expected Conclusions: This paper studies interest rate risk management within banking holding companies and documents cross hedging among subsidiaries. It extends literature regarding internal capital market about capital transfer among subsidiaries to risk management tasks transfer among subsidiaries. Derivatives trading is an effective way in hedging interest rate risk, but not all the subsidiaries trade derivatives. This paper provides a new perspective in explaining the differences in the hedging level among commercial banks, in addition to net worth (Rampini et al. 2011, 2013), and capital structure (Vuillemey 2016 b WP). This paper shows that parent companies manage aggregate risk born by all subsidiaries. Specifically, I find that large subsidiaries in a banking holding company, which is usually more capable at risk management, trade interest rate swap for hedging for other smaller subsidiaries in the same group. The findings are 22

robust to propensity score matching using hedging level of standalone banks as reference. Large subsidiaries in the banking company have higher net hedge sensitivity one year after new (small and nonderivatives trading) banks who do not trade derivatives or trade little were acquired into the banking holding company, as there are more interest rate exposure in the group after mergers. This paper defines cross hedging efficiency as the difference between net hedge ratios of large and small subsidiaries. The larger the difference, the higher the cross hedging efficiency. I expect to find parent banking holding companies with a risk committee have higher cross hedging efficiency, as it serves as central information collector and decrease information asymmetry between subsidiaries, so that large subsidiaries to hedge for small subsidiaries at a higher sufficient level. Moreover, consist with Rampini et al. (2011, 2013), I expect to find parent companies with higher net worth have higher cross hedging efficiency. Capital rich banking holding companies are able to support large subsidiaries through providing collateral not only hedge for themselves, but also for other smaller subsidiaries. Lastly, I expect to see public bank holding companies with high cross hedging efficiency with lower volatilities in stock price and operating income. 23

References: Ashcraft, Adam B., 2005, Are Banks Really Special? New Evidence from the FDIC-Induced Failure of Healthy Banks, The American Economic Review, Vol.95, No.5, 1712-1730. Campello, Murillo., 2002, Internal Capital Markets in Financial Conglomerates: Evidence from Small Bank Responses to Monetary Policy, The Journal of Finance, Vol. LVII, No.6, 2773-2805. Cremer, K.J.Martijn., Huang, Rocco., and Sautner, Zacharias, 2011, Internal Capital markets and corporate Politics in a Banking Group, The Review of Financial Studies Vol 24 No.2, 358-401. Cebenoyan, A.Sinan., Strahan, Phillip., 2004, Risk Management, capital structure and lending at banks, Journal of Banking & Finance, Vol 28 Issue 1, 19-43. Huston, Joel., James, Christopher, James., Marcus, David., 1997, Capital market frictions and the role of internal capital markets in banking, Journal of Financial Economics 46, 135-164 Lamont, Lamount, 1997, Cash Flow and Investment: Evidence from Internal Capital Markets, The Journal of Finance, Vol LII, No.1, 84-109. Rampini, Adriano A., Viswanathan, S., 2013, Collateral and Capital Structure, Journal of Financial Economics (109), 466-492. Rampini, Adriano A., Viswanathan, S., 2014, Dynamic Risk Management, Journal of Financial Economics(111), 271-298. Rampini, Adriano A., Viswanathan, S., Vuillemey, Guillaume., 2016,Risk Management In Financial Institutions, working paper. Stein, Jeremy C., 1997, Internal Capital Markets and the Competition for Corporate Resources, Journal of Finance, Vol. LII, No.1,112-113. Vuilemey, Guillaume., 2016 a, Interest Rate Risk in Banking: A Survey, working paper Vuilemey, Guillaume., 2016 b, Bank Interest Rate Risk management, working paper 24

Table 1 This is table shows the summary statistics in number of observations where commercial banks trade derivatives for hedging purposes. Sample is divided by the relation to a parent company. If a commercial bank is not related to any holding company, then the bank is a standalone; if it relates to a holding company while the parent company only holds one commercial bank, then the bank belongs to BHC- One ; if the related parent company holds multiple commercial banks, then the bank belongs to BHC- Multiple. Column 1 shows the number of commercial bank observations from call reports. Column 2 shows number of observations that trade derivatives for non-trading purposes. The third column shows the percentage that observations has derivatives trading for non-trading purposes. Time period is from 2001Q1 to 2013Q4. # of obs (1) # of obs that trade derivatives for non-trading purposes (2) Type Standalone Banks 136,969 4,672 BHC-One 314,039 24,162 BHC-Multiple 130,757 31,907 Total 581,765 60,741 Percentages (3)=(2)/(1) 3.41% 7.69% 24.40% 10.44% 25

Table 2 This table provides the comparison summary statistics for two groups of commercial banks: one report interest rate swap trading for hedging purpose as paying at a fixed rate, the other report zero or do not report. Panel A includes total number of observations of the two groups, and subgroups categorized by the status of whether belong to a banking holding company (standalone, belongs to One-bank Banking Holding Company (BHC), belongs to Multi-bank BHC), and the relative size within a banking holding company (Top: one of the largest three subsidiaries in a BHC, Bottom: one of the smallest one third of the subsidiaries in a BHC). Panel B includes number of year-quarter observations that banks from the two groups have. Panel C summarizes the statistics of balance sheet information of different groups. All the repressors are divided by bank total assets. Panel D displays the equity and income source distribution of groups. Panel E presents the maturity gap schedule of different groups. Maturity gap is calculated from the amount difference between Assets and Liability that maturity in respective years over total assets. Time period is from 2001Q1 to 2013Q4. Panel A Interest rate swap pay at a fixed rate (RCFD 1589) N Standalone One-BHC Mult-BHC TOP Bottom Medium Zero/Missing 46,887 3,317 17,803 25,767 26,048 11,280 9,697 Non-Zero 9,807 886 4,709 4,212 2,721 407 1,084 Panel B Mean Median Min p10 p90 Max Number of quarters Zero/Missing 28.02 29 1 9 52 52 Non-Zero 33.54 33 1 12 52 52

Table 2 cont. Tot asset Panel C Standalone Zero/Missing 1,388,887 Non-Zero 6,285,461 TOP Zero/Missing 3,474,382 Non-Zero 4,381,931 Bottom Zero/Missing 848,256 Non-Zero 5,976,735 One-BHC Zero/Missing 1,452,353 Non-Zero 5,223,879 Mult-BHC Zero/Missing 3,344,381 Non-Zero 7,663,658 Log (Tot asset) C&I (%) Real Estate (%) Consumer Loan (%) Construction Loan % Core Deposit (%) Wholesale funds (%) 13.07 3.74 54.91 4.03 54.91 69.57 17.90 13.99 5.38 43.35 11.73 43.35 54.23 29.79 13.35 7.29 48.47 4.04 48.47 72.44 18.07 15.293 11.10 42.73 5.45 42.73 61.80 18.87 12.08 2.58 42.08 4.93 42.08 66.33 16.79 14.34 7.60 47.49 5.34 47.49 66.67 24.10 13.15 6.64 50.32 3.27 50.32 74.06 18.16 14.18 9.88 47.08 4.07 47.09 65.84 22.16 12.89 5.58 43.49 5.81 43.5 67.29 17.34 15.58 11.76 38.20 7.28 38.20 57.71 19.18 27

Table 2 cont. Tier 1 Ratio ROA STD(ROA) Interest Income perc Non-interest Income perc Interest Expense Panel D (%) Standalone Zero/Missing 9.70 0.36 0.23 0.60 0.40 0.84 Non-Zero 9.95 0.46 0.28 0.61 0.39 0.85 TOP Zero/Missing 8.60 0.52 0.31 0.59 0.41 0.77 Non-Zero 8.30 0.48 0.30 0.58 0.42 0.74 Bottom Zero/Missing 8.80 0.48 0.33 0.59 0.41 0.70 Non-Zero 8.33 0.39 0.22 0.60 0.40 0.68 One-BHC Zero/Missing 8.78 0.50 0.29 0.59 0.41 0.75 Non-Zero 8.58 0.45 0.28 0.60 0.40 0.71 Mult-BHC Zero/Missing 8.36 0.54 0.34 0.60 0.40 0.78 Non-Zero 7.85 0.50 0.30 0.61 0.39 0.77 Panel E Maturity gap (%) <1 year 1-3 years 3-5 years 5-15 years >15 years Standalone Zero/Missing 19.83 7.35 7.73 5.72 2.34 Non-Zero 23.68 8.30 7.26 2.81 0.03 TOP Zero/Missing 25.78 10.67 8.83 0.87 0.46 Non-Zero 23.78 8.20 7.81 2.82 0.47 Bottom Zero/Missing 22.75 8.73 7.95 1.41 0.26 Non-Zero 16.01 7.36 8.94 6.34 3.41 One-BHC Zero/Missing 25.28 10.88 9.11 0.74 0.47 Non-Zero 22.67 9.46 9.11 2.21 0.21 Mult-BHC Zero/Missing 25.38 9.36 8.05 1.56 0.25 Non-Zero 24.10 6.76 6.46 3.78 1.14 28