Risk Management and Bank Loans

Transcription

1 Risk Management and Bank Loans Iftekhar Hasan Fordham University and Bank of Finland 5 Columbus Circle, 11 th floor New York, NY Telephone: ihasan@fordham.edu Mingsheng Li College of Business Bowling Green State University 210 Business Administration Bowling Green, OH Telephone: mli@bgsu.edu Liuling Liu College of Business Bowling Green State University 206 Business Administration Bowling Green, OH Telephone: liulinl@bgsu.edu Yun Zhu The Peter J. Tobin College of Business St. John's University Queens, NY Telephone: zhuy@stjohns.edu 1

2 Risk Management and Bank Loans Abstract In this paper, we examine the risk governance structure at BHCs in the US, and investigate whether differences in the loan price and non-price terms can be explained by differences in the strength of banks risk management functions. We find that BHCs with strong risk governance is more likely to select risky borrowers. However, they charge higher loan spreads to compensate themselves for the additional default risk and design more stringent loan contracts to facilitate ex-post intensive monitoring needs. Banks with strong risk governance indeed commit more diligence in monitoring, which is supported by the concentrated syndicate structure. We also find that the monitoring efforts devoted by BHCs with good risk governance are effective because their loans granted have lower likelihood of covenant violation and downgrading. In sum, we document that appropriate risk governance brings value-enhancing loans to the bank. 2

3 1. Introduction Risk controls at [Merrill Lynch], then run by CEO Stan O Neal, were beginning to loosen. A senior risk manager, John Breit, was ignored when he objected to certain risks Merrill lowered the status of Mr. Breit s job Some managers seen as impediments to the mortgage-securities strategy were pushed out. An example, some former Merrill executives say, is Jeffrey Kronthal, who had imposed informal limits on the amount of CDO exposure the firm could keep on its books ($3 billion to $4 billion) and on its risk of possible CDO losses (about $75 million a day). Merrill dismissed him and two other bond managers in mid-2006, a time when housing was still strong but was peaking. To oversee the job of taking CDOs onto Merrill s own books, the firm tapped a senior trader but one without much experience in mortgage securities. CDO holdings on Merrill s books were soon piling up at a rate of $5 billion to $6 billion per quarter Wall Street Journal (April 16, 2008) reports Banks are in the business of taking risk. A bank s ability to effectively measure and manage risk determines its success. There is widespread agreement that the financial crisis can be to an important extent attributed to a failure of risk management at banks: 1) risk managers mis-measure risk exposures due to the increasing complexity of security products and flawed historical risk model, 2) risk managers were powerless and ineffective in restraining top management who are profit-pressured and tended to take excessive risk. This led to a growing need for comprehensive risk assessment techniques and respective risk governance structures to control the risk-takers (e.g. Basel Committee on Banking Supervision, 2008; IIF, 2007). 3

4 In theory, the ideal banking risk governance highlights the need of having 1) the presence of high-status chief risk officer (CRO), and 2) dedicated and independent risk committee in board. Interestingly, according to Ellul and Yerramilli (2010), there are considerable cross-sectional differences in the importance and attention given to risk management organizations at the top 74 BHCs before crisis. For instance, as of 2006, 33% of the banks do not have a chief risk officer (CRO). Among banks with CROs, 57% of which the CRO was considered as an executive officer, and 15% of which the CRO was among the five highest paid employees. Mongiardino and Plath (2010) surveyed 20 large banks and document limited extent of improvement in risk governance after the crisis despite increased regulatory pressure. For instance, most banks have inactive risk committee and lack of enough independent and financially knowledgeable members. And most large banks had a CRO but whose positions are not high enough to report to the CEO and board. A small but growing empirical literature attempts to understand the impact of risk governance to risk outcomes and firm performance. For instance, Kashyap (2010) and Keys et al. (2009) find that CRO centrality (measured as the ratio of CRO compensation to CEO compensation) is associated with lower implied volatility and better loan performance before the crisis. Lingel and Sheedy (2012) find that better board oversight of risk and having a high-status CRO leads to lower risk. Aebi, Sabato, and Schmid (2012) find that banks with CROs directly report to the board not to the CEO have better stock and accounting performance during the crisis. Ellul and Yerramilli (2013) show that a strong and independent risk management function before the crisis is associated with lower tail risk, lower nonperforming loans, and better financial performance during the 4

5 crisis period. None of these papers have explained through which channels risk governance affects bank performance, keeping a black box view of the documented relationship. Our paper aims to explain how risk governance was carried out in the loan granting process. Specifically, in this paper, we examine if differences in lending behavior and loan contracts among bank holding companies (BHCs) in the United States can be explained by differences in the risk governance functions. To understand the impact of risk governance on bank loans, we need first to understand the risk management involvement in the loan granting process. The standard approach in practice followed by most banks is that banks hire two agents--- loan officer and risk manager, to perform the screening function (Udell, 1989; Berg, 2015). The screening process for green applications (i.e. loan applications with good rating) and red applications (i.e. loan applications with very poor rating) is purely determined by loan officers. However, the yellow applications (i.e. loan applications with poor rating) requires risk manager approval. Loan officers and risk managers have different incentives: whereas loan officer is volume incentivized and tend to be in favor of granting a risky loan; risk manager receives fixed salary and is responsible for containing the loan default rates, therefore, they tend to against granting a risky loan. In a bank with a strong incentive to manage risk effectively, appropriate risk governance can help in creating incentives for a constructive collaboration between risk managers and loan officers. Better risk governance should lead to better risk-taking, but not necessarily low risk. In line with this view, better governance should encourage the risk managers and loan officers as a whole to grant risky loans that were expected to be rewarding for shareholders and monitor risks appropriately against default. We expect that BHCs with 5

6 strong and independent risk management function is more likely to grant loans to risky borrowers and charge high interest rate correspondingly. In addition, those BHCs are expected to commit more monitoring effects and use stringent loan contracts to facilitate intensive ex-post monitoring. Finally, we expect that effective risk governance structure should help banks to control risk and reduce the default probability of those risky loans. Our main alternative hypothesis is that the risk governance function does not have any real effect on the bank loan contracts. This may be because risk management only targets the level of risk to take for the organization and provide guidance to the top risk manager of the bank, the CEO, not the CRO. It is widely agreed among policymakers, bank supervisors and academics that bank executives tended to take excessive risk. With a top management that being pressured for profits, risk managers are likely to become increasingly powerless and be deficiencies in risk oversight even if they are aware of inaccurate risk measures and false incentive. Therefore, the real power rests with loan officers and bank executives who control the bank s risk exposure. To capture the variation of risk governance function within each BHC, we adopt an innovative risk management index (RMI) from Ellul and Yerramilli (2013), which measures the importance of risk management function and quality of risk oversight provided by the BHC s board of directors. Using a sample of 13,013 loans issued to 2,878 firms by 42 bank holding companies from 1988 through 2013, we find that loans made by banks with higher risk management index (i.e. higher level of risk management) have higher loans spread, higher fees, shorter loan maturity, higher likelihood of using collateral and more restrictive covenants. The results imply that banks with strong risk management are more comfortable to take higher risk by lending to risky borrowers. But 6

7 they manage risk properly by charging higher interest rate, fees, and construct stringent loan contract terms to facilitate monitoring. We further provide direct evidence to support above implication. We find that banks with strong risk management indeed lend to riskier borrowers (i.e. low rating borrowers) while their loans are less likely to be downgraded and violate the covenants, suggesting the strong monitoring role performed by high RMI banks. In sum, we find that high RMI provide efficient risk-return trade-off by allowing banks to take riskier loan while manage the default risk properly. Our paper contributes to the exiting literature in several ways. This study reveals that risk governance is essential for risk management because effective risk governance align the interests of risk-takers and risk managers. We find that right governance can set up right incentives for loan officers and risk managers. They together will grant loans to risky borrowers but with default risk that is manageable. Under the control of good risk governance, bank designs proper loan contracts to facilitate the intensive monitoring needs and charge proper risk premium. Despite the ongoing debate on the impact of risk governance and bank performance, our study is the first paper to attempt to reveal the black box view of the observed relationship between risk governance and bank risk and performance. Further, in the context of bank loan literature, our research extends the line of research focusing on how borrower characteristics affects the loan contracts (e.g., Strahan, 1999; Graham, Li, and Qiu, 2008) and shows that certain feature of credit supplier also play significant role in shaping loan contracts. Our paper also has policy implication. We provide evidence that effective risk governance ensure banks take profitable risks and properly monitor the risks, and hence create shareholder values. 7

8 2. Literature and Hypothesis To understand the impact of RMI on bank loans, we need first understand the function/responsibility of risk management. To control risk, large banks usually have risk management organizations that employ risk managers that are headed by a chief risk officer (CRO), or/and create dedicated committee to oversee risks. Though titled as risk managers, the real risk-takers in the bank are loan officers, desk traders, and the CEO (the ultimate top risk manager). Risk manager doesn t manage risk directly by themselves; instead they have a monitoring and advising role to make sure those risks stays within the limits that have been set by the bank (i.e. the banks appetite). It is top management s job to set up a bank s risk appetite (Stulz, 2016). A potential weakness of such setting is that the risk managers may be powerless in controlling the risk-takers, which is evidenced in the recent financial crisis. A respective risk governance structure is, therefore, essential for the success of risk management. With regards to grant loans, the standard approach in practice followed by most banks involves a risk manager as monitor to control the actions of the loan officer (Udell, 1989; Berg, 2015). Specifically, banks hire two agents--- loan officer and risk manager, to perform the screening function: accepting good borrowers who will pay back and rejecting bad borrowers who will default. Berg (2015) provides detail description for the loan-granting process that has risk management involvement. The approach is socalled traffic light approach : loan applications with good rating and sufficient equity ( green applications ) are granted whereas loan applications with very poor rating and significantly insufficient equity ( red applications ) are rejected by loan officers on their own without risk-management approval. However, applications with poor rating and 8

9 insufficient equity ( yellow applications ) require risk management approval, where a risk manager reviews the loan application and communicate the final acceptance/rejection decision to the loan officer. Loan officers and risk managers have different incentives. Loan officers compensation consists of fixed salary and loan volume-related bonus. Therefore, they tend to be in favor of granting a risky loan. On the other hand, risk managers salary is fixed and they are responsible for containing the loan default rates. Hence, they tend to be against granting a risky loan. Rajan (2008) conjectured that a misalignment of internal incentives may distort the loan quality. In a bank with a strong incentive to manage risk effectively, appropriate risk governance can help in setting incentives right. What is good risk governance? With effective risk governance, the risk will not simply be managed through limits. Instead, the risk management organization will ensure that the risk capacity of the bank will be used efficiently through the price mechanism rather than rationing. In other words, with right risk governance, risk-takers will not necessarily take low risk, but good risk that is fairly rewarded. Hence, the good risk governance will create more shareholder values. Putting in the loan origination context, we expect that banks with effective risk governance will grant risky but profitable loans. Empirically, a small but growing literature attempts to understand the impact of risk governance to risk outcomes and firm performance. Kashyap (2010) finds that CRO centrality (measured as the ratio of CRO compensation to CEO compensation) is associated with lower implied volatility before the crisis. Keys et al. (2009) document a positive relationship between CRO centrality and loan performance. Using a sample of 60 publicly traded banks from 2004 to 2010, Lingel and Sheedy (2012) find that better 9

10 board oversight of risk and having a high-status CRO leads to lower risk. Aebi, Sabato, and Schmid (2012) find that banks with CROs directly report to the board not to the CEO have better stock and accounting performance during the crisis. Ellul and Yerramilli (2013) innovatively construct a risk management index (RMI) for U.S. BHCs by combining a set of risk governance attributes and show that higher RMI (stands for a strong and independent risk management function) before the crisis is associated with lower tail risk, lower nonperforming loans, and better financial performance during the crisis period. 3. Research design, sample selection and summary statistics 3.1. Measures of risk management index (RMI) We obtain the risk management index for Bank Holding Companies (BHCs) from Ellul and Yerramilli (2013) which suggest that the variations in risk-taking can be explained by the differences in the strength and independence of the risk management function at bank holding companies. The RMI index is constructed year-by-year using the principal component analysis. It consists of two major components 1) the importance of the Chief Risk Officer (CRO), 2) the quality of the board of directors risk oversight. CRO s job responsibility exclusively involves managing enterprise risk for all business segments of the BHC. Several variables such as CRO Present, CRO Executive, CRO Top5, and CRO Centrality, are adopted to measure its importance in the BHC. Specifically, CRO Present is a dummy that equals one if there is CRO position or an equivalent position within the organization, zero otherwise. CRO Executive equals one if the CRO is considered as an executive officer within the BHC, zero otherwise. CRO Top5 equals one if the CRO is among the five highest paid executives at the BHC, zero 10

11 otherwise. CRO Centrality is the ratio of CRO s total non-stock and non-option compensation to the CEO s total compensation, measures the relative power of CRO within the organization. The Risk Committee is designated to oversee and manage risk of BHC. Its quality is measured by Risk committee experience and Active risk committee. To be specific, Risk committee experience is a dummy variable that equals one if at least one independent director in the risk committee has banking and finance experience. Active risk committee equals one if the committee of a given BHC meet more frequently than the average across all BHCs in a given year Data and Sample selection We start with the BHC list in the Ellul and Yerramilli (2013) and obtain their consolidated financial statement information from the Federal Reserve Call Report. We then collect bank loans originated by those BHCs from the Loan Pricing Corporation (LPC) DealScan database, which contains detailed information on individual loan facilities such as loan spreads, maturity, collateral, covenants, loan types, and loan purposes. We retrieve accounting data for the loan borrowers from the Standard and Poor s Compustat database. Loans to the financial services (SIC code ) and utilities (SIC code ) industries are excluded because these two industries are highly regulated. After merging the databases and removing observations with incomplete information, our final sample consists of 13,013 loans issued to 2,878 firms by 42 bank holding companies from 1988 through To mitigate the effect of outliers or misrecorded data, all firm-level continuous variables are winsorized at the 1% and 99% levels. 11

12 3.3. Baseline regression model We use the following empirical model as our baseline model to examine the effect of risk governance on the loan price charged by a bank: Log (AISD) = f (RMI t 1, bank features t 1,borrower attributes t 1,loan characteristics t, borrower dummies, and year dummies), (1) where the dependent variable is the natural logarithm of loan price recorded as all-in spread drawn (AISD), which is the amount the bank charges in basis points over the London Interbank Offered Rate (LIBOR) for each dollar drawn down for a loan the firm obtains in year t. RMI measures the importance of the Chief Risk Officer (CRO) and the quality of the board of directors risk oversight, obtained from Ellul and Yerramilli (2013). Following prior studies (e.g., Strahan, 1999; Graham, Li, and Qiu, 2008), we include three sets of control variables to capture various bank features, borrower attributes and loan characteristics that are important determinants in loan price. RMI, bank features, and firm attributes are computed using information from the year immediately prior to the year in which a firm obtains a bank loan (i.e., year t-1), while the loan characteristics are computed using information for a loan a firm obtains in year t. Large banks or banks with less liquidity concerns are more comfortable to take risk. Therefore, we control for bank size and bank deposit ratio. Bank size is measured as the natural logarithm of the book value of bank assets. Bank deposit ratio is the ratio of all short term and long term deposits funding to total assets. We control for following firm characteristics in our model. Firm size is the natural logarithm of the book value of assets. Profitability is the ratio of net income to total assets. Leverage is the ratio of total debt to total assets. We calculate the modified 12

13 Altman s (1968) Z-score (= [1.2 Working Capital Retained Earnings EBIT Sales]/Total Assets) to gauge the likelihood of default of the borrowing firms. In addition, Tangibility is the ratio of tangible assets to total assets; Market to book, the ratio of the market value of assets to the book value of assets, measures a firm s growth opportunities; Cash flow volatility is defined as the standard deviation of a borrower s quarterly cash flows in the previous three years scaled by the average book assets. Rating is the numeric measure of a firm s S&P debt rating a higher value indicates higher credit risk. We further control for loan characteristics. Log(Loan size) is the natural log of the loan amount in millions of dollars. Log(Loan maturity) is the natural log of maturity in months. Relationship loan is a binary variable which takes the value of one if there is any prior lending by the same lead banks in the five-year window preceding the loan origination (Bharath et al., 2011). In addition, Performance pricing is a dummy variable which takes the value of one if banks agree to adjust loan price based on borrowers future performance, and zero otherwise. Lastly, we control for year effects, borrowing firm effects, loan purposes, and loan types in the regression models. Appendix A provides detailed definitions and measurements for all variables Summary statistics Table 1 reports the summary statistics of RMI, bank features, firm characteristics, and loan characteristics. On average, the natural log of bank size is 19.43, bank deposit ratio is 83.3%, the natural log of borrowing firm size is The average profitability is 0.133, average leverage ratio is 0.315, average modified Altman s (1968) Z-score is 3.749, average tangibility is 0.311, average market to book ratio is 1.828, and average 13

14 cash flow volatility is 0.049,. In terms of loan variables, we find that the average AISD is basis points (median 175 basis points), and the standard deviation is basis points. In our sample, the average loan size is $ million, with a mean maturity of months. The sample statistics of firm and loan variables are similar to prior studies (e.g., Strahan, 1999; Graham, Li, and Qiu, 2008). [Insert Table 1 here] 4. Empirical Results 4.1. Baseline regression results We start with ordinary least square (OLS) regressions with firm-clustered, heteroskedasticity-robust standard errors in column 1 of Table 2. The estimated coefficient on RMI is and is significant at the 1% level (t-value = 5.175). To mitigate this concern of time-invariant omitted firm-level factors driven the results, in column 2 of Table 2 we run our baseline model using firm fixed effects. The coefficient on RMI drops slightly to and remains significant at the 1% level, suggesting that our findings are not seriously plagued by omitted variable bias. Overall, the results in Table 2 are consistent with our consistent with our hypothesis that effective risk governance is positively associated with bank loan price. The coefficients for the control variables are consistent with existing literature (e.g., Graham, Li, and Qiu, 2008). The results are economically meaningful. For instance, based on the fixed effect regression results, given that the average loan spread of the sample firms is basis points, one unit increase in RMI increase loan spreads by about basis points (= 11.7% 191.8). Given that the mean sample loan size is $ million and the average loan s time to maturity is around 3.9 years, one unit increase in RMI results in an average 14

15 $3.55 million (= $ million ) interest expense increase for borrowing firms. Our estimate is higher than those reported in prior studies. For example, Bharath, Sunder, and Sunder (2008) find that a one-standard-deviation increase in accounting quality in their respective samples reduces bank loan spread by 6.65 basis points. Thus, RMI has an economically significant effect on bank loan costs. [Insert Table 2 here] Fees charged by banks are also important components of the total cost of borrowing (Berg, Saunders, and Steffen, 2016). If higher RMI increases bank loan spread, we expect it has a similar effect on the fees charged by banks. In Table 3, we find that RMI is also positively related to various bank fees (including the all-in-spreadundrawn (AISU), facility fee, commitment fee, upfront fee and letter of credit fee). [Insert Table 3 here] 4.2. Risk management and non-price loan contract terms Bank loan contracts contain both price and non-price provisions that cannot be split and traded separately. These non-price terms help to mitigate banks risk exposures and enhance their monitoring ability during the life of each loan. Specially, banks can control risk through limits on maturity, through the use of covenants, and through collateral requirements. For instance, by shortening the contractual maturity of loans, a lender can periodically evaluate a borrower s pay-off ability and maintain stronger bargaining power in the renegotiation processes (Barclay and Smith, 1995). By imposing covenants in the contract, a lender can call a loan if a covenant is violated without having to wait until conditions worsen (Rajan and Winton, 1995). By requiring collateral, a lender can commit to more rigorous monitoring since the collateral value is sensitive to 15

16 borrower behaviors (Rajan and Winton, 1995). In other words, non-price contract terms serve as mechanism to implement on-balance sheet risk management, and banks are more likely to use a short contractual maturity, to impose stringent covenants, and to require collateral when borrowers are risky and need more monitoring efforts (Berger and Udell, 1990; Strahan, 1999; Bradley and Roberts, 2003). In column 1 of Table 4, we regress the natural log of Loan Maturity on RMI, controlling for various bank features, firm characteristics and loan characteristics. The result shows that the coefficient of the risk governance measure is negative and significant at the 10% level, suggesting that holding all else constant, the loan maturity decreases by 8.4% for each unit increase in the RMI index. In column 2 of Table 4, we investigate the effects of RMI on collateral pledges using a Probit model. The dependent variable, Collateral, equals 1 if the loan is secured and 0 otherwise. 2 The result suggests that lenders with higher RMI index are indeed more likely on average to require collateral from borrowers. In order to examine the effects of RMI on covenants, we create Covenant Intensity, the sum of total number of general covenants (e.g., prepayment covenants, dividend covenants, and voting-rights covenants) and total number of financial covenants (e.g., interest-coverage ratio, current ratio, leverage, and net worth covenants). Columns 3 of Table 4 show Poisson regressions with Covenant Intensity as dependent variable. The results indicate that on average banks with higher RMI impose more covenant restrictions on borrowers than banks with lower RMI index, all else equal. In sum, the results show that high RMI bank employ more restrictive loan contracts to 2 In the reported results, we treat the loans with missing collateral information in DealScan as unsecured. 16

17 strengthen their monitoring abilities and thereby mitigate potential moral-hazard problems. [Insert Table 4 here] 4.3. Risk management and syndicate structure Because most of the loans in our sample (92 %) have more than one lender, our sample allows us to test how RMI impact the syndicate structure empirically. Consistent with the moral-hazard theory of syndicate structure (Holmstrom and Tirole, 1997), Simons (1993) finds empirical evidence that lead arrangers can syndicate a larger portion of loans when the borrower s quality is good. Recent studies find similar results showing that syndicates are more concentrated when the borrower is opaque and requires more intense monitoring (Dennis and Mullineaux, 2000; Lee and Mullineaux, 2004; and Sufi, 2007). Therefore, concentrated syndicate structure is a mechanism that banks commit more intense monitoring. In Table 5 we construct three syndicate structure variables: Syndicate size (in columns 1), HHI (Syndicate share) (in columns 2), and Lead bank share (in columns 3) to capture the extent of syndicate concentration. Specifically, the Syndicate size is the number of lenders in a syndicated loan. HHI (Syndicate share) is the sum of the squared percentage of each syndicate member s share of the loan scaled by 10,000. This measure takes into account of all syndicate members regardless of lead arrangers or participants and captures any effects of joint monitoring (Sufi, 2007; Bharath et al., 2011). The larger the HHI value, the more concentrated the syndicate is. Lead bank share measures the share of loan held by lead banks in the syndicate. The results are presented in Table 5. We find that the concentration of loan ownership is higher; that is, the number of lenders 17

18 is smaller, the HHI is higher, and lead bank share is larger when the bank has more effective risk governance. Our finding supports the view that banks with effective risk government devote more due diligence in monitoring the borrowers. [Insert Table 5 here] 4.4. RMI and ex-post monitoring We further examine how RMI affects ex-post loan performance i.e. likelihood of deterioration after the loan is granted. To measure the likelihood of loan deterioration, we adopt two dummy variables: Covenant violated that equals one if the borrower violated any financial covenant before the loan matures, zero otherwise, and Downgraded that equals one if the borrower got downgraded during the life of the loan, zero otherwise. We run probit models in Table 6. The results show that loans granted by banks with higher RMI are less likely to violate the covenants and to be downgraded. To be more specific, a one unite increase in RMI in the data reduces the probability of covenant violation by 41.7% and the probability of downgrading by 12.8%. [Insert Table 6 here] 4.5. RMI and ex-ante screening So far, our results are consistent with the view that banks with more effective risk governance tend to lend to risky borrowers while those risks are rewarded. In this section, we aim to provide direct evidence that what types of borrowers are more likely to be selected by high RMI banks. We, therefore, aggregate the loans in DealScan to form a sample of banks yearly loan portfolio. Then we calculate the total dollar value of and total number of below investment grade loans. Two dependent variables are created and 18

19 used in the Probit models of Table 7. Below investment grade loans_amt takes the value of one, if a bank has higher than median level below investment grade loans in terms of dollar value in a given year, and zero otherwise. Below investment grade loans_n takes the value of one, if a bank has higher than median level below investment grade loans in terms of number of loan in a given year, and zero otherwise. We model the likelihood of selecting risky borrowers as a function of RMI and borrower characteristics such as bank size, bank deposit ratio, roa and loan loss provision ratio. In both models, we find that RMI has statistically significant impact on banks screening decision. To be specific, banks with better risk governance, all else equal, are more likely to select risky loans measured as below investment grade loans. [Insert Table 7 here] 5. Conclusion The success of banks is critically determined by how they take risks. The recent failure in the finance system has convinced the policymakers, bank supervisors, and academics that there are important flaws in risk management, conflicts of interests between risk takers and risk managers in particular. On the one hand, banks top management are under pressure to create value for shareholders, therefore, they have incentive to take excessive risks. On the other hand, risk managers constrained by the risk limits, tend to adopt hedging decisions that lower the risks but sacrificing the returns. In addition, the role of risk managers might be significantly undervalued in some banks, which make the risk managers too powerless to communicate to the risk takers. Therefore, it is important to establish effective risk governance structure to align the conflicted incentives, which ensures the banks take profitable risks to enhance shareholder values. 19

20 In this paper, we examine the risk governance structure at BHCs in the US, and investigate whether differences in the loan price and non-price terms can be explained by differences in the strength of banks risk management functions. We find that BHCs with strong risk governance is more likely to select risky borrowers. However, they charge higher loan spreads to compensate themselves for the additional default risk and design more stringent loan contracts to facilitate ex-post intensive monitoring needs. Banks with strong risk governance indeed commit more diligence in monitoring, which is supported by the concentrated syndicate structure. We also find that the monitoring efforts devoted by BHCs with good risk governance are effective because their loans granted have lower likelihood of covenant violation and downgrading. In sum, we document that appropriate risk governance brings value-enhancing loans to the bank. 20

21 Reference Altman, E.I., Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance 23, Aebi, V., G. Sabato, and M. Schmid Risk Management, Corporate Governance, and Bank Performance in the Financial Crisis. Journal of Banking and Finance 36: Basel Committee on Banking Supervision, June Principles for sound liquidity risk management and supervision. Barclay, M. J., Smith Jr., C. W., The maturity structure of corporate debt, Journal of Finance 50, Berg, Tobias. "Playing the devil's advocate: The causal effect of risk management on loan quality." The Review of Financial Studies 28, no. 12 (2015): Berger, A., Udell, G., Relationship Lending and Lines of Credit in Small Firm Finance, Journal of Business 68, Bharath, Sreedhar T., Sandeep Dahiya, Anthony Saunders, and Anand Srinivasan. "Lending relationships and loan contract terms." The Review of Financial Studies 24, no. 4 (2009): Bradley, M., Roberts, M.R., The structure and pricing of corporate debt covenants. Working paper, Duke University. Dennis, S. A., Mullineaux, D. J., Syndicated loans. Journal of Financial Intermediation 9, Ellul, Andrew, and Vijay Yerramilli. "Stronger risk controls, lower risk: Evidence from US bank holding companies." The Journal of Finance 68, no. 5 (2013): Graham, J. R., Li, S., Qiu, J., Corporate misreporting and bank loan contracting. Journal of Financial Economics 88, Holmstrom, B., Tirole, J., Financial intermediation, loanable funds, and the real sector. Quarterly Journal of Economics 112, Institute of International Finance (IIF), March Principles of liquidity risk management. Kashyap, A. K Lessons from the Financial Crisis for Risk Management. Paper prepared for the Financial Crisi Inquiry Commission. 21

22 Keys, B. J., T. Mukherjee, A. Seru, and V. Vig Financial Regulation and Securitization: Evidence from Subprime Loans. Journal of Monetary Economics 56, no. 5 (July): Lee, S.W., Mullineaux, D.J., Monitoring, financial distress, and the structure of commercial lending syndicates. Financial Management 33, Lingel, A., and E. Sheedy The Influence of Risk Governance on Risk Outcomes International Evidence. Unpublished paper, Macquarie University. Mongiardino, A., Plath, C., Risk governance at large banks: have any lessons been learned? Journal of Risk Management in Financial Institutions 3, Rajan, R., A View of the Liquidity Crisis. University of Chicago, Mimeo. Rajan, R., Winton, A., Covenants and collateral as incentives to monitor. Journal of Finance 47, Strahan, P., Borrower risk and the price and nonprice terms of bank loans, Federal Reserve Bank of New York. Staff Paper No. 90. Stulz, R. M Risk Management Failures: What Are They and When Do They Happen? Journal of Applied Corporate Finance 20, no. 4 (Fall): Stulz, René M. "Risk management, governance, culture, and risk taking in banks." (2016). Simons, K., 1993, Why do banks syndicate loans? New England Economic Review of the Federal Reserve Bank of Boston, Sufi, A., Information asymmetry and financing arrangements: evidence from syndicated loans. The Journal of Finance 62, Udell,G Loan quality, commercial loan review, and loan officer contracting. Journal of Banking&Finance 13:

23 Table 1 Summary statistics. Variable N Mean std Median P25 P75 AISD (bps) RMI Bank size Bank deposit ratio Firm size Profitability Leverage Z-score Tangibility Market-to-book Cash flow volatility Rating Loan maturity (months) Loan size ($ mil) Relationship lending

24 Table 2 Baseline regressions: The relation between risk management and bank loan cost. (1) (2) VARIABLES Log(AISD) Log(AISD) RMI 0.198*** 0.117*** [5.175] [3.052] Bank size *** ** [-2.646] [-2.139] Bank deposit ratio *** *** [-3.200] [-2.588] Firm size *** *** [-7.913] [-5.188] Profitability *** *** [-9.494] [ ] Leverage 0.783*** 0.305*** [17.254] [5.525] Z-score *** [-0.486] [-3.402] Tangibility [0.874] [-0.607] Market-to-book [-0.910] [-0.296] Cash flow volatility ** [-2.060] [-0.640] Rating 0.029*** 0.021*** [11.474] [5.522] Log(Loan maturity) 0.168*** 0.039*** [12.016] [3.688] Log(Loan size) *** *** [-7.591] [-5.066] Relationship lending *** *** [-4.218] [-3.007] Performance pricing ** [-1.133] [-2.126] Constant 5.854*** 6.771*** [25.590] [20.842] Observations 13,055 13,055 Adjusted R-squared Loan type Y Y Loan purpose Y Y Year effects Y Y Industry effects Y N Firm effects N Y Clustered standard errors Firm Firm 24

25 Table 3 Effects of risk management on the fees charged by banks. VARIABLES (1) (2) (3) (4) (5) (6) Log(AISU) Log(Facility Log(Commitment Log(Upfront fee) fee) fee) Log(Letter of credit fee) Log(Total cost of Borrowing) RMI 0.116*** 0.115** 0.110* 0.524* 0.452*** 0.105*** [2.664] [1.995] [1.919] [1.861] [5.659] [2.795] Bank size ** * [0.247] [2.199] [0.475] [0.240] [0.522] [-1.888] Bank deposit ratio *** ** [0.139] [2.658] [0.399] [-0.144] [0.157] [-2.150] Firm size *** *** * *** [-3.070] [-4.134] [-0.346] [1.209] [1.704] [-5.521] Profitability *** *** *** * *** [-7.134] [-2.652] [-5.560] [-1.046] [-1.943] [-9.755] Leverage 0.342*** 0.651*** 0.261*** *** 0.296*** [4.634] [3.973] [3.316] [0.495] [2.706] [5.424] Z-score ** [-0.235] [0.457] [-1.194] [-0.314] [-0.568] [-2.294] Tangibility * ** [-1.934] [-1.074] [-0.145] [-0.661] [-2.080] [-0.617] Market-to-book *** *** [-0.014] [-2.722] [0.501] [0.376] [-3.367] [-0.604] Cash flow volatility ** [-0.252] [1.219] [0.241] [0.695] [2.226] [-0.935] Rating 0.023*** 0.031*** *** 0.015** 0.022*** [5.109] [4.174] [1.271] [2.599] [2.507] [5.702] Log(Loan maturity) 0.136*** 0.128*** 0.065*** 0.106** *** [11.524] [11.018] [2.856] [1.994] [0.087] [3.460] Log(Loan size) *** ** *** *** *** [-5.681] [-2.163] [-3.965] [-1.602] [-4.492] [-4.160] Relationship lending Performance pricing * * *** [-1.831] [-0.414] [-1.859] [-1.359] [-0.926] [-2.979] 0.043** 0.104*** * [2.027] [3.296] [-0.790] [1.842] [1.018] [-0.106] Constant 3.765*** 3.369*** 3.597*** *** 6.298*** [10.219] [5.080] [8.398] [0.449] [8.093] [19.682] Observations 8,651 3,281 5,575 2,212 4,880 13,055 Adjusted R squared Loan type Y Y Y Y Y Y Loan purpose Y Y Y Y Y Y Year effects Y Y Y Y Y Y Firm effects Y Y Y Y Y Y Clustered standard errors Firm Firm Firm Firm Firm Firm 25

26 Table 4 The relation between risk management and other loan contract terms. (1) (3) (4) VARIABLES Log(Loan maturity) Collateral Covenants instensity RMI * 0.200** 0.116** [-1.880] [2.269] [2.327] Bank size [-0.888] [-0.841] [-1.517] Bank deposit ratio [0.165] [-1.009] [-0.484] Firm size *** *** *** [-3.988] [-7.960] [ ] Profitability *** *** [1.638] [ ] [-4.057] Leverage *** 0.491*** [-1.494] [8.845] [9.054] Z-score 0.001* [1.706] [-1.607] [0.307] Tangibility * * [-1.668] [1.059] [-1.772] Market-to-book ** [0.687] [0.813] [-2.350] Cash flow volatility * 1.426** *** [-1.856] [2.153] [-3.219] Rating *** 0.010*** [-0.428] [4.057] [3.825] Log(Loan maturity) 0.341*** 0.219*** [10.693] [9.953] Log(Loan size) 0.111*** *** [11.359] [-3.124] [1.514] Relationship lending ** *** [-2.567] [-3.549] [0.108] Performance pricing 0.180*** 0.497*** 0.617*** [10.047] [12.997] [24.859] Constant 3.720*** [13.481] [1.305] [0.509] Observations 13,055 13,055 13,055 Adjusted R-squared Loan type Y Y Y Loan purpose Y Y Y Year effects Y Y Y Firm effects Y N N Clustered standard errors Firm Firm Firm Industry effects N Y Y 26

27 Table 5 The relation between risk management and syndicate structure. (1) (2) (3) VARIABLES Log(Syndicate size) HHI(Syndicate share) Lead bank share RMI *** 0.098*** 0.148*** [-9.896] [7.691] [8.431] Bank size *** 0.011*** 0.016*** [-7.928] [4.555] [4.713] Bank deposit ratio 0.656*** *** *** [6.027] [-7.410] [-6.862] Firm size 0.134*** *** *** [5.677] [-4.235] [-3.795] Profitability * * [1.450] [-1.900] [-1.912] Leverage [-0.693] [-0.480] [0.426] Z-score [-0.410] [0.775] [0.728] Tangibility * ** [0.765] [-1.919] [-2.208] Market-to-book [0.479] [-0.653] [-1.356] Cash flow volatility [-0.478] [0.431] [-0.643] Rating *** 0.001* 0.003** [-3.179] [1.667] [2.270] Log(Loan maturity) 0.072*** *** *** [5.524] [-5.839] [-3.123] Log(Loan size) 0.107*** *** *** [13.094] [ ] [-7.003] Relationship lending 0.048*** *** *** [3.508] [-4.526] [-5.765] Performance pricing 0.232*** *** *** [14.757] [-7.964] [-5.341] Term loan *** 0.025*** [1.226] [3.122] [5.316] Constant 1.988*** 0.545*** 0.663*** [5.667] [5.694] [5.481] Observations 13,055 13,055 13,055 Adjusted R-squared Loan type Y Y Y Loan purpose Y Y Y Year effects Y Y Y Firm effects Y Y Y Clustered standard errors Firm Firm Firm 27

28 Table 6 Post-loan performance. (1) (2) VARIABLES Covenant Violation Downgrading RMI *** ** [-3.074] [-2.043] Bank size [-0.666] [0.090] Bank deposit ratio 0.470* * [1.649] [-1.895] Firm size *** ** [-4.164] [-2.350] Profitability ** [-0.298] [-2.410] Leverage *** [-1.330] [5.875] Z-score ** [0.741] [-2.400] Tangibility 0.534** [2.275] [0.734] Market-to-book *** [-1.327] [-3.718] Cash flow volatility ** [-1.565] [-2.575] Rating * *** [-1.701] [ ] Log(Loan maturity) *** [-1.207] [24.687] Log(Loan size) 0.071*** 0.070*** [2.882] [4.275] Relationship lending 0.303*** 0.055* [5.457] [1.931] Performance pricing 0.333*** 0.116*** [6.064] [3.870] Term loan *** 0.121*** [-5.049] [3.596] Constant ** *** [-2.408] [-2.869] Observations 11,838 12,972 Pseudo R Loan type Y Y Loan purpose Y Y Year effects Y Y Industry effects Y Y Credit spread Y Y Term spread Y Y Clustered standard errors Firm Firm 28

29 Table 7 Risky v.s. safe borrowers. (1) (2) VARIABLES Below investment grade loans_amt Below investment grade loans_n RMI 0.365*** 0.166*** [4.402] [2.897] Bank size [-0.272] [1.498] Bank deposit ratio [-0.170] [-0.298] Bank roe [-0.115] [-0.648] Bank loan loss provision ratio * [-1.923] [-1.256] Constant [1.092] [0.256] Observations R-squared Loan type Y Y Loan purpose Y Y Year effects Y Y Clustered standard errors Bank Bank 29

30 Appendix A Variable definition and measurement. Variable Firm attributes Log(Asset) Profitability Market-to-Book Leverage Tangibility Z-score Earnings volatility Definition The natural log of the book value of the borrower s total assets. The ratio of EBITDA to total assets. The ratio of the market value of assets to the book value of assets and measures a firm s growth opportunities. The ratio of total debt to total assets. The ratio of tangible assets to total assets. The modified Altman s (1968) Z-score (= (1.2 Working Capital+1.4 Retained Earnings+3.3 EBIT Sales)/Total Assets). The standard deviation of the borrower's quarterly ROA in the previous five years. Loan attributes Log(AISD) Log (Loan size) Log (Loan maturity) Syndicate size Relationship Rating Loan type dummies Loan purpose dummies The natural logarithm of the amount the borrower pays in basis points over LIBOR for each dollar drawn down. The natural log of the loan facility amount in millions of dollars. The natural log of maturity in months. The number of lenders in the loan syndicate. Dummy variable equals to one if there was prior lending relationship between the lead bank and borrower in the past five years. The numeric measure of S&P debt rating. A higher value indicates higher credit risk. Dummy variables for loan types, including term loan, revolver, and miscellaneous. Dummy variables for loan purposes, including corporate purpose, debt repayment, working capital, takeover, and miscellaneous. 30