What do a bank’s legal expenses reveal about its internal controls and operational risk?

Accepted Manuscript Title: What do a bank’s legal expenses reveal about its internal controls and operational risk? Author: James E. McNulty Dean’s Distinguished Research Fellow and Professor of Finance Aigbe Akhigbe Moyer Chair and Professor of Finance PII: DOI: Reference:

S1572-3089(16)30122-X http://dx.doi.org/doi:10.1016/j.jfs.2016.10.001 JFS 486

To appear in:

Journal of Financial Stability

Received date: Revised date: Accepted date:

4-8-2015 30-9-2016 4-10-2016

Please cite this article as: McNulty, James E., Akhigbe, Aigbe, What do a bank’s legal expenses reveal about its internal controls and operational risk?.Journal of Financial Stability http://dx.doi.org/10.1016/j.jfs.2016.10.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

What do a bank’s legal expenses reveal about its internal controls and operational risk?

James E. McNulty1 [email protected], Aigbe Akhigbe2 [email protected] 1

Dean’s Distinguished Research Fellow and Professor of Finance Florida Atlantic University Boca Raton, FL 33431 2

Moyer Chair and Professor of Finance University of Akron Akron, OH 44325

Highlights • • • • • • • •

Consistent strong performance in banking requires a good system of internal control. High legal expense is indicative of weak internal control. Bank legal expense is not shown on regulatory reports. We find legal expense (measured by a proxy) is a strong determinant of loan losses and stock returns. Bank regulators should require reporting of legal expense on call reports to identify weak internal control. An analysis of litigation patterns is crucial in this process Current reporting requirements are inadequate and improvements are easy to implement. Current reporting leads to mispricing of bank securities.

Abstract Excessive (substantially above peer) litigation against a bank is indicative of operational risk because it often suggests failure to maintain a strong system of internal control. We examine the relation between bank performance and weak internal control using legal expense as a proxy. We find that legal expense is a strong determinant of loan losses and stock returns. Bank regulators should require reporting of legal expense on call reports to help identify institutions with weaknesses in internal control. Current reporting creates unnecessary information asymmetries because investors are not well informed about operational risk, leading to mispricing of bank securities.

*We thank Allen Berger, Scott Barnhart, Lamont Black, Steve Dennis, Bob DeYoung, Curt Hunter, Ed Kane, Steven Kane, Tom Lindley, Melinda Newman, Patricia Richardson, James Thompson, Larry Wall, other participants at the Financial Stability Conference and Annual Meetings of the Eastern Finance Association and Southern Finance Association, and two anonymous referees for very helpful comments and discussion. Professor McNulty also thanks the College of Business at Florida Atlantic University for Summer Research Grant support. We are solely responsible for any errors or omissions.

1. Introduction

Banks are more likely to fail from operational risk than from credit or market risk, and such risk has increased dramatically in recent years because of rapid technological change (Moosa, 2007). The due diligence failures that contributed to the 2007-09 US financial crisis reflect operational risk (Robertson, 2011). Weaknesses in internal control at banks create operational risk losses, and many institutions with such losses are repeat offenders (Chernobai et al., 2011). Internal control weaknesses such as inadequate training for employees and a failure to adhere to well-established policies and procedures are revealed in litigation against the bank. Consistently high legal expenses thus likely reflect internal control weaknesses, which are classic causes of banking problems. We test the hypothesis that high legal expenses in 2002-06 are reflected in weak bank performance during the 2007-09 US financial crisis. Because legal expense data are not reported on bank call reports, we use a unique data set reflecting the best available measure of bank legal expenses (our legal expense proxy) to test the hypothesis. These data are hand collected from annual 10K reports for 84 bank holding companies (BHCs). We expect a lag between excessive legal expenses and deteriorating bank performance. In the short run, aggressive and risky behavior may bolster bank earnings, but litigation may point to weakness in internal control. Weak internal controls such as violations of loan policies and aggressive lending practices may lead to loan delinquencies, but banks often restructure loans. Eventually provisions for loan losses increase, and the income statement is affected. We examine the effect of legal expenses on three measures of bank credit quality for both 2007 and 2008 – non-performing loans, loan charge-offs and loan loss provisions. We also analyze the effect on bank stock market performance – buy-and-hold returns (BHRs), and abnormal BHRs for 2007-08 combined. We also consider the effect on operating performance – return on assets and return on equity.

We provide several empirical tests, and the results are clearly consistent with our hypothesis. First, we find that loan losses and stock returns are substantially different for banks in the high-legal-expense decile relative to the low-legal-expense decile. This is consistent with the point that internal controls vary considerably from one BHC to another, and operational risk is similarly diverse (Moosa, 2007). Hence, these univariate results support the notion that if a BHC has very high legal expense relative to peer BHCs, there is a substantial likelihood of some potential weaknesses in internal control, leading to loan losses during the crisis. Virtually all the univariate results are consistent with the hypothesis and are economically significant, with differences in loan loss rates between the two groups of BHCs ranging from 21% to 69%, and differences in the two measures of stock returns of 92% and 203%. While economically significant, some of the differences in loan loss rates are not statistically significant. (We attribute the lack of statistical significance to the high variance in performance among banks during the crisis, and the small sample of BHCs in each decile.) Second, in the ordinary least squares (OLS) regression results using averages for the 2002-06 period for the independent variables, legal expense is significant with the expected sign in three of the seven regressions. Third and most importantly, in our panel regressions legal expense is significant at the one percent level with the expected sign in each of the three loan loss equations as well as the two stock returns regressions. There is also some evidence that high legal expense in 2002-06 is associated with lower return on assets for 2008, but at a lower level of significance. Most of 2007 results are similar. In summary, the legal expense proxy for 2002 through predicts credit quality and stock returns for 2007 and 08 with significance often at the one percent level.

This is the first research to find a direct relation between legal expenses and bank performance. The findings are consistent with the results of two recent studies on enforcement actions, a measure of corporate wrongdoing. BHCs with better corporate governance receive fewer enforcement actions from regulators (Nguyen et al., 2015). Enforcement actions are associated with negative stock market reactions (Zeiden, 2013). Hence, BHCs that conduct business in an ethical fashion – less enforcement actions and less lawsuits – have better performance. Both the theoretical reasoning and the empirical results strongly suggest that bank regulators should require separate reporting of total legal expenses. Regulators can strengthen market discipline with respect to internal controls and the resulting operational risk by the simple act of requiring disclosure of total legal expenses on publicly available bank call reports and BHC Y9 reports, and by incorporating legal expenses into the Uniform Bank Performance Reports (UBPRs) for both banks and BHCs. Of course, not all legal expense is associated with weakness in internal controls. Therefore, we suggest a two-step procedure for bank regulators and investors using these data. First, identify the banks and BHCs with consistently higher legal expense than peer banks. Second, determine which of these institutions has an ongoing and systematic pattern of similar litigation (or some very large litigation) that suggests weakness in internal control. This introduction is followed by Section 2, which discusses the literature on corporate culture, bank corporate governance, internal control and operational risk. Section 3 explores the relations among these four concepts and bank legal expenses to develop the hypothesis. Section 4 describes the data, and Section 5 presents descriptive statistics and empirical tests of the relation between bank performance and legal expenses. Section 6 presents the policy

implications and describes our related Working Paper which provides case study evidence consistent with our hypothesis, while Section 7 concludes.

2. Literature Review There is relevant economics and finance literature in four closely related areas: (a) corporate culture (CC); (b) bank corporate governance (CG); (c) the system of internal control (SIC); and (d) operational risk (OR). We suggest that weaknesses in the first two result in weak SICs; these in turn create OR and excessive (i.e., substantially above-peer) bank litigation (EBL). Figure 1 summarizes the hypothesized chain of causation. Reading Figure 1 in reverse, establishes our hypothesis, as discussed in Section 3. Codes of conduct (Brickley et al., 2002) and internal controls are the product of corporate culture. Firms have an economic interest in developing a culture (Lazear, 1995; Akerlof and Kranton, 2000, 2005; Akerloff, 2007). A firm selecting employees considers fit with the firm’s culture (Lazear, 1995; Van den Steen, 2005; Cronqvist et al., 2007). Empirical studies confirm that corporate culture affects financial decisions, likely because of persistent norms and beliefs possibly inherited from the firm’s founder (Cronqvist et al., 2007; Lemmon et al., 2008). The founder of Countrywide had a strong influence on that firm’s aggressive approach to mortgage lending before the financial crisis. In banking there is a crucial difference between a valuesdriven credit culture, marked by concern for loan quality, bank soundness, stability, and consistency, and a current-profit-driven credit culture characterized by a high tolerance for risk (Koch and MacDonald, 2010). Banking consultants recognize the significance of corporate culture in dealing with risk management (Hall, 2012). Studies find a positive relation between socially responsible behavior and financial performance (Frooman, 1997; Preston and O’Bannon, 1997; Roman et al 2007; Waddock and Graves, 1997). In banking, socially responsible behavior

improves return on assets and other measures of financial performance (Simpson and Koehrs, 2002). Boards of directors are responsible for determining the strategic direction of a corporation (Adams et al. 2010) and ensuring the institution has a good system of internal control. The board audit committee is responsible for identifying deficiencies in internal control and enforcement of a code of ethical conduct (Rezaee, 2009). Consequently, bank holding companies with better corporate governance were more profitable and suffered fewer real estate loan losses during the 2007-09 US financial crisis (Peni et al., 2012, 2013). The crisis has heightened regulators’ expectations of bank directors (Institute of International Finance, 2009; Comptroller of the Currency, 2014; Nguyen et al., 2015). As of August 8, 2013, the Federal Deposit Insurance Corporation (FDIC) had filed 76 lawsuits against officers and directors of banks that failed during the 2007-09 US financial crisis, with individual damage claims up to $600 million (Cornerstone, 2013). A bank with an independent board has a much lower probability of committing misconduct (Nguyen et al., 2015). Enforcement actions do not have a significant negative effect on operating performance (Zeiden, 2012), but there is a negative stock market reaction to announcement of an enforcement action against a bank (Zeiden; 2013). Corporate governance, the system of internal control and corporate culture are closely related (Rezaee, 2009).1 But internal control research is often not conducted in a corporate governance framework and there is disagreement about what internal controls really are (Power, 1997; Maijoor, 2000). Nonetheless, this literature does consider broad monitoring mechanisms, as well as bonding mechanisms and reward systems (Maijoor, 2000). There is some consensus that the key concepts involve people, lines of authority, segregation of duties, organizational structure, measures to protect assets, and policies,

precisely the concepts that are used to control banking risk. This literature recognizes that culture can be an important control mechanism (Maijoor, 2000). Regulators state “good” internal controls exist when “no one person is in a position to make significant errors or perpetuate significant irregularities without timely detection” (Comptroller of the Currency, 1998, p. 2). They also emphasize adequate training, competent internal audit, and adherence to managerial policies. Studies of bank failures involve evaluating weaknesses in systems of internal control. Internal control at a very large BHC with investment banking subsidiaries should be much more complex than at a small BHC with a community bank as its major subsidiary, because of differences in the number of hierarchical layers (Abdel-Khalik, 1993). Regulators need to be particularly aware of weakness in internal control at large and systematically important BHCs (e.g., Comptroller of the Currency, 2014). These institutions have many layers, and the consequences of weak internal control are very serious. Operational risk has attracted much more attention because of technology, competition, and globalization as well as new lines of business. This risk has increased substantially due to rapid technological change; a major bank is perhaps more likely to fail from operational risk than from credit risk or market risk (Moosa, 2007). Operational risk depends on the culture of the business units and can result from groupthink (Rao and Dev, 2006; Moosa, 2007. Operational risk is “the risk of loss from inadequate or failed internal processes, people and systems, or from external events” (Robertson, 2011, p. 1). The due diligence failures that led to the financial crisis are a form of operational risk. Importantly, he sees the entire crisis as “born of operational risk” (p. 4) because internal controls were not in place. Securitization transmitted operational risk from one institution to another, creating a domino effect throughout the international financial system. Hence, the financial crisis clearly highlights the importance of effective operational risk

management. There is a close relation between operational risk and credit risk. Persistent failure to have in place or to follow good procedures for evaluating credit is a form of operational risk (Moosa, 2007). The explosion of mortgage credit in the pre-crisis period in areas where income was actually declining, relative to more affluent and positive income growth areas (Mian and Sufi, 2009, 2010) highlights a lack of due diligence in the pre-crisis period, which led to considerable litigation. The weakness in due diligence and internal control described in the Countrywide, National City and Wachovia cases in McNulty and Akhigbe (2016), and the resulting litigation, reflect this type of operational risk. The classic operational risk case is the failure of Barings Bank in 1995 where a single trader created a $1.3 billion loss, partly because the appropriate policies were not in place and supervision was inadequate. The near failure of a French bank in 2007-08 is also attributed to very weak internal controls (Bisserbe, 2016). A failure to adhere to standard banking practices and internal policies as part of a strong system of internal control, including proper supervision of employees, is a common cause of operational risk losses. Our argument that operational risk reflects the corporate culture and that the causes of OR are thus internal to the firm is consistent with Chernobai et al. (2011). Based on OR events involving US financial institutions from 1980 through 2005, they conclude that factors internal to the firm contribute to operational risk losses. Consistent with our hypothesis, most operational losses in their study reflect a breakdown in internal control, and most resulted in litigation. Mergers and acquisitions are a cause of operational risk as previously separate information systems need to be integrated (Cummings et al., 2006). As discussed in McNulty and Akhigbe (2016), Wachovia, formed through a very large number of mergers over many years, experienced considerable unusual OR-related litigation failed because of its merger with Golden West Financial. Wachovia’s CEO commented “this [merger] will either cement my

reputation or get me fired” (Lowenstein, 2010, p. 70). This frank admission suggests a lack of sufficient due diligence before a large merger, a form of OR. Hence the notion that corporate culture and ethics, corporate governance, the system of internal control and operational risk are linked as shown in Figure 1 is well grounded in the literature. An aggressive corporate culture with an emphasis on short-term earnings is counterproductive to a good system of internal control, and often leads to excess bank litigation. Hence, excess bank litigation is an indicator of weaknesses in internal control. This last point forms the basis for our hypothesis and our empirical strategy. We emphasize that the relations summarized in Figure 1 apply to most banks in most countries in virtually all periods of time. The breakdown in internal controls and the resulting operational risk is most apparent in precrisis and crisis periods, but the concepts are much broader.

3. Hypothesis development and regression equations 3.1. Hypothesis development Figure 1, introduced above, shows our hypothesized links among the four areas of the literature discussed above and excess bank litigation (EBL). Corporate Governance (CG) has a major influence on the corporate culture (CC) because it establishes the framework in which managers operate and affects the system of internal control (SIC). Boards of directors are responsible both for establishing a code of ethical conduct and for ensuring that the institution has a good SIC (e.g., Rezaee, 2009). Further, an independent board of directors can reduce enforcement actions (Nguyen et al. 2015). Enforcement actions involving significant and repeated violations clearly reflect a weak SIC. Internal control systems arise from this complex interactive process; they determine the level of operational risk (OR) and the level of bank litigation; weak internal control systems

create greater operational risk. OR has long been considered to include legal risk (Moosa, 2007; Koch and MacDonald, 2010), but we show OR and Excess Bank Litigation (EBL, defined as high legal expense relative to peer banks) separately here. This suggests that an aggressive CC may result in a weaker SIC. A greater emphasis on current profits in the CC may cause the bank to short-circuit needed training programs; weaknesses in policies and procedures may be tolerated when high-risk loans (with high yields) are placed on the books. The case studies in McNulty and Akhigbe (2016) show how weaknesses in the SIC at three large BHCs that failed or were merged during 2007-08 contributed to the crisis. Reading Figure 1 in reverse, the novel part of our argument is that excess bank litigation is an indicator of a weak SIC. Clearly, a bank without a good SIC is more likely to be sued than other banks. Litigation against banks can frequently be traced to the perpetuation of irregularities by one person or a few people in the organization without detection for a significant period of time; this is a weak system of internal control by definition. Hence, excessive legal expense should serve as a red flag to help regulators and investors identify differences among banks in the adequacy of internal controls. It is likely that, other things equal, a bank that is sued much more than peer banks, and sued repeatedly for the same reasons, has weaker internal controls than other banks. This should lead to significantly weaker bank financial performance after a period of time. It takes time after risky or poorly constructed loans are made for them to go bad. Hence, we posit a lag between excessive legal expense and the deterioration in bank performance. In the short run, aggressive and risky behavior may bolster bank earnings. Complex bank litigation often goes on for many years. The full extent of deficiencies in the system of internal control are

often evident only when litigation is resolved, which may be many years later, often when new managers are in place.2 Given this reasoning, the basis of our empirical analysis is the following relationship:

N

PERFORM jT = α0 + α1LEGAL jt + ∑αiCONTROLijt + µ j

(1)

i =2

PERFORMjT represents the financial performance of bank j in period T (T = 2007 and 2008), which we measure by credit quality, operating performance and stock returns. LEGALjt is our legal expense proxy for bank j in period t (t= 2002 through 2006); CONTROLijt represents a vector of N control variables (i = 2…N) for bank j in period t (t = 2002 through 2006). The hypothesis we test in our regression analysis is that higher legal expense from 2002 through 2006 is associated with weaker financial performance in 2007 and 2008. It is the nature of the banking business that the lag is long and variable, and hence virtually impossible to identify a priori. These long and variable lags can be up to six years (2002 to 2008). We measure financial performance (PERFORM) by credit quality (loan losses), operating performance and stock returns. Following Peni et al. (2012, 2013) we are interested in the determinants of loan losses, because loan losses were the primary cause of the crisis. Finance is ultimately about value, so we also consider stock returns. We also consider operating performance to place our study in the context of the other bank performance literature. We expect a positive relation between LEGAL and loan losses for 2007-08; we expect a negative relation between LEGAL and both stock returns and operating performance for 200708.

3.2. Regression Equations

Based on the above reasoning, we estimate two regression equations:

LOANLOSS j = α0 + α1LEGAL j + α2 ASSETS j + α3CAPITAL j + α 4 FHCO j + α5 HHI j + α6 MB j + α7 MERGE j + α8 R (2)

RETURN j = α0 + α1LEGAL j + α2 ASSETS j + α3CAPITAL j + α4 FHCO j + α5 HHI j + α6 MB j + α7 MERGE j + α8 ROE (3)

For ease of exposition we do not show the lags in the regression equations. Among the dependent variables LOANLOSSj represents three measures of credit quality for bank j for 2008 relative to end-of-period assets for the same year: (a) LOAN CHARGE-OFFS/ASSETS08, (b) LOAN LOSS PROVISIONS/ASSETS08, and (c) NON-PERFORMING LOANS/ASSETS08. We also run the regressions using the same loan quality data for 2007. RETURNj represents two measures of stock returns for bank j measured from January 1, 2007 to December 31, 2008: (a) abnormal-buy-and-hold returns (ABHR07-08, i.e., the difference between bank buy-and-hold returns and market-buy-and-hold returns), and (b) buy-and-hold returns (BHR07-08). RETURNj also represents accounting returns (operating performance), namely return on assets and return on equity for 2007 and 2008. The explanatory variables consist of legal expense and the control variables computed separately for each bank for the period 2002-06. LEGALj is our legal expense proxy/assets for

bank j; ASSETSj is the natural logarithm of total assets for bank j; CAPITALj is the ratio of total equity capital to total assets for bank j; FHCOj is an indicator variable equal to one if the BHC represented by bank j is a financial holding company, and zero otherwise; HHIj is the sum of the squared market shares for bank j, a measure of local market concentration; MBj is the market value of total assets for bank j divided by their book value for 2002 through 2006; MERGEj is an indicator variable equal to 1 for BHCs that were involved in mergers and acquisitions in the 2002 - 2006 period, and zero otherwise; ROEj is the ratio of net income to the book value of equity for bank j for 2002 through 2006. ASSETSj and MBj are the Fama-French (1993) factors commonly used to analyze stock returns. 3.3. Rationale for control variables and other econometric issues

The reasons for the control variables are as follows: ASSETS is included because banks of different sizes often have different lending strategies; these may produce a different loan loss experience and different stock returns. ASSETS is also a Fama-French (1993) factor. We include the Fama-French variables, ASSETS and MB in equation (4) to be consistent with equation (5). CAPITAL and the HHI have an important influence on bank performance in a large number of studies. BHCs that formed a financial holding company (FHCO) after passage of the GrammLeach Bliley Act in 1999 may also have a more aggressive business strategy. Mergers (MERGE) are included because banks involved in a merger or acquisition may have a different loan loss experience than other banks. More importantly, they would have higher legal and accounting expenses as a result of the merger. Data on accounting expense is included in our legal expense proxy, so we need to control for the higher proxy that would be reported by BHCs involved in mergers. ROE is included because banks may be highly profitable in one period because of an

aggressive lending strategy that may produce losses or lower profits in later periods. We also include state dummy variables. These control variables are similar to those used to analyze bank performance and risk in other studies (e.g., Berger and DeYoung, 1997; Berger and Mester, 1997; Akhigbe and Martin, 2008; Peni et al., 2012, 2013). We include ASSETS, CAPITAL, FHCO, HHI, MB, and ROE in Equation (3) for the same reasons as in equation (2). At first glance there may appear to be simultaneous equation bias in these relations. We posit that nonperforming loans depend on legal expense because legal expense is one measure of the system of internal control. But legal expense depends on nonperforming loans because the expenses of collection often involve legal fees. However, the model is a lagged relationship, which eliminates this problem. The lags are long and variable and can be up to six years (2002 to 2008). The model is not based on an accounting relationship.

4. Data We draw our data from four sources: Legal expense. We examine annual 10K reports for over 150 BHCs for 2002-06, the period prior to the financial crisis. We are able to hand collect usable data on the legal expense proxy for 102 institutions. Only 84 of these institutions have stock return data available from CRSP3. The proxy includes payments to law firms for all BHCs; these will definitely be higher when a bank is sued more frequently. As shown in Table 2, we identify two accounting models used to report non-interest expense in BHC 10K reports. Accounting Model 1, described in Ryan (2007), has six categories under total non-interest expense: personnel; occupancy; technology and communications; deposit insurance; advertising; and other. “Other” expense includes an extremely large number of items

in addition to legal expense. Ryan reports that this format meets all accounting and disclosure requirements. Citigroup produces a 10K report using Accounting Model 1. Peer analysis of “other expense” from this accounting model would be meaningless. We exclude all BHCs following Accounting Model 1 from our analysis. The banks we include in our sample all follow the more detailed Accounting Model 2. The BHCs we include in the regressions and in the rankings are the institutions that have data available from CRSP to compute stock returns and also report a separate item under non-interest expense, most often generally entitled “professional fees.” This is the legal expense proxy. McNulty and Akhigbe (2016) reports how these 84 BHCs report legal expense. There is a high degree of reporting consistency; most use the same or very similar terminology. Considering the point that many BHCs use the same accounting firms, these data can be used in both the regression analysis and the case study rankings with assurance that the same or very similar items are being reported across the sample. The data include payments to law firms for all BHCs. This measures the first step in the litigation process, and payments to attorneys would be an ongoing expense until the matter is resolved. Settlements are reported as other operating expense in the 10K reports and hence are not included in the proxy. Importantly, the within-firm variation from year to year in the proxy is over 2.5 times the variation in total non-interest expense; a good deal of this variation can be attributed to litigation, which of course can vary substantially from year

to

year

as

cases

arise

and

are

ultimately

resolved.
class="xps_endnote">4 Financial Crisis Buy-and-Hold Returns. We use simple buy-and-hold returns (BHR0708) and abnormal returns (ABHR07-08, as defined above) as additional measures of bank

performance during the 2007-09 financial crisis. We measure stock returns for the two-year period ending December 31, 2008, the period where the crisis had the most impact. BHC Balance Sheet and Income Statement Data. Data for non-performing loans, assets, book value, net income, and financial holding company come from the Federal Reserve Bank of Chicago’s BHC database. The ratios were computed separately for each year and then averaged. Mergers and Acquisitions. We use Lexis/Nexis to identify those BHCs that were involved in a merger or acquisition during the period 2002-06. 5. Descriptive statistics and regression results5 Table 3 shows descriptive statistics for the sample. Abnormal buy-and-hold returns for 2007-08 (ABHR07-08) average -5.11% and range from -82.66% to +79.99%. Simple Unadjusted BHRs (BHR07-08) average -43.67% and range from -96.12% to +39.99%. LOAN CHARGEOFFS/ASSETS07

averages

0.42%

and

ranges

from

zero

to

2.75%.

LOANCHARGEOFFS/ASSETS08 averages a much higher 1.08% and ranges from zero to 8.43%. The means and distributions of the other loan quality measures for both years are similar to the first one. The independent variables are for 2002 to 2006. Our legal expense proxy/assets (LEGAL)

averages

0.13%

and

ranges

from

zero

(rounded)

to

0.77%.
class="xps_endnote">6 Since the median (0.11%) is fairly close to the mean, the data have

some

of

the

characteristics

of

a

normal

distribution.
class="xps_endnote">7 Total ASSETS of the BHCs average $56.24 billion. The BHCs range in size from $269 million to almost $1.5 trillion. The CAPITAL ratio averages 8.91% and ranges from 4.75% to over 22%. 38.23% of the BHCs are part of a financial holding company. The HHI averages 20.72% and ranges from 12.07% to 71.47%. MB averages 266% and ranges from 108% to 1,030%. 83.09% of the BHCs were involved in a merger or acquisition

during the sample period. NPL averages 0.59% and ranges from zero to 5.71%. Return on equity averages 18.60% and ranges from -43.78% to +47.66%. Table 4 shows average performance measures for the BHCs in the top decile (substantially above peer) on the basis of LEGAL and those in the bottom decile (substantially below peer), and significance tests, absolute differences and percentage differences. All but one of the differences are consistent with our hypothesis, but since the samples are small (34 to 40 observations) and the variance in performance during the crisis is high, many of the differences that are quite substantial are not statistically significant. LOAN CHARGE-OFFS/ASSETS07 averages 43.2% higher for the high-legal-expense banks (those that would be expected to have weaker systems of internal control) and LOAN CHARGE-OFFS/ASSETS08 is 29.7% higher, but the differences are not statistically significant. LOAN LOSS PROVISIONS/ASSETS07 is 69.1% higher for the first group; this difference is significant at the five percent level. LOAN LOSS PROVISIONS/ASSETS08 is 21% higher and the difference is not statistically significant. NONPERFORMING

LOANS/ASSETS07

is

41.1%

higher,

and

NONPERFORMING

LOANS/ASSETS08 is 31.7% higher, but these differences are not statistically significant. Two of the operating performance results show smaller positive differences (9.9% and 18.1%) which are not statistically significant. The difference in average ROA08 of -15.9% is contrary to the hypothesis and is not significant. The large +115% percent difference in ROE08 is attributable to the small base (2.23%) for the denominator; even this difference is not statistically significant. In contrast, the differences in the two stock return measures are both economically and statistically significant. BHR07-08 is 91.6% higher (i.e., less negative) and the difference is significant at the one percent level. ABHR07-08 is 202.5% higher and that difference is significant at the one

percent level. Interestingly, in this last comparison abnormal returns for the low legal expense banks (those that, based on our reasoning, had better internal controls than the others) are actually positive during the crisis. Internal controls vary widely from one BHC to another and operational risk is similarly diverse (Moosa, 2007). The results in Table 4 indicate, consistent with our reasoning throughout this paper, that LEGAL does indeed identify those banks with likely weaknesses in internal controls. The resulting operational risk and loan losses became apparent only during the crisis. These results constitute an important finding supporting our argument that the BHCs with very high legal expense relative to peer BHCs (reflecting potential weaknesses in internal control) need to be identified by regulators early before the banks become serious problems for the financial system. (The two-step procedure discussed in Section 6.1 should be used to identify these banks.) Table 5 shows Pearson correlation coefficients between legal expense and all the bank performance variables for 2007 and 2008. All the correlations have the expected signs. The correlation between LEGAL and the loan loss measures is positive while for the return measures it is negative. The correlations between LEGAL and the three loan loss measures, LOAN CHARGE-OFFS/ASSETS07,

LOAN

LOSS

PROVISIONS/ASSETS07,

and

NON-

PERFORMING LOANS/ASSETS07, are all significant at the one percent level with the expected

positive

sign.

For

LOAN

CHARGE-OFFS/ASSETS08,

LOAN

LOSS

PROVISIONS/ASSETS08 and NON-PERFORMING LOANS/ASSETS08 the correlation coefficient are significant at the five percent level. The correlation between LEGAL and ABHR07-08 is significant at the ten percent level.

Table 6 shows the OLS results for the estimation of equations (2) and (3). All the independent variables are averages for 2002-06. The legal expense proxy for 2002-06 predicts all three measures of loan quality for 2008 with statistical significance at the ten percent level in two equations and at the five percent level in the third. We also run the same regression equations using all three 2007 credit quality measures as dependent variables. LEGAL predicts CHARGEOFFS/ASSETS07, LOAN LOSS PROVISIONS07 and NON-PERFORMING LOANS/ASSETS07 at the five percent level of significance, all with the expected positive sign. These results are not shown here to conserve space. There are no variables that are measured contemporaneously with the dependent variable in these regression equations. Thus, the point that our legal expense proxy averaged for 2002 through 2006 predicts all three measures of loan quality for 2007 and 2008 demonstrates empirically the relation between bank legal expense and future bank performance. Nonetheless, in the OLS regressions, LEGAL does not predict either stock returns variable, return on assets or return on equity for 2007 or 2008. LEGAL does predict stock returns in the panel regressions reported below. Many other factors in addition to loan losses affect ROA and ROE, especially during the crisis years. Operating performance in 2008 would be especially difficult to predict with 2002-06 data, no doubt due to the difficult economic environment which affected many components of bank profitability in that crisis year. For example, personnel expense was affected in a major way as many banks restructured operations, and some had major layoffs. Considering the control variables in Table 6, the coefficient of ASSETS is negative and significant at the five percent and ten percent levels in the two stock returns equations, indicating that smaller banks had less favorable returns in 2008. These banks had less chance of being

rescued by regulators and would thus have been considered more likely to fail or be merged at low stock prices. CAPITAL is only significant in the ABHR07-08 equation. FHCO is not significant throughout. The HHI is positive and significant at the five percent level in two of the three loan loss equations, indicating that banks in concentrated markets had greater loan losses in 2008. MB is negative and significant in the ROE equation at the ten percent level. MERGER is not significant in any of these equations, while ROE02-06 is significant at the one percent level in the ROE08 equation. The disadvantage of the averaging technique used to construct the data in Table 6 is that it does not make use of all the scarce information on legal expense for individual years. We noted earlier the high within-firm variation in legal expense. Internal controls and operational risk are diverse; they reflect multiple causes that differ substantially from bank to bank in terms of their nature, magnitude and timing. Hence, it is possible that a large legal expense ratio in one or two years would indicate such weaknesses, but this would be obscured, at least in part, by averaging LEGAL over a five year period. We take advantage of all the data in the panel regressions in Table 7, which are perhaps the most important empirical results in the paper. We use a random effects procedure. Table 7 shows that LEGAL predicts all three measures of loan losses and the two stock returns measures with statistical significance at the one percent level. It also predicts ROA08 at the ten percent level. All six of these coefficients have the sign that is consistent with our hypothesis. High legal expense is associated with higher loan losses, lower stock returns and lower return on assets for 2008. The OLS and panel results for 2007 are similar and are not shown here to conserve space.8

Among the control variables, the coefficient of ASSETS is positive in the loan loss equations and negative is the stock returns equations. Larger banks had greater loan losses and lower stock returns. Banks with higher CAPITAL had higher stock returns. BHCs that were part of a financial holding company (FHCO) had lower loan losses and higher returns on assets and equity. Banks in less competitive markets, i.e., those with higher HHIs, have higher loan losses and lower returns on assets and equity, but stock returns are not affected. The market to book (MB) and MERGER variables have only marginal significance. In contrast, high ROE02-06 is associated with higher loan losses in 2008. This is consistent with the argument noted previously that banks with more aggressive strategies in the pre-crisis period, as reflected in higher return on equity, may experience greater problems during the crisis. This result is significant at the ten percent level in two of the three loan loss equations. These results are mixed as higher ROE0206 is also associated with higher abnormal stock returns (ABHR07-08) at the ten percent level of significance. The case studies in McNulty and Akhigbe (2015) (2016) provide additional evidence supporting the hypothesis that high legal expense and a pattern of litigation against the bank predicts weak future bank financial performance. Additional empirical tests which reveal the same relation in the context of corporate governance are in McNulty and Akhigbe (2015). We find that, if we attempt to explain which banks had greater loan losses during the crisis, our legal expense proxy has the expected sign and much higher statistical significance that the conventional corporate governance measures used by Penni et al (2012) (2013). This paper also provides four additional case studies illustrating the importance of legal expense. 6. Policy implications and suggested implementation

6.1. Policy implications In McNulty and Akhigbe (2016) we show that during 2002-12 less than 15% of BHCs reported legal expense on Y9 reports, and in recent years there has been no reporting on call reports.

This is an important omission because one indicator of weaknesses in internal control

is legal expense significantly above peer institutions. Further, no recent early warning model of bank financial distress published by economists in the bank regulatory community that we could identify contains legal expenses as a predictor (Guenther and Moore, 2003; Jagtiani et al., 2003; Whalen, 2010). Because banks are not required to report legal expense, managers can mislead investors, regulators, and themselves about such risk. Further, bank managers can use protracted litigation to hide a corporate culture characterized by weak internal controls (e.g., Wachovia). Operational risk can be mitigated by market discipline. The results of managerial actions should be transparent so managers are discouraged from taking actions, or creating and perpetuating a corporate culture, detrimental to the long-run interests of stakeholders. In banking this could work through the stock, bond, and money markets. Effective market discipline does not require that all investors in bank securities understand and act on legal expense data; it only requires that the marginal investor do so. To facilitate market discipline, regulators should require reporting of bank legal expense (item 4141) on both bank call reports and BHC Y9 reports. The ratio of legal expense to assets and to revenue should then be incorporated into the Uniform Bank Performance Reports (UBPRs) for both banks and BHCs, which are publically available. (The UBPR shows what percentile the institution is in relative to peer institutions for over one hundred financial ratios.) Institutions consistently in the top percentiles would be easily identified; securities analysts and investors could examine individual cases to see if there is a pattern of litigation that reflects

weaknesses in internal control. This operational-risk-related market discipline would improve the functioning of bank securities markets, enhancing overall economic welfare. It should also improve banking performance by providing greater incentives for managers to improve systems of internal control. The data should be incorporated into early warning models. If weak systems of internal control are identified by bank examiners through the analysis of legal expense, this should be considered when determining the “Management” component of the bank’s rating in the CAMELS system.9 Federal bank examiners are required to determine if there is a pattern of excessive litigation that puts the bank at risk (Comptroller of the Currency, 2000). But how can examiners have a good measure of how much litigation is normal for a given peer group of banks when the legal expense data are not reported? Bank compliance officers and directors could also use the data to evaluate their internal controls relative to peer banks. Securities analysts, investors, regulators, bank compliance officers and bank directors all have responsibilities with respect to internal controls and operational risk. We suggest a two-step procedure. First identify the banks with consistently high legal expense relative to peer banks using the UBPR. Second, for the banks with consistently high expense, evaluate the litigation against that bank to determine if there is an ongoing and systematic pattern of similar litigation that suggests weakness in internal control. The cases in McNulty and Akhigbe (2015 (2016) provide concrete examples of the connection between internal control, operational risk and bank litigation and illustrate how adverse litigation patterns at these BHCs could have provided an early warning to bank regulators in the period before the crisis. To help identify which banks can best handle mergers, analysts at bank regulatory

agencies could study legal expense and litigation patterns after previous mergers have taken place, using the two-step procedure. 6.2. Value enhancing vs. value destroying litigation Certainly, not all bank legal expense reflects operational risk. Some legal expenses, such as the cost of preparing loan documents and defending the bank against unfounded lawsuits, are normal, and some mistakes by bank personnel that lead to litigation may be inevitable, even in the best managed banks. Further, a BHC may be in litigation when it has done nothing wrong. In addition, in the process of financial innovation a bank often needs to involve attorneys to be sure the rights and claims of all parties to a transaction are protected. Much major financial innovation over the last 50 years or more would have required significant input from attorneys. Much of this would be normal legal expense. Our concern is legal expense that is substantially above normal, i.e., well in excess of that of peer institutions. (The regression procedure automatically identifies legal expense that is above the mean.) While some legal expense is normal and necessary (i.e., value enhancing) our empirical results indicate that a significant portion of excess expense reflects value-destroying bank operational risk. Further, high legal expense alone should not subject a bank to criticism, and should not necessarily cause investors to sell bank securities. Extremely high legal expense on a consistent basis provides a signal – a red flag – that securities analysts and bank examiners should look further by examining litigation patterns. What is the bank being sued for, is there a pattern of similar litigation, and do the bank’s alleged actions involve weak internal controls and operational risk?10

6.3. Alternatives for legal expense reporting

There are at least two choices for reporting: (a) require BHCs to fill in the blank space on the current call report for item 4141 (“legal expenses and fees”) or (b) break legal expense into components. One simple breakdown would be to divide legal expense into two categories: (a) expense associated with litigation and (b) other legal expense.

7. Conclusions

Excessive (substantially above peer) litigation against a bank is indicative of operational risk since it often suggests failure to maintain a strong system of internal control. We examine the relation between weak internal control and future bank performance using legal expense as a proxy for weak internal control. Using a unique hand collected data set representing legal expense for 84 BHCs for the pre-crisis period, we find evidence of a statistically significant relation between legal expense and our measures of performance. Specifically, we find BHCs with high legal expense in 2002-06 have higher loan losses and lower stock returns during the crisis period 2007-08. The OLS and panel regression results are both consistent with the hypothesis and the correlations and the comparison of the performance measures for the low- and high-legal-expense deciles provide additional empirical support. Our paper is the first to find that legal expense (as a measure of weak internal control) predicts future bank performance. But the findings are consistent with two recent studies considering enforcement actions by bank regulators as indicator of corporate wrongdoing. BHCs with better corporate governance receive fewer enforcement actions from regulators (Nguyen et al., 2015) and enforcement actions are associated with negative stock market reactions (Zeiden, 2013). Putting the three studies together indicates that BHCs that conduct business in an ethical fashion with less enforcement actions and less lawsuits have better performance.

Regulators should require legal expense reporting and use a two-step procedure – first identify the institutions with consistently high legal expense than peers; second, determine which has an ongoing and systematic pattern of similar litigation suggesting possible weakness in internal control. Current reporting creates unnecessary information asymmetries because investors are not as informed as they should be about operational risk, no doubt leading to mispricing of bank securities. The logic of the relation between high legal expense and weak internal control applies at all times to banks in all countries, so legal expense should be a useful measure globally. Appendix Table A1 This table presents the definitions of the dependent and independent variables. Dependent Variables: ABHR07-08 = abnormal buy-and-hold returns (the difference between the bank buy-and-hold return and the market buy-and-hold return) for 2007-08; BHR07-08 = bank buy-and-hold returns for 2007-2008; LOAN CHARGEOFFS/ASSETS07 = loan charge-offs as a percent of total assets for 2007; LOAN CHARGEOFFS/ASSETS08 = loan charge-offs as a percent of total assets for 2008; LOAN LOSS PROVISIONS/ASSETS07 = loan loss provisions as a percent of total assets for 2007; LOAN LOSS PROVISIONS/ASSETS08 = loan loss provisions as a percent of total assets for 2008; NON-PERFORMING LOANS/ASSETS07 = non-performing loans as a percent of total assets for 2007; NON-PERFORMING LOANS/ASSETS08 = non-performing loans as a percent of total assets for 2008; ROA07 = return on assets, the ratio of net-income to assets for 2007; ROA08 = return on assets, the ratio of net-income to assets for 2008; ROE07 = return on equity, the ratio of net-income to equity for 2007; ROE08 = return on equity, the ratio of net-income to equity for 2008. Independent variables (all measured for 2002-06): ASSET = the book value of total assets for 2002-06 ($billions); CAPITAL = the ratio of total equity capital to total assets for 2002-06 FHCO = an indicator variable equal to one for a BHC that is a financial holding company in 2002-06, and zero otherwise; HHI = the Hirschman-Herfindahl index for 2002-06; LEGAL = our legal expense proxy as a percent of total assets for 2002-06; MB = the ratio of the market value of equity to its book value for 2002-06; MERGE = an indicator variable equal to one for BHCs that were involved in mergers and acquisitions in the sample period, and zero otherwise; ROE = return on equity, the ratio of net-income to equity, for 2002-06. The sample consists of 408 bank-year data points for 84 bank holding companies (BHCs) for the period 2002-06 for the independent variables (data is missing for some BHCs for some years) and comparable observations on the dependent variables for the same BHCs for 2007-08. The ratios are computed separately for each year and then averaged.

References Abdel-Khalik, A.R., 1993. Why do private companies demand auditing? A case for organizational loss of control. Journal of Accounting, Auditing and Finance 8, 31-52.

Adams, R.B., Hermalin, B.E., Weisbach, M.S., 2010. The role of boards of directors in corporate covernance: A conceptual framework and survey. Journal of Economic Literature 48, 58-107.

Akerlof, G.A., 2007. The missing motivation in macroeconomics. American Economic Review 97, 5-36.

Akerlof, G.A., Kranton, R.E., 2000. Economics and identity. Quarterly Journal of Economics 110, 715-53.

Akerlof, G.A., Kranton, R.E., 2005. Identity and the economics of organizations. Journal of Economic Perspectives 19, 9-32.

Akhigbe, A., McNulty, J.E., 2003. The profit efficiency of small US commercial banks. Journal of Banking and Finance 27, 307-325.

Akhigbe, A., Martin, A.D., 2008. Influence of disclosure and governance on risk of US financial services firms following Sarbanes Oxley. Journal of Banking and Finance 32, 2124-35.

Arjoon, S., 2005. Corporate governance: An ethical perspective. Journal of Business Ethics 61, 343-52.

Basel Committee on Bank Supervision, 2006. Basel II: International Convergence of Capital Measurement and Capital Standards: A Revised Framework – Comprehensive Version. Bank for International Settlements.

Berger, A.N., DeYoung, R., 1997. Problem loans and cost efficiency in commercial banks. Journal of Banking and Finance 21, 849-70.

Berger, A.N., Mester, L.J., 1997. Inside the black box: What determines differences in the efficiency of financial institutions? Journal of Banking and Finance 21, 895-947.

Bisserbe, N., 2016. SocGen ordered to pay rouge trader. Wall Street Journal (June 8).

Brickley, J.A., Smith, C.W., Zimmerman, J.L., 2002. Business ethics and organizational architecture. Journal of Banking and Finance 26, 1821-35.

Chernobai, A., Jorion, P., Yu, F., 2011. The determinants of operational risk in US financial institutions. Journal of Financial and Quantitative Analysis 46, 1683-1725.

Comptroller of the Currency, 1998. Comptroller’s handbook for national bank examiners. Washington, D.C.

Comptroller of the Currency, 2000. Comptrollers handbook: Litigation and other legal matters. Washington, D.C.

Comptroller of the Currency, 2014. OCC proposes formal guidelines for its heightened expectations for large banks. Press release (January 16).

Cornerstone Research. (2013). Characteristics of FDIC lawsuits against directors and officers of failed financial institutions. (Available at www.cornerstone.com.)

Cronqvist, H., Low, A., Nilsson, M., 2009. Persistence in firm policies, firm origin, and corporate culture: Evidence from corporate spin-offs. Robert Day School of Economics and Finance Research Paper No. 2009-2. (Available at SSRN: http://ssrn.com/abstract=954791.

Cummings, J.D., Lewis, C.M., Wei, R., 2006. The market value impact of operational loss events for US banks and insurers. Journal of Banking and Finance 30, 2605-2634.

Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56.

Frooman, J., 1997. Socially irresponsible and illegal behavior and shareholder wealth: A meta-analysis of event studies. Business and Society 36, 221-249.

Guenther, J.W., Moore, R.R., 2003. Early warning models in real time. Journal of Banking and Finance 27, 1979-2001.

Hall, L.T., 2012. A toxic culture can be hazardous to your bank’s balance sheet. American Banker (November 27).

Institute of International Finance, 2009. Governance of financial institutions. New York, NY.

Jagtiani, J., Kolari, J., Lemieux, C., Shin, H., 2003. Early warning models for bank supervision: simpler could be better. Federal Reserve Bank of Chicago Economic Perspectives 27, 49-60.

Koch, T.W., S. Scott MacDonald, S.S., 2000, 2010. Bank Management. Fourth and Seventh Editions, Dryden, Mason, OH.

Lazear, E.P., 1995. Corporate culture and the diffusion of values. In Siebert, H., Trends in Business Organization. JCB Mohr, Tubigen, Germany.

Lemmon, M.L., Roberts, M.R., Zender, J.F., 2008. Back to the beginning: Persistence and the crosssection of corporate capital structure. Journal of Finance 63, 1575-1608.

Lowenstein, R., 2010. The end of Wall Street. Penguin Press, New York, NY.

Maijoor, S., 2000. The internal control explosion. International Journal of Auditing 4, 101-109.

McNulty, J.E., Akhigbe, A., 2015. Toward a better measure of bank corporate governance. International Corporate Governance: Advances in Financial Economics, 18, 81-124. [available at http://ssrn.com/abstract = 2843639].

McNulty, J.E., Akhigbe A., 2016, What do a bank’s legal expenses reveal about its internal controls and operational risk? With appendicies. [available at http://ssrn.com/abstract = 2848812.]

Merton, R.C., 1977. An analytic derivation of the cost of deposit insurance and loan guarantees: An application of modern options pricing theory. Journal of Banking and Finance 1, 3-11.

Merton, R.C., 1978. On the cost of deposit insurance when there are surveillance costs. Journal of Business 51, 439-52.

Mian, A., Sufi, A., 2010. The great recession: Lessons from microeconomic data. American Economic Review: Papers and Proceedings 100, 1-10.

Mian, A., Sufi, A., 2009. The consequences of mortgage credit expansion: Evidence from the US mortgage default crisis. Quarterly Journal of Economics 124, 1449-1496.

Moosa, I.A., 2007. Operational risk: A survey. Financial Markets, Institutions, and Instruments 16, 167194.

Nguyen, D.D., Hagendorf J., Eshraghi, A., 2015, Can bank boards prevent misconduct? Review of Finance, forthcoming, 1-36.

Peni, E., Vahamaa, S., 2012. Did good corporate governance improve bank performance during the financial crisis? Journal of Financial Services Research 41, 19-35.

Peni, E., Smith, S.D., Vahamaa, S., 2013. Corporate governance and bank performance during the financial crisis. Journal of Real Estate Research, 35, 313-343

Power, S., 1997. The audit society: Rituals of Verification. Oxford University Press, Oxford, England.

Preston, L.E., O’Bannon, D.P., 1997. The corporate social – financial performance relationship: A typology and analysis. Business and Society, 36, 419-429.

Rao, V., Dev A., 2006. Operational risk: Some issues in Basel II AMA implementation in US financial institutions. In Davis, E. (ed)., The advanced measurement approach to operational risk. Risk Books, London, England, 273-294.

Rezaee, Z., 2009. Corporate governance and ethics. John Wiley & Sons, Hoboken, NJ.

Robertson, D., 2011. So that’s operational risk! Economics Working Paper 2011-1. Office of the Comptroller of the Currency, Washington, DC.

Roman, R.M., Hayibor, S., Agle, B.R., 1999. The relationship between social and financial performance: Repainting a portrait. Business and Society 38, 109-125.

Ryan, S.G., 2007. Financial instruments and institutions: Accounting and disclosure rules. Wiley, Hoboken, NJ.

Simpson, W.G., Koehrs, T.W., 2002. The link between corporate social and financial performance: Evidence from the banking industry. Journal of Business Ethics 35, 97-109.

Van den Steen, E., 2005. Organizational beliefs and managerial vision. Journal of Law, Economics and Organization 21, 256-283.

Waddock, S.A., Graves, S.B., 1997. The corporate social performance – financial performance link. Strategic Management Journal 18, 303-319.

Whalen, G.W., 2010. Are early warning models still useful tools for bank supervisors? OCC Economics Working Paper 2010-3. Office of the Comptroller of the Currency, Washington DC.

Zeiden, M.J., 2013. Effects of illegal behavior on the financial performance of US banking institutions. Journal of Business Ethics 112, 313-324.

Zeiden, M.J., 2012. The effects of violating banking regulations on the financial performance of the US banking industry. Journal of Financial Regulation and Compliance 20, 56-71.

Figures caption Figure 1 Hypothesized Relationships

Tables Table 1 Hypothesized effect of corporate culture on firm performance1 Based on the literature summarized below, we consider two types of corporate cultures and the expected difference in loan quality. Type of Corporate Culture2

Indicators of the corporate culture

Legal expense (a measure of the ethical component of corporate culture3)

Loan Quality

Market Return

Conservative

• Strong internal controls • Comprehensive policies and procedures4 • Lower operational risk5 • Values driven credit culture6 • Lower risk tolerance

Lower

Higher

Higher

Aggressive

• Weaker internal controls • Less comprehensive policies and procedures • Current profit driven credit culture • Higher risk tolerance • Other managerial weaknesses

Higher

Lower

Lower

Notes: 1. These relations are the basis for our hypothesis; they represent tendencies, not absolute differences. 2. Cronqvist et al (2007) distinguish between conservative and aggressive corporate cultures in banking. 3. This term reflects adherence to a set of values and customs that place the stability and longevity of the organization above monetary benefits to individuals. (Many BHCs post codes of conduct on company websites and expect employees to adhere to them.) Litigation expense is one measure of the ethical climate in a banking organization, not necessarily the only measure. 4. This category includes both the existence of policies and procedures in all areas of operations as well as adherence to these policies and procedures. 5. Chernobai et al (2011) find that most operational risk cases involve repeat offenders. We suggest that this finding is consistent with our hypothesis and must reflect differences in the corporate culture among the banks in their sample. 6. We adopt the terminology used by Koch and MacDonald (2010) discussed in Section 2.1.

Table 2 The two accounting models used to report non-interest expense in BHC 10K reports As discussed in Section 4, BHCs report non-interest expense in two formats. Only Model 2 has separate reporting of professional fees. Accounting Model 1

Accounting Model 2

Personnel

Personnel

Occupancy

Occupancy

Technology and Communications

Equipment

Deposit Insurance Advertising

Marketing Professional Fees

Other

Amortization of Intangibles

Total non-interest Expense

Data Processing Telecommunications Other General Operating Merger and Restructuring Charges Total non-interest Expense

Table 3 Summary statistics for variables used in the regressions This table presents summary statistics for the dependent and independent variables. The variables are defined in Table A1. Dependent Variables (2007-08): ABHR07-08 BHR07-08

Mean

Median

Standard Deviation

Minimum

Maximum

-0.0511 -0.4367

-0.0744 -0.4744

0.3188 0.3240

-0.8266 -0.9612

0.7999 0.3999

LOAN CHARGE-OFFS/ASSETS (07) LOAN CHARGE-OFFS/ASSETS (08) LOAN LOSS PROVISIONS/ASSETS (07) LOAN LOSS PROVISIONS/ASSETS (08) NON-PERFORMING LOANS/ASSETS (07) NON-PERFORMING LOANS/ASSETS (08) ROA (07) ROA (08) ROE (07) ROE (08) Independent Variables (2002-06): LEGAL ASSETS ($ Billion) CAPITAL FHCO HHI MB MERGE ROE

0.0039 0.0097 0.0046

0.0023 0.0052 0.0025

0.0080 0.0143 0.0089

0.0000 0.0000 0.0000

0.1883 0.2054 0.1530

0.0136

0.0079

0.0175

0.0000

0.2389

0.0047

0.0028

0.0092

0.0000

0.2313

0.0105

0.0061

0.0152

0.0000

0.2444

0.0124 -0.0007 0.1354 -0.0372

0.0132 0.0040 0.1486 0.0478

0.0068 0.0182 0.0728 0.2693

-0.0049 -0.0466 -0.0522 -0.8231

0.0239 0.0178 0.2652 0.2008

0.0013 56.240 0.0891 0.3823 0.2072 2.6580 0.8309 0.1860

0.0011 4.8100 0.0881 0.0000 0.1791 2.4154 1.0000 0.1907

0.0011 181.7500 0.0210 0.4866 0.1207 1.0967 0.3753 0.0839

0.0000 0.2690 0.2261 0.0000 0.0641 1.0760 0.0000 -0.4378

0.0077 1463.6800 0.0475 1.0000 0.7147 10.2963 1.0000 0.4766

Table 4 Univariate tests for performance differences between top and bottom deciles This table reports differences in bank performance between banks in the top decile for legal expense and banks in the bottom decile. The variable definitions are in Table A1. For loan losses and operating performance (ROA and ROE) the number of observations is 35 for 2007 and 34 for 2008. For the stock return variables the number of observations is 40. ***denotes significance at the one percent level; **denotes significance at the five percent level. Highest decile for LEGAL

Lowest decile for LEGAL

t-test of difference in means

Absolute Difference (high decile minus low decile)

Percentage Difference (high decile vs. low decile) 11
LOAN CHARGEOFFS/ASSETS07

0.0063

0.0044

1.38

+0.0019

+43.18%

LOAN CHARGEOFFS/ASSETS08

0.0144

0.0111

1.00

+0.0033

+29.73%

LOAN LOSS PROVISIONS/ ASSETS07

0.0071**

0.0042**

2.06

+0.0029

+69.05%

LOAN LOSS PROVISIONS/ ASSETS08

0.0202

0.0167

0.90

+0.0035

+20.96%

span>

NON-PERFORMING LOANS/ASSETS07

0.0079

0.0056

1.38

+0.0023

+41.07%

NON-PERFORMING LOANS/ASSETS08

0.0158

0.0120

1.12

+0.0038

+31.67%

ROA07

0.0141

0.0155

-0.83

+0.0014

+9.93%

ROA08

0.0044

0.0037

0.23

-0.0007

-15.91%

ROE07

0.1410

0.1665

-1.54

+0.0255

+18.09%

ROE08

0.0223

0.0480

-0.68

+0.0257

+115.25%

BHR07-08

-0.4750***

-0.2479***

-3.34

+0.2271

+91.61%

ABHR07-08

-0.0751***

0.1521***

-3.34

+0.2272

+202.53%

Table 5 Pearson coefficients of correlation between legal expense and the performance measures This table shows Pearson correlation coefficients between legal expense (Legal) and each of the bank performance measures. ***denotes significance at the one percent level; **denotes significance at the five percent level; * denotes significance at the ten percent level. Number of Observations 367

Correlation Coefficient 0.1727***

Prob > |r|

351

0.1337**

0.0122

367

0.2052***

0.0001

351

0.1257**

0.0185

367

0.1743***

0.0008

351

0.1494**

0.0050

367

0.0450

0.3902

ROA08

351

0.0614

0.2514

ROE07

367

-0.0225

0.6669

ROE08

351

0.0447

0.4034

BHR07-08

351

-0.0814

0.1281

ABHR07-08

351

-0.0982*

0.0662

LOAN CHARGEOFFS/ASSETS07 LOAN CHARGEOFFS/ASSETS08 LOAN LOSS PROVISIONS/ ASSETS07 LOAN LOSS PROVISIONS/ ASSETS08 NON-PERFORMING LOANS/ASSETS07 NON-PERFORMING LOANS/ASSETS08 ROA07

Table 6. Ordinary Least Squares Regressions

0.0009

This table shows the effect of legal expense and other explanatory variables for 2002-06 on nonperforming loans, loan charge-offs, and loan loss provisions in 2008 as well as the effect on stock returns and operating performance. Non-performing loans is the dependent variable in model 1, charge-offs is the dependent variable in model 2, and loan loss provisions is the dependent variable in model 3. Stock returns are the dependent variables in Models 4 and 5 and operating performance is the dependent variable in Models 6 and 7. ***denotes significance at the one percent level; **denotes significance at the five percent level; * denotes significance at the ten percent level. Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Model 7

LOAN CHARGEOFFS/ASSETS08

LOAN LOSS PROVISIONS/ ASSETS08

NONPERFORMING LOANS/ ASSETS08

BHR07-08

ABHR0708

ROA08

ROE 08

LEGAL

2.6204*

2.9949*

2.9790**

-42.2171

-48.8650

1.5590

18.2207

ASSETS

0.0011

0.0018*

0.0011

-0.0574**

-0.0472**

-0.0023*

-0.0268

CAPITAL

-0.0545

-0.0711

-0.0535

0.8679

2.8341*

0.0941

1.3028

FHCO

-0.0012

-0.0001

-0.0011

0.0752

0.0399

0.0053

0.0919

HHI

0.0256**

0.0184

0.0249**

-0.1857

0.0754

-0.0215

-0.4379*

MB

-0.0028

-0.0017

-0.0032

-0.0250

0.0057

-0.0044

-0.0736*

MERGE

0.0004

-0.0017

0.0006

-0.0458

-0.0271

0.0022

-0.0120

ROE

0.0292

0.0250

0.0334

0.9199

0.8353

0.1137

1.7159***

N

84

84

84

84

84

84

84

Adjusted R2

0.5194

0.5773

0.5429

0.6141

0.0110

0.0208

0.0473

F-Value

10.86***

13.46***

11.84***

17.71***

1.12

1.19

1.45

Table 7 Panel regressions This table shows panel regression estimates of the effect of legal expense and other explanatory variables for 2002-06 on non-performing loans, loan charge-offs, loan loss provisions, buy and hold returns, abnormal buy and hold returns, return on assets and return on equity in 2008 using a random effects procedure. Non-performing loans is the dependent variable in model 1, charge-offs is the dependent variable in model 2, and loan loss provisions is the dependent variable in model 3. State dummy variables are also included in the regressions. ***denotes significance at the one percent level; **denotes significance at the five percent level; * denotes significance at the ten percent level. Model 1

Model 2

Model 3

LOAN CHARGEOFFS/ ASSETS08

LOAN LOSS PROVISIONS/ ASSETS08

NONPERFORMING LOANS/ ASSETS08

Model 4

Model 5

Model 6

Mode 7

BHR07-08

ABHR07-08

ROA08

ROE08

LEGAL

1.5986***

2.4113***

1.7020***

-52.0868***

-70.3799***

-1.5017*

-21.4896

ASSETS

0.0012***

0.0019***

0.0012***

-0.0443***

-0.0399***

-0.0006

-0.0112

CAPITAL

0.0049

0.0027

0.0061

2.0009***

2.7633***

0.0158

-0.1911

FHCO

-0.0045***

-0.0049***

-0.0046***

0.0697*

0.0615*

0.0090***

0.1424***

HHI

0.0574***

0.0512***

0.0564***

-0.0878

-0.0055

-0.0326***

-0.5580***

MB

-0.0009

-0.0006

-0.0010*

-0.0005

0.0123

-0.0003

-0.0032

MERGE

0.0016

-0.0010

0.0023

-0.0546

-0.0437

-0.0057*

-0.0679

ROE02-06

0.0134*

0.0127

0.0140*

0.3293

0.3988*

-

-

N

72

72

72

83

83

72

72

Time Series Length

5

5

5

5

5

5

5

R-Square

0.7319

0.7768

0.7400

0.7922

0.4543

0.5239

0.4761

1The large accounting literature on internal control over financial reporting is not the focus of this paper.

2We state the hypothesis in terms of excess legal expense and excess bank litigation (EBL) is the theoretical concept. In contrast, LEGAL refers to the data. All banks have some normal legal expense associated with such activities as diverse as drafting loan documents, pursuing collections, defending lawsuits that have no or limited merit, developing new financial products, and protecting depositors against internet fraud. Of course, the regression procedure determines which banks are well above the mean in the ratio of legal expense to assets. If regulators follow the recommended two-step procedure, they will only identify banks that are at the high end of their peer group in the ratio of legal expense/assets. The litigation patterns of these banks should then be scrutinized.

3We begin with a list of the top 150 BHCs for 2006 from the American Banker. We add as many smaller BHCs with annual 10K reports as we can find, and we also search for 10K reports for earlier years. There are many additional, generally small, BHCs in the industry, but these institutions do not have stock return data on CRSP, and/or they do not publish a 10K report showing the legal expense proxy. These two factors limit the size of our sample. The sample would have a reporting bias if BHCs that have high legal expense systematically choose Accounting Model 1 where legal expense is combined with a large number of other items. However, the three BHCs with high

legal expense discussed in the case studies all use Accounting Model 2, with its more detailed reporting. Hence, reporting bias does not appear to be a problem. It is difficult to test any hypothesis about reporting bias; there are no complete data because of the lack of regulatory reporting. Nonetheless, if this were an accurate description of actual BHC reporting (i.e., if separate legal expense data were suppressed at some institutions to hide operational risk from investors and regulators), it would support the argument made here for legal expense transparency.

4The coefficient of variation (computed separately for each BHC and then averaged for the sample) is 0.291 for the legal expense proxy and 0.110 for total noninterest expense. The other components of the proxy such as accounting and auditing expense would be expected to be relatively stable from year to year, suggesting that legal expense contributes significantly to the variance in the proxy.

5Additional empirical tests which reveal the same relation in a corporate governance framework are in McNulty and Akhigbe (2015). This paper shows that legal expense has much higher statistical significance that conventional governance measures used by Penni et al (2012) (2013) to explain which banks had greater loan losses during the crisis

6Legal and professional expense of 0.77% is clearly high relative to the mean of 0.13%. By way of comparison, return on assets (ROA) for all US banks during the period 2002 through 2006 ranged from 1.28% to 1.38% (Federal Deposit Insurance Corporation, 2007). To illustrate that the difference between 0.13% and 0.77% is an economically significant difference, assume that a bank had a legal expense proxy 0.50% (or even 0.25%) higher than necessary. This would clearly cause a significant reduction in that bank’s ROA. However, in our analysis, the main link is not an accounting relationship. As illustrated in Table 1, the hypothesized link is behavioral (high litigation expense reflects weak internal control) and the hypothesized relation is lagged.

7As shown in Table 4, the standard deviation of legal expense/assets is also 0.0011. The difference between the mean (0.0013) and the median (0.0011) is 0.0002. Thus, the median is 0.18 (.0002/.0011) standard deviations from the mean. The data are distributed as follows: 25th percentile: 0.0007; median: 0.0011; 75th percentile: 0.0016; 90th percentile: 0.0023. The maximum is 0.0077, indicating that there are a few outliers in the data. Table 6 indicates that one of the case study banks (National City) ranks fourth out of the 83 institutions in the sample for 2006 with a ratio of 0.0020. Thus, in 2006 NCC ranks between the 75th and 90th percentile for the entire distribution. (There are 408 individual bank-year observations. There would be 415 (83 times 5) individual bank-year observations if every BHC reported in every year.)

8The 2007 results are available from the authors on request.

9CAMELS represents the rating system used in the bank examination process; it is an acronym for capital adequacy, asset quality, management quality, earnings, liquidity and sensitivity to market risk (Koch and MacDonald, (2010).

10The case studies in the Working Paper provide examples of the types of bank litigation that and deserve such scrutiny and reflect operational risk.

11Absolute difference relative to the absolute value of the average for the high-legal-expense decile.