Board structure, ownership, and financial distress in banking firms

Transcription

1 International Review of Economics and Finance 8 (1999) Board structure, ownership, and financial distress in banking firms W. Gary Simpson a, *, Anne E. Gleason b a Deparment of Finance, College of Business Administration, Oklahoma State University, Stillwater, OK , USA b Department of Finance, College of Business Administration, The University of Central Oklahoma, Edmund, OK 73034, USA Received 22 August 1997; accepted 5 May 1998 Abstract This investigation pursues a new direction in the analysis of financial distress in banking firms. The research was inspired by recent research on corporate governance and the need to understand the internal processes behind the financial decisions that result in bank failures. The analysis examined the relationship between the ownership and structure of the board of directors and the internal control mechanism that influences the survival of the firm. The following aspects of ownership and governance are investigated: ownership by directors and officers, ownership by the CEO, number of directors, percentage of inside directors, and CEO duality. The influence of board structure and ownership on the probability of financial distress was explored with a sample of approximately 300 banking firms. The empirical tests indicated a lower probability of financial distress when one person is both the CEO and chairman of the board, but the other factors did not have a significant effect Elsevier Science Inc. All rights reserved. JEL classification: G21; G28 Keywords: Corporate governance; Internal control system; Financial distress; Banks 1. Introduction The emergence of the banking industry from the most serious financial crisis since the 1930s has resulted in a strengthened deposit insurance fund and a relaxation of regulatory concern for bank failures. The large amounts of risky credit card debt recently incurred by banks, however, underscore the point that financial distress in banking is always an issue (Federal Deposit Insurance Corporation, 1997). * Corresponding author. Tel.: ; fax: address: simpson@okway.okstate.edu (W.G. Simpson) /99/$ see front matter 1999 Elsevier Science Inc. All rights reserved. PII: S (99)00026-X

2 282 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) Financial distress in banking remains a significant issue for owners, managers, and the public. The incentives for risk-shifting from equity owners to depositors exist in banking similar to the agency problems caused by the conflict between owners and debt-holders in other corporations. John, John, and Senbet (1991) argue that riskshifting incentives in depository financial institutions arise from the existence of limited liability for owners and the associated convexity of the levered equity pay-off produced by the limited liability. The incentives for risk-shifting will exist in spite of risk-adjusted deposit insurance premiums according to John, John, and Senbet (1991). As a result, banks with managers closely aligned with the owners would seek risk that is shifted to depositors and the public insurance fund. An analysis of bank failures prepared by the Office of the Comptroller of the Currency (OCC) identified the major immediate cause of many bank failures as poor asset quality that eventually impaired the bank s capital position (Office of the Comptroller of the Currency, 1988). The OCC investigation further concluded that the primary reason banks encounter asset quality and capital problems is the failure of the board of directors and management. According to the OCC, the ultimate causes of bank failures are an uninformed or inattentive board of directors and/or management, overly aggressive activity by the board and/or management, problems involving the chief executive officer, and other problems related to board oversight and management deficiencies. Failure of the board to monitor the activities of management and staff resulted in poorly followed loan policies, inadequate compliance with internal policies and banking laws, inadequate problem loan identification, and ineffective asset/liability management, according to the OCC. Jensen (1993) argues that the board of directors is crucial to effective internal control systems: The problems with corporate internal control systems start with the board of directors. The board, at the apex of the internal control system, has the final responsibility for the functioning of the firm. Most importantly, it sets the rules of the game for the CEO (p. 862). The ultimate consequence of a dysfunctional corporate internal control system is the failure of the firm. A rich and important body of research that addresses the prediction of financial distress in commercial banks and the classification of banks based on financial stability has evolved (Demirguc-Kunt, 1989). The present investigation pursues a new direction in the analysis of bank survival inspired by the research on corporate governance and the need to understand the processes behind the financial decisions that result in bank failures. 1 The purpose of this analysis is to examine the relationship between the board structure and ownership of a commercial banking firm and the occurrence of financial distress in that firm. A set of testable hypotheses was developed from the model governance structure in Jensen (1993). The hypotheses were tested empirically by regressing measures of the board structure and ownership of a group of approximately 300 banking firms on an indicator of the probability of financial distress. 2. A model governance structure Jensen (1993) contends that few boards in the recent past have functioned properly in the absence of external crises and he provided several proposals that should cause

3 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) the board to become an effective control mechanism. First, board cultures must be changed to emphasize frankness and truth instead of politeness and courtesy so that CEOs do not have the influence to control the board and escape scrutiny. Second, board members must have free access to all relevant information and not just the information selected by the CEO. Then the board members must have the expertise to evaluate this information. Third, legal liabilities must be altered so that directors have the appropriate incentives to take actions that create value for the company, not reduce the risks of litigation. Fourth, management and board members should have significant equity holdings in the company to promote value maximization for shareholders. Fifth, boards should be kept small (seven or eight members) so they can function more efficiently and not be controlled by the CEO. Similarly, the CEO should be the only insider because other insiders are too easily influenced by the CEO. Sixth, the board should not be modeled after the democratic political model that represents other constituencies in addition to shareholders. Seventh, the CEO and the chairman of the board should not be the same person. Finally, the role of investors that hold large debt or equity positions in the company and actively seek to participate in the strategic direction of the company should be expanded. Jensen (1993) suggested that LBO associations and venture capital funds provide a model governance structure that has effectively resolved some of the problems associated with current corporate control systems. The problems addressed are not firm failure but slow growth/declining firms and high growth entrepreneurial firms. The concepts apply to faulty internal control systems in banking firms that result in financial distress. The characteristics of LBO associations and venture capital funds that provide a model for efficient internal corporate controls are: 1. limited partnership agreements at the top level that prohibit headquarters from cross-subsidizing divisions, 2. large equity ownership by both managers and directors, 3. directors that represent a large fraction of the owners, 4. a small number of directors on the board (eight or less), 5. no management insiders on the board other than the CEO, and 6. the CEO is not the chairman of the board. The framework offered by Jensen (1993) provides a conceptual framework drawn from observation that was intuitively appealing and provided the basis for a set of empirically testable hypotheses. 3. Hypotheses 3.1. Hypothesis I: Management and board member equity ownership Jensen (1993) suggests that many problems occur because neither managers nor directors normally own a substantial proportion of the firm s equity, which decreases the incentives of directors and officers to pursue the shareholders interests. Saunders,

4 284 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) Strock, and Travlos (1990) provide evidence that banks controlled by stockholders have incentives to take higher risk than banks controlled by managers. If stockholders prefer more risk than non-owner managers and stock ownership aligns managers and directors with owners, then the probability of financial distress in a bank would be higher when managers and directors own a higher proportion of the equity. H A : Banks with higher proportions of equity ownership by directors and managers have higher probabilities of experiencing financial distress, ceteris paribus Hypothesis II: Board size Jensen (1993) proposed that a smaller number of board members produces a more effective control mechanism. Changanti, Mahajan, and Sharma (1985) also suggested that smaller boards play a more important control function whereas larger boards have difficulty coordinating their efforts which leaves managers free to pursue their own goals. However, a smaller board might be easier for the CEO to influence and a larger board would offer a greater breadth of experience. The impact of board size on the corporate control mechanism is not obvious, but the strongest arguments suggest that a smaller board would result in closer alignment with shareholder interests, which would increase risk taking. H A : Banks with smaller boards have higher probabilities of financial distress than banks with larger boards, ceteris paribus Hypothesis III: Insiders on the board Jensen (1993) argued that corporate officers who report to the CEO cannot be effective monitors because the possibility of retribution is high. Therefore, the officers of the corporation should not serve on the board. Kesner, Victor, and Lamont (1986) referred to this point as the outsider dominance perspective. To the contrary, outsiders sometimes do not understand the complexities of the company and are technically ineffective monitors. When outsiders represent a large number of diverse interests, they may restrict the economic flexibility of the firm and produce conflicts between the board and management. It is often suggested that the participation of outsiders on the board influences the effectiveness of the control function. Weisbach (1988) and Brickley, Coles, and Terry (1994) present empirical evidence which suggests that outside directors represent shareholder interests better than inside directors. H A : Banks with higher percentages of inside directors on the board have lower probabilities of financial distress, ceteris paribus Hypothesis IV: CEO duality Jensen (1993) argued that the CEO should not have a dual position as chairman of the board because the CEO may not separate personal interests from shareholder interests. The function of the chairman of the board is to conduct board meetings and supervise the evaluation and compensation of the CEO (Jensen, 1993). The dual CEO/chairman of the board probably has significantly increased power over the

5 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) board and corporation. This would probably reduce the effectiveness of the control mechanisms of the governance structure. The issue of CEO duality has received considerable attention because the practice is commonly observed in many large corporations (Kesner, Victor, & Lamont, 1986). Supporters argue that CEO duality provides better strategic vision and leadership than an independent chairman. H A : The probability of financial distress is lower for a banking firm with a dual chairman of the board and CEO, ceteris paribus Hypothesis V: CEO equity ownership A major premise of Jensen (1993) is that the CEO should pursue the interests of the shareholders. The argument against a combination of the chairman of the board and the CEO is that the manager will be too powerful and not have interests aligned with shareholders. The fact that a CEO would be able to control other officers who were on the board follows the same line of reasoning. A parallel consideration is the equity ownership position of the CEO. The amount of equity a CEO holds should increase the alignment of the interests of the CEO with the interests of shareholders. H A : A banking firm where the CEO has a lower equity ownership position has a lower probability of financial distress, ceteris paribus. 4. Statistical methodology 4.1. Sample design and data sources The sample consisted of those banking firms listed in the SNL Quarterly Bank Digest (SNL Securities, 1993), which also had proxy statements available for The sample included only banking firms that were publicly traded because these were the only firms with publicly available ownership data. The SNL Quarterly Bank Digest provides data on most publicly traded banking firms and includes approximately 375 firms. Only firms that did not have complete financial data or a proxy statement were omitted. The following ownership and board structure measures were taken from 1989 proxy statements: 1. the percentage equity ownership of all officers and directors as a group, 2. the number of directors on the board, 3. the percentage of insiders on the board, 4. the combination of the CEO and the chairman of the board into one position, and 5. the percentage equity ownership of the CEO. A surrogate for financial distress and the control variables were taken from the SNL Quarterly Bank Digest for the end of the year This procedure produced a sample of 287 banking firms with complete information. The time structure of the regression equations reflects the proposition that the

6 286 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) effect of ownership and board structure will not be observed immediately in bank performance but will take three to five years to present. The measures of ownership and board structure were taken as of the end of the first quarter 1989 because the proxy information was prepared at that time. The ownership and board structure in place at the beginning of 1989 was expected to influence the probability of financial distress at the end of 1993, approximately five years later. The regression equations are cross-sectional with one lagged independent variable, the measure of ownership or board structure Tests of hypotheses The hypothesized relationships were tested with the following ordered logistic equation: logit (p 2 p 3 p 4 ) 0 GOV i x i i where p 1 Prob(Y i 1 GOV i, x i ), p 2 Prob(Y i 2 GOV i, x i ), p 3 Prob(Y i 3 GOV i, x i ), p 4 Prob(Y i 4 GOV i, x i ), Y i a variable representing the SNL rating of the ith banking firm (1 no risk of financial distress, 2 little risk of financial distress, 3 some risk of financial distress, and 4 strong risk of financial distress), GOV i an indicator of ownership or board structure for the ith banking firm, x i a vector of control variables that will impact the probability that Y i n, 0 a parameter to be estimated, a vector of parameters to be estimated, an intercept term, and i the error term. The term logit(p 2 p 3 p 4 ) represents cumulative probabilities and the model predicts the probability of more financial distress with changes in the relevant effects variables. The logit term on the left hand side of the equation equals log{(p 2 p 3 p 4 )/(1 p 2 p 3 p 4 )}, which is the log of the ratio of the cumulative probabilities that a particular banking firm will have a high level of risk to the cumulative probabilities that the firm will have no risk of financial distress. The estimation procedure assumed a common slope parameter associated with the relevant effects variables and used maximum likelihood regression. 2 The relevant effects vector x i is composed of the variables described in Table 1. The coefficient of primary interest is 0. The hypothesized relationships between ownership and board structure and the probability of financial distress in terms of the regression coefficients are: H 0 : H A : I. Management and board equity ownership II. Board size III. Insiders on the board IV. CEO duality V. CEO equity ownership

7 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) Table 1 Variable definitions and descriptive statistics Definition of the variable Mean Minimum Maximum SD Sign Y SNL Safety Rating GOV 1i common shares owned by directors and officers/total common shares GOV 2i number of directors GOV 3i number of insiders on board/total board members GOV 4i 1 if CEO and COB same person, 0 otherwise GOV 5i common shares owned by the CEO/total common shares X 1i book value total assets $1.714 bil $70.6 mil $187.6 bil $5.2 bil X 2i nonperforming assets/ total assets X 3i market value per share/ book value per share X 4i book value of total equity capital/total assets Sign, the hypothesized sign of the regression coefficient in the estimated equations. SD, Standard Deviation of the variable Empirical variables One indicator that measures the potential for financial distress for banking firms is the CAMELS rating developed by federal regulators. Unfortunately, this indicator is not publicly available. However, SNL Securities calculates an indicator called the SNL Safety Rating, which is similar to a CAMELS rating. The SNL Safety Rating measures the risk of each banking firm based on capital adequacy, asset quality, the risk profile of the loan portfolio, earnings, and value assessed by the stock market. The SNL Safety Rating goes from A to D, similar to a bond rating. The SNL Safety Rating was used to proxy the probability of financial distress as follows: A, A, and A 1 indicating no risk; B, B, and B 2 indicating little risk; C, C, and C 3 indicating some risk; and D, D, and D indicating strong risk. The terminology no risk, little risk, some risk, and strong risk follows that used by SNL Securities. The SNL Safety Rating is highly correlated with the probability of default measure developed by Thomson (1992). 3 The financial distress indicator was hypothesized to be a function of the ownership and board structure variables in addition to the following control variables: 1. the size of the banking firm measured by total assets, 2. the default risk of the asset portfolio measured by the ratio of nonperforming assets to total assets,

8 288 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) the risk evaluation of the equity markets measured by the market value/book value ratio, and 4. financial leverage measured by the book value of equity to book value of total assets ratio. The number of control variables was parsimonious by design but the equations show that most of the variables had high explanatory power. The calculation of the governance structure variables was straightforward except for the percentage of insiders on the board. A strict definition of insiders was applied which included current officers of the banking firm, former officers of the banking firm, and corporate counsel. Board members were considered insiders only if it was obvious from the proxy statements. Table 1 provides a list of all empirical variables with descriptive statistics and the expected sign of the regression coefficients. The banks with publicly traded stock are much larger than the average bank as indicated by the average total assets of $1.714 billion for the sample banking firms. All of the firms in the sample were bank holding companies. 5. Empirical results 5.1. Tests of hypotheses The maximum likelihood estimates of the ordered logistic regression parameters reported in Table 2 reveal that the null hypothesis was rejected for Hypothesis IV but could not be rejected in the other relationships. The parameter estimate indicates that CEO duality (i.e., when the same person is both the CEO and chairman of the board) has a significant effect on the future probability of financial distress in a banking firm. The regression coefficient 0 for Hypothesis IV is positive which indicates that banking firms where the same person is both the CEO and chairman of the board have a lower probability of financial distress five years later. This result is consistent with the theory that a dual CEO-chairman of the board is more likely to have the ability to pursue his/her personal interests and is less likely to be aligned with the interests of shareholders who prefer greater risk taking by the firm. The regression estimates give no indication that ownership is an important influence on the probability of financial distress in the future. The combined equity ownership of directors and officers and the individual equity ownership of the CEO did not have an effect. The percentage of insiders on the board and the number of directors on the board does not appear to impact future financial distress. 4 The relative magnitudes of the standardized regression coefficients suggest that the governance variables are less important influences on risk than the control variables representing bank size, the riskiness of the loan portfolio, and the banking firm s use of financial leverage Explanatory power of the equations. The signs of all significant regression coefficients were correct and all of the control variables were highly significant except the market value to book value ratio. The 2

9 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) Table 2 Maximum likelihood estimates of the ordered logistic equation parameters Parameters Hypothesis I Hypothesis II Hypothesis III Hypothesis IV Hypothesis V (constant) (0.00) (0.00) (0.00) (0.00) (0.00) 0 (GOV 1i management and board equity owner- [0.1094] ship) (0.37) 0 (GOV 2i board size) [0.1831] (0.19) 0 (GOV 3i insiders on the board) [0.1160] (0.38) 0 (GOV 4i CEO duality [0.3833] (0.01) 0 (GOV 5i CEO owner ship of equity) [0.1512] (0.19) 1 (X 1i book value total assets) [ ] [ ] [ ] [ ] [ ] (0.00) (0.00) (0.00) (0.00) (0.00) 2 (X 2i nonperforming assets/total assets) [1.1785] [1.8184] [1.1971] [1.2631] [1.1819] (0.00) (0.00) (0.00) (0.00) (0.00) 3 (X 3i market value per share/book value per [0.2454] [0.2377] [0.2084] [0.1676] [0.2167] share) (0.39) (0.41) (0.56) (0.67) (0.55) 4 (X 4i book value of total equity capital/ [ ] [ ] [ ] [ ] [ ] total assets) (0.00) (0.00) (0.00) (0.00) (0.00) Standardized regression coefficients are in brackets. Probabilities for a Wald Chi-square test are in parentheses. log likelihood statistic which is distributed as a chi-square distribution was used to test the null hypothesis that all regression coefficients in the equation are zero. The results reported in Table 3 confirm that the null hypothesis was rejected. The explanatory power of the equations as revealed by the R-square statistics was solid with generalized R-square statistics of approximately 0.55 and adjusted generalized R-statistics of approximately The predicted probabilities and observed responses indicate the model was correct in approximately 85 percent of the cases, incorrect in approximately 2 percent of the cases, and indeterminate in approximately 13 percent of the cases. The Sommers D, which is a summary measure of the predicted probabilities and observed responses, indicates good explanatory power. The Chi-square test for the proportional odds assumption indicates the computation

10 290 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) Table 3 Statistics for the ordered logistic equations Statistics Hypothesis I Hypothesis II Hypothesis III Hypothesis IV Hypothesis V Generalized R-square Adjusted generalized R-square Proportional odds [14.100] [13.654] [14.800] [12.858] [13.007] assumption Chi- (0.1685) (0.1894) (0.140) (0.232) (0.223) square test a 2 log likelihood [ ] [ ] [ ] [ ] [ ] Chi-square test a (0.000) (0.000) (0.000) (0.000) (0.000) Predicted probabili- 84.5% 86.1% 84.5% 86.2% 84.5% ties and observed responses: Concordant Predicted probabili- 2.3% 2.1% 2.2% 1.9% 2.2% ties and observed responses: Discordant Predicted probabili- 13.2% 11.7% 13.4% 11.9% 13.3% ties and observed responses: Tied Sommers D a The Chi-square statistic is in brackets and the probability is in parentheses. that assumed a common slope parameter for each explanatory variable was reasonable, that is, the null hypothesis of common slope parameters could not be rejected at a normal level of confidence. 6. Conclusion This investigation indicates that the combination of the CEO and chairman of the board into one position may influence the internal control system of a banking firm in such a way as to reduce the probability of financial distress in the firm. The result that a single powerful manager will reduce the probability of financial distress is consistent with theory and previous empirical evidence. A manager with significant control over both operations and the board would not be as susceptible to the influence of outside directors, and other monitors, that would cause the interests of management to be more closely aligned with shareholders. A dual CEO chairman of the board would be capable of pursuing his/her own interests, which could mean taking less risk to protect unique human capital. These results have obvious implications for the regulation of banking firms because shareholder dominated banks would be more likely to engage in risk-shifting to depositors and ultimately the FDIC fund. In other words, banks with a combined CEO-chairman of the board are less likely to require FDIC assistance. Furthermore, regulatory efforts to influence risk taking by controlling other aspects of basic board

11 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) structure and ownership would not be effective. The evidence of this analysis and the previous evidence of Saunders, Strock, and Travlos (1990) suggest that the nature of the internal control mechanism is an important issue for bank regulation. The fact that other basic indicators of the internal control system were not found to be significant suggests that the impact of board structure and ownership on banking firm behavior may be too complex to be captured by simple structural characteristics. The results of this investigation cannot be seen as the final answer to the question of how internal control systems function in banking firms, but only a first step. Future research should strive to penetrate inside the black box of the internal control system for banking firms to better understand the complex dynamics of corporate decisions. Acknowledgments We are grateful to Carl Chen, editor, and an anonymous referee for their patience and many helpful comments. The authors bear full responsibility for any remaining errors or omissions. Notes 1. Brickley and James (1987) empirically tested the proposition that board structure is related to the effectiveness of the outside market for takeovers in the banking industry. Saunders, Strock, and Travlos (1990) investigated the relationship between risk taking in banking firms and their ownership structure. Baysinger and Butler (1985) found that board independence had a small impact on the future relative financial performance of a sample of industrial firms. Changanti, Mahajan, and Sharma (1985) found that the number of directors on the board is inversely related to failure in a sample of retail firms but no relationship existed with other measures of governance structure (e.g., percentage of outside directors and chairman-ceo duality). 2. The SAS System was used to compute the ordered logit equations. A proportional odds model was considered to be appropriate. Refer to Peterson and Harrell (1990) and Greene (1997) for a discussion of ordered logit regressions. 3. The measure of the probability of financial distress in banking firms developed by Thomson (1992) was an adjusted capital ratio he called NCAPTA. NCAPTA was calculated for each of the firms in this sample and the correlation coefficient between the SNL Safety Rating and NCAPTA was The SNL Safety Rating was used because it is partially based on the market value of equity while NCAPTA considers only the book value of equity. Sinkey (1978) proposed a measure similar to NCAPTA. 4. Changanti, Mahajan, and Sharma (1985) did not find a significant relationship between the number of insiders on the board or CEO duality and failure in retailing firms. They did find that the number of directors was related to failure. 5. The reported R-square is a generalization of the normal coefficient of determina-

12 292 W.G. Simpson, A.E. Gleason / International Review of Economics and Finance 8 (1999) tion from ordinary least squares regressions (Magee, 1990). Nagelkerke (1991) developed the adjustment to the generalized R-square. Green (1997) indicates there is no exact counterpart to the standard regression R-square in the generalized regression model and points out the limitations of the generalized coefficients of determination. References Baysinger, B. D., & Butler, H. N. (1985). Corporate governance and the board of directors: performance effects of changes in board composition. Journal of Law, Economics, and Organization 1, Brickley, J. A., Coles, J. A., & Terry, R. L. (1994). Outside directors and the adoption of poison pills. Journal of Financial Economics 35(3), Brickley, J. A., & James, C. M. (1987). The takeover market, corporate board composition, and ownership structure: the case of banking. Journal of Law and Economics 30, Changanti, R. S., Mahajan, V., & Sharma, S. (1985). Corporate board size, composition and corporate failures in the retailing industry. Journal of Management Studies 22, Demirguc-Kunt, A. (1989). Deposit-institution failures: a review of empirical literature. Economic Review, Federal Reserve Bank of Cleveland. Quarter 4, Federal Deposit Insurance Corporation. (1997). The FDIC Quarterly Banking Profile: First Quarter Washington, DC: FDIC Public Information Center. Greene, W. H. (1997). Econometric analysis, 3rd ed. Upper Saddle River, NJ: Prentice-Hall. Jensen, M. C. (1993). The modern industrial revolution, exit, and the failure of internal control systems. Journal of Finance 48(3), John, K., John, T.A., & Senbet, L. W. (1991). Risk-shifting incentives of depository institutions: a new perspective on federal deposit insurance reform. Journal of Banking and Finance 15, Kesner, I. F., Victor B., & Lamont B. T. (1986). Board composition and the commission of illegal acts: an investigation of fortune 500 companies. Academy of Management Journal 29, Magee, L. (1990). R 2 measures based on wald and likelihood ratio joint significance tests. American Statistician 44, Nagelkerke, N. J. D. (1991). A note on a general definition of the coefficient of determination. Biometrika 78, Office of the Comptroller of the Currency. (1988). Bank Failure: An Evaluation of the Factors Contributing to the Failure of National Banks. Washinton, DC: OCC. Peterson, B. & Harrell, F. (1990). Partial proportional odds models for ordinal response variables. Applied Statistics 39, Saunders, A., Strock, E., & Travlos, N. (1990). Ownership structure, deregulation, and bank risk taking. Journal of Finance 45(2), Sinkey, J. F. (1978). Identifying problem banks: how do the banking authorities measure a bank s risk exposure? Journal of Money, Credit and Banking 10(2), SNL Securities. (1993). The SNL Quarterly Bank Digest (September). Charlottesville: SNL. Thomson, J. B. (1992). Modeling the bank regulator s closure option: a two step logit regression approach. Journal of Financial Services Research 6(1), Weisbach, M. S. (1988). Outside directors and CEO turnover. Journal of Financial Economics 20(1/2),