Risk-taking behavior and management ownership in depository institutions.

I. Introduction

Depository institution management focuses on three types of risk: liquidity, credit, and interest rate. Each institution must make decisions regarding its exposure to these risks. For liquidity risk management, the institution can choose an asset structure composed of either

liquid or illiquid assets and/or a liability structure that relies significantly or insignificantly on purchased funds and long- or short-term deposits. For credit risk management, the institution can choose conservative, moderate, or liberal evaluative standards in assessing loan applications. For interest rate risk management, each institution can adjust its balance sheet by reducing the mismatch between the duration of assets and the duration of liabilities. The institution may also choose various derivative securities, including interest rate futures, options, and swaps to hedge the exposure.

The level of risk chosen, however, can be influenced by agency problems. Since depositors can only monitor equityholders' actions imperfectly, equityholders can increase the value of their equity call options by increasing the risk of the underlying assets of the depository (Saunders, Strock, and Travlos (1990)). This is referred to as the wealth transfer effect. Furthermore, the old system of levying fixed-price deposit insurance premia (before passage of the Federal Deposit Insurance Corporation Improvement Act (FDICIA) of 1991) created a put-option-like subsidy to equityholders, which also encouraged risk-taking activities (Merton (1977)). Therefore, the risk taking of the institution depends on the alignment of managers' interests with shareholders' interests. As the managers' ownership increases, their interests become more closely aligned with shareholders' interests, creating a strong incentive to maximize the value of their call and put options by increasing the risk level.

Smith and Stulz (1985) present an alternative argument. I They contend that as managerial ownership increases, managers become increasingly risk averse and are more likely to pursue hedging and other risk-reduction strategies. This is because the managers may not hold well-diversified portfolios and will, therefore, have incentives to reduce the riskiness of the firm's returns. Thus, while both schools of thought agree that ownership structure is an important determinant of risk, they disagree on the effect of management ownership on the risk-taking activities of managers. We examine these conflicting explanations for depository institutions.

Although risk taking is a central topic in research on depository institutions (e.g., Furlong and Keeley (1989), Duan, Moreau, and Sealey (1992), Davies and McManus (1991)), few studies examine the effect of ownership structure on depository risk taking. The exceptions are Saunders, Strock, and Travlos (1990), who examine the relation between management ownership and risk taking in the banking industry, and Cebenoyan, Cooperman, and Register (1995), who study the relation between ownership structure and insolvency risk for savings and loans (S&Ls). Saunders, Strock, and Travlos (1990) study large banks from 1978 through 1985 and conclude that stockholder-controlled banks (with high managerial equity positions) exhibit greater risk-taking behavior than manager-controlled banks (with low managerial equity positions) between 1979 and 1982. Cebenoyan, Cooperman, and Register (1995) find that S&Ls with high managerial ownership engage in greater risk-taking behavior than other S&Ls during regulatory leniency and forbearance (1988), and lower risk-taking behavior during regulatory stringency and nonforbearance (1991). However, Cebenoyan, Cooperman, and Register focus on insolvency risk (measured as a dichotomous variable based on the S&L's capital ratio) as opposed to the market-based measures of risk employed in this study. Since our study is more directly related to that of Saunders, Strock, and Travlos (1990), our discussion focuses on their analysis.

We re-explore this issue because of the unsettled empirical and theoretical conclusions regarding ownership structure and risk taking, and because of methodological and environmental considerations. Saunders, Strock, and Travlos's study contains several methodological weaknesses. Their sample incorporates 38 firms per year, all of which are large bank holding companies. Furthermore, their empirical models of bank risk postulate a linear relation between management ownership and risk. They also use an unmodified interest beta to perform their analysis. We contend that the absolute value of the interest rate beta is the appropriate measure since both positive and negative interest rate betas imply some exposure to interest rate risk. In contrast, we employ a sample of 302 depository institutions per year over a six-year period for a total of 1,812 observations. The sample includes both commercial banks and savings institutions, which allows for analysis of the behavioral differences between the two types of institutions. Additionally, we examine both linear and nonlinear models, and use the appropriate interest rate beta.

The sample period environment of the Saunders, Stock, and Travlos study is substantially different from the more recent environment in which depository institutions operate. First, Saunders, Stock, and Travlos focus on 1978 to 1985. This period is characterized by deregulation in the banking industry, beginning with the passage of the Depository Institutions Deregulation and Monetary Control Act (DIDMCA) of 1980 and the Garn St. Germain Act of 1982. Since that time, the Basle Accord (1988) introduced risk-based capital standards and the FDICIA of 1991 introduced more careful examinations of depositories with regard to risk-taking practices and created risk-based deposit insurance premia. Thus, our sample period is one of relative reregulation compared with Saunders, Strock, and Travlos's more deregulated environment. Second, the interest rate environment of the Saunders, Strock, and Travlos study is one of unprecedented interest rate volatility. Between 1978 and 1985, the mean and standard deviation of the absolute value of the monthly change in the short-term government bond rate were 0.52 percent and 0.53 percent, respectively. These same statistics for 1988 to 1993, the sample period of the current study, are 0.22 percent and 0.14 percent, respectively.

These environmental changes may affect the relation between managerial ownership and risk taking in several ways. For example, Saunders, Strock, and Travlos find the relation to be positive and argue this is due to the managers' attempts to maximize the value of the call and put options they derive from their equity position. During their sample period, deregulation was beginning and interest rate volatility was high; consequently, these call and put options had the potential to deliver strong gains. However, during our sample period, the environment was characterized by increased regulatory efforts to control depository risk taking. In addition, the interest rate level and volatility were greatly reduced, which, ceteris paribus, reduces the value of options. In this later environment, the initiatives to create managerial wealth through risk transfers may be dominated by managerial risk aversion, producing a negative relation between managerial ownership and risk taking.

We re-examine the relation between risk and managerial ownership among depository institutions in the current environment using a large, complex sample. We find that the relation between the level of managerial ownership and various measures of depository institution risk is significantly negative, suggesting that as managerial ownership increases the level of risk taking decreases. This finding is consistent with the risk aversion arguments advanced by Smith and Stulz (1985) and the illustrative model presented in this paper, but it is inconsistent with the empirical evidence provided by Saunders, Strock, and Travlos. Further, we provide evidence that the optimal model of bank risk taking is nonlinear with respect to managerial ownership.

The results for savings institutions are different from those for commercial banks. Specifically, we find that when systematic market risk, unsystematic risk, or total risk is used as the proxy for risk, S&Ls exhibit greater risk taking than commercial banks. However, banks and S&Ls are comparable in terms of their exposure to interest rate risk. The differences in results relative to Saunders, Stock and Travlos's findings may be attributable to several factors including differences in the regulatory environment and methodological issues addressed earlier.

II. A Model of the Relation Between Hedging and Managerial Ownership

Consider a one-period model, based on Smith and Stulz (1985), and define the following terms:

U = utility function of the manager;

[V1.sup.*] = value of the firm assuming state 1 occurs - state 1 occurs with a 50 percent probability;

[V2.sup.*] = value of the firm assuming state 2 occurs - state 2 occurs with a 50 percent probability;

E([V.sup.*]) = expected value of the firm = .5([V1.sup.*]) + .5([V2.sup.*]);

F = face value of debt - debt is assumed to be issued at par;

a = percentage of the firm owned by managers;

a[V1.sup.*] = value of manager's equity positions in state 1;

a[V2.sup.*] = value of manager's equity positions in state 2;

E(a[V.sup.*]) = expected value of manager's equity positions;

VH = value of a hedged firm; and

CE = certainty equivalent.

The following analysis explores managerial compensation and decision making. We assume the manager only receives compensation from an equity-holder position and makes decisions to maximize this compensation; however, the manager's equity position is undiversified. The manager owns (aV) of the firm.

Figure I presents the utility function of the manager over firm value and the equity valuation of the manager's position. The value of a increases with managerial ownership; thus, the slope of the equity valuation function increases ({[Delta] aV[where][Delta] V} [greater than] 0). Figure I is the low-ownership position. The upper quadrant of Figure I maps the managers' utility function against their level of wealth. The shape of the function assumes risk aversion on the part of managers. The lower quadrant maps the firm's value and the managers' wealth. The value of the managers' wealth for any firm value below F (the value of debt) is equal to zero. The slope of the function in the lower quadrant increases with the percentage of the equity held by managers.

Two values for the firm are possible, [V1.sup.*] and [V2.sup.*], where [V2.sup.*] [greater than] [V1.sup.*]. In turn, for an unhedged firm, the value of the managers' compensation is (a[V1.sup.*]) and (a[V2.sup.*]) with corresponding levels of utility. Consider the following two scenarios. If the managers do not hedge the firm, the expected value of the firm is equal to E([V.sup.*]). The corresponding expected value of the managers' wealth is equal to E(a[V.sup.*]) with a corresponding level of utility and a certainty equivalent level of wealth equal to CE. CE represents the level of managerial wealth that is equivalent in dollars to the gamble of receiving (a [V1.sup.*]) or (a [V2.sup.*]). If the managers choose to hedge, the value of the firm is then VH with certainty and the managers' wealth is (aVH) with certainty. Since CE = aVH, the managers are indifferent between hedging the firm and the gamble given this low ownership percentage.

Figure II shows the same utility function in the upper quadrant. The lower quadrant shows a function representing the relation between firm value and managerial wealth. This function, however, is steeper than in Figure I, representing a higher percentage ownership by managers. Two values for the firm are possible: [V1.sup.*] and [V2.sup.*], where [V2.sup.*] [greater than] [V1.sup.*]. If the managers choose not to hedge the firm, the certainty equivalent level of wealth is equal to CE. If the managers choose to hedge the firm, the managers' wealth is (aVH) with certainty. Since CE [less than] aVH, the managers prefer the hedge to the gamble given this higher ownership percentage.

The results of this analysis show that as the managers' equity position in percentage terms increases, they are increasingly interested in hedging the firm's risk. This is consistent with the risk aversion hypothesis developed by Smith and Stulz (1985).

III. Data and Methodology

The initial sample of institutions is formed based on a review of the Standard Industrial Classification (SIC) manual to identify the three-digit SIC codes for commercial banks, savings institutions, and bank holding companies. We obtain daily returns for all institutions with the appropriate SIC codes (602, 603, or 671) from the Center for Research in Security Prices (CRSP) return files for the New York Stock Exchange (NYSE), American Stock Exchange (AMEX), and the over-the-counter (OTC) market for 1988 to 1993. This process results in a final sample of 1,812 observations, or 302 firms per year. The sample includes 1,566 observations for banks and 246 observations for savings institutions.

We estimate several risk measures for each sample depository institution using daily returns data taken from CRSP. For example, we calculate risk measures for bank j in year t using daily market return and market interest rate data from year t. We use the two-index model to generate various risk measures.(2) The two-index model is represented as:

[Mathematical Expression Omitted] (1)

where

[R.sub.jt] = return on the jth bank's stock at time t;

[R.sub.mt] = return on the CRSP equally weighted index at time t;

[Mathematical Expression Omitted] = daily changes in yield on a thirty-day Treasury bill (bond equivalent yield) used to proxy the short interest rate series - the data were obtained from Data Resources Inc. (DRI);(3) and

[e.sub.jt] = random error terms.

We derive the following risk measures from the two-index model:

[[Sigma].sub.s] = total risk measured as the standard deviation of the daily returns on each institution's stock;

[Mathematical Expression Omitted] = unsystematic risk measured as the standard deviation of the residual error term from the two-index model using the short interest rate series;

[Mathematical Expression Omitted] = the coefficient on the short interest rate series in the two-index model - the absolute value of [Mathematical Expression Omitted] measures the short interest rate beta; and

[Mathematical Expression Omitted] = the coefficient on the market portfolio returns when the short interest rate series is used in the two-index model.

While Saunders, Strock, and Travlos (1990) use the unmodified interest rate beta, we employ the absolute value of the interest rate beta. If institutions are operating in a short-funded mode, the interest rate beta is expected to be negative. However, the interest rate beta may be negative, zero, or positive depending on the durations of assets relative to liabilities and the hedging strategies employed by the institutions. A zero interest rate beta implies an immunized position for interest rate risk; yet, either a positive or a negative interest rate beta implies exposure to interest rate risk. Therefore, the sign of the interest rate beta is not of major concern. Rather, the magnitude of the interest rate beta determines the interest rate risk exposure for the financial institution. As such, we use the absolute value of the interest rate beta in the analysis.(4)

Management decisions affect both systematic and unsystematic risk. Banks are simply a portfolio of primarily financial assets. For example, a bank that invests heavily in Treasury securities would be relatively low risk and may have a low market beta. Thus, systematic risk can be influenced by management decisions.(5) Moreover, as a portfolio of financial assets, depositories may be able to diversify completely the unsystematic risk, and, as such, this measure would be unimportant. However, if depositories find it difficult to diversify this risk away because they concentrate lending in certain regions or certain products, unsystematic risk can be an important measure.

We formulate two empirical models to measure the effect of managerial ownership on risk taking in depositories.

Model 1:

[RISK.sub.jt] = [[Gamma].sub.0] + [[Gamma].sub.1][PROP.sub.jt] + [[Gamma].sub.2]C[A.sub.jt] + [[Gamma].sub.3]F[A.sub.jt] + [[Gamma].sub.4]LT[A.sub.jt] + [[Gamma].sub.5][YEAR.sub.jt] + [[Gamma].sub.6][TYPE.sub.it] + [[Epsilon].sub.jt] (2)

Model 2:

[RISK.sub.jt] = [[Gamma].sub.0] + [[Gamma].sub.1][LPROP.sub.jt] + [[Gamma].sub.2][CA.sub.jt] + [[Gamma].sub.3][FA.sub.jt] + [[Gamma].sub.4][LTA.sub.jt] + [[Gamma].sub.5][YEAR.sub.jt] + [[Gamma].sub.6][TYPE.sub.i] + [[Epsilon].sub.jt] (3)

where

[RISK.sub.jt] = one of four risk measures obtained using the two-index model for firm j in year t;

[PROP.sub.jt] = percentage of the jth institution's stock owned by officers and directors in year t;

[LPROP.sub.jt] = the natural logarithm of the percentage of the jth institution owned by officers and directors in year t;

[CA.sub.jt] = the capital-to-asset ratio (shareholders' equity/total assets) of the jth institution in year t;

[FA.sub.jt] = the fixed asset ratio (property, plant, and equipment/total assets) of the jth institution in year t;

[LTA.sub.jt] = the natural logarithm of total assets of the jth institution in year t;

[YEAR.sub.jt] = a dummy variable for firm j equal to one if the observation is from year t and zero otherwise (t = 1988-93);

[TYPE.sub.i] = a dummy variable equal to one if the jth institution is a savings institution and zero if it is a commercial bank; and

[[Epsilon].sub.jt] = a random error term.

We use several control variables in the models: CA, FA, LTA, and two kinds of dummy variables. CA, a measure of financial leverage, should be negatively related to the risk measures: when the capital ratio is higher, the institution's risk is lower. FA should be positively related to the risk measures. FA provides a measure for both operating leverage and the liquidity of the asset portfolio. Ceteris paribus, when operating leverage is higher and liquidity is lower, risk is higher. We expect LTA to be negatively related to the risk measures. We contend that larger depositories will tend to hold more diversified asset portfolios, resulting in a lower level of risk. The parameter estimate on TYPE should be positive, reflecting higher risk for savings institutions relative to commercial banks. We test two hypotheses on the relation between management ownership and depository risk:

[H.sub.1]: Since managers are risk averse and become less diversified as their ownership of the institution increases, managerial ownership is inversely related to the degree of risk bearing by financial institutions.

[H.sub.2]: The functional form between risk measures and managerial ownership is not linear. A nonlinear relation between the risk-taking behavior of managers and managerial ownership may be due to the importance of the risk aversion effect relative to the wealth transfer effect.

IV. Empirical Results

Table 1 reports descriptive statistics for the sample. The mean level of total risk, measured as the standard deviation of the return series, is 3.26 percent. The mean level of unsystematic risk (the standard deviation of the residuals of the market model) is 3.15 percent. The mean absolute value of the interest rate beta for the thirty-day Treasury series is 1.64 with a standard deviation of 2.61. The mean market beta is positive but less than one. The mean level of total assets is approximately $6.1 billion. The average capital ratio for the institutions in the [TABULAR DATA FOR TABLE 1 OMITTED] sample is 7.7 percent. The fixed-assets-to-total-assets ratio is generally small with a ratio of 1.7 percent. The maximum fixed-asset-to-total-assets ratio is 9 percent, while the smallest is 0.1 percent. Finally, the average level of officer and director ownership is 15.67 percent with a standard deviation of 13.6 percent.

Table 2 presents results obtained by estimating the two empirical models shown above. The first two columns show the results obtained when total risk, [[Sigma].sub.s], is used as the dependent variable. Results of both linear and nonlinear transformations are reported. PROP is negatively and significantly related to total risk at the 1 percent level in Model 1. This result conflicts with the empirical findings of Saunders, Strock, and Travlos (1990) but supports our model. CA is negatively and significantly related to total risk, which is consistent with the contention that financial leverage is positively related to risk. LTA has a significantly negative sign on the parameter estimate, as expected.(6) FA has a negative yet insignificant sign, which is also reported by Saunders, Strock, and Traylos. Thus, operating leverage is not a significant determinant of bank risk. This is plausible because fixed operating assets normally represent a small portion of a financial institution's total assets. Operating leverage risk, therefore, is more important for a manufacturing firm than for a financial institution.(7) The institutional dummy variable (TYPE) is significantly positive at the 1 percent level, which implies S&Ls are generally more risky (higher total risk) than commercial banks. The model explains 19.67 percent of the variation in the risk measure.

The second model specifies PROP in logarithmic form, which is the nonlinear variation.The variable is again significantly negative at the 1 percent level with a t-statistic of -4.36. This suggests a diminishing slope at higher values of PROP, which supports the notion that the relation between risk measures and managerial ownership may not be linear. The model explains 20.01 percent of the variation in total risk.

The last two columns in Table 2 present the results from modeling residual risk, [Mathematical Expression Omitted]. In each model, the ownership variable is significantly negative at the 1 percent level. Among the three control variables employed, CA and LTA are both significant and carry the expected negative sign while FA is again not statistically significant. The institutional dummy is statistically significant. Therefore, the residual risk assumed by S&Ls is higher than that assumed by commercial banks. Both the linear and the nonlinear models explain approximately 21 percent of the total variation in residual risk.

The results in Table 2 support hypothesis l, which states that because of risk aversion, managerial ownership is inversely related to risk. Consequently, the results are also consistent with the arguments advanced by Smith and Stulz (1985) and the model in this paper. Furthermore, the results support hypothesis 2, which predicts a nonlinear relation between managerial ownership and risk because the risk aversion effect is partially offset by the wealth transfer effect.

The first two columns in Table 3 report the results when [Mathematical Expression Omitted] (interest rate risk) is used as the risk measure. The ownership variable both in linear and log transformation form is significantly negative. Similar to the models employing total risk and residual risk measures, LTA and CA are significant at the 1 percent level with the expected signs. FA, again, is not statistically significant. This time, however, TYPE, is not statistically significant. This suggests the level of risk borne by S&Ls is comparable to the level of risk borne by commercial banks. The results in columns 1 and 2 provide further support that managers display risk-averting behavior as their ownership share increases but that this behavior diminishes as managerial ownership increases, possibly because of the wealth transfer effect.

Columns 3 and 4 in Table 3 present results from modeling [Mathematical Expression Omitted] (the market beta of the firm's stock returns), which measures systematic risk.(8) PROP again is significantly negative. The log transformation of PROP (LPROP) is also significant with a stronger t-statistic; this indicates the relation between managerial ownership and systematic risk is negative but the slope is decreasing. In this final model, the parameter estimate on CA is significantly negative, which is consistent with the findings reported earlier for other risk measures. The parameter estimate on LTA, [TABULAR DATA FOR TABLE 3 OMITTED] however, is significantly positive,(9) while the parameter estimate on FA is significantly negative.(10) These results are different from those found for other risk measures. Finally, the systematic risk of S&Ls exceeds that of commercial banks in the sample. Moreover, to further investigate if the relation between managerial ownership and risk is different for S&Ls relative to commercial banks, we retest the models, including an S&L interactive term. The interactive variable (PROPSVG) is derived by multiplying the variable LPROP by a binary variable equal to one for an S&L and zero otherwise. The S&L interactive variable is significantly negative at the 1 percent level for the market beta and significantly negative at the 10 percent level for the interest rate beta and total risk models. It is not significant for the residual risk model. These results suggest greater risk aversion by S&Ls.(11) Overall, the results of Tables 2 and 3 provide support for the risk aversion argument in hypothesis 1 as well as the nonlinearity argument in hypothesis 2.(12)

V. Conclusions

We explore the relation between managerial ownership and bank risk taking. Saunders, Strock, and Travlos (1990) use a much smaller sample over a significantly different regulatory and interest rate environment (1978-85). They observe a positive relation between managerial ownership and depository risk taking, which they attribute to the management's attempts to increase the value of their call and put options. We employ a larger sample over a period following much re-regulation during which the interest rate environment was less volatile (1988-93). Our findings of a negative relation between managerial ownership and various market-based risk measures suggest managers increasingly engage in risk-averting behavior as their percentage ownership of the institution increases. As such, our empirical evidence supports the arguments proposed by Smith and Stulz (1985) and the illustrative model developed in this paper.

The conflicting conclusions between this research and Saunders, Strock, and Travlos may be because, first, Saunders, Strock, and Travlos use only 38 bank holding companies. Variations among these 38 homogenous banking firms are limited,(13) as evidenced by the lack of a significant relation between bank size (total assets) and most of the risk proxies. On the other hand, we find a strong relation between these two variables. Our results are more consistent with the strong size effect documented in the finance literature. Second, we document a possible nonlinear relation between managerial ownership and risk proxies, while Saunders, Strock, and Travlos assume a linear functional form. Our results are strong enough to conjecture that the risk aversion coefficient does not remain constant over wide ranges of managerial ownership. Moreover, if Saunders, Strock, and Travlos's argument that increased risk taking is motivated by subsidies similar to call and put options is accurate, our findings of increasing risk aversion at a decreasing rate as managerial ownership increases may be because these effects partially offset risk-aversion effects. Third, the regulatory and interest rate environments changed substantially by the late 1980s and early 1990s. Regulatory initiatives attempted to control risk taking by modifying regulatory standards and monitoring firms directly. We postulate that our results are consistent with these environmental changes. In effect, the value of management's call and put options derived from their equity position is affected by these regulatory initiatives. Moreover, the lower interest rate volatility during this more recent period provides less incentive for management to use the options. Consequently, we find risk aversion to be a dominant consideration in risk choices among depositories between 1988 and 1993. For regulators, our findings imply effectiveness of risk-based capital standards.

The authors would like to thank the participants at the 1995 Financial Management Association meetings and an anonymous referee for many helpful comments.

1 This argument does not preclude the existence of the Saunders, Strock, and Travlos (1990) contention. The relation between managerial ownership and risk taking may be a function of the dominance of one effect over the other across different levels of managerial ownership.

2 Several studies model depository stock returns as a function of both a market index and an interest rate index in a two-factor model, including Flannery and James (1984) and Chen and Chan (1989).

3 We also use the five-year Treasury note series to generate the risk measures out of the two-index model. The results from modeling these risk measures are similar to those presented in this paper. This additional analysis is available from the authors.

4 The mean absolute value of the interest rate beta is 1.64 (reported in Table 1). The mean level of the unmodified interest rate beta is -.12 with 948 observations above this mean and 864 observations below it.

5 While Beaver, Kettler, and Scholes (1970) use accounting information to estimate beta, Rosenberg and McKibben (1973) and Rosenberg and Marathe (1975) employ fifty-four factors to estimate beta. Therefore, to some extent, systematic risk can be affected by the management.

6 While our results show the size variable to be significant in all models irrespective of the specification of risk measure, Saunders, Strock, and Travlos (1990) find it to be significant in only one risk measure.

7 FA is only significant in modeling one of the risk measures in our study. The signs on the parameter estimates of FA in our study across these dependent risk variables are similar to Saunders, Strock, and Traylos.

8 Saunders, Strock, and Travlos argue that since systematic risk is nondiscretionary, it is expected to bear no relation to the ownership structure. This, however, may not be true. As we point out in the previous section, existing research supports the notion that systematic risk can be affected by both market price data and a firm's specific information.

9 The Pearson correlation coefficient between LTA and the market beta is positive while the correlation between LTA and the interest rate beta, the residual risk, and the total risk is negative. These same relations emerge from the regression analysis. Nevertheless, these results are consistent with Saunders, Strock, and Travlos, who argue that larger banks appear to be more sensitive to general market movements but are better able to diversify their interest rate risk exposure.

10 Saunders, Strock, and Travlos also report a negative relation between FA and market risk.

11 The complete results are available from the authors.

12 Because our methodology of using a panel data set excludes firms that are not traded for the entire sample period, a survivorship bias may be created. To assess this possible bias, we test the results again using a pooled data set that does not exclude firms that are not publicly traded in all sample years. These results are consistent with respect to parameter estimates and statistical significance.

13 In an attempt to proxy the Saunders, Strock, and Travlos sample, we exclude institutions that are not bank holding companies and re-estimate the empirical models. The results using this restricted sample continue to support the risk aversion hypothesis in which managerial ownership is negatively related to the risk measures.

References

Beaver, W., P. Kettler, and M. Scholes, 1970, The association between market determined and accounting determined risk measures, Accounting Review 45, 654-82.

Cebenoyan, A. S., E. S. Cooperman, and C. A. Register, 1995, Deregulation, reregulation, equity ownership, and S&L risk taking, Financial Management 24, 63-76.

Chen, C. R. and A. Chan, 1989, Interest rate sensitivity, asymmetry, and stock returns of financial institutions, Financial Review 24, 457-73.

Davies, S. M. and D. A. McManus, 1991, The effects of closure policies on bank risk-taking, Journal of Banking and Finance 15, 917-38.

Duan, J. C., A. F. Moreau, and C. W. Sealey, 1992, Fixed-rate deposit insurance and risk-shifting behavior at commercial banks, Journal of Banking and Finance 16, 715-42.

Flannery, M. J., and C. M. James, 1984, The effect of interest rate changes on the common stock returns of financial institutions, Journal of Finance 38, 1141-53.

Furlong, F. T. and M. C. Keeley, 1989, Capital regulation and bank risk taking: A note, Journal of Banking and Finance 13,883-92.

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

Rosenberg, B. and V. Marathe, 1975, Tests of the capital asset pricing hypothesis, Working paper no. 32, University of California, Berkeley.

Rosenberg, B. and W. McKibben, 1973, The prediction of systematic and specific risk in common stocks, Journal of Financial and Quantitative Analysis 8, 317-33.

Saunders, A., F. Strock, and N. Travlos, 1990, Ownership structure, deregulation, and bank risk taking, Journal of Finance 45, 643-54.

Smith, C. W. and R. M. Stulz, 1985, The determinants of firms' hedging policies, Journal of Financial and Quantitative Analysis 20, 443-57.

Stulz, R., 1984, Optimal hedging policies, Journal of Financial and Quantitative Analysis 19, 127-40.