Does a better education make for better managers? An empirical examination of CEO educational quality and

firm performance

Aron A. Gottesman Matthew R. MoreyDepartment of Finance Department of Finance Lubin School of Business Lubin School of BusinessPace University Pace UniversityOne Pace Plaza One Pace PlazaNew York, NY 10038 New York, NY 10038agottesman@pace.edu mmorey@pace.edu

April 11, 2006

Abstract

This paper represents the first attempt, to our knowledge, to empirically examine the relationship between the quality of Chief Executive Officer (CEO) education and firm performance. This is an important question as many papers in the management literature have postulated that managers with higher educational attainment will be more adaptive and innovative, and more likely to possess other characteristics that may improve firm performance. We find four results in our analysis. First, using the mean entrance scores as proxies for the prestige of undergraduate and graduate programs, we find no evidence that firms with CEOs from more prestigious schools perform better than firms with CEOs from less prestigious schools. Second, we find that firms managed by CEOs with MBA or law degrees perform no better than firms with CEOs without graduate degrees. Third, we find some limited evidence that firms led by CEOs with non-MBA, non-law graduate degrees have slightly better risk-adjusted market performance than other firms. Fourth, we find that compensation is somewhat higher for CEOs who attended more prestigious schools.

Does a better education make for better managers? An empirical examination of CEO educational quality and

firm performance

This paper represents the first attempt, to our knowledge, to empirically examine the relationship between the quality of Chief Executive Officer (CEO) education and firm performance. This is an important question as many papers in the management literature have postulated that managers with higher educational attainment will be more adaptive and innovative, and more likely to possess other characteristics that may improve firm performance. We find four results in our analysis. First, using the mean entrance scores as proxies for the prestige of undergraduate and graduate programs, we find no evidence that firms with CEOs from more prestigious schools perform better than firms with CEOs from less prestigious schools. Second, we find that firms managed by CEOs with MBA or law degrees perform no better than firms with CEOs without graduate degrees. Third, we find some limited evidence that firms led by CEOs with non-MBA, non-law graduate degrees have slightly better risk-adjusted market performance than other firms. Fourth, we find that compensation is somewhat higher for CEOs who attended more prestigious schools.

1I. IntroductionThis paper attempts to answer the question of whether educational background of the Chief Executive Officer (CEO) matters to firm performance. As such, we are extending a stream of research that has examined the impact mutual fund managers educational background on mutual fund performance (see Golec (1996), Chevalier and Ellison (1999), and Gottesman and Morey (2006)) to CEOs. This research generally finds that educational background actually does matter in terms of mutual fund manager performance as managers with stronger educational profiles

(managers from better schools) produce better risk-adjusted performance than other managers. The question this paper tries to answer is whether this same effect occurs with CEOs.

Why should CEO educational background matter in terms of firm performance? One obvious reason is that education background may proxy for intelligence, and more intelligent

managers may be better managers. Indeed, there are a number of studies which find that CEOswith higher educational attainment have a greater capacity to process information and to innovate than CEOs with lower educational attainment.1 Another reason is that more highly educated CEOs may be more likely to use sophisticated methodologies that may improve firm

performance. For example, it has been demonstrated by Graham and Harvey (2002) that top executives are more likely to use sophisticated methodologies when conducting capital budgeting and when estimating the cost of capital. Finally, it may be that CEO education is positively related to the CEOs social capital. That is, CEOs with higher educational profiles

enjoy more social ties to other CEOs and government officials, which may improve the performance of the firm.

On the other hand, there are strong reasons to suspect that the educational background of a CEO bears little or no relation to their performance. For instance, it may be that since all CEOs

have achieved a high level of success in their work lives, their educational backgrounds do not matter in terms of firm performance. That is, since all the CEOs have attained a great deal of success is does not matter to firm performance if they graduated from Harvard Business School or are a high school drop out. Furthermore, it may be that personality traits that are not directly

1 Kimberly and Evansiko (1981), Bantel and Jackson (1989), Hitt and Tyler (1991), Thomas et. al. (1991),

Wiersema and Bantel (1992), and Wally and Baum (1994) have found that more educated CEOs are better able to process information and are more receptive to change than CEOs with lower educational attainment. In addition, Barker and Mueller (2002, p. 787) state in a survey of the literature that the prevailing pattern of results suggests that more educated CEOs will have preferences towards higher R&D spending as part of being more receptive to innovation.

2developed by education such as charisma, collegiality, effort, etc. may be more important in producing superior firm performance.

How does one measure the quality of CEO education? The answer is not straightforward

as a high quality education can be attained at any school and even without attending a formal academic institution. Our approach is to define educational quality in two ways. First, we define educational quality by the level of educational attainment. Hence, CEOs with graduate degrees are assumed to have higher quality educations than CEOs without graduate degrees. Second, we

define educational quality by the prestige of the schools from which the CEO graduated. To define prestige we use an approach similar to Chevalier and Ellison (1999) and Palia (2000) in which we extract the mean SAT, GMAT, and LSAT scores of the undergraduate and graduate schools from which the CEO graduated. Our assumption is that since more prestigious schools

generally have higher entrance requirements, the use of mean test scores captures the educational prestige of the CEO education. Using this logic, a CEO who graduated from a high mean SAT undergraduate institution would be said to have a higher quality education than a CEO that attended a low mean SAT undergraduate institution. Similarly, a CEO who graduated from a

high mean GMAT business school would be said to have a higher quality education than a CEO who graduated from a low mean GMAT business school.

To conduct the study we survey all firms listed on the New York Stock Exchange and gather those firms for which the CEO of the firm has at least an undergraduate degree. Then,

using the measures of educational quality described above, we examine how educational quality is related to five common measures of firm performance as well as to CEO compensation.

The rest of this paper is organized as follows. Section II presents a brief literature reviewof the previous research on the relation between CEO education and its impact on the firm.

Section III describes the data and Section IV explains the methodology used in the study. Section V presents the results and we conclude with Section VI.

II. Literature Review In the academic press there has been some interest in discerning how CEO education influences the firm. Most of this interest has come from the management area and generally examines how the CEOs education influences the organizational structure of the firm. For example, one area of research finds that the educational attainment of the CEO is a reasonable proxy of the cognitive

3abilities of the CEO. Indeed, several papers in the management literature (Kimberly and Evansiko (1981), Bantel and Jackson (1989), Hitt and Tyler (1991), Thomas et. al (1991), Wiersema and Bantel (1992), and Wally and Baum (1994)) find that CEOs with graduate degrees have a greater capacity to process information and are more receptive to change than CEOs with lower educational attainment. While these papers do not explicitly examine firm performance, they do implicitly argue that firms with highly educated CEOs will perform better than other firms. For example, they do find that more innovative companies are led by CEOs

with higher levels of educational attainment. Another reason for the interest in CEO education is it may uncover certain behavioral

patterns of the manager that, in turn, can impact firm performance. For example, Tyler and Steensma (1998), Finkelstein and Hambrick (1996), and Barker and Mueller (2002) find that companies run by CEOs with degrees in technical fields have significantly higher funding of research and development (R&D). Conversely, CEOs with educational backgrounds in business or law tend to be more risk-adverse with regard to R&D. They also find that if the CEO came up the ranks via technical or marketing channels then they were more supportive of R&D than

CEOs that advanced through the accounting, finance or legal channels. In another study, Graham and Harvey (2002) find that chief financial officers (CFOs) holding MBAs were more likely than other CFOs to use such learned techniques as net present value for capital budgeting and the capital asset pricing model in cost of capital calculations. These results suggest that the type of

educational training can affect the managerial behavior of the CEO.Yet another reason to examine CEO education is that it provides a measure, to some

extent, of the firms social capital. It is well known that education can be a strong indicator of social prestige and class status. Indeed, one can surmise that a large part of why the CEO rose to

his or her position is due to their social network. In addition to using social capital for personal advancement up the corporate ladder, research by Belliveau et. al (1996) and Burt (1992) finds that CEOs with higher educational profiles enjoy more weak-ties to government officials and other decision makers that can improve firm performance. For example, a CEO with strong

social linkages to politicians and policy makers can help the company receive government contracts or favorable tax treatment.

Finally, CEO education may be related to firm type. In a paper that uses somewhat similar methodology to ours, Palia (2000) measures the quality of CEOs by the prestige of their

4educational backgrounds. That is, CEOs who went to more prestigious schools (as determined by school rankings) are assumed to be of higher quality. Using this logic, he finds that firms in non-regulated industries disproportionately hire CEOs from more prestigious academic programs as

compared to firms in regulated industries. Palia provides two possible explanations for this result. First, since non-regulated firms have fewer restrictions, the impact of the CEO on the firm may be greater for a firm in a non-regulated industry than it would be for a firm in a regulated industry. As a result, the non-regulated firms will tend to more hire high-quality CEOs, i.e., those

with more prestigious educational backgrounds, as compared to regulated firms. Second, non-regulated industries have significantly higher CEO compensation than regulated industries Consequently, those high quality executives with prestigious educational backgrounds will be drawn to non-regulated industries and away from regulated industries. Whatever the actual

explanation, the results of Palias paper indicating that underlying firm characteristics can influence the education background of the CEO are very relevant to this paper. While we do not ask whether certain types of firms will hire certain type of CEOs, we do ask whether or not these high quality CEOs, i.e., those CEOs with higher quality educations, imply better performance for

the firms that hired them.

III. Data

To facilitate the description of the data and methodology, this section is divided into three subsections: CEO selection criteria and firm data; the out-of-sample period; and survivorship issues.

III.A. CEO Selection Criteria and Firm DataWe extract CEOs from the EXECUCOMP database. We select all CEOs with U.S. undergraduate degrees that managed NYSE-listed firms as of January 1, 2000. Moreover, we require each of the firms managed by the CEO to have three years of stock return data prior to

January 1, 2000, i.e., 1997 through 1999, as some of our performance measures require in-sample data (see Section IV.B. for more on this issue).

For each of these CEOs we then extract biographical information from the Register Executives publication provided by Standard and Poors NetAdvantage database. This

information includes the CEO tenure, age, gender, and educational background. The educational

5background information provides the name of the educational institution where each CEO received their undergraduate and graduate degrees, and whether the graduate degree was an MBA, law degree, or other graduate degree.2 Unfortunately, information on the undergraduate

field of study is unavailable. For each of the CEOs we then identify several other education variables. First, we

identify the mean composite SAT score (math and verbal combined) associated with each CEOs undergraduate school using the latest available data from the schools. Hence, while the CEO may

have graduated as many as 50 years earlier, we use the current mean SAT scores for their school. In this way, we are implicitly making the assumption that the prestige of the school today (as measured by SAT scores) is similar to that when the CEO graduated. While there are certainly examples of schools that have improved or receded in terms of prestige, most schools that were

considered prestigious years ago are still so today. Second, similar to the SAT extraction, for each CEO that completed an MBA we identify the latest mean GMAT score of the graduate business school attended by the CEO. Likewise, for each CEO that completed a law degree we identify the latest mean composite LSAT score of the law school attended by the CEO. We also

identify whether the CEOs undergraduate school was a liberal arts college. The process through which we identify these educational variables can be found in Appendix A.

After selecting the CEOs and identifying their educational backgrounds, we next gather firm level data. Using the above listed EXECUCOMP database, as well as the COMPUSTAT

and CRSP databases, we gather three types of firm level data over the period 1997-2002. First, we gather the monthly returns of all of the firms. Second, we collect annual operational performance measures such as total sales, Tobins q, ROA, ROE, liquidity and leverage.3 Third,

2 For some CEOs for which the graduate field of study was not easily identifiable, we performed

a general Internet search to identify each CEOs field of study, focusing primarily on business-oriented publications such as Forbes.com, biographical sketches provided by the CEOs firm, or information available through the CEOs alma maters alumni affairs departments.3 Total sales is the net annual sales as reported by the company. ROA is return on assets,

calculated as the net income before extraordinary items and discontinued operations divided by total assets. ROE is return on equity, calculated as the net income before extraordinary items and discontinued operations divided by total common equity. Total sales, ROA, and ROE are extracted from the EXECUCOMP database. Tobins q is calculated for each firm using the Chung and Pruitt (1994) approximation, where all data is extracted from COMPUSTAT. The liquidity ratio is calculated as cash and short-term investments divided by total assets. Leverage

6we extract CEO specific variables from the firm such as annual CEO compensation, the percentage of the firms stock that is owned by the CEO (defined as OWNERSHIP), the CEOs age, and tenure as CEO.4

III.B. The Out-of-Sample PeriodOur study is constructed using an out-of-sample approach. As stated in Section III.A., all the CEOs are chosen as of January 1, 2000. We then evaluate these CEOs over the out-of-sample

period 2000-2002. In this way, all CEOs who meet our criteria are included in the sample. We use the data from 1997-1999 as in-sample data to help construct some of our performance measures, and as lagged variables in some of our regressions.

The reason we use the out-of-sample approach is that it allows us to measure

performance over the relatively lengthy period of three years (2000-2002). If we were to instead gather all CEOs each year and examine their annual performance, we would be limiting our measure of performance to an unnecessarily small window of time that may not be long enough to accurately measure a CEOs impact on the firm.

III.C. Survivorship IssuesSince most of the CEOs in our sample retained their position for the entire out-of-sample period, and most firms survived the entire period, obtaining the data required to measure their out-of-

sample performance is a simple extraction from EXECUCOMP and CRSP. However, some CEOs either retired or resigned during the out-of-sample evaluation period, and some firms disappeared during this time. If we were to simply exclude these firms, it would create a survivorship bias, as we would only be including those CEOs and firms that survived throughout

the entire out-of-sample period. To avoid this survivorship bias problem we proceed in the following fashion. If we identify that the CEO retired or resigned during the out-of-sample

is calculated as the sum of total long-term debt and debt in current liabilities, divided by total assets. The data used to calculate the liquidity ratio and leverage is from COMPUSTAT.

4 CEO compensation is total current compensation comprised of salary and bonus. The

percentage of firms stock that is owned by the CEO is the aggregate number of shares held by the named executive officer, excluding stock options, divided by the number of common shares outstanding as reported by the company. Age and tenure are as of January 2000. The data used to calculate all of these measures is from EXECUCOMP.

7period, or that the firm disappeared, we identify the month and year of the resignation, retirement, or disappearance. For those months in the out-of-sample period that precede the exit, we extract data as we do for CEOs that retained their positions and firms that survived. For the

subsequent months, we replace the missing data with the average of all firms in our sample that share the same two-digit SIC industry code as the given firm.

For example, consider an observation where the CEO had left the firm in July 2001. The monthly returns that we use for the period up to and including June 2001 are the firms own

returns. The monthly returns from July 2001 through December 2002 are the average monthly returns of the surviving firms that share the same industry classification. Similarly, the ROE of a firm whose CEO left in July 2001 is the average ROE for each of the three years, where the 2000 ROE value is the value for the original firm itself. The 2001 ROE value is the weighted average

of the ROE of the original firm and the ROE associated with a portfolio of surviving firms with identical industry classification as the observation, where weighting is based on the proportion of months in 2001 during which the CEO remained CEO. The 2002 ROE value is the ROE associated with the average firm with identical industry classification.

IV. MethodologyTo examine performance, we use five different performance measures: out-of-sample simple excess monthly returns, a modified version of the four-index alpha, Tobins q, return on assets

(ROA), and return on equity (ROE). Both the simple excess returns and the 4-index alpha are market measures of performances while Tobins q, ROA, and ROE can be seen as operationalmeasures of performance.

IV.A. Out-of-Sample Simple Excess Returns Simple excess returns are excess mean monthly returns less the one-month T-Bill rate during the out-of-sample evaluation period (2000-2002), where monthly returns for each firm are extracted from the CRSP database.

IV.B. A Modified 4-index Alpha

8Four-index alpha is a modified version of the Fama-French-Jagadeesh-Titman-Carhart four-index alpha, which represents a risk-adjusted measure of performance.5 To estimate the four-index alpha, the following time-series regression model is used:

ittitititiiftit UMDHMLBSMBRMRFRR +++++= 4321 (1)where ftit RR is the excess total return (net of the 30-day T-bill return) for firm i in the in-

sample Month t; i is the alpha for firm i; tRMRF is the value weighted market return on all

NYSE/AMEX/NASDAQ firms in excess of the risk-free rate; tSMB is the difference in returns across small and big stock portfolios controlling for the same weighted average book-to-market

equity in the two portfolios; tHML is the difference in returns between high and low book-to-

market equity portfolios; tUMD is the momentum factor, the average return on two high prior

return portfolios minus the average return on two low prior portfolios, computed by Fama and French.6

To actually estimate the four-index alphas we use a methodology similar to Elton, Gruber

and Blake (1996). Specifically, for each firm, we utilize a time series period of excess monthlyreturns going back three years (1997 through 1999), and forward to the end of the out-of-sample evaluation period (December 2002), to obtain an estimate of the intercept from equation (1). As noted earlier, to be included in the sample, each firm requires three years of in-sample returns.

To obtain the alphas, we add the average monthly residual during the evaluation period to the intercept. For example, we run equation (1) on firm returns starting in January 1997 and ending in December 2002 (six years) to obtain an estimate of the intercept. We then add the average of each firms residuals during the three years after the selection date (2000-2002) to the estimated intercept (from the regression on 1997-2002) to obtain the firms modified four-index alpha.

To obtain alphas for firms for which the observation disappears, either due to the CEOs retirement or resignation, or due to the firms disappearance, we proceed as follows. First, we

run two regressions: (1) a regression using the firms returns beginning in January 1997 andending in the month prior to the observations disappearance; and (2) a regression run over the

5 See Fama and French (1993) Jagadeesh and Titman (1993) and Carhart (1997).

6 The data for the four-index alpha were obtained from Kenneth Frenchs webpage.

9entire regression period (1997-2002) using the returns of an equally weighted portfolio formed each month from the surviving observations in the sample that share the same two-digit SIC. We then form a weighted average of: (1) the firms estimated intercept plus the firms average residual during the time it is non-missing in the evaluation period and (2) the estimated intercept plus the average residual during the remaining time in the evaluation period of the equally weighted portfolio, where the firms weight is the fraction of the evaluation period it survived and the equally weighted portfolios weight is the remaining fraction. This provides a

performance measure for an investor who buys a random surviving same two-digit SIC firm if the original firm changes CEO or otherwise disappears.

IV.C. Tobins q, ROE, ROATobins q, ROA, and ROE, the extraction of which we described in Section III, are constructed as the average annual value for the period 2000-2002. As stated in Section III, if a CEO retires or resigns, or the firm disappears before the end of the out-of-sample period, we do not eliminate the observation. Instead, the value for the observation is the weighted average of the value for the

original firm and for a portfolio of the firms in the sample that share the same two-digit as the firm itself, where weighting is based on the proportion of each year before which the CEO retired or resigned, or before which the firm disappeared.

V. ResultsV.a. Summary StatisticsSummary statistics are presented in Table I. The table reports several noteworthy findings. The mean composite SAT score of undergraduate institutions of the CEOs is approximately 1,219 (out of a possible 1600). Moreover, approximately 33.6 percent of firms have CEOs who hold MBA degrees, and the mean GMAT score of these MBA programs is approximately 656.7 (out of a possible 800). Approximately 12.2 percent of firms have CEOs who hold law degrees, and the mean LSAT score of these law programs is approximately 161.8 (out of a possible 180). Table I also reports that approximately 10.2 percent of firms have CEOs with graduate degrees other than an MBA, and approximately 10.4 percent of the CEOs received their undergraduate degrees from liberal arts institutions.

10

To fully examine the effect of MBA and law degree quality on performance, we create five dummy variables, GMAT1, GMAT2, GMAT3, LSAT1, and LSAT2 to rank-order thequality of the MBA and law programs by their GMAT and LSAT scores, respectively.

Specifically, GMAT1, GMAT2, and GMAT3 are dummy variables that are equal to unity if the CEO completed an MBA from a graduate school with mean GMAT scores 700 or greater, between 600-699, and 599 or less, respectively.7 Similarly, LSAT1 and LSAT2 are dummy variables that are equal to unity if the CEO completed a law degree from a graduate school with

mean LSAT scores equal to or greater than 165 and below 165, respectively. The reference group is the group of firms with CEOs who do not hold graduate degrees.

Table I reports that approximately 15.4, 11.6, and 6.6 percent of firms in our entire sample are run by CEOs holding MBAs from GMAT1, GMAT2, and GMAT3 schools,

respectively. Approximately 5.6 and 7.1 percent of the firms are run by CEOs holding law degrees from LSAT1 and LSAT2 schools, respectively. Table I also provides summary statistics for other CEO characteristics, measures of firm performance, and other firm characteristics that we use in our tests.

V.b. CEO Performance and Education CharacteristicsTables II through V provide the results of tests of the relation between CEO education and performance. In each of these four tables, we estimate CEO performance over the period 2000

through 2002 using excess simple (mean monthly) returns, the modified 4-index alpha, Tobins q, ROE, and ROA. For each of these five measures of performance, we run two types of regressions. First, using ordinary least squares (OLS) we run a regression using CEO education variables, age, tenure, ownership, compensation, sales (as a proxy for size), and dummy variables for the 2-digit SIC of the firm. The purpose of these regressions is to examine the effect of CEO educational quality on firm performance while controlling for other issues. The age and tenure variables are used to control for the presence of any age or tenure effects. Ownership is used to control for the fact that some CEOs may have more control over the firm than others and

as a result may have more influence over performance. Compensation is used to control for the

7 For example, the business schools from which fund managers in our sample graduated from in

the GMAT1 group included Wharton, Columbia, Harvard, Stanford, NYU, MIT, Northwestern, UCLA, Duke and UC-Berkeley.

11

possibility that more highly compensated CEOs perform differently than less well compensated CEOs. Sales is used to control for the size of the firm and dummy variables based on the 2-digit SIC codes are used to control for any industry effects.8

For the second regression, we run an instrumental variable (IV) regression where we use the same variables as in the OLS regressions but also include out-of-sample leverage and the out-of-sample liquidity ratio as endogenous variables. We include both of these variables as endogenous variables as the CEO may be able to influence these variables. Lagged values of

leverage and liquidity, specified as the 1999 values of these measures, as used as instruments in the regression. 9

The results of CEO education on performance are organized as follows. Table II presents the results of tests where the education variables include SAT (divided by 100), the liberal arts dummy, and indicator variables equal to unity if the CEO completed an MBA, law degree, or other graduate degree, respectively.10 In Table III we replicate the same regressions as estimated in Table II but replace the MBA and law degree indicator variables with the three GMAT and two LSAT indicator variables, GMAT1-3 and LSAT1-2. In Table IV, we present results for the

sample of firms whose CEOs hold MBA degrees. Because all observations in this sample have a GMAT score, we include the GMAT score as a regressor. Similarly, Table V displays the results for the sample of firms whose CEO hold law degrees. Since all observations in this sample have a LSAT score, we include the LSAT score as a regressor.

The performance regressions show three interesting findings. First, we find no evidence that firms run by CEOs with from higher prestige undergraduate institutions perform any better than firms lead by CEOs with educations from less prestigious undergraduate schools. All ten of the SAT coefficients in Table II are negative and generally insignificant, indicating that

undergraduate educational prestige has little impact on firm performance. In the regressions

8 The results of these industry dummies are not reported. These results are available upon request.

9 Note that we conducted the analysis presented in Tables II-V using lagged performance

measures to control for past performance of the firm. In all cases, that the inclusion of these variables did not significantly impact our results.

10 The reference group for the MBA and Law dummies is firms with CEOs who do not hold

graduate degrees.

12

where the dependent variable is the firms ROA, or the firms Tobins Q, we actually find that the SAT score has a negative and significant relationship to firm performance, suggesting that firms with CEOs from lower prestige undergraduate institutions actually perform better than

other firms. While these results for ROA and Tobins Q are confirmed in our sample of CEOs that hold law degrees (Table V), they do not hold up when we examine a sample of CEOs that all hold MBA degrees (Table IV).

Second, the results of Tables III, IV, and V illustrate that there is no evidence that firms

run by CEOs with from higher prestige graduate business or law schools perform any better than firms lead by CEOs with educations from less prestigious graduate business or law schools. Table III shows that the graduate quality dummy variables, GMAT1, GMAT2, GMAT3, and LSAT1, LSAT2, are always insignificant across all the regressions, indicating that the

performance of CEOs has little do with the quality of the graduate program. Similarly, Tables IV and V illustrate that the GMAT and LSAT variables are also insignificant when the tests are conducted on samples of CEOs who all hold MBA or law degrees. Moreover, the point estimates of these variables are often negative, indicating that if anything, there is a negative relationship

between educational prestige of the CEOs graduate school and firm performance.Third, we find mixed results when examining the relationship between CEO educational

attainment and firm performance. We find that firms managed by CEOs with MBA or law degrees perform no better than firms with CEOs without graduate degrees. As can be seen in

Table II, the dummy variables, MBA and LAW, are both insignificant. Hence, while much of the management literature (as briefly described in Section II) suggests that CEOs with higher levels of education attainment will have management styles and characteristics that may help the firm, we do not find that greater CEO educational attainment (in terms of MBA and law degrees) improves firm performance. On the other hand, we do find some limited evidence that firms led by CEOs with graduate degrees in fields other than business and law, i.e., other graduate degrees, do perform significantly better (at the 10 percent level) in terms of risk-adjusted performance (the 4-index alpha). Assuming that these other graduate degree holding CEOs have technical graduate degrees, this last result may be consistent with Barker and Muller (2002) who find that firms with CEOs with technical degrees spend more on R&D and hence may have better performance.

13

V.c. CEO Compensation and Education CharacteristicsWhile Tables II-V illustrate the results of educational quality on CEO performance, another issue we wanted to examine was whether the quality of the education of the CEO influenced their

compensation. As mentioned in Section II, one of the findings of Palia (2000) was that CEOs in non-regulated industries received greater compensation than CEOs in regulated industries. Since Palia also found that CEOs of firms in non-regulated industries had had more prestigious educations, his results suggest that CEOs with better educational backgrounds extract higher

compensation. To more directly examine this issue, we examine the relationship between CEO educational quality and CEO compensation.

To estimate CEO compensation, we use the average annual compensation for each CEO in our sample for the period 2000 through 2002. Using OLS only, we then examine the linkage

between educational quality and CEO compensation. As control variables we use age, tenure, ownership, compensation, sales, and 2-digit SIC variables as independent variables, though the results for 2-digit SIC are unreported. As well, we include a single performance measure in each regression (simple mean monthly returns, the 4-index alpha, Tobinq, ROE or ROA) to control for performance of the firm as the literature has shown that CEO compensation is related to firm performance (see Jensen and Murphy (1990) for example).

In the first five regressions, columns 1 through 5 of Table VI, we limit the educational variables to the ones we use in Table II: SAT, the liberal arts indicator variable, and indicator

variables that are equal to unity if the CEO completed an MBA, law degree, or other graduate degree, respectively. In the subsequent five regressions, columns 6 through 10, we replace the MBA and law degree indicator variables with the three GMAT and two LSAT indicator variables, GMAT1-3 and LSAT1-2, respectively, similar to the refinement we presented in Table

III.

The results of Table VI indicate that the quality of CEO education is related to CEO compensation. In regressions 1 through 5, we find a positive and significant (at the 10 percent level) coefficient associated with the SAT variable, indicating that CEOs from higher prestige undergraduate institutions receive significant more compensation than CEOs from lower prestigeundergraduate institutions. For the regressions in columns 6 through 10, we find a significantly negative coefficient associated with the LSAT2 variable, significant throughout at the one

14

percent level. Hence, CEOs who graduated from low LSAT law schools have significantly worse compensation.

VI. ConclusionsIn this paper, we empirically relate the quality of CEO education to firm performance and CEO compensation. We measure educational quality by examining the CEO educational attainment and by measuring the prestige of the schools that the CEO graduated from by using mean entrance exam scores. We find four results. First, we find no evidence that firms run by CEOs

from more prestigious schools (undergraduate or graduate) perform better than firms run by CEOs from less prestigious schools. If anything, we find that the prestige of the CEOs school is negatively related to firm performance as the point estimates of the coefficients are generally negative. Indeed, in the cases where performance is measured by Tobins Q and ROA, the SAT coefficients are actually negative and significant, indicating that there is a negative relationship between the prestige of the CEOs undergraduate education and firm performance. Second, we find that firms managed by CEOs with MBA or law degrees perform no better than firms with CEOs without graduate degrees, suggesting that the impact of a graduate business or law degree

is minimal on CEO performance. Third, we find some limited evidence that firms run by CEOs holding graduate degrees in other fields besides that of law or an MBA have slightly better risk-adjusted market performance than other firms. Assuming that these other graduate degree holding CEOs have technical graduate degrees, this last result is consistent with Barker and

Muller (2002) who find that firms with CEOs with technical degrees spend more on R&D and hence may have better performance. Fourth, and finally, we find that that compensation is somewhat higher for CEOs who attended high mean SAT undergraduate institutions and lower for CEOs that attended low LSAT law schools.

What are some explanations for our results? First, consider our findings that the firm performance is not impacted by the prestige or quantity of the CEOs undergraduate or graduate education. There are several reasons why prestige may not matter. For example, even if mean entrance test scores are a satisfactory measure of educational quality, the amount of time

between the CEOs completion of the degree(s) and the attainment of the position of CEO may be sufficiently lengthy to diminish any benefit that can flow from a superior education. Indeed, anyone who becomes a CEO of a NYSE firm likely has certain skills, developed over a lifetime,

15

which enabled them to achieve their position. Consequently, the quality of CEO education, particularly education that was completed years earlier, may have little to do with CEOs current performance. Furthermore, the lack of a relationship between CEO educational prestige and firm

performance may be due to the fact that CEOs with less prestigious educations work harder or longer than CEOs with high prestige educational backgrounds. As a result, any positive effect of prestige, i.e., better education, better social networks, etc., may be offset by a competing negative relation that results from CEOs from weaker schools overcompensating through superior

performance. After all, completing a degree at a less prestigious school leaves a student at a reputational and social capital disadvantage relative to the students competitors from better schools. This disadvantage represents a barrier to career progression, beginning with the initial job search and continuing as the individual pursues promotions. To overcome this disadvantage and eventually achieve the position as CEO, the individual from a weaker school must demonstrate superior performance capabilities, to compensate for this disadvantage. This rationale can explain why we find only weak evidence of any relation between educational quality and performance. Finally, our results showing that CEO educational prestige has little

impact on firm performance may be simply because mean entrance test scores represent a weak proxy for prestige. As is well known, entrance into an educational institution is a function of several factors besides entrance test scores. Further, while simply gaining entrance into a degree program is an achievement, the acceptance alone does not provide full information regarding the

nature of the education and social capital that the student acquires.Finally, consider the issue of compensation. Similar to Palia (2000), we do find some

evidence that attending more (less) prestigious undergraduate (law) schools implies higher (less) compensation. These results suggest that even though performance may not be influenced by

education, the ability of a CEO to extract higher compensation is enhanced by a more prestigious educational background. In a sense, companies are willing to pay more for a CEO with a prestigious educational background even though our previous results suggest that this prestigious educational background may not be worth the price.

What are the implications of our results? Assuming that we are measuring the quality of

education correctly, our results suggest, for the most part, that CEO educational quality is unrelated to firm performance. We do find some limited evidence that CEOs holding graduate degrees other than MBAs and law degrees perform better on a risk-adjusted basis; however,

16

these results are only weakly significant and do not hold up across all measures of firm performance. Such results indicate that companies in their search for the right CEO should not focus too much attention on the educational background of a CEO candidate if better firm

performance is the goal.

17

Appendix A: Further Description of Education Quality Variables. SATWe obtain up-to-date SAT scores for the undergraduate schools through initially searching

Collegeboard.com for the SAT I Verbal and SAT I Math test score ranges for the middle 50% of first-year students.11 The mean values of the verbal and math score ranges are calculated, and the average of the verbal and math scores is calculated to identify a single SAT score for each school. For all schools for which SAT scores were not identified on Collegeboard.com, we then

search the Princeton Guide to Colleges (2004) for mean SAT scores. For a select few schools, while SAT scores are unavailable, ACT scores are reported. In these cases, the ACT scores are converted into SAT scores using the SAT ACT score comparisons provided on Collegeboard.com.12

GMATWe obtain GMAT scores for the MBA schools through initially searching MBA.com for the mean GMAT score of new entrants.13 For all schools for which GMAT scores are not identified

on MBA.com, we then search Businessweek.coms 2003 Full-Time MBA Profiles for the mean GMAT scores. For all schools for which GMAT scores are not identified on either of the above two sources, we search Petersons Guide to MBA Programs (2004) for the mean GMAT scores.14

11 Collegeboard.com is the website of the College Board, the organization that administers the SAT tests, among other activities. The search was performed December 2003.

12 For a few of the CEOs, the undergraduate school reported is actually a system of schools. In

these cases, the SAT score identified is the average of the SAT scores for the schools within the system for which SAT scores are reported on Collegeboard.com.

13 MBA.com is the website of the Graduate Management Admission Council, the organization that administers the GMAT tests. The search was performed December 2003. 14

As for the SAT scores extraction, the graduate school reported is occasionally a system of schools. In these cases, we proceeded in an identical fashion as described above for SAT scores.

18

LSATWe obtain LSAT scores for the law schools through searching LSAT.org for test score ranges for the middle 50% of first-year students.15 The mean values of these ranges are calculated to identify a single LSAT score for each law school.

Liberal ArtsUsing USnews.coms 2004 list of liberal arts colleges we determine whether the CEOs

undergraduate institution is a liberal arts college or not. We choose this variable since many educators argue that the individual attention given to students at liberal arts schools is superior to that in larger, research-oriented institutions.

15 LSAT.com is the website of the Law School Admission Council, the organization that administers the LSAT tests, among other activities. The search was performed December 2003.

19

ReferencesBantel, K.A. and S.E. Jackson, 1989, Top Management Innovations in Banking: Does the Composition of the Top Team Make a Difference? Strategic Management Journal, 10, 107-124.

Barker, V.L., and G.C. Muller, 2002, CEO Characteristics and Firm R&D Spending, Management Science, 48, 6, 782-801.

Belliveau, M.A., C.A. OReilly III, and J.B. Wade, 1996, Social Capital at the Top: Effects of Social Similarity and Status on CEO Compensation, Academy of Management Journal, 39, 6, 1568-1593.

Burt, R.S., 1992, Structural Holes: The Social Structure of Competition, Cambridge, MA: Harvard University Press.

Carhart, Mark, 1997, On Persistence in Mutual Fund Performance, Journal of Finance, 52, 1, 57-82.

Chevalier, Judith and Glenn Ellison, 1999, Are Some Mutual Fund Managers Better than Others? Cross-Sectional Patterns in Behavior and Performance, Journal of Finance, 54, 3, 875-899.

Chung, K. H. and S.W. Pruitt, 1994, A Simple Approximation of Tobins q, Financial Management 23, 70 74.

Elton, Edwin J., Martin J. Gruber and Christopher R. Blake, 1996, The Persistence of Risk-Adjusted Mutual Fund Performance, Journal of Business, 69, 2, 133-157.

Fama, Eugene and Kenneth R. French, 1993, The Common Risk Factors in the Returns on Stocks and Bonds, Journal of Financial Economics, 33, 1, 3-56.

Finkelstein, S. and D.C. Hambrick, 1996, Strategic Leadership: Top Executives and their Effects on Organizations, West Publishing Company, St. Paul, MN.

Golec, Joseph H., 1996, The Effects of Mutual Fund Managers Characteristics on Their Portfolio Performance, Risk and Fees, Financial Services Review, 5, 133-148.

Gottesman, Aron, and Matthew R. Morey, 2006, Manager Education and Mutual Fund Performance, Journal of Empirical Finance, 2006, 13, 2, 145-182.

Graham, John and Campbell Harvey, 2002, "How do CFOs make capital budgeting and capital structure decisions?, Journal of Applied Corporate Finance 15, 1, 8-23.

Hitt, M.A., and B.B. Tyler, 1991, Strategic Decision Models: Integrating Different Perspectives, Strategic Management Journal, 12, 327-351.

20

Jegadeesh, Narasimhan and Sheridan Titman, 1993, Returns to buying winners and selling losers: Implications for Stock Market Efficiency, Journal of Finance, 48, 65-91.

Jensen, M.C. and Murphy, K.J., 1990, Performance Pay and Top-Management Incentives, Journal of Political Economy, 98, 225-264.

Kimberly, J.R. and M.J. Evansiko, 1981, Organizational innovation: The influence of individual, organizational, and contextual factors on hospital adoption of technical and administrative innovations Academy of Management Journal, 24, 689-713.

Palia, D., 2000, The Impact of Regulation on CEO Labor markets, Rand Journal of Economics, 31, 1, 165-79.

Petersons Guide to MBA Programs, 2004 edition, Lawrenceville, NJ, Peterson Publishing.

Princeton Guide to Colleges, 2004 edition, Random House, New York, NY.

Thomas, A.S., R. J. Litschert, K. Ramaswamy, 1991, The performance impact of strategy-manager coalignment: An empirical examination, Strategic Management Journal, 12, 509-522.

Tyler, B.B., and H.K. Steensma, 1998, The effects of executives experiences and perceptions on their assessment of potential technical alliances, Strategic Management Journal, 19, 939-965.

Wally, S. and J. Baum, 1994, Personal and Structural Determinants of the Pace of Strategic Decision Making, Academy of Management Journal, 37, 932-956.

Wiersema, M.F. and Bantel, A.K., 1992, Top Management Team Demography and Corporate Strategic Change, Academy of Management Journal, 35, 91-121.

21

Table I: Summary StatisticsSAT is the mean composite SAT score of the CEOs undergraduate school. LIBERAL ARTS is a dummy variable that is equal to unity if the CEO attended a liberal arts school. MBA is a dummy variable that is equal to unity if the CEO completed an MBA degree. GMAT is the mean GMAT score of the CEOs MBA graduate school. GMAT1-3 are dummy variables that are equal to unity if the CEO completed an MBA from a graduate school with mean GMAT scores equal or greater than 700, between 600-699, and 599 or less, respectively. LAW is a dummy variable that is equal to unity if the CEO completed a law degree. LSAT is the mean LSAT score of the CEOs law school. LSAT1-2 are dummy variables that are equal to unity if the CEO completed a law degree from a graduate school with mean LSAT scores equal to or greater than 165 and below 165, respectively. OTHER SECOND DEGREE is a dummy variable that is equal to unity if the CEO completed a graduate degree that is not identifiable as a law or MBA degree. AGE is the CEOs age. TENURE is the tenure of the CEO at the firm. The remaining variables are calculated over the period 2000-2002. OWNERSHIP is the average annual percentage of the firms stocks that the CEO holds. COMPENSATION is the average annual total compensation of the CEO. SIMPLE EXCESS RETURNS are the mean monthly excess returns of the CEOs firm. 4-INDEX ALPHA is the 4-index alpha of the CEOs firm. TOBINS Q is the average annual Tobins q. ROE is the average annual return on equity for the CEOs firm. ROA is the average annual return on assets for the CEOs firm. SALES is the average annual sales of the CEOs firm. LEVERAGE is the average annual debt ratio of the CEOs firm. LIQUIDITY RATIO is the average annual liquidity ratio of the CEOs firm.

Name Number Mean STDCEO Education Variables:SAT/100 482 12.191 1.475MBA 482 0.336 0.473GMAT/100 162 6.567 0.655GMAT1 (700 and above) 482 0.154 0.361GMAT2 (600-699) 482 0.116 0.321GMAT3 (599-below) 482 0.066 0.249LAW 482 0.122 0.328LSAT/100 59 1.618 0.066LSAT1 (165 and above) 482 0.056 0.23LSAT2 (164 and below) 482 0.071 0.256LIBERAL ARTS 482 0.104 0.305OTHER GRAD DEGREE 482 0.102 0.303Other CEO Characteristics:AGE/10 482 5.628 0.899TENURE 482 6.743 6.359OWNERSHIP 482 0.021 0.048COMPENSATION/1,000 482 1,657.711 1,465.701Firm Performance:SIMPLE EXCESS RETURNS 482 0.007 0.0164-INDEX ALPHA 482 0.008 0.016TOBINS Q 469 1.689 1.278ROE 481 11.004 33.161ROA 482 4.014 5.308Other Firm Characteristics:LOG SALES/1,000,000 482 6,151.228 13,255.43LEVERAGE 482 0.282 0.167LIQUIDITY RATIO 482 0.084 0.114

22

Table II: Performance Regressions Using MBA and LAW Dummy Variables. We examine the relation between firm performance and CEO education using MBA and LAW Dummy variables. Each regression is estimated using both Ordinary Least Squares (OLS) and Instrumental Variables (IV) methodologies. For the IV regressions liquidity and leverage are treated as endogenous variables where the instruments of the regression are lagged liquidity and lagged leverage. These regressions include dummy variables for 2-digit SIC codes. The results for the industry dummies are not reported. ***, **, * indicate significance at the one, five, and ten percent levels, respectively.

Dependent Variable is the Performance Measure

Independent Variables:SIMPLE

/ OLS4-INDEX

/ OLS TOBINS Q

/OLSROE/OLS

ROA/OLS

SIMPLE/IV

4-INDEX/IV

TOBINS Q/IV

ROE/IV

ROA/IV

Intercept 0.0836*** 0.0517*** 1.6972* -21.9061 6.1955 0.0867*** 0.0387** 1.3828 -20.3493 5.4222

SAT/100 -0.0004 -0.0006 -0.0608** -0.3786 -0.3697** -0.0004 -0.0005 -0.0560** -0.4026 -0.3556**LIBERAL ARTS -0.0003 -0.0007 -0.0101 -6.3077 -0.9294 0.0002 -0.0005 -0.0295 -6.1707 -1.0233MBA -0.0005 0.0000 -0.0687 1.7144 0.0255 0.0001 0.0003 -0.0916 1.8727 -0.0848LAW -0.0008 -0.0014 0.0864 -2.6925 -0.4367 -0.0013 -0.0023 0.0895 -2.8096 -0.3470OTHER GRAD DEGREE -0.0004 0.0053* 0.1028 -4.4459 -0.0287 0.0003 0.0032 0.0451 -4.1326 -0.1990AGE/10 -0.0007 -0.0013 -0.0431 1.8695 0.0072 -0.0012 -0.0011 -0.0186 1.7140 0.1045TENURE

0.0001 0.0001 0.0057 -0.4554 -0.0667 0.0001 0.0001 0.0064 -0.4605 -0.0634OWNERSHIP

0.0120 0.0210 1.7590 13.2730 17.6360*** 0.0278 0.0046 0.9137 18.6500 14.4411**log(COMPENSATION)

0.0006 0.0014 0.0911 1.6374 0.9027*** 0.0009 0.0007 0.0692 1.7645 0.8349***LOG SALES -0.0037*** -0.0026*** -0.0488 0.1430 -0.5258** -0.0043*** -0.0017** -0.0122 -0.0758 -0.3994*LEVERAGE

0.0261*** 0.0013 -1.0967*** 7.4691 -4.8787**LIQUIDITY RATIO

0.0003 0.0384*** 0.4348 -1.8489 0.5074# OBS.

482 482 469 481 482 482 482 469 481 482Adj-R2

0.213 0.109 0.567 0.090 0.195 0.191 0.188 0.597 0.083 0.227

23

Table III: Performance Regressions using GMAT1-3 and LSAT1-2 Dummy Variables.The table examines the relation between firm performance and CEO education using GMAT1-3 and LSAT1-2 dummy variables. Each regression is estimated using both Ordinary Least Squares (OLS) and Instrumental Variables (IV) methodologies. For the IV regressions liquidity and leverage are treated as endogenous variables where the instruments of the regression are lagged liquidity and lagged leverage. These regressions include dummy variables for 2-digit SIC codes. The results for the industry dummies are not reported. ***, **, * indicate significance at the one, five, and ten percent levels, respectively.

Dependent Variable is the Performance Measure

Independent Variables:SIMPLE

/ OLS4-INDEX

/ OLS TOBINS Q

/OLSROE/OLS

ROA/OLS

SIMPLE/IV

4-INDEX/IV

TOBINS Q/IV

ROE/IV

ROA/IV

Intercept 0.0819*** 0.0523*** 1.8357** -22.2806 7.0863 0.0858*** 0.039** 1.4998* -19.8392 6.2943

SAT/100 -0.0002 -0.0006 -0.0712** -0.2899 -0.4398*** -0.0003 -0.0005 -0.0642** -0.3319 -0.4163**LIBERAL ARTS -0.0001 -0.0007 -0.0179 -5.6812 -0.9312 0.0004 -0.0005 -0.0369 -5.5641 -1.0238GMAT1 -0.0023 0.0004 0.0231 -1.7083 0.3986 -0.0013 0.0007 -0.0129 -1.4809 0.2260GMAT2

0.0020 0.0022 -0.0619 5.9684 0.0106 0.0028 0.0020 -0.0970 6.1938 -0.1303GMAT3 -0.0010 -0.0042 -0.2706 2.1800 -0.6924 -0.0013 -0.0030 -0.2432 1.9535 -0.6193LSAT1 -0.0028 -0.0024 0.1823 3.1226 1.0976 -0.0024 -0.0036 0.1511 3.3445 1.0210LSAT2

0.0006 -0.0002 0.0148 -6.5478 -1.4873 -0.0005 -0.0007 0.0438 -6.8112 -1.2815OTHER GRAD DEGREE -0.0003 0.0054** 0.1041 -3.9454 0.0070 0.0005 0.0034 0.0471 -3.5213 -0.1495AGE/10 -0.0007 -0.0014 -0.0547 1.6567 -0.0672 -0.0012 -0.0012 -0.0276 1.4748 0.0386TENURE

0.0001 0.0001 0.0059 -0.4858* -0.0679 0.0000 0.0001 0.0066 -0.4914* -0.0647OWNERSHIP

0.0147 0.0226 1.6954 13.6495 16.8655*** 0.0301 0.0064 0.8864 19.5317 13.9483**log(COMPENSATION)

0.0008 0.0015 0.0843 1.3670 0.8148** 0.0010 0.0008 0.0653 1.5089 0.7636**LOG SALES -0.0038*** -0.0026*** -0.0415 0.3239 -0.4490** -0.0044*** -0.0018** -0.0080 0.0883 -0.3417LEVERAGE

0.0257*** 0.0013 -1.0701*** 6.9153 -4.7409**LIQUIDITY RATIO -0.0002 0.0379*** 0.4140 -3.9636 0.2296# OBS.

482 482 469 481 482 482 482 469 481 482Adj-R2

0.213 0.110 0.567 0.089 0.197 0.191 0.187 0.595 0.083 0.227

24

Table IV: Performance Regressions MBA Only. The table examines the relation between firm performance and CEO education using a sample of CEOs that hold MBA degrees. GMAT is the mean GMAT score of the CEOs MBA graduate school. Each regression is estimated using both Ordinary Least Squares (OLS) and Instrumental Variables (IV) methodologies. For the IV regressions liquidity and leverage are treated as endogenous variables where the instruments of the regression are lagged liquidity and lagged leverage. These regressions include dummy variables for 2-digit SIC codes. The results for the industry dummies are not reported. ***, **, * indicate significance at the one, five, and ten percent levels, respectively.

Dependent Variable is the Performance Measure

Independent Variables:SIMPLE

/ OLS4-INDEX

/ OLS TOBINS Q

/OLSROE/OLS

ROA/OLS

SIMPLE/IV

4-INDEX/IV

TOBINS Q/IV

ROE/IV

ROA/IV

Intercept 0.0332 -0.0476 -3.3455 1.2000 -13.3084 0.0290 -0.0543 -3.6237 -9.0078 -13.4908

SAT/100 0.0006 0.0001 -0.0026 0.0955 0.3071 0.0003 0.0009 0.0420 -0.5553 0.4065

LIBERAL ARTS 0.0005 -0.0002 -0.0115 -18.5834* -0.5917 0.0006 0.0009 0.0241 -18.3992* -0.4926

GMAT -0.0012 0.0028 0.1454 -0.9666 0.1946 -0.0003 0.0016 0.0232 1.2285 -0.0183AGE/10 -0.0014 -0.0005 -0.0418 1.2644 -0.3666 -0.0019 -0.0007 -0.0300 0.1488 -0.3321TENURE

0.0007** 0.0005 -0.0039 -0.2446 -0.0096 0.0005* 0.0005 0.0009 -0.6655 0.0077OWNERSHIP -0.0549 -0.0831 -1.6855 -77.6869 -23.1318 -0.0063 -0.1032 -4.4209 39.3825 -30.0078log(COMPENSATION)

0.0075* 0.0071* 0.2934 4.4319 2.6889** 0.0075* 0.0065* 0.2978 4.5856 2.6221**LOG SALES -0.0058*** -0.0031* -0.0191 -2.3916 -1.1079** -0.0062*** -0.0028* -0.0049 -3.2684 -1.0409**LEVERAGE

0.0350** 0.0107 -0.4295 83.8844* -2.5893LIQUIDITY RATIO

0.0000 0.0681** 2.8801* -1.2892 6.4131# OBS.

161 161 137 161 161 161 161 137 161 161Adj-R2

0.306 0.160 0.334 0.066 0.116 0.280 0.283 0.390 0.046 0.121

25

Table V: Performance Regressions LAW Only. The table examines the relation between firm performance and CEO education using a sample of CEOs that hold law degrees. LSAT is the mean LSAT score of the CEOs law school. Each regression is estimated using both Ordinary Least Squares (OLS) and Instrumental Variables (IV) methodologies. For the IV regressions liquidity and leverage are treated as endogenous variables where the instruments of the regression are lagged liquidity and lagged leverage. These regressions include dummy variables for 2-digit SIC codes. The results for the industry dummies are not reported. ***, **, * indicate significance at the one, five, and ten percent levels, respectively.

Dependent Variable is the Performance Measure

Independent Variables:SIMPLE

/ OLS4-INDEX

/ OLS TOBINS Q

/OLSROE/OLS

ROA/OLS

SIMPLE/IV

4-INDEX/IV

TOBINS Q/IV

ROE/IV

ROA/IV

Intercept 0.0060 0.0351 -3.3195 -14.268 -62.8724* 0.0552 0.1063 -2.4345 -100.0169 -77.8987*

SAT/100 -0.0031* -0.0028 -0.0828 -2.6146* -1.8895*** -0.0030 -0.0024 -0.0724 -2.8318* -1.8869***LIBERAL ARTS -0.0033 0.0022 -0.5057 -5.2366 -3.9767 -0.0027 0.0020 -0.5325 -6.3373 -4.4850*LSAT

0.0335 0.0353 2.6415 41.0710 45.8591** 0.0152 -0.0004 2.0252 73.083* 48.857**AGE/10

0.0002 -0.0013 0.9317*** 0.8302 1.0177 -0.0015 -0.0049 0.8204** 3.7407 1.1795TENURE -0.0013** -0.0003 -0.0472 -0.0785 -0.2002 -0.0010 0.0005 -0.0230 -0.6727 -0.2204OWNERSHIP

0.0818 -0.0390 -11.3356** -45.2724 2.1275 0.0957 -0.0188 -10.9711** -69.4206 -2.0598log(COMPENSATION)

0.0003 -0.0018 -0.1517* -0.5156 0.0613 0.0006 -0.0015 -0.148* -1.0455 -0.0518LOG SALES -0.0024 -0.0011 0.0491 0.0040 0.5079 -0.0031 -0.0021 0.0264 1.1818 0.6893LEVERAGE -0.0354 -0.0659 -1.7869 61.9972* 6.7179LIQUIDITY RATIO -0.0133 0.0088 0.9513 22.7224 11.9729# OBS.

59 59 49 59 59 59 59 49 59 59Adj-R2

0.385 0.304 0.725 0.898 0.553 0.339 0.297 0.711 0.908 0.566

26

Table VI: Compensation Regressions The table examines the relation between CEO compensation and CEO education. Each regression is estimated using Ordinary Least Squares (OLS). The dependent variable is log (average annual CEO compensation over the period 2000-2002). Regressions 1-5 use MBA and LAW dummies. Regressions 6-10 use GMAT1-3 and LSAT1-2 dummies. To control for performance, regressions 1 and 6 use the mean monthly excess returns of the CEOs firm over the period 2000-2002; regressions 2 and 7 use the 4-index alpha; regressions 3 and 8 use Tobins q; regressions 4 and 9 use ROE; regressions 5 and 10 use ROA. These regressions include dummy variables for 2-digit SIC codes. The results for the industry dummies are not reported. ***, **, * indicate significance at the one, five, and ten percent levels, respectively.

Dependent Variable is log(COMPENSATION)Independent Variables: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)Intercept

6.2858*** 6.233*** 6.2608*** 6.3999*** 6.1755*** 6.2771*** 6.2398*** 6.3076*** 6.4179*** 6.215***SAT/100

0.0406 0.0417* 0.0454* 0.0417* 0.0472* 0.0273 0.0289 0.0326 0.0288 0.0350LIBERAL ARTS

0.0485 0.0502 0.0490 0.0554 0.0670 0.0514 0.0534 0.0534 0.0574 0.0685MBA -0.0399 -0.0407 -0.0332 -0.0363 -0.0406LAW -0.1794 -0.1754 -0.1867 -0.1706 -0.1680GMAT1

0.0029 -0.0031 -0.002 0.0031 -0.0094GMAT2 -0.0614 -0.0647 -0.0542 -0.0544 -0.0567GMAT3 -0.0589 -0.0462 -0.0451 -0.0544 -0.0467LSAT1

0.2205 0.2224 0.1898 0.2182 0.1907LSAT2 -0.4175*** -0.414*** -0.4320*** -0.4036*** -0.3818***OTHER GRAD DEGREE

0.0142 -0.0038 0.0077 0.0279 0.0139 0.0275 0.0083 0.0184 0.0401 0.0264Age/10 -0.0211 -0.0181 -0.0215 -0.0238 -0.0220 -0.0326 -0.0290 -0.0329 -0.0350 -0.0322TENURE

0.0256*** 0.0252*** 0.0253*** 0.0259*** 0.0267*** 0.0245*** 0.0241*** 0.0242*** 0.0248*** 0.0256***OWNERSHIP -5.9679*** -5.9948*** -6.0461*** -5.9394*** -6.2168*** -6.0433*** -6.0692*** -6.0944*** -6.008*** -6.2529***LOG SALES

0.3214*** 0.3224*** 0.3171*** 0.3152*** 0.3207*** 0.3326*** 0.3325*** 0.3259*** 0.3251*** 0.3289***SIMPLE

1.6780 2.01904-INDEX

3.2799 3.4274TOBINS Q

0.0724 0.0661ROE

0.0009 0.0007ROA

0.0214*** 0.0192**# OBS.

482 482 469 481 482 482 482 469 481 482Adj-R2

0.336 0.338 0.306 0.339 0.348 0.345 0.347 0.315 0.347 0.354