Shareholder value and the number of outside board seats held by executive officers

Shareholder value and the number of outside board seats held by executive officers by Tod Perry a and Urs C. Peyer b Preliminary Draft Comments Welcome 3/14/2002 Abstract We find that shareholders react negatively to the announcement that one of their executives has accepted a position on the board of another publicly traded company. These shareholder losses are even greater in firms where the executive holds more outside directorships. Our results are consistent with an agency problem interpretation because the individual executives gain in terms of prestige or wealth by accumulating board seats at the expense of shareholders. Our tests indicate that these problems are severe and statistically significant in companies with high growth opportunities and in firms where the board does not have a majority of independent directors. We also document a negative relation between the number of directorships held by the nominee and the announcement return for the firm nominating the executive to the board. We find limited support for the hypothesis that the number of directorships is a signal of managerial quality, even after controlling for the predicted number of directorships and the investment opportunities of the firms. We thank Hal Bosher and Swaminathan Kalpathy for research assistance. Tod Perry would also like to thank the Investor Responsibility Research Center for help in obtaining some of the board data used in this study. a Tod Perry, Department of Finance, Arizona State University, Tempe AZ. Tel. (480) 965 1682, email: Tod.Perry@asu.edu b Urs Peyer, Department of Finance, INSEAD, Boulevard de Constance, 77305 Fontainebleau, France. Tel. +33 1 6072 4178, email: Urs.Peyer@insead.edu

1 Introduction In 1996, the National Association of Corporate Directors (NACD) issued guidelines relating to director professionalism and conduct, including a recommendation that corporate executives should not hold more than three outside directorships. Outside directorships require significant time commitments by executives -- Blackhurst (2000) estimates as many as two full workdays per board meeting plus another two days per committee meeting. Managerial ability is a scarce resource, and the time and effort expected of managers to monitor firms will generally reduce the time committed to the executive s own firm. Furthermore, Core, Holthauser, and Larcker (1999) find that CEO pay is excessive in firms where board members are busy, i.e., hold three or more outside directorships, suggesting that busy boards might not monitor management effectively. The purpose of this paper is to test the relation between the number of outside directorships held by executives and firm value in a more direct way. We examine the cumulative abnormal returns (CAR) around the announcement that an executive accepts an additional directorship, controlling for the number of prior outside directorships. Based upon the NACD recommendation and evidence in Core et al. (1999), we would expect to find a negative association between the number of directorships and firm value. On the other hand, researchers have argued that the number of directorships can signal managerial quality. Fama s (1980) model predicts that directors who perform well are rewarded by the market in the form of additional board seats. Kaplan and Reishus (1990), Gilson (1990), Shivdasani (1993) and Brickley, Coles, and Linck (1999) find evidence consistent with a positive relation between the number of directorships and director quality. If executives signal their 1

quality by accepting outside directorships (Fama and Jensen, 1993), we would expect a positive relation between the number of directorships and firm value. We collect 216 announcements of new director appointments in the years 1995-1996 where both the sending firm, i.e., the firm where the executive or Chairman is employed, and the receiving firm, i.e. the firm that the individual is nominated to join, are publicly traded. In this sample, approximately 10% of the events involve individuals with three or more outside directorships, a number that is generally consistent with what Ferris and Jagannathan (2001) report for the year 1995. We find that the average announcement return for the three-day period around the earlier of the press release or proxy filing date is significantly negative for both sender and receiver firms. Rosenstein and Wyatt (1994) also find negative average announcement returns for sender and receiver firms, but Rosenstein and Wyatt (1990) find a positive announcement return for receiving firms, including board nominees from publicly as well as non-publicly traded sender firms. Economically, we show that the mean (median) dollar loss to shareholders of the sender firms at the announcement is $33 ($3.5) million. This result indicates that shareholders of sending firms on average lose a substantial amount of money when one of their executives joins the board of another company. The question now becomes whether the number of board seats the nominee already holds is related to the cross-sectional variation in the announcement return. If outside directorships constrain the time available for executives to manage their own firm, we expect a negative relation between the announcement return for sender firms and the number of directorships held by the executive. According to Smith and Watts (1992), this relation should be even greater in firms with more investment decisions since the marginal product of an executive is higher as a 2

decision maker than as a monitor. Alternatively, based on Fama and Jensen (1983), the signaling hypothesis suggests a positive relation between the announcement return and the number of directorships, because investors update their beliefs about the individual s quality based on the number of directorships the individual holds. On the one hand, we find that shareholders of high-growth sender firms (high capital expenditure to assets ratio or market-to-book ratio) suffer greater losses when their executives sit on more outside boards. Also, the losses are greater in sender firms without independent boards. In addition, we find that individuals at the center of the decision making process, namely the CEOs, are associated with more negative announcement returns if these individuals sit on more outside boards. On the other hand, our tests show that for firms with below-average growth opportunities, there is a positive relation between the number of outside directorships and announcement returns when executives of those firms accept additional outside directorships. This evidence supports the signaling hypothesis if executives of firms with fewer growth opportunities find it more difficult to signal their quality by making investment decisions. The signaling argument should be particularly true if we observe that an executive takes on more than the expected number of directorships. In contrast, a director accepting more than the expected number of directorships could indicate a director that is too busy. We try to distinguish between these two arguments by estimating a two-stage analysis. In the first stage we predict the optimal number of directorships using a framework similar to Booth and Deli (1996). In the second stage, we replace the actual number of directorships held with the difference to the predicted number (excess number of directorships). We find that the excess number of directorships is negatively related to the announcement returns only for executives of high-growth firms. However, there is no longer evidence of a 3

positive relation between the number of directorships and the announcement returns for low-growth firms. Consistent with the findings reported in Baysinger and Butler (1985), Weisbach (1988), Rosenstein and Wyatt (1990), Byrd and Hickman (1992), Brickley, Coles, and Terry (1994), and Cotter, Shivdasani, and Zenner (1997), we find that receiver firms benefit from the announcement of an outside director nomination if their board has less than a majority of independent directors and if the prior receiver firm stock performance is poor. This raises the question of whether a fully diversified shareholder would care that an executive joins the board of another firm given the potential that the shareholder s losses in the sender firm are offset by gains in the receiver firm or firms. Using the market-value-weighted sum of the announcement returns per event, we find that executives of high growth sender firms with more directorships on average cause negative announcement effects across the portfolio of firms in which he or she is involved as a board member. We begin in section 2 with a description of the sample and data. In section 3 we describe the hypotheses and empirical tests and present the results. We conclude in section 4. 2 Sample Selection and Data Description 2.1 Sample Selection We select a sample of new director appointments in the following manner. First, we identify a subset of publicly traded firms listed on Compact Disclosure in both 1995 and 1996 that also have available data on Execucomp and from a database provided by the Investor Responsibility Research Center database (approximately 1,000 firms). Next, we identify individuals that were listed as directors on Compact Disclosure in 1996 but were not listed in 1995. Using Lexis-Nexis, we search each 4

prospective director appointment to determine if the individual was actually added to the Board of Directors of the receiver firm during the prior year. For each observation, we collect both the proxy filing date and the date of the earliest announcement of the nomination or election of the new board member. We investigate each event and collect additional information relating to the receiver firm and the individual director. We eliminate events where the new director is an employee of the receiver firm, and events where the director represents a large ownership position or blockholder, as indicated in the news announcement or proxy statement. We also exclude announcements of new appointments that are related to a takeover, merger, or major restructuring. However, we do not exclude announcements where more than one director appointment is announced on the same day because the main focus of this paper relates to the impact of the announcement on the sender firm, i.e., where the director is otherwise employed. 1 The selection process results in a sample of 471 events where we can identify the main occupation of the director appointed or elected to the board. Of the 471 events, 216 involve individuals that are also employed by another firm that is publicly traded and has information available on Compustat and CRSP. We designate the firm for which the individual is either an executive officer or Chairman of the Board as the sender firm for each director appointment. Almost all of the remaining director appointments are individuals that are either retired, lawyers, consultants, academics or executives of private companies. The main focus of our analysis involves the 216 events where we have data available for the sender firm, the individual and the receiver firm. 1 Our final sample contains 14 events where two individuals, and one event where three individuals join the same receiver firm and the announcements were made on the same day. Excluding those events does not change the findings reported below. 5

2.2 Data Description The distribution of the events by month of the year is shown in Figure 1, with 79 events occurring in 1995 and 137 events occurring in 1996. The event dates reported in Figure 1 reflect the earlier of either the proxy filing date or the announcement date. In our sample, 61% of the events have a distinct announcement date prior to the proxy filing date. 2 We observe the highest concentration of events in March (55 observations), which corresponds to the typical proxy season for firms with a December fiscal year-end. Otherwise, the event dates are distributed evenly throughout the remainder of the year. In panel A of Table 1, we report the individual s position in the sender firm. 52% of the new directors in our sample are CEOs, 43% of the individuals hold the position of Chairman of the Board in the sender firm, and 39% hold other executive officer positions at the sender firm, such as president, CFO, or COO. In 34% of the sender firms, the individual being appointed to the additional board holds the doublefunction of CEO and Chairman in the sender firm. The average age of a nominated director in our sample is 54, and 12% of the new directors are between age 62 and 66 (not shown), which corresponds to the typical retirement age for many CEOs and other executive officers. The characteristics of the sender and receiver firms in our sample are summarized in panel B of Table 1. We find that the average sender firm is significantly larger in terms of market capitalization, however, 41% of the events involve an individual from a smaller firm joining the board of larger firm. Sender firms also have significantly higher stock returns prior to the event, as measured by 2 Shivdasani and Yermack (1999) report that 76% of their director appointment events have distinct announcement dates before the filing date. If we restrict our sample to those events where we have both dates available, we find that 75% of our events have an announcement date before the proxy filing date. 6

the cumulative abnormal return (CAR) over the prior year. The CAR is estimated using a market model with an equally weighted market return over the days 250 to 20, where day 0 is the event day. Rosenstein and Wyatt (1994) find similar results for a sample of sender and receiver firms from 1981 to 1985. 3 Our results are also consistent with the findings of Brickley, Coles, and Linck (1999), who show that retiring CEOs have a higher likelihood of obtaining subsequent outside directorships if their own firm performed well. However, the average return on assets (ROA), measured as the ratio of operating income before depreciation to total assets, is not significantly different between sender and receiver firms. Neither are differences with respect to the level of investment, as measured by the capital expenditures to assets ratio, or their growth opportunities, measured by the market-to-book ratio. The market value is computed as the market value of equity plus the book value of assets minus the book value of common equity. We find that the receiver firms held an average of 7.6 board meetings in the fiscal year prior to the event, and that nearly 60% of the sender firms do not have a majority of independent outside directors, which is consistent with the proportions reported by Shivdasani and Yermack (1999). Kroszner and Strahan (2001b) conclude that connections established by bank executives on the board of non-financial firms and non-financial executives on the board of banks reflect efforts to reduce the costs of conflicts of interest. Booth and Deli (1999) report that the presence of bankers on the board of directors is related to the use of debt. In our sample, we find that 16% (15%) of the receiver (sender) firms are financial firms that operate in the one-digit SIC of 6. In subsequent tests, we will 3 Should these abnormal returns be caused by a misspecified returns model, we would expect to have an upward bias in our event study returns. However, our results are not affected if we use simple netof-market returns instead of market model abnormal returns. 7

attempt to control for the impact of directors from financial institutions by including indicator variables representing financial sender firms. 2.3 Number of Directorships A key variable for our tests is the number of directorships an individual holds prior to the appointment under consideration. We collect this information from the proxy statements of the receiver firm and the news announcements. While the SEC only requires firms to reveal directorships in publicly traded companies 4 some firms volunteer additional biographical information about their directors. 5 Therefore, we calculate the number of directorships using two different methods. In the first one, we include all directorships in companies listed in the proxy statement or the news announcement, but excluding directorships in charitable organizations, foundations and other clearly nonprofit firms. This number includes directorships in the sender firm, but excludes the new directorship. We call this variable forprofit. In panel A of Table 1, we report that the average new appointee already holds 2.24 directorships. We also limit the count to only the number of directorships held in publicly traded firms 6 outside the sender firm. Using this second definition of directorships, which we call otherpub, the average individual only holds 0.97 outside directorships. Approximately 10% of the events involve individuals that hold 3 or more outside directorships in publicly traded firms, however 20% of the individuals hold 3 or more directorships in forprofit companies. We find that 16% of the individuals become 4 Pursuant to Item 401(e)(2) of Regulation S-K, firms are required to Indicate any other directorships held by each director or person nominated or chosen to be a director in any company with a class of securities registered pursuant to section 12 of the Exchange Act. 5 For example, many firms will list non-public and foundation directorships in the biographical information required for each director in the proxy statement. 6 We call a firm publicly traded if it has information listed on CRSP. The match is based on the name of the company listed in the proxy statement or news announcement. 8

directors of a receiver firm without holding any other directorship in a for-profit firm, and 46% do not hold any outside directorships in publicly traded firms. For the remainder of this paper, we focus on the variable otherpub as our measure of other directorships held by the individual because firms are only required to report the other publicly traded directorships held by their directors or nominees. Similar results obtain, however, if we use the variable forprofit in our analysis. 7 Comparing our results to Ferris and Jagannathan (2001), who document the distribution of the number of directorships held by directors of firms listed on Compact Disclosure in 1995, we find a higher proportion of directors with three or more directorships. In their sample, only 4% of the directors hold three or more directorships. This difference is more than likely attributable to our requirement that the sender firm have available data on CRSP, Compustat, Execucomp, and the IRRC database, thus restricting our sample to larger firms. Ferris, Jagannathan, and Pritchard (1999) report that 19% of the directors have 3 or more directorships (including the sender firm) in 1995 for a sample of Forbes 500 firms, which are firms closer in size to the firms in our sample. 3 Shareholders Reaction to Appointments In this section, we discuss alternative hypotheses and testable implications that arise from the hypotheses relating to the expected relation between shareholder value and the number of outside directorships held by an executive or Chairman. We test these hypotheses by examining investors reaction to the announcement that an 7 It is interesting to note that the fact that firms may report more than required did not alter the conclusions of the study. In particular the difference between forprofit and otherpub as a separate variable never significantly affected the CARs, although the results are more significant using the forprofit instead of the otherpub variable. 9

executive joins the board of another company, conditioned on the number of outside directorships of the individual. 3.1 Hypotheses If agency problems allow managers to consume perquisites through the acceptance of additional directorships at the expense of shareholders, then a negative relation should exist between shareholder value and the number of outside directorships. Due to these concerns, the NACD recommends that executives do not hold more than three outside directorships because of the time commitment required to properly monitor management. Blackhurst (2000) estimates that a directorship in a company with a monthly board meeting corresponds to approximately 24 annual workdays, with an additional two days per committee meeting. The NACD even recommends to budget at least four full 40-hour weeks of service for every board on which [the directors] serve. 8 While this argument predicts that any directorship takes away valuable time from the executive s main job, Smith and Watts (1992) assert that the marginal product of an executive is higher as a decision maker than as a monitor. This argument has two implications. First, an executive of a company with high investment requirements or high growth opportunities should have a higher marginal product as a decision maker than an executive of a company with fewer investment activities. Second, the marginal product should be higher for individuals at the center of the decision making process, such as the CEO. The first insight leads to the prediction that the announcement return for high growth companies will be more negative if their executives accept additional outside directorships. The second insight leads to 8 As cited in Byrne, J. and R. Melcher, The best & worst boards, Business Week, November 25, 1996. p 102. Business Week also assigned lower corporate governance rankings for firms where outside employed directors hold more than three board seats. 10

the prediction that a more negative announcement effect will be associated with CEOs relative to other executives. An alternative view is based on the labor market for directors introduced in Fama (1980) and Fama and Jensen (1983). Fama and Jensen (1983) argue that in an imperfect labor market, the good managers want to signal their quality. One way of doing so is to take on outside directorships. If taking on outside directorships signals quality, we would expect the sender and receiver firms to have a more positive reaction to the new appointment the more outside directorships an individual holds. A similar relation could exist based on Fama (1980), where the labor market awards successful managers with better jobs or, in our case, with additional outside directorships. Thus, higher quality individuals should hold more outside directorships. To the extent that investors cannot update their belief about the quality of the individual purely based on past performance, a new nomination can carry positive information. 3.2 Testable Implications We first study the relation between the number of directorships and the sender firm s announcement return (CAR). 9 A negative relation would support the agency cost and busy director hypotheses, a positive relation would support the signaling hypothesis. Since the hypotheses have opposite implications for the relation between the number of directorships and shareholder returns, we refine our tests using the insights from Smith and Watts (1992). We test the hypotheses related to Smith and Watts (1992) and Fama and Jensen (1983) in two ways. The first test is based on the following multivariate regression: CAR = f(otherpub; GrowthD; Otherpub GrowthD; other controls), (1) 9 Please refer to section 3.4 for a more precise definition of CAR. 11

where CAR is the cumulative abnormal return of the sender firm caused by the announcement that an executive or Chairman of a sender firm joins an outside board; Otherpub is the number of outside directorships currently held by the executive; GrowthD reflects the growth opportunities or investment activity levels of a company. As a proxy for the growth opportunities of the sender firm (GrowthD), we use both the ratio of capital expenditures to assets and the market-to-book ratio. In both cases, we separate the firms into groups based on whether they are above or below the median for each ratio. In our sample, we hypothesize that more investment activities or growth opportunities require more attention and time commitment by the executives of the firm. If our two measures are highly correlated with firm growth opportunities, we would expect sender firms to have a more negative CAR the more outside directorships an individual accepts. On the other hand, individuals of firms with low growth opportunities may wish to signal his or her quality by taking on additional outside directorships because they find it more difficult to signal their quality by making investment decisions. This would result in a positive association between the number of directorships and investor reaction for low-growth firms. Alternatively, the market-to-book ratio could already incorporate management quality. In this case, individuals from high market-to-book companies would be higher quality managers that create more value than low-quality managers. Using this rationale, we would expect a more positive relation between the number of directorships and abnormal returns for individuals from high market-to-book companies. The second test is based on the following multivariate regression: CAR = f(otherpub; PositionD; Otherpub PositionD; other controls), (2) 12

where PositionD is a dummy for the individual s position in the sender firm corresponding to the positions listed in Table 1, i.e, CEO, other executives, Chairman and the double function CEO and Chairman. On the one hand, if a CEO or Chairman s time spent at the sender firm is more valuable than the time of other executives, then the position of CEO or Chairman should be associated with a negative relation between the CAR and the number of directorships. On the other hand, if outside directorships signal the quality of the individual, then CEO and Chairman could be associated with a positive relation between the CAR and the number of directorships. Alternatively, it may be more difficult for the investor to assess the quality of a lower ranked executive than of a CEO. Therefore, lower ranked executives may want to signal their availability in the labor market by accepting outside directorships. In this case, the signal would have a more positive impact for executives other than CEOs and Chairmen, and we should see a positive relation between the CAR and the number of directorships held. We also test the agency story by controlling for potential agency problems in the sender firm using the following model: CAR = f(otherpub; AgencyD; Otherpub AgencyD; other controls), (3) where AgencyD is a proxy for the agency conflicts in the sender firm. We use two different proxies for AgencyD, board composition and age. Hermalin and Weisbach (1998) present a model of a bargaining process determining board composition where the proportion of independent directors is negatively related to the power held by the CEO. We use a dummy variable equal to one if less than 50% of the directors on the board are independent outside directors. We predict that executives can more easily consume perquisites in the form of outside directorships in 13

these firms than in firms with better monitoring by the board. Thus, we expect a more negative CAR for sender firms without an independent board unless the market already discounts the value of such firms. Alternatively, a firm with fewer outside directors might want to attract more outside directors. If the market for board members is an in-kind barter market (e.g., Lorsch and MacIver (1989)), then taking on more outside directorships could be positive news for shareholders. The second proxy is age. The relationship between the individual s age and the impact on sender firms could also go both ways. If the acceptance of outside directorships is normal behavior for executives in the typical succession process, individuals nearing retirement from managing day-to-day operations may not be as time constrained as younger individuals and therefore should have a more positive association between CAR and the number of directorships held. On the other hand, the likelihood of directors consuming perquisites at the expense of shareholders can be greater near the end of an executive s career due to decreased career concerns and the horizon problem becoming more severe with age (Fama, 1980; Brickley, Coles, and Linck 1999). In this case, we expect a more negative association between the number of directorships held and CAR for older individuals. 3.3 Control Variables In the previous section we have indicated that our tests include other control variables. In this section, we describe the variables and our motivation to include these variables in our empirical tests. Expertise or firm-specific human capital. We expect a negative relation between receiver firm R&D intensity and the sender firm CAR if the receiver board requires a greater time commitment to monitor the investment activities associated with evaluating new technologies. Alternatively, joining the board of a firm where 14

more R&D is performed could be good news for the sender firm if the sender firm can benefit from the knowledge. Cooperation between firms. To control for potential synergies or complementarities through cooperation between the sender and receiver firms, we include the absolute value for the correlation in stock returns between the sender and receiver firm (e.g., Morck, Shleifer and Vishny, 1990). Financial firms. Rosenstein and Wyatt (1990), Booth and Deli (1999), and Kroszner and Strahan (2001a) provide evidence that board relationships with financial institutions are potentially important because firms may have better access to capital or alternatively because conflicts may arise between creditors and shareholders. We control for financial firms by including dummy variables equal to one if the sender or receiver firm is operating in the one-digit SIC of 6. Prestige. We use a dummy variable equal to one if the receiver firm s equity market capitalization is higher than the sender firm s equity market capitalization. We expect a positive relation between this variable and the CAR if an outside directorship in a larger firm is interpreted as a signal of the individual s quality, however it is also possible that the CAR will be negative if accepting a directorship in a larger firm represents perquisite consumption by the individual (Mace, 1986) or greater time commitment. Sender firm size. We use the logarithm of sales as our proxy for sender firm size. Prior performance. We measure prior performance by the CAR over the period of days [-250,-20] for the sender and the receiver firm. Consistent with the labor market for directors outlined in Fama (1980), a number of researchers including Kaplan and Reishus (1990), Booth and Deli (1996) and Brickley, Coles, and Linck (1999) provide evidence that managers of sender firms with better 15

performance have more opportunities to serve on outside boards. In addition, it is important to control for the receiver firm s prior performance because poor prior performance may be negatively correlated with the sender announcement return if the market perceives that the appointment will require a significant commitment of time in order to fix the past problems. Agency problems in receiver firm. Shivdasani and Yermack (1999) find that when CEOs of receiving firms are directly involved in the director nomination process, stock price reactions for independent director appointments in receiver firms are significantly lower. We control for the potential agency problems associated with receiver firm CEOs that appear to have greater power in determining the board structure by including a dummy variable equal to one if the CEO sits on the nominating committee or if there is no nominating committee. Time commitment in receiver firm. As an additional control for the market perception of the time commitment required to serve on the receiver board, we include a dummy variable equal to one if the board of the receiver firm met more than 7 times (sample median) in the prior year. Vafeas (1999) reports a median number of meetings of 7 and an inverse relation between firm value and the number of meetings. 3.4 Empirical Method To study the effect of an appointment on the sender and receiver firms, we perform an event study using the earlier of the news announcement date or the proxy filing date. We compute cumulative abnormal returns (CAR) over the three-day window [-1,1], following the standard event-study methodology of Dodd and Warner (1983) and Brown and Warner (1985). The market model parameters are estimated using the days [-250,-20] before the event date. We exclude five receiver firms from 16

the event study due to either missing returns data in the window [-1,1] or less then 100 observations in the parameter estimation window. We use an equally weighted average return including distributions as our benchmark. The significance levels are computed using the cross-sectional standard deviation of the CARs (p 1 -value). We also report the significance levels based on the standardized prediction error (p 2 - value), using the time-series of abnormal returns prior to the event, separately, if they differ from the cross-sectional results. Medians are tested using the Wilcoxon ranksum test and proportions are tested using a binomial test. 3.5 Univariate Analysis The cumulative abnormal returns (CAR) over the announcement window of [-1,1] are significantly negative for both sender and receiver firms as shown in panels B and C of Table 2. The average announcement CAR for the sender firm is 0.41% with a p 1 -value of 0.07 and a p 2 -value of 0.05. The median CAR is 0.52% and is statistically significant at the 1% level. In addition, only 41.7% of the events have a positive CAR and this fraction is significantly different from 50% with a p-value of 0.02. It is also interesting to note that the dollar loss to the average sender firm is $33.9 million. The median loss is $3.51 million. Under the assumption that individuals trade off the personal benefits of accepting a directorship with the costs, the data is consistent with the existence of agency problems because shareholders do not seem to benefit from the individual s action (Jensen and Meckling, 1976). For the receiver firm, we find an average drop in shareholder value of 0.46% or $55.7 million. At the median, the losses are 0.49% and $5.08 million. Again, the binomial tests support the finding that the typical receiving firm loses value. It is interesting to note that our event study results for the receiver firms differ from the results reported in Rosenstein and Wyatt (1990), but are consistent with the results in 17

Rosenstein and Wyatt (1994) who find an insignificant negative CAR for firms that appoint a new director who is also CEO of another publicly traded firm. Our sample is closer to the latter in that we require the new director appointees to be affiliated with a publicly traded sender firm. If we relax this requirement, we find an insignificant CAR of 0.21% for our entire sample of 471 events (not shown). 10 We next turn to our main question of whether the sender firm CARs differ once we condition on the number of directorships an individual already holds. Panel A of Table 2 shows that the announcement of the first outside directorship in a publicly traded firm is already bad news for the sending firm shareholders. We find an average CAR of 0.89%, which is significantly negative at the 1% level. The average CARs for events where the new appointee already holds other outside directorships in publicly traded firms are mostly negative but insignificant. However, we do not find a monotonic decrease in the CARs. The conclusions are similar if we use the variable forprofit that includes the directorship in the sender firm as well as directorships in private firms. The data suggest that sending firm shareholders on average begin losing value with the first outside directorship appointment of one of their executives. These univariate statistics are consistent with the argument that outside directorships represent perquisite consumption, however our hypotheses require a multivariate analysis in order to determine whether our inability to find a monotonic decrease in the CAR is due to the offsetting signaling effects. 10 An additional difference to Rosenstein and Wyatt (1990, 1994) is that we use the earlier of the proxy filing and news announcement date and do not exclude all events with contaminating information related to the receiver firm. For example, in news announcements that follow the proxy filing date, we find several instances where dividend and other information about the receiver firm potentially contaminate the event information. Concentrating on the events where the announcement date is before the filing date we find an average CAR for the receiver firms of 0.42%, which is significant at the 10% level (127 events). 18

3.6 Multivariate Analysis In this section, we discuss the results of the multivariate regressions examining the cross-sectional distribution of sender firm CARs. For purposes of Tables 3 and 4, we use the variable otherpub (other publicly-traded directorships) as our measure for the number of outside directorships held by the individual. However, the conclusions drawn from the regressions using the forprofit variable are similar and omitted for brevity. Table 3 presents the multivariate regression results. We report coefficients of ordinary least squares regressions with p-values based on White-adjusted standard errors. The dependent variable is the sender firm s cumulative abnormal return (CAR) over the event window [-1,1]. Confirming the univariate analysis, we do not find a significant relationship between otherpub and CAR in model 1. This is consistent with two interpretations. First, on average, costs and benefits to shareholders might both increase in lockstep with the number of directorships held by the individual. Second, some shareholders lose while others win. In order to distinguish between the two interpretations, we proceed below by testing equations 1 to 3. 3.6.1 Investment levels and opportunities In models 2-4 of Table 3, we test equation 1 by employing the sender firm s capital expenditures to assets ratio as a proxy for investment opportunities ( GrowthD ), and in models 5 and 6 we use the market-to-book ratio as the proxy for investment opportunities. In each case, the firms are split into above the median and below the median groups. In all the specifications, we find a significantly positive coefficient on otherpub and a significantly negative coefficient on the interaction variable between otherpub and GrowthD. Furthermore the sum of the two coefficients 19

is at least marginally negative and significant in all cases. These findings are consistent with both Fama and Jensen (1983) and Smith and Watts (1992). The CAR in low growth firms is positively related to the number of directorships held, which is consistent with investors making a positive update of the individual s quality. For firms with high growth opportunities, however, the announcement returns are negatively related to the number of directorships. This is consistent with the hypothesis that an individual s marginal product is higher as a decision maker in a high growth company. Since we only have relatively few observations with a high number of outside directorships, we test the robustness of our findings by replacing otherpub with a dummy variable equal to one if the individual holds three or more outside directorships (many dirships). 11 The results are reported in models 4 and 6. For high growth firms we find robust and economically significant results. For example, in model 6 we find that the CAR is 2.01% higher for low market to book firms if the individual holds three or more outside directorships. For sender firms with high growth opportunities, the coefficient on the interaction variable is 3.92%. Thus, high growth firms experience a drop in value of almost 2% if an individual holds three or more outside directorships at the announcement date. In addition, the coefficient on many dirships in model 4 is not statistically significant, indicating that the signaling value might not be as robust as shown before. 12 We further examine these findings by creating a series of dummy (OtherpubdumX) and interaction variables (OtherpubdumX GrowthD), where X 11 While this definition of the variable does not directly correspond to the recommendation of the NACD, it allows us to have a sufficient number of observations in both groups. 12 Core et al. (1999) conclude that busy directors, i.e., those with more than three directorships, monitor CEOs less effectively because CEOs with busy boards are paid excess compensation. 20

represents the number of outside directorships, otherpub. For reasons of sample size, we create a variable Otherpubdum4-6, which is one if otherpub is 4, 5 or 6. 13 Table 4 displays the results for the two definitions of GrowthD. Surprisingly, none of the dummy variables is significant while all the interaction variables are negative and significant. These findings support the previous interpretation that high growth firms incur a significant drop in shareholder value that is related to the number of directorships. Given the limited number of observations, we refrain from making strong conclusions about the validity of the signaling hypothesis, however the data clearly supports the notion that shareholder value is destroyed in high growth firms. With respect to the other control variables, the regressions in Table 3 and 4 provide similar insights. We find a significantly negative relation between the receiver firm s R&D intensity and the sender firm s CAR. One interpretation is that directorships in R&D intense companies require more attention of the director thus reducing the time spent at the sender firm. We further find that firms with higher stock returns over the past year (days 250 to 20 relative to the event) experience a higher drop in share price. If better performance is positively correlated with the manager s ability and time devoted to the job, the coefficient is consistent with the interpretation that this nomination distracts the manager from his or her primary task at the employer firm. With the exception of the R&D to sales ratio of the receiver firm, other control variables relating to the receiver firm are insignificant. 3.6.2 Individual s position in sender firm Additional evidence supporting the arguments of Smith and Watts (1992) comes from our test of equation 2, where we study the announcement return conditioning on the individual s position in the sender firm. The results are reported in Table 5. We 13 Note that we suppress the intercept in order to estimate this regression and get coefficients that resemble averages. 21

find that shareholders in sender firms lose more value if the CEOs hold more outside directorships relative to other executives, such as COOs, CFO, and divisional executives. The negative relation supports the notion that CEOs are too busy and too valuable as decision makers to take on more outside directorships. Our findings cannot reject an interpretation that investors update their belief about the quality of other executives more than for CEOs based upon the number of directorships held. However, full support for the signaling hypothesis would have required a positive coefficient on otherpub for other executives. The robust results so far suggest that individuals of high growth companies who take on more outside directorships reduce sender firm value. The loss to sender firm shareholders is even greater if the individual accepting the outside directorship is at the center of the decision making process and holds more outside directorships at the time they join yet another board. 3.6.3 Agency problems The above findings would seem to be consistent with an agency problem argument. The individual s cost of effort by taking on one additional outside directorship must be lower than the benefits in terms of reputation and remuneration. If this is true, we expect that such behavior should be more prevalent in firms with greater agency problems. In Table 3, models 7 and 8, we show results for our equation 3. The proxy for agency conflicts in the sender firm is a dummy equal to one if the sender board consists of less than a majority of independent directors, assuming that independent directors monitor management more effectively (e.g., Weisbach, 1988; Brickley, Coles and Terry, 1994). We find that the coefficient on otherpub is insignificantly positive but find a highly significant negative coefficient on the interaction variable. The data is 22

consistent with an interpretation that agency conflicts play a major role in the market s perception of an individual s accumulation of outside directorships. In Table 5, we also measure potential agency problems by the individual s age. As explained above, individuals closer to retirement might have different career concerns which affect the level of perquisite consumption. Our specification uses a dummy variable equal to one for individuals age 62 and older and a dummy variable for individuals age 61 and younger. We omit the intercept from this regression in order to estimate the coefficients on both dummies individually. Neither the coefficients on the age dummies nor the interaction variables are significant. However, the interaction variables do significantly differ from one another with a more negative relation between CAR and otherpub for older individuals. This supports the interpretation that age reflects horizon problems while it rejects the notion that older individuals have more time because they are in the process of retiring. 3.7 Two stage Regression Given the evidence against an interpretation that the number of directorships is a signal of quality, the question arises whether we have appropriately accounted for the market s expectations about the number of directorships an individual will accept (or that the director will be offered). In this section we use a two-stage regression approach in order to control for market expectations related to the announcement of a director appointment. The first stage regression is a Poisson maximum likelihood model with independent variables closely related to those used by Booth and Deli (1996) and Shivdasani and Yermack (1999). 14 The sample for the first stage 14 We also report a Tobit model for comparison with Booth and Deli (1996). 23

regression is based on the sample of firms covered in the IRRC database and Compact Disclosure. Details of the sample selection are reported in the Appendix. 3.7.1 Empirical Results of the First-Stage Regression The results of the first-stage Poisson regressions are reported in Table 6. We find that individuals hold more directorships if they are older and if they hold the position of Chairman or the double position of CEO and Chairman. The number of additional directorships is also positively related to the size of the sender firm, the number of independent directors on the sender firm s board, and the market-to-book ratio of the sender firm. Non-CEO executives and individuals from financial institutions hold significantly fewer directorships. Most of the results are consistent with the results reported by Booth and Deli (1996) for a sample of firms in 1989-1990. However, our regression results show a positive association between the market-to-book ratio and the number of outside directorships held. 15 This is consistent with Ferris and Jagannathan (2001) who study a sample using 1995 data. Since the market-to-book ratio can proxy for growth opportunities as well as the market s assessment of the managerial quality, we estimate a second regression and include the ratio of R&D to sales 16 to represent growth opportunities and ROA as a measure of performance. 17 We find an insignificant negative coefficient on the R&D to sales variable and a significant positive coefficient on ROA. Since ROA was insignificant in the market-to-book regression, it is possible that the positive coefficient on the market-to-book ratio in model 1 reflects performance rather than growth opportunities. 15 We also use the Tobit regression specification and find a positive coefficient on the market-to-book ratio, as shown in the third regression in Table 6. 16 For firms with a missing value of R&D, we assume that they have the industry average R&D, measured at the 2-digit SIC level. 17 We do not use a long-term performance measure as in Brickley et al. (1999) because we do not know for how long the individual has already been with the current employer. 24

In untabulated regressions we also include ownership dummies as reported by IRRC. We find a non-linear relationship between the number of outside directorships and ownership in the sender firm. Only individuals with ownership levels between 5-10% and those with more than 20% hold significantly fewer directorships. However, for many non-ceo executives this variable was missing, thus potentially introducing bias into the sample. 3.7.2 Empirical Results of the Second-Stage Regression In Table 7 we report the results of the second-stage regression where the dependent variable is the CAR of the sender firm. The results discussed below are based on the first-stage Poisson model using the market-to-book specification. However, similar results obtain if we use the specification with R&D or if we use the Tobit specification. The expected number of directorships held by an individual is called E(Otherpub) and is computed as follows: E(Otherpub) = exp( ˆ' β x), where x is the vector of variables used in the first-stage regression and βˆ is the coefficient-vector of the Poisson maximum likelihood regression. 18 We continue to find a significant negative relation between the number of directorships and the CAR for sender firms with greater growth opportunities, even after correcting for the expected number of directorships. However, compared to the results in Table 3, firms with lower investment opportunities no longer benefit when their executives are appointed to additional boards. Assuming that the results of our first stage regression positively relate with the market s expectation about how many additional directorships an individual will accept (or be offered), the second stage results suggest no benefit to sender firm shareholders for individuals wanting to signal 18 Qualitatively similar results obtain if we round E(Otherpub) to the closest integer value. 25