The Market for Non-executives: Takeover Performance and the Subsequent Holding of Directorships

Transcription

1 The Market for Non-executives: Takeover Performance and the Subsequent Holding of Directorships Svetlana Mira Cardiff Business School Marc Goergen Cardiff Business School and European Corporate Governance Institute (ECGI) Noel O Sullivan Loughborough University ABSTRACT This study examines whether the holding of directorships by non-executives, subsequent to their company acquiring another company, is sensitive to the acquirer s post-acquisition performance for 488 non-executives involved in 147 acquisitions completed between 1994 and We relate subsequent changes in directorships held by these non-executives to a range of performance measures of the acquiring firms. We find that the holding of directorships by nonexecutives is not sensitive to the post-acquisition performance of acquirers. This suggests that the market for non-executives does not punish non-executives on the basis of the quality of their previous decisions with fewer board positions. Keywords: Market for non-executive directors, corporate governance, mergers and acquisitions JEL code: G34 1

2 Over the past 20 years the board of directors has been emphasised as one of the most important instruments of corporate governance, with regulators continually seeking to strengthen the quality of board monitoring. Central to these reforms has been the monitoring potential of non-executive directors 1 (Cadbury, 1992; Combined Codes, ; UK Corporate Governance Code, 2010). As a result of these initiatives, a majority of board positions in large UK companies are now frequently held by non-executive directors who also almost exclusively make up the composition of key board sub-committees. Structurally, therefore, non-executive monitoring has been placed at the heart of corporate decision-making with the expectation that this will ensure management actions are continually being undertaken in the interests of shareholders. Parallel to these regulatory developments, researchers have sought to investigate the governance role of non-executive directors, in particular seeking to ascertain whether a greater non-executive presence on boards is associated with improved shareholder wealth. However, the available evidence is mixed with few studies reporting a positive relationship between performance and the proportion of non-executives directors (Bhagat and Black, 2002; Hermalin and Weisbach, 2003; Adams, Hermalin and Weisbach, 2010). 2 The apparent absence of a direct link between non-executive representation and firm performance has encouraged researchers to pursue other avenues in seeking to understand the value of non-executives on boards. One such initiative has been research seeking to understand 1 The terms non-executive director and outside director are used interchangeably in this paper with both terms meant to signify directors who are not full-time employees of the firm. 2 Bhagat and Black (2002) and Hermalin and Weisbach (2003) are reviews of prior research examining the relationship between outside directors and firm performance while Adams et al. (2010) is a broader review of the governance role of boards. 2

3 the operation of the market for non-executives. This has its roots in much earlier work by Fama (1980) and Fama and Jensen (1983) who argued that the labor market for directors should serve as an incentive mechanism for directors to pursue shareholder wealth continuously in their management and monitoring activities. A number of US studies have sought to test this thesis and have produced results broadly consistent with the notion that the market for directors does seek to differentiate on the basis of a non-executive s prior performance. For example, Gilson (1990) finds that non-executives who resign from the boards of financially distressed firms subsequently hold fewer non-executive directorships at other firms; Farrell and Whidbee (2000) find that directors who remove underperforming CEOs are subsequently rewarded with a greater likelihood of an additional directorship; Ferris et al. (2003) find that the past performance of firms for which an individual serves as a director correlates with the number of directorships subsequently held by that individual; Yermack (2004) reports a positive relationship between the performance of firms and the likelihood of outside directors serving during this period obtaining more outside directorships in subsequent periods while Fairchild and Li (2005) find that directors on the boards of above-average performing takeover targets hold more board seats afterwards. This study contributes to this line of enquiry by investigating whether the market for directors differentiates between non-executive directors on the basis of their monitoring ability. Specifically, the study investigates whether the holding of directorships by non-executives subsequent to their company taking over another company is sensitive to their company s postacquisition performance. Focusing on the holding of directorships in firms other than the acquiring firm subsequent to takeover is potentially useful for two reasons. First, takeovers are 3

4 one of the most important strategic decisions made by boards and consequently represent a useful environment in which to ascertain the quality of non-executive monitoring and decisionmaking. Second, existing research highlights significant variation in the performance of acquiring companies, with a large proportion experiencing weak post-bid performance (see e.g. Agrawal and Jaffe, 2000; Tuch and O Sullivan, 2007). Therefore, from a shareholder perspective, it is possible to assess the wealth created (or destroyed) by an acquisition. Consequently, the impact of acquisitions on shareholder wealth represents a discrete and clearly identifiable opportunity to investigate the link between the quality of a non-executive s monitoring and the consequent demand for his/ her services by other companies. In summary, our study undertakes a direct test of whether the quality of non-executive decision-making surrounding a specific corporate event, with likely substantial impact on the firm and therefore requiring particularly high levels of non-executive monitoring, has an impact on the subsequent board careers of the non-executives involved. The implicit assumption behind our main research question is that such a market for nonexecutives does exist. As we shall discuss in the next section, it is far from clear that such a market exists. Hence, we also develop a competing hypothesis which claims that there is no such market and hence the career prospects of non-executives are determined by factors other than their past track record as monitors. In order to address our main research question we identify all successful takeovers in the UK between 1994 and 2003 as well as identifying the individuals holding non-executive positions on the acquirers boards on completion of the bid. We track the number of directorships held by 4

5 each of these individual non-executives for up to five years after the bid. We then measure the market, operating and accounting performance as well as dividend changes of the acquiring firms over the five years following the bid and investigate whether the subsequent holding of directorships is sensitive to post-bid performance. Our empirical analysis reveals that the holding of directorships by non-executives is not sensitive to the post-acquisition performance of acquirers. We find that non-executives on the boards of the worst performing acquirers are just as likely to hold non-executive directorships in other listed companies as their counterparts in the best performing acquirers, up to five years after completion of the acquisition. Furthermore, we find no consistent evidence that the subsequent performance of the acquirer has an impact on the number of non-executive directorships held by the non-executives involved in the acquisition decision. These findings are especially interesting as they suggest that the market for nonexecutives does not punish (nor reward) non-executives on the basis of the quality of their previous decisions with fewer (or more) board positions. Our paper is structured as follows. The next section reviews existing research on the market for non-executives and develops our three hypotheses. Section II describes the sample, variables and research methodology. Our empirical analysis is presented in Section III while the following section reviews a number of robustness checks. Finally, conclusions are discussed in Section V. I. The Market For Non-Executive Directors Underlying the arguments supporting the monitoring potential of non-executives is a belief in the operation of a market for directors (Fama, 1980; Fama and Jensen, 1983). This market should serve as an incentive mechanism for directors to pursue shareholder wealth in their management 5

6 and monitoring activities. A key question addressed in existing research is how prior performance of their firms influences the subsequent holding of board seats by directors. Essentially, this strand of enquiry seeks to ascertain whether there is a reputation effect in the market for non-executives whereby those non-executives perceived as being good (bad) monitors are rewarded (penalized) with additional (fewer) board appointments. Some of the earliest insights into the operation of a market for non-executives have emerged from studies focusing on the link between firm performance and the holding of outside directorships by CEOs. Brickley et al. (1999) show that the probability of retired CEOs holding non-executive appointments in other firms is positively related to their performance as CEOs while Ferris et al. (2003) find that the prior performance of a CEO s firm has a positive influence on the number of outside directorships the CEO holds. Similarly, Fich (2005) finds that CEOs are more likely to receive outside directorships when their own firms are performing well. She also reports that the cumulative abnormal returns around the announcement of the appointment of a CEO as a nonexecutive director are positively related to the financial performance of the CEO s own firm. Finally, Kaplan and Reishus (1990) provide evidence that executives in firms that reduce dividends hold fewer non-executive positions three years afterwards. This evidence is broadly consistent with a market for directors where past performance determines the demand for executives to serve as non-executives elsewhere. In terms of non-executive directors, the existing evidence suggests that the quality of their monitoring influences the likelihood and number of other board seats they hold. Gilson (1990) finds that non-executives who resign from the boards of financially distressed firms subsequently 6

7 hold fewer board positions at other firms. Farrell and Whidbee (2000) find that directors who remove underperforming CEOs are rewarded with additional directorships in other firms. Yermack (2004) reports a positive link between the performance of companies and the number of outside directorships subsequently held by directors. Similarly, Srinivasan (2005) finds that non-executives of firms that restate earnings hold significantly fewer other directorships three years after the restatements. These reputational penalties are especially severe for non-executives associated with downward restatements. Overall, the evidence is consistent with non-executives being subject to an active market for their services with additional appointments being significantly influenced by their past performance as non-executives elsewhere. As highlighted by Fairchild and Li (2005), separating out non-executive performance from that of management is difficult because management is responsible for the day-to-day operation of the company whereas non-executives are appointed to supervise and oversee management s performance. Consequently, the function of non-executives and the performance of management can be entangled in the firm s performance and there are obvious difficulties in disentangling the two. In order to overcome this difficulty researchers have ascertained the impact of nonexecutive behavior in the context of takeovers on their subsequent holding of directorships. For example, Harford (2003) reports that non-executives in poorly-performing firms that reject takeover bids are subsequently punished by holding fewer board seats, but those facilitating such takeovers are not similarly penalised. Fairchild and Li (2005) find that directors on the boards of above-average performing takeover targets hold more board seats subsequent to the takeover. Our study also uses takeovers to provide insights into the non-executive labor market. 7

8 However, unlike Harford (2003) and Fairchild and Li (2005), we focus on the performance of non-executives in acquiring firms. Focusing on acquirers helps to study the operation of the market for non-executives for at least three reasons. First, the acquisition of another firm represents a discrete event making fairly robust before and after performance comparisons possible. Second, acquisitions are one of the most important strategic decisions made by boards and hence represent a useful context in which to ascertain the quality of non-executive monitoring. Third, a vast body of research has investigated the post-acquisition performance of acquirers with mixed findings regarding the gains to shareholders from such transactions (e.g. Agrawal and Jaffe, 2000; Tuch and O Sullivan, 2007). Hence, takeovers are an ideal context in which to ascertain whether non-executives involved in acquisitions are rewarded or punished in light of the impact of the transaction on shareholder wealth. By undertaking a comparison of the number of directorships held by non-executives involved in an acquisition decision in the years after the acquisition is completed, and relating this to the subsequent performance of the acquirer over a similar time period, we can ascertain how such non-executives fare in the non-executive labor market. Therefore, we propose the following two hypotheses: HYPOTHESIS 1: There is a positive link between the likelihood of a non-executive remaining in the market for non-executives and the post-bid performance of the acquiring firm. HYPOTHESIS 2: There is a positive link between the number of non-executive directorships held by a non-executive after a takeover and the post-bid performance of the acquiring firm. However, a case can be made that there is no labor market for non-executives, especially as it 8

9 is not clear who the buyers in the market might be. The answer to this question is not straightforward. Some of the earliest papers on the theory of the firm that mention the existence of a labor market for directors, in particular Alchian (1969) and Alchian and Demsetz (1972), highlight the important role of shareholders in disciplining badly performing directors. While these papers acknowledge the existence of a labor market for directors as well as a market for corporate control, the role of these markets is limited to helping shareholders in their monitoring and disciplining role rather than performing these roles themselves. Jensen and Meckling (1976) doubt that shareholders will ever fulfill these roles given the dispersion of corporate ownership and the resulting separation between ownership and control. Similarly, Manne (1965) disputes that the shareholders are those who buy and fire in the market for directors. He argues for the importance of the market for corporate control in disciplining directors. The buyers in the market for corporate control are expert investors who are good at identifying firms with weak corporate governance and bringing about management and strategic changes. Given that it is difficult to identify the buyers in the labor market for directors the existence of this market can be questioned. If there is no such market, then it is unlikely that nonexecutives career prospects are influenced by their past performance as monitors. These arguments allow us to propose the following hypothesis: HYPOTHESIS 3: There is no labor market for non-executives. II. Sample Selection and Methodology A. Sample Selection Our dataset has three dimensions. We have bidders (dimension 1) that carry out takeovers 9

10 (dimension 2) that have been approved by their non-executive directors (dimension 3). While we focus on the latter, the former two dimensions still influence our research design. Our reference point in time is year 0 which is the year the takeover was completed. We obtain the list of UK takeovers completed by non-financial firms between 1 January 1994 and 31 December 2003 from Thomson One Banker. There are a total of 177 takeovers carried out by 154 bidders during that period. As we require data on firm performance for up to five years after the takeover, the period of study ends in Data related to individual non-executive directors is collected from various editions of the Corporate Register, which is published quarterly. Given the study s focus on the bidder s performance subsequent to the takeover decision and its impact on the nonexecutive s career, we expect that any consequences (or rewards) arising from the subsequent quality of such takeovers are dealt out to those non-executives in post when the takeover is completed. Therefore, our sample comprises those non-executives on the board of the bidder when the takeover is completed. 3 We then track the number of non-executive directorships held by each of these individual non-executives for up to five years after completion of the takeover. We use the issue of the Corporate Registers in the same quarter that the takeover is completed as well as the Corporate Register for each of the five subsequent years. The Corporate Registers enable us to trace non-executive data for 140 bidders engaged in 160 takeovers from the initial sample of 154 bidders and 177 takeovers. We have identified 535 non-executives on the board of these 140 bidders in the quarter that the takeover is completed. Six bidders are listed on the Alternative Investment Market (AIM). Given the relatively small 3 The decision to acquire another firm might have been taken up to a year before the takeover completion date. Hence, as a robustness check we focus on the sample of non-executives on the board of the bidder in the year before the completion of the takeover. As discussed in Section IV, similar results are obtained for this alternative sample. 10

11 number of these firms in the sample and the less stringent listing requirements for AIM, we exclude these firms from the sample. A further six takeovers were dropped because three bidders were involved in two takeovers in the same year. 4 Further, 11 of the non-executives participated in two takeovers carried out by separate bidders in the same year. Eight out of the 11 nonexecutives participated in only these two concurrent takeovers. Given the difficulty in measuring the annual performance of non-executives involved in more than one takeover in the same year, we drop these observations from the sample. Further, 13 directors switch from a non-executive position sometime after year 0 to an executive position and are therefore excluded from the sample. Finally, two non-executives with tenure greater than 30 years were also dropped from the sample as such exceptionally long tenure probably made the non-executives concerned immune from any non-executive labor market censure arising from acquisition activity. The final sample, therefore, comprises 488 non-executive directors sitting on the board of 131 bidders involved in 147 successful takeovers. Given that some non-executives participated in more than one takeover completed in different years, the total number of observations in the final sample is 567 takeovers for the 488 non-executives. Performance data was sourced from Datastream whereas the remaining data used in the paper was collected from the Corporate Registers. Methodology We test the validity of Hypothesis 1, which takes the extreme position that non-executives associated with bad takeover decisions will no longer hold any non-executive directorships five 4 We do not exclude repeat takeovers by the same bidder as long these repeat takeovers occur in different years. Out of the remaining 132 bidders, 14 were involved in two takeovers and another bidder was involved in three takeovers. 11

12 years following the completion of the acquisition, as follows. We focus on year 5 as the typical tenure in the UK for board members is three years and a five-year period includes at least one incidence of reappointment of the non-executive and/ or the chance of a new appointment to another company s board. Focusing on year 5 also makes sense as the extant literature suggests that taking a longer post-acquisition perspective provides a better assessment of both the accounting and stock performance of acquirers (Tuch and O Sullivan, 2007). We proceed by estimating logit regressions predicting the likelihood that a given non-executive director no longer holds any non-executive directorships in year 5 (the dependent variable equals zero) versus the likelihood of the non-executive still holding directorships in year 5 (the dependent variable equals one). Hypothesis 2 focuses on changes in the number of non-executive directorships held subsequent to the acquisition. Here we explain the change in the number of non-executive directorships from year 0 to year 5. We expect that non-executive directors of poorly performing bidders experience a drop in the number of their non-executive directorships whereas those of all other bidders see no change or an increase in the number of non-executive directorships they hold. We run OLS regressions which estimate the change in the number of non-executive directorships from year 0 to year 5. A shortcoming of this approach is that it ignores the busyness of the non-executive director at the start of the period. Non-executives who hold a lot of directorships in year 0 may not want to take on additional directorships even though there is further demand from the labor market for their services given the good acquisition decision(s) they are associated with. Alternatively, busyness may undermine the non-executive s monitoring 12

13 capacity (Fich and Shivdasani, 2006) and hence the demand for his/ her services. What both arguments have in common is the persistence in the number of non-executive directorships, once a certain threshold has been reached. Hence, there is a need for a second approach to explain changes in the number of non-executive directorships (Hypothesis 2), which adjusts for busyness. This approach consists of estimating the total number of non-executive directorships via the following dynamic model: Non executive directorships it = α + βnon executive directorships it 1 + γcontrol variables it + δ i + ε it (1) where Control variables is a vector of variables including bidder and non-executive characteristics, δ i is unobserved non-executive specific fixed effects and ε it is an error term. The main advantage of a dynamic model is that it is better at adjusting for complex behavioral patterns than is a single cross-section or time series (see Hsiao, 2007). In particular, it adjusts for situations where the current status of an individual, in our case the current number of nonexecutive directorships held, is a perfect predictor of the individual s future status. The extant literature suggests that there is a busyness effect in addition to the reputation effect and that each of these two effects may have an impact on the number of directorships held by an individual. According to the reputation effect, some non-executives have multiple directorships because they are good directors and the holding of additional board seats does not impact their monitoring performance (Harris and Shimizu, 2004). In contrast, the busyness hypothesis suggests that too many directorships reduce the monitoring effectiveness of the director concerned (Fich and Shivdasani, 2005). The empirical literature suggests that the two effects coexist. At first good 13

14 non-executives may earn further directorships given their reputation. Once they reach a certain number of directorships, they may decline further directorships to protect their reputational capital and/ or the labor market may refrain from offering them further directorships as concerns about the directors future performance may mount. However, this threshold likely varies across individuals reflecting both observable characteristics, such as age, and unobservable characteristics, such as family life. Including the lag of non-executive directorships on the right hand side adjusts for these complex dynamics. It also adjusts for heterogeneity in these dynamics across individuals, which could not be achieved via a cross-section (Hsiao, 2007). Nevertheless, the dynamic model also raises methodological issues as OLS likely results in an inconsistent estimate of β, the coefficient on the lagged dependent variable, given the correlation of the non-executive specific fixed effects δ i with the error term ε it (see e.g. Bond 2002). As a result, the estimated coefficient on β will be upward biased. A way of dealing with these unobserved fixed effects is to take the time mean from each variable, i.e. run a fixed-effects regression. Hence, equation (1) becomes: Non executive directorships i. = α + βnon executive directorships i,. 1 + γcontrol variables i. + ε i. (2) As δ i is time invariant, this process results in differencing away the unobservable nonexecutive characteristics δ i. However, if the number of time periods is small, which is the case in our study (T= 6), the transformed lagged dependent variable will be highly correlated with the transformed error term. As a result, the fixed-effects estimator of β is also inconsistent and 14

15 biased. However, the bias will now be downward (Bond 2002). 5 While the coefficient on the lagged dependent variable is not of direct interest, the biased OLS and fixed-effects coefficient estimates come hand-in-hand with a reduced consistency of the coefficient estimates on the other variables, including those for the performance measure (Bond, 2002). These estimates are clearly of interest to us. The way forward is to use the Generalised Method in Moments as a System (GMMsys), the estimation technique developed by Blundell and Bond (1998). 6 This estimation technique is more efficient than Anderson and Hsiao s (1981, 1982) first-differenced 2-stage least squares (2SLS) as it uses more instruments and also performs better when T is small compared to Arellano and Bond s (1991) estimation technique. B. Definition of Variables All the dependent variables in this study are derived from the number of non-executive directorships held by the 488 non-executives on the board of the 131 bidders up to five years after the takeover was completed. In the logit regressions, we use a dummy variable that is set to zero if the non-executive no longer holds any non-executive directorships in year 5, and one otherwise. In the OLS regressions, the dependent variable is the change in the number of nonexecutive directorships from year 0 to year 5. In the GMMsys estimation technique we use the number of non-executive directorships held by the non-executive in year t. The number of nonexecutive directorships is the sum of non-executive directorships held by the non-executive. 5 Even if one ignores this issue, most of our bidder characteristics and non-executive characteristics are either time invariant or increase by a value of one each year (e.g. the non-executive s age). Hence, using fixed-effects regressions would difference away the impact of these (virtually) time invariant variables. 6 This technique is based on a system of two sets of equations, consisting of equations in levels and equations in first differences. It follows an instrumental variable approach whereby the first differences of the dependent variable and independent variables are used as instruments in the equations in levels and lagged levels of the dependent variable and independent variables are used as instruments in the equations in first differences. 15

16 All our regressions include non-executive characteristics as well as bidder characteristics as control variables. The former include tenure, age, the non-executive s ownership and the fact whether he/ she chairs the board. Tenure is defined as the number of years the non-executive has served on the board of the bidder at the completion of the takeover. Age is the age of the nonexecutive at the completion date of the takeover. In the logit and OLS regressions both variables are measured in year 0; in the GMMsys regressions these variables are measured in each year t. We also include the proportion of firm equity held by the non-executive in year 0 and year t, respectively. Finally, we include a dummy variable that takes a value of one if the non-executive is the chairman of the bidder s board in year 0 for the logits and OLS regression or year t for the dynamic regressions. It takes the value of zero otherwise. Market value, long-term debt-equity ratio and institutional ownership are bidder characteristics also included in the regressions. We use two different measures of accounting performance, i.e. return on assets and return on equity, dividend increases and dividend cuts and omissions, as well as various measures of operating, i.e. cash flow based performance. We use cumulative abnormal returns (CARs) for various event windows as a measure of market based performance. Return on assets is earnings before interest, taxes and depreciation divided by total assets. Return on equity is earnings after interest and tax divided by common equity. Dividend increases and dividend cuts and omissions are dummy variables set to one for dividend increases and dividend cuts or omissions, respectively, and zero otherwise. Operating cash flow is widely used to measure post-takeover performance. 7 This study uses 7 An alternative measure of performance is earnings. However, Erickson and Wang (1999) show that managers of bidding firms practice earnings manipulation. Therefore this approach has clear drawbacks. 16

17 two measures of operating cash flow performance. 8 The first one is based on cash flow adjusted for changes in working capital, defined as pre-depreciation profit adjusted for changes in working capital. The second one is cash flow measured as simply pre-depreciation profit. Following Powell and Stark (2005), we deflate both cash flow measures by total assets. 9 All the above performance measures, except the dividend dummies, are industry adjusted using the Fama and French industry classification. 10 We also use cumulative abnormal returns (CARs) as a market based performance measure. CARs are calculated based on monthly data for the UK based on the Fama and French fourfactor model. The event month is the month of the takeover announcement, not the completion date. The parameters of the four-factor model are estimated over the period ranging from month - 66 to month -6. The bidder CARs are calculated for the following 6 windows: CAR[1,12], CAR[13,24], CAR[25, 36], CAR[37, 48], CAR[49, 60] and CAR[1,60]. 11 III. Empirical Analysis and Discussion Of The Results A. Descriptive Statistics Table I reports the distribution across time (Panel A) as well as the distribution across industries (Panel B) of the 147 takeovers. Panel A shows that, similar to extant studies on 8 Healy et al. (1992), Linn and Switzer (2001) and Ghosh (2001) use an accruals definition of operating cash flow. However, this measure of performance is likely to be distorted by some accounting practices (Powell and Stark, 2005). Therefore, this measure should be adjusted for changes in working capital to ensure the operating cash flow is free from interest, tax payments and bad debts and is not affected by various accounting policies (Lawson, 1985). 9 Similar to Powell and Stark (2005) we also use total sales and total market value as deflators. Total market value is the sum of the market value of equity, and the book values of debt and preferred stock of the bidder. The results are discussed in the robustness section. 10 The industry classification is based on the 10 industries as defined by Kenneth French and Eugene Fama. Details can be found at 11 CAR[1,60] is the CAR from month 1 (the month immediately after the takeover announcement) to month

18 mergers and acquisitions, there is a peak in takeover activity during 1997 and 2000 (see e.g. Martynova and Renneboog, 2006). 12 Slightly more than 31% of the bidders are from the Other Industries: most of these firms are operators of non-residential buildings, real estate investment trusts, water transportation services, amusement and recreation services, and general contractorsresidential buildings. The second most represented industry is the wholesale and retail industry with just under 25% of the takeovers. INSERT TABLE I ABOUT HERE Table II provides descriptive statistics for the bidders (Panel A) and their non-executives (Panel B) for year 0, the year of the takeover completion. As expected, bidders are large with an average market capitalization of 4.65 billion. However, the median market value is much lower with billion. 13 The average value for total assets is approximately 3 billion and is also highly skewed given the much lower median of billion. The average percentage of equity held by institutional investors is 20.3%, with a median of 15.5%. The industry-adjusted ratio of long-term debt over equity is high with an average of 2.0%, but a median value of only 7.4%. The industry-adjusted return on assets (ROA) and the industry-adjusted return on equity (ROE) have an average (median) of 10.0% (2.3%) and 8% (5.1%), respectively. As for the nonexecutives, their tenure with the firm is on average 5.5 years while the median is just under 4 years. Their average age is about 60 years, with a median age of approximately 61 years. 12 Multiple bids conducted by the same bidder (not reported in the table) as well as multiple bids involving the same non-executive (included in the last column of the table) are also more frequent during that period. 13 Further analysis (not reported in the table) suggests that larger bidders are more likely to be involved in multiple takeovers. The median market capitalization for bidders that participated in two takeovers in a given year is almost twice as large as that for bidders that completed only one takeover a year ( 1.6 billion compared to 843 million). This compares to an even larger market capitalization of 3.4 billion for bidders that participated in three takeovers. 18

19 Typically, non-executives hold little equity in their firm as suggested by the mean and median values (0.2% and 0.002%, respectively although69.6% of non-executives in the sample hold some equity in the bidder. Finally, 13% of non-executives are chairs of the bidder s board. INSERT TABLE II ABOUT HERE Table III reports descriptives for the number of non-executive directorships for the 488 nonexecutives for each of the years 0-5. Panel A reports the descriptive separately for all of the 488 non-executives, those that still hold a directorship in year 5, those that are younger than 60 years and those that are older than 60 years 14 in year 0. The average number of non-executive directorships held by the 488 non-executives in our sample is two in year 0. This number drops to one in year 5. While older non-executives tend to have on average one additional nonexecutive directorship compared to younger non-executives in year 0, amounting to a total of roughly three directorships, this number drops to an average of less than two directorships in year 5. This drop in directorships is significant at the 1% level. In contrast, younger nonexecutives experience little change over time in the number of non-executive directorships they hold. Panel A also shows that most non-executive directorships (roughly 93%) are in firms that are part of the Official List of the London Stock Exchange. When non-executive directorships held in firms listed on the AIM are excluded, the above-mentioned trends remain. For the approximately 7% of the non-executives of the bidder firms that hold at least one non-executive directorship in firms listed on the AIM in year 0 the number of non-executive directorships on the board of these firms significantly increases (at the 10% level or better) rather than decreases 14 The cut-off point of 60 years corresponds to the median age of the 488 non-executive directors. When a cut-off point of 65 years is used, we end up with only 21 non-executives aged 65 years or more which makes inferences drawn from the statistical tests problematic. 19

20 from year 0 to year 5. This is the case for all non-executives, whatever their age. This increase likely reflects the substantial growth in the AIM: there were 1,683 admissions to the AIM from 1995 to During that period, the number of initial public offerings (IPOs) also exceeded that on the Official List (Espenlaub et al., 2012). Panel B of Table III reports, for each of the years 0-5, the number of directors who do not hold any non-executive directorships. 44.9% of the nonexecutives are not in the labor market in year 5. Exits are particularly frequent in years 3-5. While the vast majority of non-executives who exit the labor market are older than 60 years in year 0, the percentage of younger non-executives who exit the labor market is nevertheless nonnegligible, hovering at around 30%. INSERT TABLE III ABOUT HERE Table IV shows the mean and median performance for years 0-5 for the various performance measures. There is little evidence of an improvement in the ROA and ROE post-acquisition. However, there is some evidence of an increase in the percentage of bidders that omit or cut their dividend in year 5 compared to year 0. There is also a decrease in the percentage of bidders that increase their dividend in year 5 (51.1%) compared to year 0 (79.7%). As to the operating, cash flow based measures of performance, there is no evidence that the bidders see their performance improve over the five years following the acquisition. This confirms extant research which suggests that, post-acquisition, bidders experience either no change in performance or significant losses. Finally, the mean (median) CARs are negative for four (three) out of the five years. INSERT TABLE IV ABOUT HERE There are 34 bidders among the 131 bidders in the sample that do not survive until year 5. 20

21 Thirty one of the firms that leave the sample are taken over and the remaining three go bankrupt. 15 Is this attrition driven by poor post-acquisition performance? In other words, is it the case that bidders that made bad acquisition decisions in turn had to be taken over to survive? Table V attempts to answer this question by comparing the 34 bidders that have dropped out of the sample with those that remain in the sample until year 5. Before attempting to answer this question, we investigate whether there are statistically significant differences in firm characteristics and non-executive characteristics between the two types of bidders. Panel A shows that there are indeed significant differences between the two types. The bidders that leave the sample are significantly smaller (both in terms of their market capitalization and their total assets value) and have significantly higher leverage. Panel B suggests that there are also significant differences in terms of the non-executive characteristics between the two types of bidders. The non-executives of the firms that leave the sample have lower tenure, have less stock ownership and are less likely to chair the board. Also, the non-executives on the boards of bidders that do not survive until year 5 have significantly fewer non-executive directorships in firms listed on the Official List, but significantly more non-executive directorships in AIM listed firms. Returning to the above question, Panel C suggests that the bidders that leave the sample have significantly lower performance as evidenced by the lower ROA, the lower ROE (for the median only), the higher percentage of dividend omissions or cuts and the lower percentage of dividend increases. Similar conclusions apply when the two measures based on cash flows are considered. In contrast, there is no significant difference in the CARs for the two types of 15 One can argue that one needs to differentiate between the firms that were taken over and those that went bankrupt. When the three firms that went bankrupt are excluded from Panel C in Table V the results are upheld. 21

22 bidders. Overall, Table V suggests that those firms that do not survive until year 5 have significantly worse post-acquisition performance than those that survive. INSERT TABLE V ABOUT HERE B. Univariate Analysis Table VI presents the results from the univariate analysis which investigates whether the non-executives of bidders performing badly after their acquisition are punished in the labor market by holding fewer seats in year 5. The table is based on all the 488 non-executives, including non-executives from bidders that do not survive until year 5. Except for Panel C, the table distinguishes between the worst performing bidders, the best performing bidders and all other bidders. The 20% of bidders with the lowest performance from year 0 to year 5 are classed as the worst performers, whereas the 20% of bidders with the highest performance from year 0 to year 5 are classed as the best performers. Panel A is based on average ROA. Panel B focuses on the average ratio of cash flow over assets. Panel D is based on the CARs from year 0 to year 5. In Panel C, bidders that increase dividends in at least one year are defined as bidders with no dividend cuts or omissions and at least one dividend increase in years 0-5. Panel A, based on the average ROA, shows that the non-executives of the worst performers are no different from the non-executives of the best performers in terms of the average number of non-executive directorships they hold in year 0. When the median is considered the difference is significant at the 10% level. However, there are no significant differences in the average and median number of non-executive directorships between all three performance categories in year 5. Further, for all three categories there is a significant drop in the number of non-executive 22

23 directorships held by the non-executives over the entire period. 16 Panels B, C and D report the results for average cash flow over assets, changes in dividends, and CARs, respectively. As was for ROA (see Panel A), there is a decrease in the number of directorships across time for all three categories of bidder performance. Also, there is no significant difference in the number of directorships held by the non-executives across the three categories when performance is measured by the average cash flow over assets ratio (Panel B) and the CARs (Panel D). In contrast, the non-executives that cut or omit their dividend have fewer directorships in year 5 whereas the opposite is the case for an increase in dividends. To summarize, the univariate analysis provides no evidence that the non-executives of the worst performing bidders hold fewer board seats in year 5 when performance is measured by ROA. However, non-executives that increase their dividends are rewarded with an increase in the number of directorships they hold. Finally, there is a decreasing trend in the number of directorships which is observed for all three categories of bidder performance as well as all three categories of bidder dividend policy. INSERT TABLE VI ABOUT HERE C. Regression Analysis We run three types of regressions: logit regressions that estimate the likelihood of surviving in the labor market until year 5 (Hypothesis 1), OLS regressions on the change in total 16 When performance is measured by average ROE (the figures are not tabulated), as for ROA, the number of directorships declines from year 0 to year 5. However, contrary to the ROA results, the difference between the number of non-executive directorships held by the non-executives of the best performing bidders and the number of non-executive directorships held by other performers is negative and significant at the 5% level in year 5. Thus, contrary to our expectations, in year 5 the non-executives of the best performing bidders seem to hold significantly fewer non-executive directorships compared to those of other bidders when performance is measured by average ROE. However, this difference is insignificant for the other two performance measures. 23

24 directorships from year 0 to year 5, and dynamic panel data regressions estimating the number of directorships in each of years 0-5 (both testing Hypothesis 2). The likelihood of surviving in the labor market until year 5 Table VII reports the results for the logit regressions which estimate the likelihood of the non-executive director still holding directorships in year 5. The dependent variable is set to one if the non-executive still holds directorships in year 5, and is zero otherwise. The independent variables are the performance (i.e. ROA), bidder characteristics (i.e. the ratio of long-term debt over equity and the market value of the firm), and non-executive characteristics (i.e. tenure, age and the dummy variable which indicates whether the non-executive chairs the bidder s board). Bidder and non-executive characteristics are measured in year 0. The first seven regressions include ROA for a single year; the next four regressions include ROA measured in two or more years from year 0 to year 5. The final regression includes the average ROA over year 0-5. The performance measures are significant in only four of the 12 regressions. When ROA is included for just one of the years 0-5, ROA is only significant for years 4 and 5. Still, while ROA 4 is significant and positive in regressions (5), (10) and (11), ROA 2 is significant and negative in regression (11). A significantly negative coefficient on the performance measure suggests that, contrary to Hypothesis 1, non-executives are punished for good acquisition performance rather than bad acquisition performance. Finally, regression (12) suggests that the average ROA across the entire period has no significant impact on the likelihood of the nonexecutive holding directorships in year 5. Hence, there is no consistent evidence that nonexecutives associated with poorly performing acquisitions are more likely to have left the non- 24

25 executive labor market in year 5. In contrast, the non-executive characteristics (except for tenure and individual non-executive ownership) as well as the bidder characteristics are significant in (virtually) all of the regressions and have consistent signs. Nevertheless, tenure is only significant and negative when its squared term which is significant and positive is also included in the regression (due to space constraints the relevant regressions are not reported in the table). This suggests that, while nonexecutives with longer tenure on the bidder s board are less likely to be in the labor market in year 5, this negative effect wears off as tenure becomes greater. Age is negative and significant at the 1% level in all 12 regressions. Age remains highly significant when we exclude tenure from the regressions (the results are not tabulated). The significantly negative coefficient on age suggests that non-executives, once they reach a certain age, retire and leave the labor market of their own will. However, when the square of age is included (due to space constraints these regressions are not reported in the table) the coefficient on age becomes positive, while remaining highly significant at the 1% level. The coefficient on the square of age is also significant at the 1% level and has a consistently negative sign. When the overall effect of age on the likelihood of remaining in the labor market in year 5 is evaluated using the range of ages observed in Table II, this effect is always positive. However, it first increases until it reaches its maximum at 56 years and then decreases. This suggests that age proxies for both experience, and hence the increased likelihood of remaining in the labor market, as well as closeness to retirement age, and hence the reduced likelihood of remaining in the labor market. Further, nonexecutives that chair the bidder s board are more likely to remain in the labor market in year 5. 25

26 The dummy variable is positive and significant at the 10% level or better in 11 of the 12 regressions, suggesting that there is more demand for non-executives with experience of chairing a board, hence the higher likelihood of them remaining in the labor market in year 5. Finally, the non-executive s ownership has no significant impact on the likelihood of staying in the labor market. However, Table II suggests that, despite about 70% of non-executives in the sample holding some shares in the bidder, the size of their holdings is very small. Hence, it is not surprising that this variable has no impact on the likelihood of remaining in the labor market. Moving on to the bidder characteristics, institutional ownership is positive and significant in only one of the regressions. This result suggests that non-executives of bidders with greater institutional ownership are able to tap into the network of the institutional investors, making it more likely for them to hold multiple directorships and stay in the labor market. In contrast, the debt-equity ratio is always negative and significant at the 1% level. 17 This suggests that nonexecutives of bidders that are highly indebted are less in demand, maybe reflecting the fact that the greater debt-equity ratio reduces the free cash flow problem (Jensen, 1986), hence reducing the monitoring role of the non-executives within the firm. Finally, the bidder s market value is positive and significant at the 10% level or better in seven of the 12 regressions. The positive coefficient suggests that non-executives of larger bidders are more attractive to other firms as potential non-executives for their boards given their experience of larger, more complex organizations and also given their higher visibility and greater professional standing. INSERT TABLE VII ABOUT HERE 17 The 34 bidders that do not survive until year 5 have significantly higher long term debt-equity ratios than the other bidders. However, when these observations are excluded, the long-term debt-equity ratio remains negative and significant at the 1% level. 26