Tournament Incentives and Acquisition Performance*

Transcription

1 Tournament Incentives and Acquisition Performance* Iftekhar Hasan Fordham University and Bank of Finland Marco Navone University of Technology Sydney Thomas To University of Sydney Eliza Wu** University of Sydney *We thank Sumit Agarwal, Si Cheng (SFS discussant), Ronald Masulis, Buhui Qiu and conference participants at the 2017 SFS Cavalcade Asia-Pacific and the 2017 FIRCG Conference at the University of Melbourne for helpful comments and discussions. All errors remain our own. **Corresponding author. Tel.:

2 Tournament Incentives and Acquisition Performance ABSTRACT This paper examines the impact of promotion-based tournament incentives on corporate acquisition performance. Measuring tournament incentives as the pay gap between the CEO and other senior executives, we show that acquirers with greater tournament incentives experience lower announcement returns. Further analysis shows that the negative effect is driven by overly risky deals, and the effect is stronger during the period when a promotion tournament is most likely to occur. Our results are robust to alternative identification strategies. Our evidence highlights that senior executives play a role in acquisition decisions in addition to the CEO. JEL classification: G34, J31, J33 Keywords: Tournament incentives, executive compensation, corporate acquisitions

3 1. Introduction Corporate acquisitions are one of the largest forms of corporate investment. As such, one would expect managers to exert a lot of effort into selecting acquisition targets and be able to generate shareholder value. Yet, the literature has found that this is often not the case (Andrade, Mitchell, and Stafford, 2001; Moeller, Schlingermannn, and Stulz, 2005; Hackbarth and Morellec, 2008). Some authors showed that the decision to make value destroying acquisitions could be due to the empire-building motive of the CEO (Jensen, 1986; Lang, Stulz, and Walkling, 1991; Masulis, Wang, and Xie, 2007; Harford, Humphery-Jenner, and Powell, 2012), while others have pointed out that managers who receive little equity-based compensation have little incentive to make good acquisitions (Lewellen and Loderer, 1985; Datta, Iskandar-Datta, and Raman, 2001). However, to the best of our knowledge, no prior study has focused on examining how incentives of senior executives other than the CEO can play an incremental role in affecting acquisition performance. This is very surprising since corporate acquisition decisions requires the combined efforts and agreements of a firm's senior executive team for a successful execution. In this paper, we aim to fill this gap in the literature by examining the effect of tournament incentives of senior executives (created by an observable pay gap between the CEO and other senior executives) on corporate acquisition performance. Over the past decade, CEO pay has received much attention from the media. 1 Due to the skewed corporate pay structure, the advancement of senior executives in a corporate hierarchy is often viewed as a tournament in which individuals compete with one another for 1 See Edmans, Gabaix, and Jenter (2017) for a review of the executive compensation literature. 1

4 the CEO position. The compensation gap between workers of various ranks as a feature of tournament incentives was first proposed by Lazear and Rosen (1981). They argue that the large compensation packages received by executives are likely to induce all lower ranked employees in the firm to work harder to increase their chances of securing a promotion. In the same vein, Bognanno (2001) focuses on CEO promotion tournaments and uses the compensation gap between the CEO and senior executives as the measure of tournament incentives. Since virtually every major corporate decision requires the combined efforts of the senior executive team, studies have found that tournament incentives are associated with improved innovative efficiency (Jia, Tian, and Zhang, 2017; Shen and Zhang, 2017) and better firm performance (Kale, Reis, and Venkateswaran, 2009; Chen, Ezzamel, and Cai, 2011; Burns, Minnick, and Starks, 2017). Due to these findings, our primary hypothesis which we label the Effort Hypothesis, predicts that firms with greater tournament incentives should make better acquisitions as the senior executives in these firms would put more effort in assisting the CEO in investigating which acquisitions would benefit the company the most in order to try and win the tournament prize (the pay gap). Anecdotal evidence suggests that the senior executive team are heavily involved in acquisition decisions. For example, in an interview with Google s corporate development VP Don Harrison, Harrison said My team, the deal sponsors, and Google s senior executives sit around the room to really decide whether or not this (the acquisition) is something we should do. (D Onfro, 2015). On the other hand, several studies have shown that higher tournament incentives also encourage greater corporate risk-taking by senior executives. Goel and Thakor (2008) theoretically model the relation between tournament incentives and corporate risk-taking. In 2

5 their model, they document that if every senior executive chooses the same level of risk as his competitors in the CEO promotion tournament, they will all have the same output at the end of the period given the risk-return relationship. The probability of getting promoted for all of the senior executives would then also be the same. Therefore, each executive can increase his own probability of promotion by taking on or supporting the CEO in pursuing riskier and larger projects as these projects can yield more extreme outcomes. Similarly, Hvide (2002) and Glilpatric (2009) theoretically demonstrate that promotion tournaments generate perverse incentives for executives to engage in excessive risk taking, while Kini and Williams (2012) provide empirical evidence to show that greater tournament incentives lead to the implementation of riskier corporate policies. Based on these established findings on the effect of tournament incentives on managers risktaking behavior, we expect that corporate senior executives would rationally support the CEO and board members in making extremely large and risky deals to enhance their own chances of winning the promotion tournament because these acquisitions can yield more extreme outcomes, and the failure to support these deals may push an executive down the rank in the tournament if one of these acquisitions eventually achieves a big success for the company. More importantly, senior executives are unlikely to receive any severe punishment if the acquisition fails as the CEO has the final say on acquisitions and is likely to be blamed. Moreover, objecting to these large and risky deals would be unlikely to benefit senior executives even if these deals fail as they are typically aligned with the longer term strategic plans set for the company and backed by the board of directors, and the board ultimately decides who the next CEO would be. Hence, our alternative Risk-Seeking Hypothesis predicts that firms with greater tournament incentives are more likely to make overly risky acquisitions 3

6 that, on average, destroy shareholder value as the senior executives in these firms would support for the extremely large and risky acquisitions to increase their own chances of securing a promotion sooner. We test these two competing hypotheses by empirically examining the relation between tournament incentives and acquisition performance using an extensive sample of 9,331 corporate acquisitions from 1993 to Controlling for performance-based incentives and risk-taking incentives of the CEO, firm and deal characteristics together with year and industry fixed effects, our baseline results indicate that tournament incentives created by pay gaps are negatively related to the firm's acquisition performance as measured by cumulative abnormal stock returns around the acquisition announcement. While our baseline results are more supportive of our risk-seeking hypothesis, a potential concern is that unobservable firm heterogeneity correlated with both a firm s pay gap and acquisition performance may be driving the results (i.e. the omitted variable concern), which makes it difficult to establish causality. We use a number of approaches to address the endogeneity issue, although we are aware of the fact that they can only alleviate but not completely solve the endogeneity problem due to the lack of direct exogenous shocks in the executive compensation literature (Adams, Hermalin, and Weisbach, 2010; Coles, Lemmon, and Meschke, 2012; Edmans, Gabaix, and Jenter, 2017). First, we employ an instrumental variable approach where we instrument tournament and CEO incentives with size adjusted industry median values, as prior studies suggests that the level and structure of managerial compensation varies by firm size and industry. The negative relationship between tournament incentives and acquisition performance remains robust to the use of these instruments. Second, we follow Shen and 4

7 Zhang (2017) and perform regression analysis based on a propensity-score matched sample to control for the systematic differences in firm and deal characteristics between firms with high and low tournament incentives. We find that our results continue to hold. Third, we conduct a modified difference-in-differences test in which we use the changes to the NYSE/NASDAQ listing rules in the early 2000s as a potential exogenous shock to tournament incentives. The new listing rules require the board of each listed company to have a majority of board members being independent. Thus, it is likely to provide a negative shock (weakening effect) to tournament incentives in the (treated) firms that were forced to increase board independence since the additional independent directors in these firms are likely to closely examine the potential acquisitions and control the excessive risk-taking behavior of senior executives. Again, we find results that supports our risk-seeking hypothesis. We rule out that our findings may be driven by an entrenchment hypothesis. This competing hypothesis arises as the pay gap between the CEO and other senior executives can also be interpreted as CEO entrenchment (Bebchuk and Fried, 2011; Chen et al., 2013), and the prior literature has found that entrenched CEOs make worse acquisitions (Masulis, Wang, and Xie, 2007; Harford, Humphery-Jenner, and Powell, 2012). To distinguish our tournament hypothesis from the entrenchment hypothesis, we first explicitly control for CEO entrenchment in all of our regressions. Second, we adopt a two-stage regression approach, where in the first stage we decompose the pay gap measure into the part that proxies for CEO entrenchment and the residual part which more likely proxy for tournament incentives and in the second stage we use the residual as our proxy for tournament incentives. We continue to find a negative relationship between tournament incentives and acquisition performance. 5

8 We then try to provide stronger evidence of our tournament hypothesis by examining situations when firms are most likely to initiate a CEO promotion tournament, and test whether the effect of tournament incentives on acquisition performance is stronger in these situations. First, the effect of tournament incentives should be most pronounced during the actual promotion tournament, i.e. during the period prior to CEO turnovers. Second, a firm that has been performing poorly is more likely to replace its CEO, and hence is likely to hold a promotion tournament to do so. Third, a firm whose current CEO was hired internally is more likely to continue its tradition and run a promotion tournament again. Consistent with our predictions, we find that the negative relation between tournament incentives and acquisition performance is stronger in firms that are more likely to foster a promotion tournament. Finally, we examine the source of value destruction. According to our risk-seeking hypothesis, we expect the negative effect of tournament incentives to be driven by extremely large and extremely risky acquisitions that, on average, destroy shareholder value, as senior executives are likely to have played a key role in supporting or even aggressively pushing these overly risky deals due to their own tournament incentives. We find evidence in support of our riskseeking hypothesis as we discover that the negative effect of tournament incentives on acquisition performance only occurs for extremely large and extremely risky deals. To check the robustness of our results, we use alternative constructs to proxy for tournament incentives, include CEO overconfidence and senior executives' equity incentives as additional controls in our regressions, and exclude utility firms from our sample. We find that our results continue to hold in each of these robustness checks. Our paper contributes to two strands of literature. First, it contributes new evidence to the extensive literature on agency problems in corporate acquisitions. Prior studies have generally 6

9 found that managers in firms with weaker corporate governance (Masulis, Wang, and Xie, 2007; Harford, Humphery-Jenner, and Powell, 2012) and those having excessive cash flows (Jensen, 1986; Harford, 1999) are more likely to make value destroying acquisitions. In contrast, firms that provide managers with a higher level of equity-based compensation (Lewellen, Loderer, and Rosenfeld, 1985; Datta, Iskandar-Datta, and Raman, 2001) are more likely to make value enhancing acquisitions. However, to the best of our knowledge, no study has examined how the difference in compensation levels between the CEO and other senior executives can lead to agency problems that affect acquisition performance. This is an important research question since almost every corporate acquisition decision requires the combined efforts and agreements of its senior executive team for a successful execution. Our study contributes to the literature by showing that tournament incentives induce senior executives to support for overly risky acquisitions that on average destroy shareholder value. Second, our paper contributes to the growing literature examining the effects of tournament incentives. Existing studies find that greater tournament incentives lead to improved innovative efficiency (Jia, Tian, and Zhang, 2017; Shen and Zhang, 2017), better firm performance (Kale, Reis, and Venkateswaran, 2009; Chen, Ezzamel, and Cai, 2011; Burns, Minnick, and Starks, 2017), riskier corporate policies (Goel and Thakor, 2008; Kini and Williams, 2012), and a higher likelihood of corporate fraud (Haß, Müller, and Vergauwe, 2015). By using a large sample of corporate acquisitions, we are able to contribute new evidence on the effect of tournament incentives on corporate acquisition performance. The rest of the paper is organized as follows. Section 2 describes the data and explains the construction of various variables used in this study. Section 3 examines that impact of 7

10 tournament incentives on corporate acquisition performance. Section 4 addresses the identification issues. Section 5 provides robustness tests, and section 6 concludes. 2. Data sources and variable construction 2.1. Data sources We obtain compensation data for CEO and other senior executives over from the Standard and Poor's (S&P) ExecuComp database. Following Kale, Reis, and Venkateswaran (2009), we classify an executive as a CEO if he is identified as the CEO of a firm in ExecuComp (CEOANN = CEO), and classify all other senior executives as VPs. We also collect numerous CEO related variables from ExecuComp, such as the tenure of the CEO, whether the CEO is also the Chair of the firm, whether the CEO was promoted from within the firm, and the number and value of options held by the CEO. We then obtain board information from RiskMetrics, financial data from Compustat and financial markets data from CRSP for our sample of firms in the ExecuComp database. To examine the effect of tournament incentives on acquisition performance, we collect acquisitions data from the Securities Data Company (SDC) M&A database from 1993 to We require a minimum deal value of $1 million and we include only deals where the acquiring firm controls less than 50% of the target's stocks before the announcement and owns 100% of the target's stocks after the transaction. After merging with the data from other databases, we end up with a final sample of 9,331 acquisitions made by 2,026 acquirers. 8

11 2.2. Tournament incentives Following Kini and Williams (2012), we measure tournament incentives as the natural logarithm of the difference between the CEO's total compensation package (ExecuComp variable TDC1) and the median VP's total compensation package. The pay gap captures the increase in a VP's compensation if he wins the tournament to become the firm's next CEO. The logarithmic transformation helps us to eliminate observations where the calculated pay gap is unlikely to be a good proxy for tournament incentives (when the median VP's salary exceeds the CEO). 2 (Insert Table 1) 2.3. CEO, firm and deal characteristics We control for CEO alignment and CEO risk taking incentives. Following Kini and Williams (2012), we include as control variables Ln(CEO delta) and Ln(CEO vega). CEO delta is computed as the dollar increase in a CEO's portfolio wealth for a 1% increase in the firm's stock price, whereas CEO vega is computed as the dollar increase in a CEO's portfolio wealth for a 1% increase in the standard deviation of the firm's stock volatility. 3 We take the natural logarithm of CEO delta and CEO vega to ensure that the scale is comparable to tournament incentives. We also include CEO entrenchment as a control variable since prior studies document that entrenched CEOs make worse acquisitions (Masulis, Wang, and Xie, 2007; Harford, Humphery- Jenner, and Powell, 2012). Specifically, we combine multiple CEO entrenchment measures 2 Kini and Williams (2012) find that most of the negative pay gaps occur in firms where the CEO was a founder and received little to no compensation. The calculated pay gap is unlikely to be a good proxy for tournament incentives in these firms. 3 We thank Lalitha Naveen for making the data on CEO delta and CEO vega publicly available. See Coles, Daniel, and Naveen (2006) for the detailed steps involved in the calculations. 9

12 into a broader index of governance quality. The measures we use include CEO/Chair duality (Jensen, 1993), whether the CEO is the only insider director (Adams, Almeida, and Ferreira, 2005), and whether the CEO is highly tenured (4 th quartile) (Berger, Ofek, and Yermack, 1997). To construct the index, we split each of the CEO entrenchment measures into two groups, with higher values indicating higher entrenchment, and cumulate the ranks (0 1). We then divide the cumulated ranks by the number of measures available for the firm-year to obtain the CEO entrenchment index score. 4 Following the acquisition literature, we control for a vector of firm and deal characteristics that may affect a firm's acquisition decisions (variable definitions are provided in Table 1). Our firm-level controls include firm size, leverage, tobin's q, cash flow, governance quality, and stock price runup. Our deal-level controls include relative deal size, method of payment, target public status, cross-industry indicator, and cross-border indicator. Following Jenter and Lewellen (2015), we combine multiple governance measures into a broader index of governance quality. The measures we use include board independence, board size, institutional ownership and the G-index. Similar to constructing the CEO entrenchment index, we split each of the governance measures into two groups, with higher values indicating better governance, and cumulate the ranks (0 1). We then divide the cumulated ranks by the number of measures available for the firm-year to obtain the governance index score. 5 (Insert Table 2) 4 Combining different measures into a single CEO entrenchment index helps us to control for multiple aspects of CEO entrenchment without sacrificing sample size, since board information is only available from 1996 from RiskMetrics. We also examine each of these measures separately in Section Similar to our CEO entrenchment index, combining different measures into a single governance index helps us to control for multiple aspects of firm governance without sacrificing sample size, especially since the data on the G-index is only available until To be consistent with our CEO entrenchment index, we divide the governance measure at the median rather than at the tercile as in Jenter and Lewellen (2015). All our results are similar if we divide the governance measure at the tercile rather than at the median. 10

13 2.4. Summary statistics Table 2 shows the summary statistics. We winsorize all continuous variables at the 1st and 99th percentiles. Tournament incentives and CEO and firm characteristics are all measured at the fiscal year-end prior to the acquisition announcement. Panel A reports that the average pay gap between the CEO and the median VP is $4.31 million, and 68% of the CEOs were promoted from within the firm. Panel B shows that an average acquirer in our sample has a leverage ratio of 21% and tobin's q of 2.16, which is consistent with the figures reported in prior studies (e.g. Yim, 2013; Huang et al., 2014). Panel C shows that the size of the average acquisition in our sample is 12% of acquirer s market capitalization. Among the acquisitions, 43% are cross-industry deals, 21% are cross-border deals, and 39% are funded entirely by cash. 3. Tournament incentives and acquirer announcement returns 3.1. Baseline results To examine how tournament incentives affect acquisition performance, we estimate the following model: CAR i,t+1 = α + βlnpaygap i,t + γz i,t + Year t + Industry j + ε i,t (1) where t donates year, i donates firm, and j denotes industries. The dependent variable, CAR i,t, is the 5-day cumulative abnormal returns centered on the acquisition announcement date. Following Masulis, Wang, and Xie (2007), the abnormal stock returns are calculated by estimating the market model for each acquirer over a 200 day period ending 11 days before the announcement date (-210, -11) with the CRSP value-weighted return as the market index. The tournament incentives measure, LnPayGap i,t, is calculated as the natural logarithm of 11

14 the difference between the CEO's total compensation and the median VP's total compensation. Z i,t is a vector of firm, CEO, and deal characteristics that are likely to affect a firm s acquisition performance. Year t and Industry j capture time and industry fixed effects respectively. Due to the presence of serial acquirers in our sample, the residuals in our regressions may be correlated and hence may overstate the t-statistics (Petersen, 2009). To control for this potential problem, we also cluster standard errors by acquiring firm in all of our regressions. We first start with a parsimonious model that regresses abnormal returns on the key variable of interest, Ln(Pay gap). Column 1 of Table 3 reports the results. We find that the coefficient estimate of Ln(Pay gap) is negative and significant at the 1% level, suggesting a negative raw association between tournament incentives and acquisition performance. Next, we add CEOlevel and firm-level controls and report the results in Column 2. The coefficient estimate continues to be negative, and is significant at the 5% level. Finally, we add deal-level controls into the regression and find that the coefficient estimate of Ln(Pay gap) continues to be negative and significant at the 5% level, as reported in Column 3. 6 In terms of economic magnitudes, we find that a one standard deviation increase in pay gap centred on its sample mean reduces CAR by 31 basis points, which translates into a loss of $40 million in shareholder value for the average acquirer in our sample. 7 6 We find similar results for the effect of tournament incentives on change in acquirer s operating performance proxied by change in ROA from year t-1 to year t+1. 7 The economic significance is easier to interpret when it is computed in terms of a one standard deviation change in pay gap rather than the natural logarithm of pay gap. To achieve this, we first calculate the level of pay gap that is 0.5 standard deviations below its mean (low pay gap) and 0.5 standard deviations above its mean (high pay gap). We then compute the difference between the natural logarithm of high pay gap less the natural logarithm of low pay gap. 12

15 The coefficients of our control variables exhibit the expected signs. For example, CEOs who receive more equity-based compensation have more incentive to make value-enhancing acquisitions (Datta, Iskandar-Datta, and Raman, 2001), managers in larger firms are more entrenched (large firm size serves as an effective takeover defense) and therefore are likely to make value-destroying acquisition (Masulis, Xie, and Wang, 2007), firms that are overvalued (higher tobin s q) are likely to reveal their true value to the market on the announcement of an acquisition (Moeller, Schlingemann, and Stulz, 2004), acquirers do worse when paying with equity due to the well-documented adverse selection problem (Myers and Maliuf, 1984), and acquirers do better when acquiring private and subsidiary targets since they capture a liquidity discount (Fuller, Netter, and Stegemoller, 2002). (Insert Table 3) 3.2. Presence of a Tournament Next, we try to provide stronger evidence for our tournament hypothesis by closely looking at certain situations when firms are most likely to hold a CEO promotion tournament. The effect of tournament incentives on acquisition performance should be much stronger in these circumstances because it is more likely that there is an actual tournament happening. On the other hand, the effects of tournament incentives are likely to be weaker or non-existent in all other circumstances because there may not be an actual tournament happening at all. To test our predictions, we examine three situations when a firm is most likely to hold a promotion tournament. First, a tournament is most likely to have been held during the period prior to CEO turnovers in order to try and pick a candidate to become the next CEO (Masulis and Zhang, 2017; Shen 13

16 and Zhang, 2017). VPs are likely to be highly induced by tournament incentives in this period because they are likely to know that the current CEO is about to step down (either through private information channels within the firm or due to poor performance of the CEO). Prior to CEO turnover is defined as the time period where the acquisition is made within 2 years before a CEO turnover event. Consistent with our prediction, Columns 1 and 2 of Table 4 show that the negative effect of tournament incentives acquisition performance is much more pronounced during the time period prior to CEO turnovers. (Insert Table 4) Second, a firm is more likely to hold a promotion tournament in order to try and find a suitable candidate to replace the current CEO if the firm has been performing poorly for the past few years, as it is well documented that a CEO is more likely to be replaced if the firm has been performing poorly (Weisbash, 1988; Morck, Shleifer, and Vishny, 1989; Jensen and Murphy, 1990). On the other hand, one wouldn t expect a firm to hold a promotion tournament if the firm has been performing great because there is no need to replace the current CEO. We classify firms as having been performing poorly (well) if the average industry-adjusted profitability (ROA) in the 3 years prior to the acquisition is below (above) the median. Columns 3 and 4 report the regression results. The negative effect of tournament incentives is more pronounced for firms that have been performing poorly prior to the acquisition. Third, a firm whose current CEO was hired internally is more likely to continue its tradition and run a promotion tournament again. For example, Hewlett-Packard was hiring CEO internally for the first 52 years of operations (Larcker and Tayan, 2011). We define a CEO to be an insider if he has been working for the firm prior to becoming the CEO. Results in columns 14

17 5 and 6 show that the negative effect of tournament incentives on acquisition performance only occurs in firms where the current CEO was promoted from within the firm The risk-seeking channel Our evidence so far suggests that tournament incentives harm acquisition performance. Our risk-seeking hypothesis predicts that tournament incentives induce senior executives to support/convince the CEO in making overly risky deals to enhance their own chances of winning the promotion tournament because these projects can yield more extreme outcomes, and there is little downside risk for doing so as they receive little punishment from their excessive risk-taking. If this is the case, we should only observe the negative effect of tournament incentives on acquisition performance in these deals. First, extremely large acquisitions can yield extreme outcomes. Hence, we expect the negative relation between tournament incentives and acquisition performance to be driven by extremely large acquisitions as VPs are likely to have supported these acquisition due to their tournament incentives. We split relative deal size into quintiles and define deals in the 5 th quartile to be extremely large deals. Consistent with our prediction, Columns 1 and 2 of Table 5 show that the negative effect of tournament incentives only occurs for extremely large deals. (Insert Table 5) Second, since we hypothesize that tournament incentives induce senior executives to support for overly risky acquisitions, we should then expect the negative effect of tournament incentives to be concentrated in extremely risky deals. We follow Agrawal and Mandelker (1987) and use the change in the standard deviation of acquirers' stock returns surrounding the acquisition to proxy for deal risk. Specifically, deal risk is calculated as the difference in the 15

18 acquirers' standard deviation of stock returns in the post-acquisition period (11 to 70 days following the effective date) compared to the pre-acquisition period (120 days to 60 days prior to the announcement date). The pre-acquisition period ends 60 days before the announcement date and the post-acquisition period begins 11 days after the effective date in order to avoid the acquisition negotiation or completion period in affecting stock returns. We sort the acquisitions into quintiles according to their level of risk and define deals in the 5 th quartile to be extremely risky deals. Columns 3 and 4 of Table 5 present the results. Consistent with our prediction, we find that the negative effect of tournament incentives only exists in deals that are highly risky (5th quintile), as VPs are likely to have played a key role in supporting these deals due to their own tournament incentives Tournament incentives or CEO entrenchment? So far we have used the pay gap between the CEO and the VPs as a measure of tournament incentives. However, prior studies (Bebchuk and Fried, 2011; Chen et al., 2013) have also interpreted the pay gap as a measure of CEO power. Given that prior research finds that entrenched CEOs tend to make worse acquisitions (Masulis, Wang, and Xie, 2007; Harford, Humphery-Jenner, and Powell, 2012), it is also possible that our findings are actually driven by the entrenchment hypothesis rather than our tournament hypothesis. In addition to explicitly controlling for CEO entrenchment in all of our regressions, in this section we adopt a two-stage regression approach to further distinguish our tournament hypothesis from the CEO entrenchment hypothesis. Specifically, in the first stage we decompose the pay gap measure into 2 parts, one part that proxies for CEO entrenchment and the residual part that more likely proxy for tournament incentives. Then in the second stage we use the residual 16

19 part as our proxy for tournament incentives to examine the relationship between tournament incentives on acquisition performance. (Insert Table 6) Table 6 presents the results. Column 1 reports the results from the first-stage regression. 8 We see that all of the three CEO entrenchment measures are significantly related to pay gap. In particular, the pay gap is larger if the CEO is also the Chair of the firm or is the only insider on the board, and the pay gap is smaller for higher tenured CEOs, which are consistent with findings of prior studies on CEO compensation (Core, Holthausen, and Larcker, 1999; Guthrie, Sokolowsky, and Wan, 2012). Columns 2 to 4 report the results from the second-stage regressions where we regress announcement returns on the residuals from our first-stage regression which more likely proxy for tournament incentives. We find results that are similar to our baseline regressions, both in significance and in magnitude, indicating that the negative relation between pay gaps and acquisition performance is indeed driven by tournament incentives rather than CEO entrenchment. 5. Identification A potential endogeneity issue that may affect our baseline results is the omitted variables bias. Even after controlling for several known CEO, firm, and deal characteristics and year and industry fixed effects, there may still be unobservable firm or CEO heterogeneity correlated with both the pay gap and acquisition performance, and these factors could bias our results. 8 Our sample size shrink as we are only able to obtain board information for 73% of the acquirers in our sample from Riskmetrics. 17

20 As both tournament and CEO incentive measures are related to managerial compensation, Ln(Pay gap), Ln(CEO delta), and Ln(CEO vega) could all be endogenous. In this section, we address the endogeneity issue using a number of approaches, although we are aware of the fact that they can only lighten but not completely solve the endogeneity problem due to the lack of natural experiments (direct exogenous shocks) in the executive compensation literature (Adams, Hermalin, and Weisbach, 2010; Coles, Lemmon, and Meschke, 2012; Edmans, Gabaix, and Jenter, 2017) Instrumental variable analysis First, we employ an instrumental variable approach to address the endogeneity problem. Specifically, we use the median values of tournament and CEO incentives for firms in the same industry and in the same size quartile as the firm to instrument for tournament and CEO incentives. The underlying economic rationale for using the instruments comes from Murphy (1999), who documents that the level and structure of managerial compensation varies by firm size and industry. Table 7 presents the two-stage least squares (2SLS) regression results. Columns 1 to 3 report the results from the first-stage regressions for the three endogenous variables. We include the same set of controls as the baseline regressions, and also include year and industry fixed effects with standard errors clustered by acquiring firm. In columns 1 to 3 we show that industry median values of tournament and CEO incentives are positively correlated the firm s tournament and CEO incentives, which confirms the relevance criteria of these instruments. (Insert Table 7) 18

21 Column 4 reports the results from the second-stage regression. Our main tournament variable Ln(Pay gap) and the two CEO incentive variables Ln(CEO delta) and Ln(CEO vega) are all replaced by its fitted values from the first-stage regressions. We find that the coefficient estimate of Ln(Pay gap) is negative and statistically significant at the 1% level, which supports our risk-seeking hypothesis. In addition to the regression results, we also report the F-statistics for our first-stage regressions and the Anderson Rubin F-statistics for our second stage regression to show that the instruments are relevant Propensity-score matching Although the instrumental variables are widely used in the literature (Kale, Reis, and Venkateswaran, 2009; Kini and Williams, 2012), one concern is that when an industry specific shock occurs it can jointly affect both M&A activities and compensation levels of CEOs and VPs within an industry. That is, even though the industry median levels of managerial compensation do not have a direct effect on a firm s acquisition performance, the occurrence of an industry shock may cause a correlation between industry median levels of managerial compensation and a firm s acquisition performance and threaten the validity of our instrumental variable analysis (Gormley and Matsa, 2014). Hence, in our second approach to address the endogeneity problem we follow the recent work of Shen and Zhang (2017) and conduct regression analysis on a propensity-score matched sample, in which firms with high tournament incentives are matched with firms with low tournament incentives to control for the systematic differences in firm and deal characteristics across these two groups. To construct the matched sample, we first estimate a Probit regression where the dependent variable is High pay gap, which equals to one if the acquirer s pay gap is above the 60 th percentile and zero if the acquirer s pay gap is below the 40 th percentile, and the independent 19

22 variables include all the firm characteristics variables. This generates a predicted probability of being a high-pay-gap firm for each observation, which is called the propensity score. Second, we match each treatment firm (a firm with high pay gap) with a matching firm (a firm with low pay gap) drawn from the same year and Fama-French 49 industry and have the closest propensity score within a caliper of 1% (and 5% as a robustness check). Using the matched sample, we conduct the baseline regressions using the new indicator variable High pay gap. Table 8 presents the regression results. The coefficients for High pay gap are negative and significant at the 5% level. (Insert Table 8) 5.3. NYSE/NASDAQ listing rule changes Our third approach to address the endogeneity problem is to conduct a modified differencein-differences (DiD) test in which we use the changes to the NYSE/NASDAQ listing rules as a potential exogenous shock to tournament incentives. In July 2002, about half a year after Enron declared bankruptcy, Congress passed the Sarbanes- Oxley Act (SOX). Shortly after, the New York Stock Exchange (NYSE) and the Nasdaq Stock Market (NASDAQ) proposed new exchange listing rules aimed to improve the governance of listed firms. The proposals were approved by the SEC with minor changes in November 2003, and went into effect in The main provisions of the rules require the board of each listed company to have (1) a majority of independent directors and (2) fully independent nominating, compensation, and audit committees. A major attraction of this natural experiment is that a large number of firms were already in compliance with the new exchange listing rules before the proposed changes, and hence these firms can act as control firms to the incompliant (treated) firms in a difference-in-differences analysis. 20

23 We hypothesize that the introduction of the new listing requirements would provide a negative shock (weakening effect) to tournament incentives in the (treated) firms that were forced to raise board independence since the additional independent directors in these firms are likely to closely examine the potential acquisitions, ask additional questions in board meetings, and ultimately vote against the overly risky acquisitions supported by senior executives who look to increase their own chances of promotion. We obtain data on board independence from RiskMetrics and stock exchange data from CRSP. To implement the modified DiD analysis, we follow Banerjee, Humphery-Jenner and Nanda (2015) and estimate the following model: CAR i,t+1 = α + β (1) LnPayGap i,t + β (2) Treated i + β (3) Post t + β (4) Treated i Post t + β (5) Treated i LnPayGap i,t + β (6) Post t LnPayGap i,t + β (7) Treated i Post t LnPayGap i,t + γz i,t + Year t + Industry j + ε i,t (2) where t donates year, i donates firm, and j denotes industries. Treated i equals to 1 for firms that were members of the NYSE or NASDAQ and did not have a majority of independent directors on the board in 2001 before the changes were proposed, and equals to 0 for (control) firms that were members of the NYSE or NASDAQ and already had a majority of independent directors on the board. Post t equals to 1 for year 2005 and onwards and zero otherwise, as listed firms were required to comply with the new NYSE and NASDAQ rules by October 31, Z i,t is a vector of firm, CEO, and deal-level characteristics that are likely to affect a firm s acquisition performance. Year t and Industry j capture time ad industry fixed effects respectively. Standard errors are clustered at the firm level. Our hypothesis predicts β (7) > 0 (i.e., an improvement in acquisition performance after a rise in board independence for firms with large tournament incentives). 21

24 The success of the DiD approach rests on the assumption that the only relevant difference between the treated and control firms is the treatment. To ensure that the treatment and control firms are similar before the listing rule changes, we follow Guo and Masulis (2015) and run the DiD regressions using a propensity-score matched sample. First, we predict the likelihood that a firm would not have a majority of independent directors on the board in 2001 (before the new listing rules were introduced) using data up to Specifically, we estimate a probit model where the dependent variable is Treated, and the independent variables include all the firm characteristics variables. Second, we match each treatment firm with a control firm drawn from the same Fama-French 49 industry and have the closest propensity score within a caliper of 1% (and also 5% as a robustness check). (Insert Table 9) Table 9 presents that regression results using the matched sample. Consistent with our hypothesis, in columns 1 and 2 we find the coefficient estimates of β (7) to be positively significant. A potential concern of our finding is that certain firm characteristics of the treated firms not controlled for could be causing the changes in acquisition performance. To address this concern, we conduct a placebo test by using the same set of treated and control firms but assume that listed firms were required to comply with the new NYSE and NASDAQ rules in a different year. We use year 2009 as the event year since Guo and Masulis (2015) show that the percentage of independent directors for the treated and control firms starts to converge in Columns 3 and 4 report the DiD estimation results that uses 2009 as the event year. The coefficient estimates of β (7) is insignificant. This result suggests that our finding is unlikely to be driven by firm characteristics specific to the treated or control firms. Overall, the baseline 22

25 regression results, instrumental variable regression results, propensity-score matched sample regression results and DiD results are all consistent with our risk-seeking hypothesis. 6. Robustness tests In this section, we discuss the results of an array of additional robustness tests. These robustness tests relate to alternative constructs to proxy for tournament incentives, the inclusion of CEO overconfidence and VP equity incentives as additional controls in our regressions, and the exclusion utility firms from our sample. First, our use of the median VPs' compensation in constructing pay gaps could overestimate tournament incentives if only one or two of the highest paid VPs have a chance of obtaining the promotion (Masulis and Zhang, 2017). We address this potential measurement error by using two alternative measures of pay gap; the natural logarithm of the difference between the CEO's pay and the highest VP's pay and the natural logarithm of the difference between the CEO's pay and the mean VP's pay. The number of observations drops by about 13% when we use highest VP s pay in constructing pay gaps since the compensation of some of the highest paid VPs exceeds the CEO, and those observations would be dropped when we take the natural logarithm of the difference in pay. We repeat our tests using the two alternative measures of pay gap and report the results in Table 10. Our prior findings remain unchanged as we continue to find that higher tournament incentives lead to lower announcement returns. (Insert Table 10) Second, an omitted factor that can possibly be driving our results is CEOs level of risk tolerance. Specifically, one may argue that firms with higher tournament incentives are more likely to employ risk-loving CEOs since the winners of these tournaments have an 23

26 exceptionally high tolerance for risk. To ensure that this alternative explanation is not driving our results, we control for CEO overconfidence in our regressions. In unreported results we also find that CEOs in firms with higher tournament incentives are not more risk-loving. This suggests that the winners of these tournaments only took more risk during the CEO selection process due to the high level of tournament incentives. We follow Banerjee, Humphery-Jenner and Nanda (2015) in calculating how deep in-the-money the CEO s vested stock options are to construct a measure of CEO overconfidence (detailed definition of the variable can be found in Table 1). Our sample size drops by about 9% as some CEOs do not receive options as a form of compensation. Columns 1 and 2 of Table 11 report the regression results after controlling for the level of CEO overconfidence. The control variables are omitted for brevity. We find that our results remain unchanged. (Insert Table 11) Third, in all of our tests so far we have assumed that the potential pay rise from becoming the next CEO is the only incentive of senior executives. However, one may argue that senior executives also have alignment and risk taking incentives associated with the firm's stock performance, and these incentives could also be driving their decisions in choosing suitable acquisition targets. We address this potential problem by controlling for the natural logarithm of one plus the median VP's delta and vega in our regressions. Columns 3 and 4 of Table 11 report the results. We continue to find a negative effect of tournament incentives on acquisition performance. Fourth, we exclude utility firms from our sample as utility firms are often heavily regulated and the linkage between compensation polices and firm performance may be different to 24

27 other industries. Columns 5 and 6 of Table 11 report the results. We continue to find that higher tournament incentives lead to lower announcement returns 7. Conclusions In this paper, we examine the effects of tournament incentives of senior executives, created by an observable pay gap between the CEO and other senior executives, on the performance of corporate acquisitions. Using a large sample of corporate acquisitions made by U.S. firms over the period from 1993 to 2015 we show that stronger tournament incentives lead to worse acquisition performance. This relationship holds in an instrumental variable regression analysis, propensity-score matched sample regressions, and a DiD analysis that uses the NYSE/NASDAQ listing rule changes as an exogenous shock to tournament incentives. Furthermore, our empirical evidence indicates that the negative effect of tournament incentives are driven by overly risky deals, and the effect is more pronounced in firms that are more likely to hold a promotion tournament. We also address an alternative interpretation of our empirical findings. Some authors have argued that the pay gap between the CEO and other senior executives can also be a measure of CEO entrenchment, and prior literature has found that entrenched CEOs make worse acquisitions. We explicitly control for CEO entrenchment in all of our regressions and also adopt a two-stage regression approach to distinguish our tournament hypothesis from the entrenchment hypothesis. We find that the negative relationship between tournament incentives and acquisition performance holds in the two-stage regressions models. Our paper contributes to the extant literature on corporate acquisitions by showing that senior executives play a crucial role in corporate acquisition decisions in addition to the CEO. 25

28 Specifically we show that tournament incentives induce senior executives to support overly risky acquisitions that, on average, destroy shareholder value. This channel of wealth destruction adds to our current knowledge on how acquiring shareholders lost significant wealth in the hands of corporate managers as documented by Moeller, Schlingemann, and Stulz (2005). Overall, our findings uncover a dark side of tournament incentives and provide boards with another important factor to consider when designing compensation structures to lift firm performance. 26

29 References Adams, R. B., Almeida, H., Ferreira, D., Powerful CEOs and their impact on firm performance. Review of Financial Studies 18, Adams, R. B., Hermalin, B. E., Weisbach, M. S., The role of boards of directors in corporate governance: A conceptual framework and survey. Journal of Economic Literature 48, Agrawal, A., Mandelker, G., Managerial incentives and corporate investment and financing decisions. Journal of Finance 42, Andrade, G., Mitchell, M., Stafford, E., New evidence and perspectives on mergers. Journal of Economic Perspectives 15, Banerjee, S., Humphery-Jenner, M., Nanda, V., Restraining overconfident CEOs through improved governance: Evidence from the Sarbanes-Oxley-Act. Review of Financial Studies 28, Bebchuk, L, Fried, J., The CEO pay slice. Journal of Financial Economics 102, Berger, P., Ofek, E., Yermack, D., Managerial entrenchment and capital structure decisions. Journal of Finance 52, Bognanno, M.L., Corporate tournaments. Journal of Labor Economics 19, Burns, N., Minnick, K., Starks, L., CEO tournaments: A cross-country analysis of causes, cultural influences and consequences. Journal of Financial and Quantitative Analysis 52, Chen, J., Ezzamel, M., Cai, Z., Managerial power theory, tournament theory, and executive pay in China. Journal of Corporate Finance 17, Chen, Z., Huang, Y., Wei, K., Executive pay disparity and the cost of equity capital. Journal of Financial and Quantitative Analysis 48, Coles, J. L., Daniel, N. D., Naveen, L., Managerial incentives and risk-taking. Journal of Financial Economics 79,