Market Valuation and Target Horizon in Mergers & Acquisitions

Transcription

1 Market Valuation and Target Horizon in Mergers & Acquisitions Tao Lin University of Hong Kong Liyan Miao University of Hong Kong First draft: March, 2006 This draft: May, 2006 We are also grateful to Prof. Eric Chang, Qiao Liu, Xianming Zhou and participants at the seminar of the University of Hong Kong. All errors are ours.

2 Market Valuation and Target Horizon in Mergers & Acquisitions Abstract Under the market-driven acquisition theory, targets tend to be underpaid by rational acquirers. One of the reasons targets agree to the transaction is that they are short-horizon oriented. Using a sample of 770 mergers and acquisitions announced between 1993 and 2002, this study empirically shows that targets are more likely to be underpaid when they have shorter horizon-oriented shareholders and CEOs. Shareholder turnover ratio, CEO age, tenure, equity-based wealth and incentive to cash out are proxied for horizon. 1

3 1. Introduction Much of the research in corporate finance has focused on the underlying motivation for mergers and acquisition. Jensen (1988) summarizes that in the long-run, takeovers generate substantial gains, which result from easier access to resources, transfer to more highly-valued use of assets, etc. The managerial hubris hypothesis posited by Roll (1986) suggests a psychological interpretation and implies that managers in bidding firms tend to overpay for the target. Agency theory focuses on the interest disparity between shareholders and executives. Grinstein and Hribar (2004) investigate the CEO compensation package, the alignment of interest between shareholders and executives and managerial incentives in mergers and acquisitions. Recent research work tries to explain the observed merger waves and the positive correlation between market valuation and merger waves. Shleifer and Vishny (2003) establish a model of mergers and acquisitions which argues that stock market misvaluation drives acquisitions. According to this model, they posit that the long-run reason for bidders to takeover the target is the undervaluation of the target. Shareholders of the acquiring firm will gain from the acquisition if the target is undervalued to fundamentals in cash acquisitions and relatively undervalued to the bidder in stock acquisitions. Overvalued firms are more likely to be bidders and undervalued firms are more likely to be targets. Target shareholder will gain in the short-run. However, shareholders holding onto shares in an overvalued market will lose in the long-run. Therefore, target managers will protect long-horizon shareholders interest if they resist the offer. The model predicts that bidders in stock acquisitions either have longer horizon than targets or pay personal deals to target managers to agree to the acquisition. Rhodes-Kropf and Viswanathan (2004) propose a different model, which separates misvaluation into firm-specific component and market-wide component. The model shows that the potential market valuation deviation from fundamental values impacts 2

4 merger activities. Their theory predicts that acquisitions are more likely to occur in overvalued markets. Because in the more overvalued market, the rational target managers with limited information will overestimate the firm-specific misvaluation, and thus overvalue the synergies from the acquisition. Method of payment will contain a higher fraction of stock in overvalued markets and higher fraction of cash in undervalued markets. The follow-up empirical research provides strong support to the theory of Shleifer and Vishny (2003) and Rhodes-Kropf and Viswanathan (2004). Dong et al. (2003) tests the hypotheses of misvaluation driven acquisitions. Their study employs two contemporaneous measures of misvaluation: pre-takeover ratio of book value of equity to price and pre-takeover ratio of residual income value to price. They find evidence that bidders are relatively overvalued to targets, bidders are more overvalued in stock acquisitions than in cash acquisitions, undervalued targets receive higher premium and are more likely to resist the offer, which is consistent with the predictions by Shleifer and Vishny (2003). Rhodes-Kropf, Robinson and Viswanathan (2005), based on the theory of Rhodes-Kropf and Viswanathan (2004), decompose M/B into three components: firm-specific error, time-series sector error and long-run value to book to measure misvaluation. The empirical results show that acquirers and targets appear to share the common misvaluation component of time-series sector. The difference in M/B between acquirers and targets is mainly attributed to the firm-specific error. High firm-specific error is positively correlated to the possibility of being involved in a merger as an acquirer and using stock as method of payment. Similar to Dong et al. (2003) and Rhodes-Kropf, Robinson, and Viswanathan (2004), Ang and Cheng (2005) find evidence on the market-driven acquisitions theory. In addition, they also find some new results which are consistent with the critical assumption made by Shleifer and Vishny (2003). Shareholders of acquiring firms do not suffer loss in the 3

5 long-run compared to those of matching firms. However, former shareholders of target firms are the losers if they hold shares in a long-term post-merger period. Although previous empirical research has found evidence on the misvaluation of firms and the long-run return for the market-driven acquisitions, we still lack evidence on the reason of targets to accept the acquisition offer which hurts the value of shareholders in the long-run. Shleifer and Vishny (2003) conjecture that target agrees to the acquisition because the targets have shorter horizons in the firm or the self-benefited target managers get private payment from the acquirer. This study intends to test several empirical predictions under the market-driven acquisition theory inspired by Shleifer and Vishny (2003). Previous literature shows that targets are undervalued compared to non-target match firms in cash acquisition and relatively undervalued compared to bidder in stock acquisition. Although bidders usually offer premium price, the premium may be not enough to compensate the undervaluation of targets in cash acquisitions and the overvaluation of bidders in stock acquisitions. It is expected that underpaid targets shareholders will lose money in the long run. Therefore, either the shareholders or the managers have short horizons in the firm. Targets with long-horizon investors and managers will be more likely to resist the offer and thus less likely to complete the acquisition. 2. Research Design 2.1 Main Hypotheses Although bidders are blamed for overpayment in mergers and acquisitions, the overvaluation of bidders and undervaluation of targets is ignored. If taking the misvaluation into consideration, we may conclude that the premium provided by acquirers is not huge enough to compensate the undervaluation of target market value in a cash acquisition and the overvaluation of bidder market value in a stock acquisition. 4

6 Based on the theory of Shleifer and Vishny (2003), I develop several hypotheses for empirical test. First, acquirers will tend to underpay targets in market-driven acquisitions, which should be found empirically. Second, if targets have shorter horizon-oriented shareholders and CEOs, they are more likely to be underpaid. Last, it is expected that firm horizon is also related to the magnitude of underpayment. 2.2 Variable Construction Measurement of underpayment Target value should be measured to see whether the deal value offered by acquirer is appropriate. Two principle valuation approaches are the Price to book (P/B) method and the Residual Income Model (RIM). P/B is the market-to-book ratio of the stock. 1 In the RIM based on Ohlson (1995), firm value is current period book value of equity plus all expected future discounted earnings in excess of required return on the equity capital. Empirically, future earnings are proxied by analyst earnings forecasts (e.g., Lee et al., 1999). Misvaluation, either overvalued or undervalued, is measured by the difference between market price and equity book value or fundamental value resulted from RIM. 2 The definition of underpayment is based on the theory of Shleifer and Vishny (2003). Targets are underpaid if the price is less than target s long-run stand-alone value in cash acquisitions. In stock acquisitions, targets are underpaid if the price is less than targets proportionate value of the short-term post-merger combined firm. To make it more empirically practical, a slightly different measure of underpayment is employed. According to the assumption of Shleifer and Vishny (2003), the misvalued market will return to efficient in the long run. Therefore, the intrinsic value of the target 1 A large amount of literature has demonstrated that P/B ratio is a good proxy to measure mispricing (e.g., Barberis and Huang (2001), Daniel et al. (2001)). 2 Dong et al. (2003) and Ang and Cheng (2005) employ both of RIM and P/B method to measure bidder and target valuation. 5

7 can be measured using the long run market to book ratio. In cash acquisitions, underpayment is the difference between offer price and long-run stand-alone value of the target. In stock acquisitions, underpayment is measured by the difference between original target shareholders value in the combined firm and the stand-alone value in the long run. Here three-year term is used for the long-run base. Since target s stand alone market performance can hardly be traced once merged, target s stand-alone value is measured by the average value of the matching firm without merger announcement. The definition of matching firms is that firms, within the same industry according to Fama-French 38 industry classification, have the same quartile of market-to-book ratio with the target in the quarter prior to the announcement date. In cash acquisitions, it is straight forward that the price offered by acquirers is the real value the targets can get, therefore, underpayment is measured by Underpayment = (P-q)*K, (1) where P is offering price per unit of capital, q is the average price per unit of capital of target s matching firms three years after the acquisition, K is target book equity. In stock acquisitions, however, acquirers pay targets using their overvalued stocks, the offered price is even more misleading especially when the target stocks are undervalued, therefore, the offer price should be adjusted to reflect the real value, Underpayment = (Pq1/Q1-q)*K, (2) where P is offering price per unit of capital, q1 is the average price per unit of capital of combined firm three years after merger completion, Q1 is the market price per unit of capital of acquirers at the end of quarter prior the announcement. Still, q is the average price per unit of capital of target s matching firms three years after the acquisition, K is target book equity. If the value of the equation is less than 0, then the target is underpaid, the more negative the value is, the more underpayment of the target. Otherwise, the target is 6

8 overpaid Proxies for Shareholder Horizon Shareholder daily turnover ratio is used to proxy for shareholder horizon. Turnover is measured as daily trading volume divided by total shares outstanding, multiplying It is commonly stated that the larger the turnover ratio is, the shorter horizon shareholders hold in the firms (e.g. Ang and Cheng (2005), Gaspar et al. (2005)). Turnover may contain too much information in addition to shareholder horizon. For example, firms with larger analyst coverage will have a relatively more active trading pattern. Firms with different size are more likely to attract different groups of investors. To subtract the noise information from turnover, we use the residual mean of target firms turnover ratio after the regression on logarithm of market value of equity, Turnover ji = β 0 + β1 Lg( ME ji ) + ε ji, (3) where Turnover ji is the turnover ratio of firm j on i days before the announcement, Lg( ME ji ) is the logarithm form of market value of firm j on i days before the announcement. The shareholder horizon variable mreone for firm j is constructed as mreone j = ε ji, (4) 365 i= 31 where ε ji is the residual of the regression (3), the period for the mean of residual is chosen as one year before one month prior the announcement of the mergers and acquisitions. We choose the period starting from one month prior the announcement to exclude the abnormal volatility and trading since the rumors of mergers and acquisitions of firms usually occur before the official announcement Proxies for CEO Horizon Age: This study employs two methods to proxy age for CEO horizon. The first method follows Dikolli et al. (2003), which assume that CEO s horizon is a decreasing and 7

9 [ ] 2 concave function of the CEO age. The variable Horizon is defined as ( CEOage /10). In addition, to study the specific impact of different age on CEO horizon in the firm, we also use several dummy variables to separate age to different groups. 3 Tenure: We classify tenure into three groups, no more than 3 years, 3 to 5 years and 5 years more. CEOs with different tenure in their position tend to have different level of career concerns and equity-based wealth in the firms. We use the cutting points of 3 and 5 year based on the study of Gibbons and Murphy (1992), which states that the CEO tenure with the highest possibility of stepping down from the position clusters in 4 and 5 years. It is expected that CEOs with the tenure between 3 to 5 years have the shortest horizons compared to other two groups. One reason is given out by Gibbons and Murphy (1992), CEOs with the tenure between 3 to 5 years are most likely to step down. In addition, CEOs with longer tenure will have more equity-based wealth in their firms. The large stake in their firm can align the interests of CEOs and shareholders, therefore partly solve the horizon problem of CEOs with longer tenure. Equity-based wealth: CEOs equity-based wealth in their firms include value of stock holding, restricted grants money value, in the money unexercised exercisable option dollar value and in the money unexercised unexercisable option dollar value. The wealth is held and valued at the end of last fiscal year prior the announcement of the mergers and acquisitions. Generally, the equity-based wealth, used as a forward-looking compensation contract, will lead CEOs to have a longer horizon. 4 To pair different horizons of CEOs in the acquirer and target, we construct a variable welredif to measure the difference of horizons between acquirer and target CEOs. Welredif is the difference of residuals of CEO equity-based wealth after regression wealth 3 Cheng (2001) uses the age of 63 to distinguish different horizons held by CEO. Survey by Murphy (1999) shows that the typical age of retirement is 64 around. Therefore the age of 64 can also be used as a flag to differentiate horizon. 4 Lambert and Larcker (1991) suggest that contract tying managers compensation to firm s market performance can partly solve the problem of managerial myopia. 8

10 on size of acquirers and targets. The larger the welredif, the acquirer CEO horizon is longer than target CEO. Incentive to cash out: CEOs trapped in the firm by a large holding of restricted stocks have strong incentive to cash out. Since restrictions on stocks and options are usually exempted upon a deal of merger and acquisition (Moeller, 2005), we use the ratio of restricted stock grants value over total equity-based wealth or market value to measure the incentive to cash out. It is expected that CEOs with larger ratio will have shorter horizons in their firms Control Variables Several other variables that may affect the transaction are controlled in the tests. Targets firm size and the relative size to acquirers will affect the deal value. 5 To control the industry factor, we include a dummy variable to indicate whether the merger and acquisition is undertaken by firms in the same industry. The impact of diversification is ambiguous. It is possible that acquirers will pay higher price to targets because they lack perfect information and knowledge to value targets. However, we can also expect that the level of misvaluation is different among industries. Diversified acquisition occurs between more overvalued industry as acquirer and relatively undervalued industry as target. Targets are actually receiving lower price and therefore more likely to be underpaid. Prior literature addresses that firms with higher valuation are associated with greater use of stock as a method of payment (Dong et al., 2003). We add an indicator variable of stock to control the impact of method of payment. Finally, we also include several market return indicators to control the influence of broad market environment. 5 Moeller (2005) finds that the larger the target relative to the acquirer, the lower the takeover premium the target receives. 9

11 2.3 Sample Selection The samples are selected from the mergers and acquisitions database of Securities Data Company (SDC). Transactions with an announcement date between 1993 and 2002 are included in the sample. We choose the period from 1993 to 2002 because CEO information from ExecuComp is available from 1992 and complete three year post-merger market performance from CRSP ends in In addition, Andrade et al. (2001) find that merger waves in different decades have different characteristics. The merger wave in 1990s has notable features distinguished from others. Therefore, here we only focus on the 1990s merger wave. Both the acquirers and targets are U.S. public firms listed in NYSE, Amex or Nasdaq. The sample is limited to listed firms so that the market valuation can be studied. The method of payment is either all cash or all stock. The status of transaction is either completed or withdrawn. Buyback, exchange offer and recapitalization are excluded from the sample. SDC database provides data on the value of transaction, the announcement date, acquirer and target ticker and Standard industrial classification (SIC) codes and status of transaction. Daily stock price, trading volume, and number of shares outstanding are obtained from Center for Research in Securities Prices (CRSP) database. Financial statement information about acquirers and targets is obtained from Compustat Industrial Quarterly. Financial information, including stockholders equity and number of common shares outstanding at the end of each quarter and closing price of stock at the 3 rd month of each quarter, is used to calculate market-to-book ratio of firms at the end of quarter. In addition, we also use all firms registered in Compustat as the base to constructing matching groups. Information of CEO characteristics and compensation is extracted from ExecuComp database, which reports CEO name, age, annual compensation and equity-based wealth including share ownership, restricted grants and options. I also check the proxy statements and other SEC filings to fill the missing values in ExecuComp database and 10

12 collect information on CEO tenure, outside blockholding, etc. Finally, observations in which targets have more than one transactions recorded in SDC are excluded. After all the data collection, the final number of observations in the base sample is Descriptive Statistics Table 1 presents summary statistics for the sample. Of all the observations during 1993 and 2002, most of the transactions cluster in 1996 to 2001, partly subject to data availability. Generally, in a merger and acquisition, acquirer size is much larger than target, with an average acquirer market value of 24, million dollars and target market value of 1, million dollars. Mean deal value of transaction is slightly larger than average target size. In the sample, 58% of all target firms are underpaid by an average value of million dollars. Under the assumption of Shleifer and Vishny (2003), the motivation for rational acquirer managers to undertake mergers and acquisitions in the misvalued stock market is to arbitrage by underpaying targets. Therefore, it is expected that empirically most firms should be underpaid. Result of binomial test shows that percentage of underpayment is larger than 0.5 at the significance level less than Prior literature has stated a greater use of stock as a method of payment in the 1990s merger wave (e.g., Dong et al. (2003) and Holmstrom and Kaplan (2001)), the summary statistics shows the increase of stock payment with a large fall in the The percentage of diversification is 32, which means that over two-thirds of transactions in our sample occur between acquirers and targets in the same industry. This result is consistent with recent studies (e.g., Andrade et al. (2001) and Gaspar et al. (2005)) which report an increase of related acquisition in the 1990s. Finally, table 1 shows that most of the observations in our sample ended with complete. Although we include both completed and withdrawn transactions, only 7% of the sample 11

13 firms are withdrawn acquisitions. <Insert Table 1 here> Further study is focused on firm characteristics related to firm horizon. Table 2 summarizes acquirers and targets characteristics and relative difference between them in different groups classified by underpayment. Panel A and Panel B reports the means of various variables of targets and acquirers respectively. Our hypotheses predict that targets with CEOs who have larger equity-based wealth in the firms will be less likely underpaid, while acquirers with CEOs who have larger equity-based wealth will more likely underpay targets. Therefore, target CEOs in Underpayment are assumed to have smaller equity-based compensation, while acquirer CEOs in Underpayment are assumed to have larger equity-based compensation. The descriptive statistics in Panel A and B show that most variables on CEO compensation and equity-related variables are larger in group of Underpayment than in group of Overpayment for both acquirers and targets. It is mainly due to the reason that firm size has not been controlled in the comparison. Although the dollar value and number of securities holding are ambiguous, statistics on share ownership percentage is consistent with hypothesis. Table 2 shows that CEOs stock ownership percentage is lower for targets and higher for acquirers in the group of Underpayment than Overpayment. The difference for acquirers is significant at 5% level. Statistics on age and tenure reported in Panel A and B shows involvement of CEOs with older age and longer tenure in the overpayment. Since age and tenure are expected to have nonlinear relation with CEO horizon, the implication of group mean difference is unclear. To control the firm size effect and impact of acquirer CEO horizon on target decision, Panel C describes the relative size, age difference and the variable welredif, which is the difference of residuals after regressing CEOs equity-based wealth on firm size between acquirers and targets. Welredif is constructed to measure difference of CEO horizons 12

14 between acquirers and targets. As a result, welredif is positive in Underpayment and negative in Overpayment, with significant difference between two groups. The result indicates that targets will accept underpaid offer when their CEOs have relatively shorter horizon than those of acquirers. However, when targets CEOs have longer horizon than those of acquirers, they will strive for a higher price even resulting in acquirers overpayment. Table 2 reports a summary statistics on all observations in the sample, including both completed and withdrawn transactions. To check the influence of different status of the deal, same test is done for the sample with only completed transactions and the results are similar to Table 2. 6 <Insert Table 2 here> In Table 2, the shareholder horizon variable mreone is significantly higher in targets underpaid then overpaid, which means that shareholders in underpaid firms have shorter horizon. This is consistent with the hypothesis. Table 3 presents a further study on mreone of targets in different classifications. Panel A categorizes all targets into four groups based on the status and whether they are underpaid or overpaid. Shareholder horizon is reasonably expected to affect final status of the transaction. Generally, if an investor has long horizon in the firm and intends to hold onto shares in a long-term period, he will be less likely to accept a merger offer. Taking underpayment and overpayment into consideration, long-horizon oriented shareholders of targets will benefit either by rejecting an underpaid offer or accepting an overpaid offer. Shareholders are expected short-horizon oriented to accept an underpaid offer, with an intention to profit from short-term stock price boost around the announcement. For targets which finally withdraw from transactions with overpaid offer price, shareholders horizons are ambiguous. The order of mreone magnitude in four groups reported in Panel 6 Results are available upon request. 13

15 A is consistent with expectation. Targets underpaid with status of competed have largest group mean of mreone, which indicates that their shareholders have shortest horizon. Targets overpaid and with status of completed and targets which are underpaid but withdraw from the deal have relatively lower group mean of mreone than the other two groups. Group mean of targets which withdraw from an overpaid deal is in between. However, F-statistics for mean difference among the four groups is not significant. Since the major difference is between Group 1 with underpaid and completed targets and other groups, Panel B and Panel C report T-test statistics of the difference between Group 1 and other two groups. It shows that shareholders in targets which competed underpaid deals have shorter horizon than those in targets which competed overpaid deals and which withdraw from underpaid deals, both significant at less than 5% level. <Insert Table 3 here> Shareholder horizon variable mreone in Table 2 and Table 3 is constructed as mean of target s turnover residuals over a period from 31 to 395 days before announcement. To exclude an impact of extreme values during a specific period, Fig. 1 and Fig. 2 describe daily residual means of targets turnover ratio after the regression on logarithm of firms market value of equity over the period of 31 to 395 days prior announcement. By comparing daily mean of underpaid targets and overpaid targets, Figure 1 presents consistently higher daily means of shareholder turnover residuals in underpaid targets than those in overpaid targets in that one year period. It is consistent with our prediction that shareholders in underpaid targets have shorter horizon than in overpaid targets. <Insert Figure 1 here> Figure 2 presents lines of daily mean of shareholder turnover residuals in targets with different status. With only 54 withdrawn observations in the sample, the difference is not so obvious as in Fig. 2. Even though mreone of targets with status of withdrawn have several extreme values in specific days, targets with status of completed still consistently 14

16 present higher daily mreone than targets with status of withdrawn. <Insert Figure 2 here> 3. Regression Analysis 3.1 Target Shareholder horizon and Underpayment Table 4 presents the relation between shareholder horizon and underpayment using different regression models. ModelⅠreports logistic regression results of the likelihood of a target being underpaid based on following model: Pr ob( up β Controls + ε, (5) i = 1) = 0 + β1shareholderhorizoni + β 2 where variable up equals 1 if underpayment is less than 0, which denotes that targets are underpaid. UP equals 0 if underpayment is larger than or equal to 0, which denotes that targets are overpaid. ShareholderHorizon i is the variable measuring target i i shareholder horizon of firm i, proxied by mreone. Controls i is a group of control variables including relative size of targets to acquirers, diversification, target equity value, method of payment and the market environment before announcement. In Model Ⅰ, the coefficient of mreone is significantly positive, which indicates that the higher the shareholder turnover, i.e. the shorter the shareholder horizon, the more likely the target is underpaid. In regard of control variables, it shows that targets are more likely to be underpaid when acquirers are in different industries. It confirms previous prediction that diversified acquisitions occur between more overvalued industry as acquirer and relatively undervalued industry as target. I use logarithm of equity value of target to control target firm size effect and find significant positive relation with the likelihood of underpayment. This is consistent with Schwert (2000) and Gaspar et al. (2005) which find that target premium is negatively related to firm size. Relative size, method of payment, and market return are not statistically significant. Model Ⅱ in Table 4 uses OLS regression to examine the relation between shareholder 15

17 horizon and the magnitude of underpayment based on: Underpay i β β β + ε = 0 + 1ShareholderHorizoni + 2Controlsi i. (6) Dependent variable in Model Ⅱ is underpay, measured by the defined methods. Since underpay is calculated as multiple of target book value of equity, equity value is employed as control variable instead of the logarithm form. Although coefficient for shareholder horizon variable mreone is not significant at 10% level, the sign is as expected. The negative coefficient between mreone and underpay implies that targets with shorter-horizon shareholders are more underpaid. Two statistically significant control variables are rsize and equityvalue. The larger the target size and size relative to acquirer, the more the target is underpaid. Moeller (2005) reports a negative relation between target relative size and takeover premium. Measurement of underpayment is partly determined by takeover premium. Therefore, statistic results of rsize and equityvalue are consistent with prior empirical research. Figure 3 plots the magnitude of underpayment and the quantile position in the overall dataset. Most of the observations gradually increase with the fraction of data and cluster closely around 0. Extreme values exist and drive the scale of underpayment. To reduce the impact of the outliers on the estimation, I transform the value of underpay into cumulative distribution function. 7 By using the percentile of underpay instead of its original value, the scale of underpay is adjusted and more consistent with the independent variables. <Insert Figure 3 here> Model Ⅲ in Table 4 reports estimation of the relation between shareholder horizon and underpayment. Generalized linear model (GLM) is used to control for potential econometric problems such as heteroskedasticity: 7 Aggarwal and Samwick (1999) transform variance of returns into cumulative distribution function to measure CEOs pay-performance sensitivity. 16

18 F( underpay) = β 0 + β1shareholderhorizon i + β 2Controls i + ε i. (7) Dependent variable F(underpay) is the cumulative distribution function (CDF) of underpay. The CDF values of zero and one correspond to the minimum and maximum of underpay in the sample. Mreone is negatively related to the CDF of underpay at 5% significant level. It shows that in the whole sample, targets with shorter horizon shareholders are more underpaid compared to other observations. Firm size and market performance before announcement also impact the level of underpayment. Smaller target size and higher market return increase the percentile of CDF, which means less underpayment. Relative size, diversification and method of payment do not have statistically significant impact on level of underpayment in Model Ⅲ. <Insert Table 4 here> 3.2 Target CEO horizon and Underpayment CEO horizon and Probability of Underpayment To study the relation between CEO horizon and probability of underpayment, similar logistic regression as model (6) is employed: Pr ob( up β Controls + ε, (8) i = 1) = 0 + β1ceohorizoni + β 2 where dependent variable is the probability of underpayment, denoted by the dummy variable of up. CEOhorizon is a vector of target CEO horizon in firm i, Controls is a i group of control variables same as those used in estimating the impact of shareholder horizons. Different sets of variables are included in Table 5 to measure CEO horizon. Model Ⅰ employs two age variables of target CEO to indicate CEO horizons. Following the method of Dikolli et al. (2003) which assume that CEO s horizon is a decreasing and concave function of the CEO age, I expect that CEO horizon has nonlinear relation with age. Therefore, I construct the variable horizon as ( CEOage /10) i i 2 [ ]. Since horizon is a i 17

19 square term of age, variable of age is also added to depict the concave function. In Model Ⅰ, the coefficients for horizon and age are negative, which demonstrates that there is a positive quadratic relation between CEO age and probability of underpayment. Either elder or younger CEOs have shorter horizon and therefore are more likely to accept underpaid offers. CEOs with age in between have the longest horizon. Following the results of Model Ⅰ, Model Ⅱ employs two dummy variables of age. It is expected that the bottom of the curve can be depicted by two cutting points. Consistent with the results reported in Model Ⅰ, results in Model Ⅱ show that target CEOs with age in between are least likely to make their firms underpaid. Two cutting points for age are 65 and 70. The dummy variables age65less and age70up classify CEO age into three groups and make targets with CEO age between 65 and 70 as a base group. The coefficients for two dummy variable age65less and age70up are significantly positive, which denotes that CEOs whose ages are smaller than 65 or elder than 70 are more likely to agree with underpaid deals. Cheng (2001) uses age of 63 to distinguish different horizons, CEOs younger than 63 are classified as holding longer horizon. Dikolli et al. (2003) assume decreasing relation between age and horizon. Model Ⅱ reports different results with prior literature. It is found that CEO horizon increases with age at first stage, and starts to decrease with age after reaching a specific range. Empirical results in Model Ⅱ show that CEOs with age between 65 and 70 have longest horizon in the firm. It is reasoned that CEOs who are elder than usual retirement age but still hold their positions in the firm may have large interests or special power in the firms, and therefore may have longer horizon. 8 However, when CEOs are too old, they may concern more about their current financial gains instead of future careers and long-term interests in the firms. Model Ⅲ of Table 5 examines tenure as proxy for CEO horizon for the probability of 8 Murphy (1999) shows that the typical age for CEOs retirement is 64 around. 18

20 target underpayment. It shows that CEOs with tenure between 3 and 5 years have significantly higher chance of accepting underpaid offers. This is consistent with the survey of Gibbons and Murphy (1992). Since CEOs with tenure between 3 and 5 years are most likely to step down, their career concern is lower. Therefore, they may care more about current financial gains. CEOs with tenure shorter than 3 years usually lack strong control of firm decisions. In addition, they are less likely to agree with the deal for their own career concern 9, which forces acquirers to provide a higher offer price. On the other hand, CEOs that stay in the top position for a very long period either have large stake of interests in the firm or strong control for career security. Both the large amount of equity-based wealth and the career concern lead CEOs to make long-horizon oriented decisions for their firms. Last column in Table 5 employs both age and tenure variable to measure CEO horizon. The results of Model Ⅳ are similar to those of models that only include either age or tenure. Regression results on control variables are consistent across all the models in Table 6 and also consistent with the logistic regression of shareholder horizons. <Insert Table 5 here> Table 6 adds several CEO compensation variables in the logistic regression models. It is expected that the larger CEOs equity-based wealth in the firms, the less likely they will have their firms underpaid in mergers and acquisitions. However, when CEOs hold a large percentage of restricted stock grants, they will have strong incentives to cash out and therefore less reluctant to agree an underpaid offer. Model Ⅰ in Table 6 adds logarithm of target CEO wealth and ratio of restricted stock grants over target size. The negative coefficient of target CEO wealth and positive coefficient of restricted ratio 9 According to the survey of Gibbons and Murphy (1992), it is reasonable to expect that CEOs with shorter tenure than typical retiring tenure have longer period left to stay in their position. However, once the firms are acquired, most of target CEOs can not maintain their position in the combined firms. Hartzell et al. (2004) finds that target CEOs experience high turnover rates both at the announcement of and years after the acquisitions. 19

21 proves the expectation, even though the coefficients are not statistically significant. In the negotiation process of merger and acquisition, both the horizons of target and acquirer CEOs will have impact on the bid price. Long-horizon oriented acquirers and short-horizon oriented targets are more likely to involve in a transaction underpaying targets. To control the impact of acquirer CEOs horizon, Model Ⅱ employs two variables. Welredif in Model Ⅱ is the difference between residuals of CEO equity-based wealth after regressing on firm size of acquirers and targets. Larger welredif indicates that acquirer CEO has longer horizon than target CEO. As shown in Model Ⅱ, the coefficient of welredif is significantly positive, which means that the longer of acquirer CEO horizon than target CEO, the more likely targets are underpaid. Model Ⅲ includes both target CEO equity-based wealth and acquirer CEO equity-based wealth. It presents that larger acquirer CEO s wealth and smaller target CEO s wealth will increase the probability of underpayment. The potential problem of using market value of CEO equity-based wealth is the bias of misvalution. Under the market driven acquisition, targets tend to be relatively undervalued while acquirers tend to be overvalued. Therefore, CEO value of equity-based wealth in the firm based on market price will also be affected. To control the impact of market misvalution on CEO wealth measurement, Model Ⅳ of Table 6 uses ratio of wealth to firm size. Variable weltosize is ratio of target CEO equity-based wealth over target firm size at the end of fiscal year prior the announcement. After dividing wealth by firm size, market value does not impact the measurement of CEO wealth, only the relative stake in the firm matters. The incentive to cash out is proxied by variable of tgresratio, which is measured by restricted stock grants over target size. Guay (1999) states the difference between stock options and stocks. Unexercisable stock options with restrictions on selling are expected to have same effect as restricted grants. The different impact of different kinds of 20

22 equity-based wealth on CEO horizon will probably contradict and mitigate each other. Model Ⅳ adds a new variable of shownpc to separate different effect of stocks from other equity-based wealth. Shownpc is the percentage of common shares held by target CEO. Model Ⅳ reports a significantly positive coefficient for weltosize and negative coefficient for shownpc. After adding a variable of share percentage owned by CEO, the relation between target CEO wealth and probability of underpayment is positive, this is contradictory with previous results on CEO wealth variables. A further study is needed to investigate the impact of stock ownership and other equity-based wealth. Since wealth is the sum of all kinds of equity-based equity value, variable weltosize can be regarded as the sum of share ownership percentage, i.e. shownpc, and stocks underlying options and restricted stocks as a percentage of total shares outstanding. The impact of common stocks owned by CEO should be determined by two coefficients for both weltosize and shownpc. Shownpc is the number of percents. To make the scale of shownpc consistent with weltosize, the coefficients should be multiplied by 100. Therefore, the total impact coefficient in the logistic regression for shownpc is ( * ), which is negatively related to the probability of underpayment. Controlling shares ownership of CEO, the coefficient for other equity-based wealth including restricted stock grants and stocks underlying both exercisable and unexercisable options over total shares outstanding is positive. It shows that CEOs with larger stock ownership are less likely to undersell their firms, however, CEOs will be more likely to agree with underpaid deals when they have large stake in other equity-based wealth. The results in Model Ⅳ denotes that stock ownership can align the interests of shareholders and CEOs and lead CEOs to make long-term oriented decisions. However, the increasing use of stock options and restricted stock grants has the opposite effect, which lends CEOs incentive to cash out and therefore makes them more short-horizon oriented. 21

23 Variables on CEO age, tenure and control variables present almost same effect in the logistic regression model as in Table 5. <Insert Table 6 here> CEO horizon and Level of Underpayment Same as for shareholder horizon, two regression models are employed to examine the relation between CEO horizon and the level of underpayment. Panel A. in Table 7 uses OLS regression to examine the relation between CEO horizon and the magnitude of underpayment based on: Underpay i β β β + ε = 0 + 1CEOhorizoni + 2Controlsi i. (9) Different sets of CEO horizon variables are included in Panel A. The positive coefficients of variables horizon and age depict a first increasing and then decreasing relation between CEO age and underpayment. The larger the value of variable underpay, the more the target is overpaid. Therefore, targets with CEOs aged in between are most overpaid; targets with CEOs either elder or younger are more underpaid. CEOs with tenure between 3 and 5 years make their firms most underpaid. The results are consistent with those reported in logistic regression. However, CEO age and tenure variables are not statistically significant in both models of Panel A. Statistical results on logarithm of target CEO equity-based wealth, acquirer CEO equity-based wealth and welredif are consistent with those in the logistic regression. Variable of tgresratio in Panel A reports contradictory sign to expectation. None of these variables are significant except welredif. Results of control variables show that relative size of targets to acquirers and target size will significantly increase the level of underpayment. In addition, targets involved in diversified transactions, paid by stock and in relatively bearish market are more underpaid. 22

24 Panel B. in Table 7 reports regression results of generalized linear model (GLM). Dependent variable F(underpay) is the cumulative distribution function (CDF) of underpay. F( underpay) = β 0 + β1ceohorizon i + β 2Controls i + ε i, (10) Panel B presents consistent results with Panel A and the logistic regression. Using cumulative distribution function of underpay as dependent variable, Panel B reduces the impact of outliers and makes the dependent and independent variables more comparable in scale. Model Ⅲ to Model Ⅴ report significant statistical results on age and tenure variables. Target CEO ratio of restricted stock grants in Model Ⅲ and Ⅳ presents expected sign. CEOs with larger ratio of restricted stock grants will make their firms more underpaid. Variables on CEO wealth in both targets and acquirers present that larger acquirer CEO s wealth and smaller target CEO s wealth will result in a more underpaid deal for targets. The larger the difference between acquirer and target CEO s wealth, the more the targets are underpaid. Model Ⅴ employs two variables weltosize and shownpc, same as Model Ⅳ in Table 6. The negative coefficient for weltosize is mainly driven by the impact of stock options and restricted stock grants, since the coefficient for stock ownership is positive. <Insert Table 7 here> Correlation between CEO characteristic variables When studying CEO horizons, I employ several CEO characteristic variables such as age, tenure and compensation structures. Since previous literature states that CEO age and tenure are highly correlated with equity-based compensation, 10 Table 8 reports the correlation between different CEO characteristic variables. Panel A of Table 8 presents 10 Lewellen et al. (1987) find that CEOs who are near retirement have a larger fraction of stock-based compensation. Gibbons and Murphy (1992) present that CEOs with longer tenure are paid more by equity-based compensation. 23

25 the correlation matrix among CEO characteristic variables. The magnitudes of correlations do not suggest the problem of multicollinearity. Statistical results of tenure in previous regressions present a curved relation with CEO horizon. CEOs with tenure between 3 and 5 years have the shortest horizons in their firms. Panel B reports mean and median of CEO compensation variables in different tenure groups. It shows that target CEO equity-based wealth, either in logarithm form or as relative ratio to firm size, increases with tenure. Share ownership percentage also increases with tenure. The only variable which presents different trend across tenure groups is tgresratio, calculated by target CEO s value of restricted stock grants over target size. Tgresratio is constructed to measure level of CEO s wealth trapped in the firm by restrictions and incentive to cash out. Group of tenure between 3 and 5 years presents highest mean of tgresratio. The median of tgresratio is same across all three groups and the group difference is not significant. Mean and median difference among different tenure groups demonstrates that the impact of tenure on CEO decision horizons is not driven by compensation structure. The major determinant for CEOs decision horizons in different tenure groups is their career concern. <Insert Table 8 here> 3.3 Target horizon and Underpayment Previous regression models focus either on CEO horizon or on shareholder horizon alone. Table 9 includes both sets of variables on CEO horizons and shareholder horizons to examine the relation between target horizon and underpayment. Panel A. reports logistic regression results of the likelihood of a target being underpaid. Panel B. uses OLS regression to examine the relation between target horizon and the magnitude of underpayment. Panel C. reports regression results of generalized linear model (GLM). Dependent variable F(underpay) is the cumulative distribution function (CDF) of underpay. 24

26 Shareholder horizon variable mreone presents expected sign in all the models. However, none of the coefficients are statistically significant. Statistical results on CEO horizon variables in Table 9 are quite similar to those in Table 6 and Table 7 that only include CEO horizon variables. It posits that the relation between target horizon and underpayment is mainly determined by CEO horizons. <Insert Table 9 here> 4. Conclusion This study examines the hypotheses inspired from Shleifer and Vishny (2003). Using a sample of 770 mergers and acquisitions announced between 1993 and 2002, I find that the number of targets underpaid in the deals is significantly larger than the number of targets overpaid. Target shareholder horizon, denoted by mean of daily turnovers over one year period before the month of announcement after controlling firm size effect, has negative relation with probability of underpayment. CEO horizon is measured by age, tenure in that position, equity-based wealth in the firm, and incentives to cash out. It is found that CEOs have short horizon when they are either younger than 65 or older than 70, or when their tenure is between 3 to 5 years. Equity-based wealth can align the interests of shareholders and CEOs and reduce the probability of underpayment. However, the overuse of restricted stock grants will induce CEOs to seek cashing out opportunities and agree with an underpaid offer. This study mainly focuses on target side horizon. It is based on the assumption that mergers and acquisitions are a form of arbitrage by rational acquirers managers. Since the benefit of acquirers for the transaction is in long-term, it is reasonable to expect that acquirers horizon also plays an important role in the deal negotiation. Here I only employ acquirer CEOs equity-based wealth to denote acquirer horizon. Further study can be done 25

27 by using more proxies for shareholder and CEO horizons in acquiring firms. 26

28 References: Andrade, G., Mitchell, M., and Stafford, E., New evidence and perspectives on mergers. Journal of Economic Perspectives 15, Ang, J., and Cheng, Y., Direct evidence on the market-driven acquisition theory. Journal of Financial Research, forthcoming. Baker, M., and Wurgler, J., The equity share in new issues and aggregate stock returns. Journal of Finance 55, Barberis, N., and Huang, M., Mental accounting, loss aversion, and individual stock returns. Journal of Finance 56, Callahan, T., and Moeller, T., Who s cheating whom in mergers and acquisitions? How managerial preferences influences attitude, target choices, and payoffs. Working paper, University of Texas-Austin. Cheng, S., The relation between R&D expenditures and CEO compensation. Working paper, University of Pittsburgh. Daniel, K., Hirshleifer, D., and Subrahmanyam, A., Overconfidence, arbitrage, and equilibrium asset pricing. Journal of Finance 56, Datta, S., Iskandar-Datta and Raman, K., Executive compensation and corporate acquisition decisions. Journal of Finance 56, Dikolli, S., Kulp, S., and Sedatole, K., The role of CEO and investor horizons in the contracting use of forward-looking performance measures. Harvard NOM research paper No D Mello, R., and Shroff, P., Equity undervaluation and decisions related to repurchase tender offers: an empirical investigation. Journal of Finance 55, Dong, M., Hirshleifer, D., Richardson, S., and Teoh, S., Does investor misvaluation drive the takeover market? Working paper, Ohio State University. Gaspar, J., Massa, M., and Matos, P., Shareholder investment horizons and the market for corporate control. Journal of Financial Economics, forthcoming. Gibbons, R., and Murphy, K., Optimal incentive contracts in the presence of career concerns: theory and evidence. Journal of Political Economy 100, Grinstein, Y., and Hribar, P., CEO compensation and incentives: evidence from M&A bonuses. Journal of Financial Economics 73, Harford, J., What drives merger waves? Journal of Financial Economics 77, Hartzell, J., Ofek, E., and Yermach, D., What s in it for me? CEOs whose firms are acquired. The Review of Financial Studies 17, Holmstrom, B., and Kaplan, S., Corporate governance and merger activity in the U.S.: Making sense of the 1980s and 1990s. Working paper, MIT. Holthausen, R., Larcker, D., and Sloan, R., Annual bonus schemes and the manipulation of earnings. Journal of Accounting and Economics 19, Jenson, M., Takeovers: their causes and consequences. Journal of Economic Perspectives 2,