Lower the basket for easy shots? Expectation management before takeovers *

Transcription

1 Lower the basket for easy shots? Expectation management before takeovers * JIE (JACK) HE TINGTING LIU TAO SHU January 2014 * Jie (Jack) He, Tingting Liu, and Tao Shu are at Terry College of Business, University of Georgia. He s is jiehe@terry.uga.edu. Liu s is ttliu@uga.edu. Shu s is taoshu@terry.uga.edu. We are grateful for helpful comments from Stuart Gillan, Sara Holland, Harold Mulherin, Jeff Netter, Bradley Paye, Julie Wu, Fei Xie, and seminar participants at the University of Georgia. We remain responsible for any remaining errors or omissions.

2 Lower the basket for easy shots? Expectation management before takeovers January 2014 Abstract Takeover bidders in stock-for-stock mergers have a strong incentive to boost their own stock prices to lower the cost of acquisition. We find strong evidence that bidders manage down analyst earnings forecasts to push up earnings surprises and thus their own stock prices prior to mergers. This finding cannot be explained by analyst pessimism or omitted variables that simultaneously drive expectation management and mergers. Analysts who participate in expectation management tend to have a stronger relation to the bidder and obtain benefits from producing more accurate earnings forecasts. Investors do not adjust bidders merger announcement returns for pre-merger expectation management, but bidders long-term returns in the post-merger period are significantly lower in the presence of pre-merger expectation management. Overall, our findings identify expectation management as a previously underexplored opportunistic behavior by takeover bidders.

3 Mergers and acquisitions are among the most important events in the corporate world. They not only directly affect investment returns, governance, and future growth of the involved parties, but also have a profound impact on the reallocation of corporate resources within an economy as well as the competitive landscape of related industries. Hence, understanding the behaviors of bidder and target firms has always been a central question in corporate finance (e.g., Mitchell and Mulherin 1996; Shleifer and Vishny 2003; Masulis, Wang, and Xie 2007; Cai, Song, and Walkling 2011). In this paper, we examine takeover bidders practices of expectation management, a previously underexplored opportunistic behavior to push up their pre-merger stock prices. It is known that bidder firms that buy target firms with their stocks (henceforth stock mergers) have a strong incentive to boost their own stock prices prior to the acquisitions. An increase in a bidder s stock price can reduce the number of the bidder s shares used for acquisition and therefore lower the bidder s cost of acquisition. As a result, bidder managers may engage in various opportunistic behaviors to push up their stock prices before mergers. One particular way of doing so is through expectation management, in which the bidder manages down the market expectation about its premerger earnings to increase the corresponding earnings surprises which results in a stock price jump. Our study is motivated by a growing literature on expectation management. This literature suggests that because earnings announcement is a predictable event associated with tremendous investor attention and strong market responses, corporate managers may attempt to increase earnings surprises by pushing analyst forecasts downward prior to earnings announcements. 1, 2 The 1 Consistent with this idea, Soffer, Thiagarajan, and Walther (2000) find that managers tend to release bad earnings news through pre-announcement shortly before the formal earnings announcement date to lower investor expectation and minimize the potential negative market reaction subsequently. Similarly, managers tend to withhold a significant part of the positive earnings news until the formal earnings announcement date. 2 Compared to earnings surprise, adjustment in analyst forecast is a less salient event. The motivation of pushing analyst forecasts downward prior to earnings announcement is also consistent with the argument that owing to limited investor attention, prices do not fully impound relevant public information, especially when such information is less salient. (e.g., Klibanoff, Lamont, and Wizman 1998; Huberman and Regev 2001; DellaVigna and Pollet 2009; Hirshleifer, Lim, Teoh 2009; Hirshleifer, Hsu, and Li 2013). 1

4 seminal paper of Matsumoto (2002) develops a novel measure of unexpected analyst forecast and shows that firms manage analyst forecast downward prior to earnings announcement to ensure positive earnings surprise. Since Matsumoto (2002), a number of studies have documented expectation management by corporate managers to boost earnings surprise (e.g., Bartov, Givoly, and Hayn 2002; Richardson, Teoh, and Wysocki 2004). Das, Kim, and Patro (2011) find that the benefit of expectation management, measured by the increase in stock price on earnings announcement date due to positive earnings surprise, outweighs the cost of expectation management, measured by the potential decreases in stock price at the time of downward forecast revision and on earnings announcement date when investors detect expectation management. Koh, Matsumoto, and Rajgopal (2008) show that since the accounting scandals and regulations in early 2000s, corporate managers propensity to use earnings management via discretionary accruals has decreased but their propensity to use expectation management has increased. In this paper, we examine whether bidder firms engage in expectation management, and in particular, whether they manage down analyst forecasts to pump up earnings surprises and thus their stock prices prior to stock mergers. Our sample includes 1,035 mergers and acquisitions between 1996 and We divide the sample into stock mergers, cash mergers (takeovers that use cash as the only method of payment), and mixed mergers (takeovers that use both cash and stock as the methods of payment). While bidders of stock mergers have strong incentives to engage in expectation management, bidders of cash mergers have very few or virtually no such incentives as they do not directly benefit from an increase in their pre-merger stock prices. Moreover, compared to bidders of stock mergers, bidders of mixed mergers have weaker incentives to adopt expectation management as their own stocks only account for part of the deal payments. As a starting point, we examine the evolvement of bidders consensus analyst forecasts for the last quarterly earnings release prior to merger announcement (henceforth Quarter -1). For stock 2

5 mergers, the bidders consensus analyst forecasts monotonically decrease over the one-year period prior to the merger announcements, suggesting that analysts adjust their forecasts for Quarter -1 earnings downward as the mergers get closer. By contrast, this downward adjustment in analyst forecasts is much smaller in magnitude for mixed mergers and cash mergers, which is consistent with weaker incentives of expectation management in mixed mergers and cash mergers. To rule out the possibility that the above results are driven by analysts pessimistic view about bidders future prospects, we further examine the evolvement of consensus analyst recommendations for bidders prior to merger announcements. For stock mergers, bidders analyst recommendations become increasingly optimistic in the one-year period prior to the merger announcement. Meanwhile, the analyst recommendations during the same window before merger announcements are almost flat for cash mergers, and only slightly improve (i.e., become more optimistic) for mixed mergers. The stark contrast between analyst forecasts and analyst recommendations suggests that the lowered analyst forecast in stock mergers is not likely to be driven by analyst pessimism. Next, we employ a formal measure of expectation management developed by previous studies. Specifically, we follow the literature (Matsumoto 2002; Koh, Matsumoto, and Rajgopal 2008) and construct a measure of unexpected analyst forecast, which is the difference between the actual consensus analyst forecast and the expected consensus forecast (estimated using industry, past earnings growth, and past stock returns). A negative value of unexpected forecast indicates expectation management. We then analyze unexpected forecasts in the one-year period prior to merger announcements, because bidders merger talks with potential target firms generally take place several quarters prior to the formal merger announcements (e.g., Boone and Mulherin 2007, 2008). For stock mergers, we observe significantly negative unexpected forecasts for bidders quarterly earnings in the one-year period prior to merger announcements, suggesting that bidders actively 3

6 engage in expectation management in stock mergers. Consistent with weaker incentives to use expectation management for mixed deals and cash deals, the negative unexpected forecasts prior to merger announcements are much smaller in magnitude for mixed deals, and are neither statistically or economically significant for cash deals. These results hold for both univariate analysis and multivariate analysis when we control for various deal and firm characteristics as well as year and industry fixed effects. To complement the analysis of expectation management, we also revisit the previously documented earnings management by bidders in stock mergers (Erickson and Wang 1999). We estimate quarterly discretionary accruals using the modified Jones model (Dechow, Sloan, and Sweeney 1995), and calculate abnormal accruals by further adjusting for historical performance (Kothari, Leone, and Wasley 2005). The quarterly abnormal accruals in the one-year period prior to stock mergers are positive, but they are significant at the 10% level for only two quarters and insignificant for the remaining quarters. This result suggests that takeover bidders incentives to engage in earnings management seem to weaken in our sample period of , consistent with the findings of Koh, Matsumoto, and Rajgopal (2008) that corporate managers in general tend to use less accrual management in recent years. It also highlights the importance of expectation management as an alternative way to pump up bidder s stock price during this period. While our findings show that bidders engage in expectation management in stock mergers, one may argue that expectation management is not triggered by stock mergers but instead caused by some unobserved characteristics of the bidder firms in stock mergers. This concern is alleviated by our controlling for a broad set of firm characteristics in our multivariate analysis. However, since it is impossible to exhaustively control for all the firm characteristics, we design three additional tests to address the endogeneity issue. First, we examine a subsample of bidders that conducted both stock mergers and cash mergers. We find that the unexpected forecasts for the same bidder are significantly 4

7 lower in its stock deals than in its cash deals. Second, we create pseudo events by shifting the merger announcements back by four quarters or eight quarters. We find little evidence of expectation management by bidder firms in the pseudo event window. Third, we directly include firm fixed effects in the regressions of unexpected forecasts, and our findings remain the same. These results provide further evidence that the observed expectation management is unlikely to be driven by unobserved firm characteristics. Our evidence on expectation management raises a natural question: Why would analysts cooperate with bidder firms and adjust down their earnings forecasts prior to mergers? Ke and Yu (2006) attribute the analysts involvement in expectation management to conflicts of interest. They find that analysts involved in expectation management are able to produce more accurate earnings forecasts and are less likely to be fired by their employers in the subsequent period. We therefore conduct two tests to examine whether the conflict-of-interest explanation also applies to our setting. We first link expectation management to the strength of relations between analysts and bidders. To the extent that analysts with a closer relation to bidders are subject to more severe conflicts of interest, they would be more likely to participate in expectation management. We measure the analyst-bidder relation using the total number of historical forecasts issued by an analyst for a takeover bidder. A larger number of forecasts indicate that the analyst is likely to have followed the bidder for a longer period of time and therefore have established a closer relation with the bidder managers. A larger number of forecasts also indicate that the analyst is likely to issue forecasts for various dimensions of the bidder and therefore can benefit more from a returning favor from the bidder managers in terms of improved forecast accuracy. 3 Consistent with the 3 In untabulated analysis, we also define the relation between an analyst and the bidder as the number of years he or she has been covering the firm. Our results are qualitatively similar if we use this alternative definition of analyst-firm relation. 5

8 conflict of interest hypothesis, we find that the analysts with a closer relation to the bidders are associated with greater downward bias in their earnings forecasts. To further explore the conflict-of-interest explanation, we also follow Ke and Yu (2006) and examine the accuracy of analyst forecasts. We find that, consistent with Ke and Yu (2006), analysts who adjust their earnings forecasts downward prior to bidders earnings announcements indeed produce significantly more accurate earnings forecasts. This finding is consistent with analysts participating in bidders expectation management in exchange for improved forecast accuracy. 4 Next, we examine market reactions to bidders expectation management practices. The results show that investors do not adjust bidders merger announcement returns downward in response to pre-merger expectation management practices. By contrast, bidders merger announcement returns are significantly lower in the presence of earnings management prior to stock mergers. These results indicate that on merger announcement dates, investors adjust for bidders earnings management in stock mergers but fail to do so for bidders expectation management. Finally, we examine the relation between expectation management and bidders long-term returns after takeovers. If bidders use expectation management as an opportunistic behavior to push up their prices temporarily prior to stock mergers and the market is reasonably efficient in the long run, then we would expect that this effect on bidders prices will reverse after mergers and result in lower long-term returns for bidders after mergers. We find that this is indeed the case. The long-run abnormal stock returns after the stock mergers are significantly lower in the presence of pre-merger expectation management. Our paper contributes to the literature on mergers and acquisitions by identifying expectation management as a way to boost bidders stock prices prior to stock mergers. Because of 4 It is also possible that bidders takeover advisors make their analysts participate in bidders expectation management in return for investment banking business. Inconsistent with this explanation, we find no difference in the extent of expectation management between the analysts affiliated with bidders advisors and the unaffiliated analysts. 6

9 the importance of mergers and acquisitions and the strong economic incentives associated with them, previous studies have examined a number of opportunistic behaviors by takeover bidders. 5 However, none of the existing studies have examined expectation management by takeover bidders. Hence, our paper fills in the gap and complements the above literature by providing evidence on a previously underexplored opportunistic behavior by bidder firms. Further, our sample is by far the most comprehensive one to examine bidders opportunistic behavior before mergers and acquisitions. 6 Our findings also extend the literature on expectation management, which has focused on whether managers in general guide down analyst forecasts to meet or beat the earnings targets. We are the first study to investigate expectation management in major corporate events. Our paper offers strong evidence that managers use expectation management to benefit from corporate events such as stock mergers. While our study focuses on mergers and acquisitions, future studies can extend our research framework to examine expectation management in other corporate events where managers incentives of engaging in opportunistic behavior are also strong. The outline of our paper is as follows. Section I discusses related literature. Section II describes sample selection and variable construction. Section III examines expectation management and earnings management in mergers and acquisitions. Section IV studies the relation between expectation management and analyst characteristics. Section V analyzes market responses to expectation management and earnings management, and Section VI concludes. 5 For example, Erickson and Wang (1999) and Louis (2004) reveal that bidder firms in stock mergers manage earnings upward prior to mergers via aggressive usage of discretionary accruals. Kolasinski and Kothari (2008) find that relative to unaffiliated analysts, analysts affiliated with acquirers are more likely to upgrade their stock recommendations of the acquirers. Ahern and Sosyura (2013) find that bidders in stock mergers originate substantially more news stories. 6 For comparisons, the samples of Erickson and Wang (1999) and Louis (2004) include 119 and 373 deals, respectively, while our sample includes 1,035 deals. 7

10 I. Literature Review A. Literature on Expectation Management Earnings announcement is a predictable event associated with great investor attention and strong market responses. As a result, corporate managers may attempt to increase earnings surprises by pushing analysts earnings forecasts downward prior to earnings announcements. This intuition is reflected by the following discussion of a Wall Street Journal article (McGee 1997): (The head of quantitative research at Merrill Lynch observes,) Investor-relations people have been making sure the hurdle of expectations remains low so their companies can clear it easily. This intuition is also reflected by Soffer, Thiagarajan, and Walther (2000), who find that managers with poor earnings news tend to release such information through pre-announcement shortly before the formal earnings announcement date to lower investor expectation. In contrast, when managers with favorable news make pre-announcement, they only release half of the news to ensure good market response to formal earnings announcement. Despite the early anecdotal evidence, it is not until recently that researchers have formally begun to examine expectation management. Matsumoto (2002) is the first comprehensive analysis of expectation management. She develops a measure of unexpected analyst forecast and shows that corporate managers push analyst forecasts downward prior to earnings announcements to ensure positive earnings surprises. The findings in Matsumoto (2002) are later confirmed by Burgstahler and Eames (2006). Richardson, Teoh, and Wysocki (2004) show that expectation management is most pronounced when firms or insiders are net sellers of stocks after an earnings announcement, suggesting that expectation management can be caused by capital market incentives. Previous studies also examine whether investors are aware of expectation management and undo the effect of expectation management. Bartov, Givoly, and Hayn (2002) find that investors seem to correct expectation management by adjusting earnings announcement returns downward, 8

11 but the correction is only partial. Das, Kim, and Patro (2011) further illustrate that the increase in stock price caused by expectation management outweighs the downward price adjustment by investors in the presence of expectation management. Koh, Matsumoto, and Rajgopal (2008) show that since the accounting scandals in early 2000s, corporate managers propensity to use earnings management via accruals has decreased but their propensity to use expectation management via downward-adjusted analyst forecasts has increased. Finally, a number of studies examine analysts incentives with respect to expectation management. Cotter, Tuna, and Wysocki (2006) observe that managers issue earnings guidance more often when initial analyst forecasts are optimistic and when forecast dispersion is low. Analysts respond quickly to earnings guidance and are more likely to issue final meetable or beatable forecasts (i.e. earnings targets ) when firm management provides such guidance. These results suggest that analysts can be influenced by managers. Malmendier and Shanthikumar (2007) report that affiliated analysts issue the most optimistic stock recommendations but issue the most pessimistic forecasts prior to earnings announcement. They conclude that affiliated analysts distort earnings forecasts downward prior to earnings releases. Ke and Yu (2006) conduct a comprehensive study of the incentives of analysts involved in expectation management. Consistent with the conflict of interest hypothesis, analysts who make downward adjustments to their earnings forecasts produce more accurate subsequent earnings forecasts and are less likely to be fired by their employers. Their findings suggest that the downward bias in analyst forecasts is due to an exchange of benefits between analysts and managers. B. Literature on Opportunistic Behaviors by Takeover Bidders Researchers have long noticed that bidder firms in stock mergers have a strong incentive to boost their own stock prices, which can potentially cause opportunistic behaviors by bidder 9

12 managers. As a result, a number of studies have investigated and found evidence of several opportunistic behaviors by bidders in stock mergers. Over the sample period between 1985 and 1990, Erickson and Wang (1999) document that bidders in stock mergers manage earnings upward in the periods prior to the merger agreement. They find that acquiring firms have positive accruals as early as three quarters before the merger. They use cash mergers as a control group and find that bidders in cash mergers do not manage earnings via accruals. Louis (2004) also finds that acquiring firms overstate their earnings in the quarter preceding a stock merger announcement. He further shows that the post-merger long-term return is negatively related to the pre-merger earnings management, which is consistent with earnings management causing overpricing for bidder firms. Researchers have also examined bidders opportunistic behaviors other than earnings management. For example, Kolasinski and Kothari (2008) examine analyst recommendations in mergers and acquisitions. They find that analysts affiliated with acquirers are more likely than unaffiliated analysts to upgrade their recommendations of the acquirers. Ahern and Sosyura (2013) find that bidders in stock mergers dramatically increase the number of press releases disseminated to financial media prior to merger announcements. They conclude that firms issue press releases as a method to raise their stock value temporarily by generating more media coverage. II. Sample Selection and Constructions of the Measures A. Sample Selection The sample examined in this paper includes mergers and acquisitions (M&As) between U.S. public firms during the period of , obtained from the Securities Data Company s (SDC s) Mergers and Acquisitions database. We require an M&A deal to be announced between 1996 and 2011, with a deal (transaction) value of above $20 million, and with both the bidder and the target 10

13 being publicly traded companies. 7 We further require the bidder to acquire at least 50% of the target s shares, and to have available stock price information from CRSP, accounting data from Compustat, and analyst forecast information from the Institutional Brokers Estimate System (I/B/E/S). We drop the deals where the target share price one day prior to the announcement date is below $5 and those deals without sufficient information about the methods of payment. The sample selection process ends up with 1,035 mergers and acquisitions. Table 1 reports the distribution of the M&A deals in our sample. Year denotes the calendar year of the deal announcement. Stock deals use stocks as the only method of payment; Cash deals use cash as the only method of payment; and Mixed deals use both stocks and cash as methods of payment. Although the deals are more or less evenly distributed over our sample period, most of them are concentrated in the first half of our sample, with a peak in the internet bubble period ( ), which is consistent with existing literature. There are a total of 294 stock deals, 377 cash deals, and 364 mixed deals in our sample. B. Variable Construction and Summary Statistics Table 2 presents the summary statistics of the deal and firm characteristics in our sample. CAR is the cumulative abnormal returns for bidders in the three-day window [-1, 1] surrounding a deal announcement, where day 0 is the announcement date. Daily abnormal returns are calculated as daily stock returns in excess of the daily CRSP value-weighted market returns. DealSize is the deal (transaction) value reported by SDC. MktCap is the bidder s market capitalization for Quarter -1, where Quarter -1 is the last quarter with an earnings announcement date preceding the merger announcement date as plotted in Figure 1. RSize is the ratio of DealSize to MktCap. MarketBook is the 7 The literature to date has no consensus on the size filter that should be used in merger analysis. For example, Boone and Mulherin (2007) examine deals with at least $50 million but Kisgen, Qian, and Song (2009) use $5 million as the cutoff. To minimize sampling bias, we apply a moderate size filter ($20 million) to our analysis. Using other size filters such as $5 million or $50 million does not affect our main results. Furthermore, we also directly control for deal size in our multivariate analysis. 11

14 ratio of the bidder s market value of equity to its book value of equity for Quarter -1. ROA of a bidder is its income before extraordinary items for Quarter -1 divided by its lagged total assets. Panel A of Table 2 reports summary statistics for deals in the full sample. Consistent with the existing literature, the 3-day announcement returns (CAR) for the merger bidders in our sample have a negative mean (-1.49 percent) and median (-0.76 percent), though the standard deviation is large (7.67 percent). An average deal in our sample has a deal value (DealSize) of $2.2 billion. An average bidder in our sample has market capital (MktCap) of $21.5 billion, market-to-book ratio (MarketBook) of 4.48, and return on assets (ROA) of 1.75 percent. On average, the deal value is 60.8 percent of the market capital of the takeover bidder. Panel B reports the means and medians of firm and deal characteristics for deals in the subsamples of different payment methods. The last two columns compare cash and stock deals. The column T-test reports the two-tail t-statistics of two-sample T-tests comparing the means of cash and stock deals. The column Wilcoxon reports the two-tail p-values of the two-sample Wilcoxon Rank-sum tests comparing cash and stock deals in a non-parametric way. Consistent with previous studies, the bidders of stock deals have big negative announcement returns (CAR) while those of cash deals have slightly positive CARs, and the difference between the two groups is statistically significant. The average CARs for bidders of mixed deals are also negative, but the magnitude of them is much smaller (i.e., less negative) than that of all-stock deals. These announcement returns are similar to those in previous studies (e.g., Travlos 1987; Houston and Ryngaert 1997; Fuller, Netter and Stegemoller 2002; Mitchell, Pulvino, and Stafford 2004). On average, stock deals are larger than cash deals in terms of deal (transaction) value, though the difference is only marginally significant. Compared to an average bidder of stock deals, an average bidder of cash deals is larger in terms of market capital, has a lower market-to-book value of equity, and has similar ROA. 12

15 III. Expectation Management by Takeover Bidders This section analyzes whether takeover bidders use expectation management to manipulate their own stock price before and during their merger talks with potential target firms, which generally take place several quarters prior to the formal merger announcements (e.g., Boone and Mulherin 2007, 2008). Previous studies have documented earnings management by takeover bidders via aggressive use of discretionary accruals. However, a more hidden way for a bidder to pump up its own pre-merger stock price is through expectation management, in which the bidder manages down the market expectation about its earnings to increase earnings surprises and realize a jump in stock price. Given the covert nature of expectation management by bidders and its lack of attention in the literature, we first establish the case of expectation management by bidders, and then formally test it using a measure of expectation management developed by the existing literature. We examine three takeover categories as the incentive of expectation management differs across them. While bidders of stock mergers have strong incentives to engage in expectation management, bidders of cash mergers have very few or virtually no incentives to use expectation management as they do not directly benefit from an increase in pre-merger stock price. Moreover, compared to bidders of stock mergers, bidders of mixed mergers have weaker incentives to engage in expectation management as their own stocks only account for part of their payments. Therefore we predict that expectation management, if any, will be stronger in stock mergers than mixed or cash mergers. A. Evolvement of Analyst Forecasts Prior to Merger Announcements As a starting point, we examine the evolvement of consensus analyst forecasts for bidder s earnings in the last quarter prior to merger announcement (Quarter -1). The consensus analyst forecast is the monthly mean earnings forecast reported by I/B/E/S, scaled by the bidder s stock price on the day I/B/E/S reports the consensus forecast. For each of the three merger categories, 13

16 Figure 2 plots average monthly consensus analyst forecasts for Quarter -1 earnings in the twelve months up to Month -1, where Month -1 is the calendar month immediately prior to the month of the earnings release of Quarter -1. We first obtain consensus forecasts for each deal and then report the average consensus analyst forecast across the deals in each merger sample. Figure 2 shows that for stock mergers, bidder s consensus analyst forecast for Quarter -1 earnings monotonically decreases, indicating that analysts adjust their forecasts downward as earnings announcement is approaching. There is also a slight decline in consensus analyst forecast for mixed and cash deals but with a much smaller magnitude than stock mergers. Table 3 presents the corresponding forecast numbers. The bidder s consensus forecast monotonically decreases from above 1.1% to below 0.9% (a decrease by 18.9%) over the one-year period prior to the merger announcement date. In contrast, the downward trend in analyst forecasts is much more modest for mixed deals (from 1.3% to 1.2%, or a decrease by 8.5%) and cash deals (from 1.4% to 1.3%, or a decrease by 7.7%), consistent with our prior that relative to cash deals, bidders of stock deals have the strongest incentives to manage down analyst forecasts before their takeover attempts. 8 However, while the decreasing trend in analyst forecasts prior to stock mergers is consistent with expectation management, it may also result from analysts negative private information or their pessimism about the future prospects of stock bidders. To address this concern, we examine the trend of consensus analyst recommendations for bidders during the same window. Monthly consensus analyst recommendation for a bidder, obtained from the I/B/E/S database, is calculated as the mean of all outstanding analyst recommendations in a given month. Analyst recommendations take on one of the five values: 1 = Strong Buy; 2 = Buy; 3 = Hold; 4 = Underperform; 5 = Sell. We calculate consensus recommendations for the twelve months ending in 8 For robustness, we also repeat this test using raw analyst forecasts (not scaled by stock prices), and observe similar results (untabulated). 14

17 Month -1, where Month -1 is the calendar month immediately prior to the month of the earnings release of Quarter We then report the average consensus analyst recommendation for each month across the deals in our sample. Figure 3 shows that for stock mergers, bidders analyst recommendations become more optimistic as the merger announcement date gets closer. In the meantime, the analyst recommendations remain almost steady for cash deals and only slightly improve (i.e. become more optimistic) for mixed deals. The corresponding numbers in Table 4 also reflect these patterns. To summarize, the sharp contrast between analyst forecasts and analyst recommendations for stock deals during the same window suggests that the decreasing trend in analyst forecasts identified by Table 3 is not likely to be driven by analysts pessimism or their negative private information about the bidders future prospects. B. Unexpected Analyst Forecasts Prior to Merger Announcements While the evidence so far indicates that the takeover bidders may have pressured analysts to adjust down their forecasts prior to their merger attempts, the raw analyst forecasts that we have examined may merely reflect unobserved firm and industry characteristics rather than the opportunistic behavior of merger bidders. To address this concern, we formally test the presence of expectation management in mergers using a measure of unexpected analyst forecast developed by the previous studies. We follow the literature (e.g., Matsumoto 2002; Koh, Matsumoto, and Rajgopal 2008) and construct a measure of expectation management based on unexpected analyst forecast. The idea is that we first construct an expected (or target ) level of a firm s earnings based on its industry, past earnings growth, and past stock price performance, and then measure the unexpected 9 In untabulated analysis, we also examine the monthly consensus recommendation in the one-year period prior to the merger announcement month (rather than Month -1 defined above) and find similar results. 15

18 consensus analyst forecast as the difference between the actual consensus analyst forecast and the expected level of the firm s earnings. For a given firm-quarter, we follow the literature and construct the expected change in earnings per share (EPS) using the model below: (1) where is the earnings per share of firm i for quarter q minus the earnings per share of firm i for quarter q 4; is the stock price of firm i for quarter q 4; is the earnings per share of firm i for quarter q 1 minus the earnings per share of firm i for quarter q 5; is the stock price of firm i for quarter q 5; and is the cumulative daily excess return of firm i in the one-year window up to the announcement date of quarter q earnings. We estimate equation (1) for each four-digit SIC industry j and year t. We then construct the unexpected analyst forecast using the following equation:, (2) where is firm i s unexpected analyst forecast for quarter q earnings; is firm i s consensus forecast for quarter q earnings in the month prior to earnings announcement; is the earnings per share of firm i in quarter q 4. The coefficients for firms in equation (2) are estimated using equation (1) in the previous year. To facilitate cross-firm comparisons, we scale the unexpected forecast,, by stock price on the summary date of the consensus forecast. A negative value of unexpected analyst forecasts indicates expectation management. Table 5 presents average unexpected forecasts for bidders five quarterly earnings announcements (Quarters -5 to -1) prior to the merger announcements for the three takeover categories. We first estimate unexpected consensus forecasts for each bidder, and then report means and medians across the deals in each category. For stock deals, we find strong evidence for 16

19 expectation management as both the mean and median unexpected consensus forecasts are significantly negative for each of the five quarterly earnings prior to merger announcement. Consistent with weaker incentives of expectation management for mixed deals and especially cash deals, the negative unexpected forecasts prior to merger announcements are much smaller in magnitude for mixed deals, and become both statistically and economically insignificant for cash deals, especially in the last three quarters prior to merger announcement. We also directly compare the mean expected forecasts between stock deals and cash deals. The results based on T-test and Wilcoxon p-values show that the unexpected forecasts for stock deals are significantly more negative than those of cash deals in the last three quarters. These results are consistent with the stronger incentives to engage in expectation management by bidders in stock deals than in cash deals. C. Earnings Management in Mergers To complement the analysis of expectation management, we also examine earnings management by bidders using the measure of abnormal accrual. Following the literature, we calculate a firm s quarterly discretionary accrual as the residual of the following regression:, (3) where is the quarterly total accrual (the change in non-cash current assets minus the change in current liabilities plus the change in debt in current liabilities) for firm i; Q j is a dummy variable that equals one if the quarter is the jth (j = 1, 2, 3, or 4) quarter in the fiscal year and zero otherwise; is the quarterly change in sales for firm i; is the quarterly change in accounts receivable for firm i; and is the error term. All the variables are scaled by assets at the preceding quarter. The model is estimated for each industry-year using quarterly data for all firms in the same two-digit SIC industry. We further follow Kothari, Leone, and Wasley (2005) and construct abnormal accruals by adjusting discretionary accruals for past performance. Specifically, for each 17

20 quarter q, we sort firms in a two-digit SIC industry into quintiles according to their ROA in q - 4, and calculate a firm s abnormal accrual as its discretionary accrual minus the average discretionary accrual of all the other firms in the same ROA quintile. Table 6 reports bidder s abnormal accruals for the five quarters before merger announcement. We first calculate the abnormal accruals for each bidder, and then report means and medians across the deals in each group. Consistent with the existing literature (e.g., Erickson and Wang 1999), we observe positive quarterly abnormal accruals in the one-year window (from quarter -4 to quarter -1) prior to stock mergers. However, the abnormal accruals are only statistically significant at the 10% level for Quarter -4 and Quarter -1, and insignificant for the rest of the quarters. Meanwhile, the abnormal accruals for cash and mixed deals are either negative or insignificantly positive. The differences in mean accruals between stock and cash deals are statistically significant in three of the four quarters prior to merger announcements. These results are consistent with earnings management by takeover bidders. In the meantime, the marginally significant abnormal accruals seem to suggest that earnings management prior to mergers is less pronounced in our sample period of This pattern is consistent with Koh, Matsumoto, and Rajgopal (2008) who suggest that the propensity of using earnings management by corporate managers has been decreasing since the accounting scandals in early 2000s. D. Multivariate Analysis of Expectation Management and Earnings Management In order to fully control for firm and deal characteristics that might affect bidders opportunistic behavior before their takeover attempts, we also examine expectation management and earnings management in a multiple regression framework. We focus on stock deals and cash deals since they are associated with an unambiguous difference in the incentives of using expectation management, though including mixed deals do not change our results qualitatively. 18

21 Panel A of Table 7 presents the multiple regressions of bidder s unexpected analyst forecast. The dependent variable is the total amount of unexpected forecasts for the bidder s four quarterly earnings announcements prior to merger announcement. The major independent variable is Stock, a dummy variable that equals one if the deal uses stocks as the only method of payment, and zero if the deal uses cash as the only method of payment. We also control for deal characteristics and firm characteristics one year prior to the merger announcement. 10 MktCap is the market capitalization one month prior to Quarter -4, where Quarter -4 is the quarter of the fourth quarterly earnings announcement prior to merger announcement. MarketBook is the ratio of market value of equity (MktCap) to book value of equity for Quarter -4. TenderOffer is a dummy variable that equals one if the deal has a tender offer, and zero otherwise. Hostile is a dummy variable that equals one if the deal is classified as hostile by SDC, and zero otherwise. AnalystCoverage is the number of analysts following the bidder prior to Quarter -4. In some specifications, we also control for year fixed effects and industry fixed effects. In Model (1), the coefficient on Stock is negative and statistically significant at the 1% level, suggesting that unexpected analyst forecasts of stock mergers are significantly more negative than those of cash mergers. In models (2) (4), we further control for firm and deal characteristics, year fixed effects, and industry fixed effects. The coefficient on Stock continues to be significantly negative at the standard levels, confirming our univariate results that bidders in stock deals have much stronger incentives to engage in expectation management. In terms of economic significance, the coefficient of in Model (4) shows that controlling for various deal characteristics and firm characteristics, the total unexpected forecast in the four quarters prior to a stock deal is percentage points lower than that prior to a cash deal. Given that the mean of the total unexpected 10 In untabulated analysis, we tried controlling for bidders characteristics right before the merger announcement (rather than their characteristics one year before the announcement) and find qualitatively similar results. 19

22 forecast for cash deals in our sample is percentage points, this coefficient suggests that controlling for deal and firm characteristics, the unexpected forecast for stock deals is 4.5 times larger than that for cash deals (0.884 / 0.195=4.53), which is economically large. Other deal and firm characteristics do not seem to significantly affect the magnitude of expectation management. Panel B of Table 7 is similar to Panel A except that the dependent variable is the total amount of abnormal accruals corresponding to the bidder s four quarterly earnings announcements prior to the merger announcement. Consistent with our univariate results, the coefficients on Stock are positive and statistically significant at the 5% or 1% levels, suggesting that bidders in stock deals have much stronger incentives to engage in earnings management via aggressive usage of discretionary accruals. Other than the method of payment, smaller firms, and especially growth firms (those with a higher market to book ratio) tend to use more abnormal accruals before merger announcements, which is consistent with previous studies (e.g., Kothari, Leone, and Wasley 2005). E. Is Expectation Management Driven by Stock Mergers or Omitted Firm Characteristics? Our finding of expectation management primarily in stock mergers is consistent with bidders using expectation management to boost their stock prices prior to merger announcement and in turn lower acquisition costs. However, one may argue that some omitted firm characteristics simultaneously drive expectation management and stock mergers. That is, instead of being triggered by stock mergers, the observed expectation management may be associated with some characteristics of the bidder firms. We partially address this concern by controlling for a broad set of firm characteristics in the multivariate analyses. Since it is impossible to exhaustively control for all firm characteristics, we design three additional tests to address the endogeneity concern. Our first approach uses 123 stock and cash acquisitions by 44 bidders that conducted both stock deals and cash deals in our sample period. We repeat the multivariate regressions of expectation management in Panel A of Table 7 using this subsample where the bidders in the stock 20

23 and the cash deals are the same set of firms. This approach effectively controls for all time-invariant firm-level characteristics. Table 8 reports the results, which show that the stock-merger dummy is significantly positive in all the models, including the last model that controls for firm fixed effects. These results indicate that our main findings so far, i.e., bidders in stock mergers engaging in expectation management, are unlikely to be driven by omitted variable bias. 11 Our second approach is the pseudo event analysis. For each merger announcement in our sample, we create a pseudo announcement for the bidder by shifting the merger announcement back by four quarters and examine the extent of expectation management prior to these pseudo events. In other words, the pseudo announcement and the corresponding actual merger announcement are made by the same bidder firm, but the pseudo announcement is not associated with any actual merger activities. To completely rule out the potential confounding effects of other merger activities by the same bidder in the pseudo event window, we restrict this test to first-time bidders. We then repeat the multivariate regressions of expectation management in Table 7 using the pseudo events. Panel A of Table 9 shows that, in a stark contrast to the results based on our original sample, the dummy variable of stock merger is insignificant in all models for the pseudo events. Further, the coefficient on the stock-merger dummy are only half as big as our original coefficient in Table 7, indicating that the insignificant results here are not purely driven by a lack of power (due to the smaller sample size). For robustness, we also create pseudo events by shifting the merger annoucements by eight quarters, and observe similar results in Panel B of Table 9. These results provide evidence that expectation management is triggered by merger activities instead of timeinvariant firm characteristics of bidders such as corporate culture or management quality We also repeat the univariate analysis (Table 5) using this subsample and find similar results. 12 We also examine univariate results (Table 5) for the sample of pseudo events and observe little evidence of expectation management in the pseudo events as well. 21

24 For the third approach, we directly include firm fixed effects into the multivariate regression of expectation management (Table 7) and find similar (and even stronger) results. For example, after controlling for firm fixed effects, the coefficient on the dummy of stock mergers in model (4) of Table 7 is significantly negative and much larger in magnitude (-2.358, t-stat -3.48) than the original model. Therefore the results from all three approaches provide strong evidence that our finding on expectation management cannot be explained by omitted firm characteristics. Overall, our analysis in this section suggests that takeover bidders in stock mergers tend to manage the market expectation of their earnings downward prior to their merger attempts and that this result is unlikely to be driven by analyst pessimism, industry and year effects, or omitted firm characteristics. IV. Why Do Analysts Participate in Bidders Expectation Management? Our results so far show that takeover bidders engage in expectation management by pushing analyst forecasts downward prior to stock mergers. A natural question is why analysts would adjust their forecasts downward for the benefits of takeover bidders. Analysts may revise their forecasts downward for two reasons. First, they might be misled by the bidders either explicitly through conference calls and earnings guidance, or implicitly through informal interactions with the management. This explanation, however, is unlikely because we have shown that over the same period when analysts adjust down their forecasts, they simultaneously adjust their recommendations upward. The second and more likely reason is conflict of interest, i.e., analysts intentionally cater to the needs of bidder managers in exchange of potential favors in return. Consistent with this explanation, Ke and Yu (2006) find that analysts who are involved in expectation management produce more accurate earnings forecasts and are less likely to be fired by their employers in the 22

25 subsequent period. We therefore conduct two tests to examine the conflict of interest explanation in our setting. A. Analyst-Bidder Relation and Expectation Management We first link the downward adjustment in analyst forecasts to the strength of the relation between analysts and bidders. To the extent that analysts with a closer relation to bidders are subject to more severe conflicts of interest, they would issue earnings forecasts with more downward bias. We measure the relation between an analyst and the takeover bidder using the total number of historical forecasts issued by the analyst for the bidder up to an earnings announcement. A larger number of forecasts indicate that the analyst is likely to have followed the bidder for a longer period of time during which the analyst could establish a closer relation with the bidder managers. A larger number of forecasts also indicate that the analyst is likely to issue forecasts for various metrics of the bidder and therefore can benefit more from a returning favor from the bidder managers such as improved future forecast accuracy. 13 Table 10 reports regressions of unexpected analyst forecast on the relation between an individual analyst and the bidder. In Panel A, the dependent variable is an individual analyst s unexpected forecasts for the bidder s four quarterly earnings prior to the merger announcement. Individual analysts unexpected forecast is constructed in the same fashion as the unexpected forecast based on consensus forecast, except that the consensus analyst forecast ( in equation (2)) is now replaced by a given individual analyst s last published forecast for the particular quarterly earnings. The major independent variable is Relation, which is the natural logarithm of one plus the 13 In untabulated analysis, we also define the relation between an analyst and the bidder as the number of years he or she has been covering the firm. Our results are qualitatively similar if we use this alternative definition of analyst-firm relation. 23