Earnings management, acquirers long-term underperformance, and investor inattention: A case of managerial intent

Transcription

1 Earnings management, acquirers long-term underperformance, and investor inattention: A case of managerial intent HENOCK LOUIS, Penn State University AMY X. SUN, Penn State University Abstract Using the timing of merger announcements and completions to infer managerial intent, we show that the negative long-term post-acquisition returns result largely from managers opportunistic behaviors. Stock-for-stock acquirers that inflate earnings the most tend to announce mergers on Fridays, while distancing most of their earnings management activities from the merger announcement date. These acquirers long-term market performance is particularly striking. They experience an average abnormal return over the year after the merger announcement of 14.30%, compared to only 3.85% for non-friday announcers. Furthermore, the differential pre-merger abnormal accruals and post-merger returns are observed mainly when the announcement date is relatively close to the completion date, with almost all the market losses (about 24%) occurring after the completion date. The acquirers actions apparently reduce the attention that they receive from investors at the merger announcement and over the approval phase, enabling them to sustain their overvaluation until at least the merger completion date.

2 Earnings management, acquirers long-term underperformance, and investor inattention: A case of managerial intent Abstract Using the timing of merger announcements and completions to infer managerial intent, we show that the negative long-term post-acquisition returns result largely from managers opportunistic behaviors. Stock-for-stock acquirers that inflate earnings the most tend to announce mergers on Fridays, while distancing most of their earnings management activities from the merger announcement date. These acquirers long-term market performance is particularly striking. They experience an average abnormal return over the year after the merger announcement of 14.30%, compared to only 3.85% for non-friday announcers. Furthermore, the differential pre-merger abnormal accruals and post-merger returns are observed mainly when the announcement date is relatively close to the completion date, with almost all the market losses (about 24%) occurring after the completion date. The acquirers actions apparently reduce the attention that they receive from investors at the merger announcement and over the approval phase, enabling them to sustain their overvaluation until at least the merger completion date. 1

3 Prior studies find that stock-for-stock acquirers experience significant long-term market losses. Jensen and Ruback (1983, p. 20) comment that the post-outcome negative abnormal returns are unsettling. After decades of investigation, this phenomenon still remains a puzzle. Fama (1998, p. 284) argues that long-term abnormal returns like the ones observed after stock swaps and stock issuances in general can reasonably be attributed to chance. One alternative explanation is that managers are generally overconfident and their hubris leads them to engage in valuedestroying mergers (Rau and Vermaelen 1998). The overconfidence would be more manifest in stock swaps because these transactions do not require external funding and, hence, do not subject the managers to the discipline associated with external financing. However, this hypothesis does not explain why the average stock price does not fully impound the effect of the acquisitions at the announcement date. Another potential explanation is that the underperformance is due to preacquisition overvaluation and managerial opportunism. 1 However, given the strong negative signal associated with stock issuances and the time it generally takes to complete a merger, it is not obvious how acquirers manage to sustain, and benefit from, pre-merger announcement overvaluation. It appears that, on average, investors should correct for the misvaluation at the announcement date or, at least, by the completion date, given that merger negotiations can take months and even years. Our study enhances our understanding of the underperformance phenomenon by addressing these three related questions: First, is the underperformance due to pre-acquisition overvaluation? Second, do the managers intentionally mislead investors? Third, how do acquirers manage to sustain, and benefit from, the pre-announcement overvaluation? One distinguishing feature of the competing explanations for the post-acquisition underperformance relates to managerial intent. Under Fama s (1998) explanation, the returns are due to randomness and, therefore, unrelated to anything that the managers did or intended to do. Under the overconfidence hypothesis, the negative returns are due to bad acquisitions by managers who overestimate their abilities. There is no presumption that the managers intended to 1 Shleifer and Vishny (2003), Louis (2004), Rhodes-Kropf and Viswanathan (2004), Jensen (2005), Rhodes-Kropf et al. (2005), Dong et al. (2006), and Savor and Lu (2009). 2

4 mislead anyone. In contrast, the overvaluation hypothesis presumes that the managers actions are deliberate and opportunistic. There are at least three versions of the overvaluation hypothesis. The first one, advocated by Jensen (2005, p. 10), maintains that overvaluation leads managers to engage in value-destroying acquisitions to project the appearance of satisfying growth expectations and postpone the day of reckoning until they are gone or start generating the expected economic profits. Under this explanation, the post-acquisition returns are related to both overvaluation and bad acquisitions. The second version, advocated by Shleifer and Vishny (2003), presumes that managers of overvalued firms use their stocks to finance acquisitions to take advantage of their overvaluation. Under this explanation, the post-acquisition returns are related to the overvaluation and not necessarily to the mergers per se. As Jensen (2005, p. 10) notes, Shleifer and Vishny allow markets to make mistakes in valuation of companies, but assume that managers are perfectly informed and rational [and] that mergers have no longrun real consequences. In both Shleifer and Vishny (2003) and Jensen (2005), overvaluation drives the decision to engage in acquisitions. A third version of the overvaluation hypothesis, formulated by Louis (2004), presumes that the acquisitions actually drive the overvaluation. Managers deliberately inflate their stocks prices, through earnings management and potentially other actions, to reduce their acquisition costs. This view can be expanded to stock issuance in general (see, e.g., Teoh et al. 1998a, 1998b; DuCharme et al. 2004). A common element of all three versions of the overvaluation hypothesis is that the managers intentionally mislead investors, which is not the case under the alternative hypotheses. Therefore, an assessment of the managers intent could enhance our understanding of the anomalous post-acquisition returns. As Dichev et al. (2012, p. 1) note, many questions are difficult to address with archival work because answers often rely on managerial intent, which is not observable and is difficult to infer. Extant studies suggest that managers inflate earnings prior to stock issuances, 2 and that 2 These include: initial public offerings (Teoh et al. 1998a; Teoh, Wong, and Rao, 1998; DuCharme et al. 2004), seasoned public offerings (Teoh et al. 1998b; Teoh and Wong 2002; DuCharme et al. 2004), reverse leverage buyouts (Chou et al. 2006), and stock swaps (Erickson and Wang 1999; Louis 2004; Gong et al. 2008a). 3

5 issuers overvaluation is partly attributable to the pre-event earnings management (Teoh et al. 1998a, 1998b; DuCharme et al. 2004; Louis 2004; Gong et al. 2008a). This evidence seems to establish managerial intent to mislead investors. However, this inference has been questioned. Some studies maintain that stock prices are not generally affected by earnings manipulations or that managers do not generally manage earnings prior to corporate events with the intent to mislead. 3 In this study, we indentify a simple but powerful approach to assess managerial intent: The timing of merger announcements and completions, and the interaction between the timing and the managers financial reporting behaviors and the post-acquisition market performance. 4 Extant anecdotal and empirical evidence suggests that investors are less attentive to Friday news announcements (Penman 1987; Bagnoli et al. 2005; DellaVigna and Pollet 2009; Louis and Sun 2010). Louis and Sun (2010), in particular, find a substantial reduction in acquirers announcement returns and trading volumes on Fridays. The belief that people are less attentive to Friday news is so common among public relation professionals that the expression bad news Friday becomes idiomatic. 5 In addition, managers often control various aspects of their mergers very meticulously. As Wayne Moore, former Vice President of Goldman Sachs, notes, the time, effort, and care that goes into announcing a deal can be enormous (Moore et al. 1998, p. 12). Hence, we posit that, if managers believe that investors are less attentive on Fridays and if they act deliberately to mislead investors, then we are likely to observe more earnings inflation prior to stock swap announcements made on Fridays. 6 We also expect the Friday 3 Narayanan (1985), Stein (1989), Erickson and Wang (1999), Shivakumar (2000), Coles et al. (2006), and Ball and Shivakumar (2008). We discuss this issue in detail in the next section. 4 An alternative approach would be to directly ask the managers, as in Brav et al. (2005), Graham et al. (2005), and Dichev et al. (2012). However, survey studies generally ask the managers about their perceptions of their peers motivations and actions and not about their own motivations and actions. Given the high incidence of class-action lawsuits associated with stock-for-stock mergers (Gong et al. 2008b), managers would probably not respond truthfully (even anonymously) to questions about their own potential misleading pre-acquisition behaviors. 5 Commenting on the Friday phenomenon, Will Clemens, CEO of Internet shopping service Respond.com, states: I think it s as simple as on Friday, people are wrapping up business for the week. On a Monday, people are all looking at the market ( 6 We could have used initial or seasoned equity offerings as alternative settings. However, the Security Data Company (SDC) provides filing dates for stock issuances but not the exact announcement dates. 4

6 announcements to be associated with more negative post-announcement long-term returns, as overvalued firms would tend to announce acquisitions when attention is low. Because the days of the week are totally exogenous, Friday can serve as a powerful instrument to identify managerial intent. Managers could be aware of factors other than Friday that can also lower investors attention, and therefore we might observe negative abnormal returns even for the non-friday announcers. However, under the inattention hypothesis, we should still observe a stronger effect for the Friday announcers than for the non-friday announcers. Consistent with extant studies, we find that acquirers report significantly positive abnormal accruals in the quarter prior to the merger announcement (Quarter-1). However, remarkably, we find that acquirers that announce mergers on Fridays report substantially more positive abnormal accruals over the three quarters prior to the merger announcement than those that announce mergers on the other business days. We also find that the differential abnormal accrual between Friday and non-friday announcers is notably more pronounced in Quarter-2 and Quarter-3. The result suggests that stock-for-stock acquirers that inflate their pre-merger announcement earnings tend not only to make the merger announcements when investors are less attentive but also to distance most of their earnings management activities from the merger announcement date. The notion that managers would distance their earnings management activities from the merger announcement date is consistent with Ke, Huddart, and Petroni (2003), who find that managers distance their trading activities from disappointing earnings announcements by as many as three quarters to conceal their opportunistic activities. We also find that the Friday announcers experience substantially more negative postmerger announcement returns than the non-friday announcers. The differential return across the two groups of acquirers is actually quite striking. The average buy-and-hold abnormal return over the year after the merger announcement is only 3.85% for the non-friday announcers but 14.30% for the Friday announcers. Furthermore, we find a much stronger negative association between the abnormal returns and the pre-acquisition announcement abnormal returns for the 5

7 Friday announcers than for the non-friday announcers. These results are consistent with the notion that the Friday announcers are more likely to have acted opportunistically. There is generally a delay between the announcement and the completion of a merger. Subsequent news, disclosures, and earnings announcements can bring renewed attention to the acquirer and dilute the potential benefit associated with the timing of the merger announcement. Therefore, managers of overvalued acquirers have incentives to make merger announcements as close to the completion dates as possible. Consistent with this view, we find that the Friday effect is driven largely by mergers with close announcement and completion dates, which is also consistent with managers anticipating delays in the completions of the mergers and acting accordingly. Not only do Friday announcers with close completion dates inflate earnings prior to the merger announcement more, but they also manage to sustain their overvaluation until at least the merger completion date. Friday announcers that complete their acquisitions within 10 weeks of the announcement date experience an average negative return of about 24% over the year after the merger announcement, with almost all the loss occurring after the completion date. Quite remarkably, we find that these results hold even for very large acquirers, with average (median) market value of over $50.5 ($9.6) billion. Taken together, the evidence strongly suggests that the negative long-term post-merger returns largely result from deliberate managerial choices to use inflated stocks as currencies. While the evidence does not rule out the overconfidence explanation, it strongly supports the overvaluation hypothesis and raises serious doubts about the notion that the negative postacquisition abnormal returns are attributable to chance. One challenge of the overvaluation hypothesis is to explain how acquirers manage to sustain, and benefit from, pre-merger announcement overvaluation, considering the strong negative signal associated with stock issuance and the time it generally takes to complete an acquisition. Our analysis addresses this issue by relating the pre-merger earnings management activities and the post-merger long-term returns to the timing of the merger announcements and completions. 6

8 The balance of the paper is organized as follows. The next section discusses the extant literature on the effects of managerial opportunistic reporting behaviors on long-term returns. Section II describes the research design. Section III describes the sample. The results are presented in Section IV. We conclude in Section V. I. Managerial opportunistic reporting behaviors and long-term returns The role of earnings management in explaining issuers long-term stock returns is an unsettled issue. Several studies suggest that the negative post-issuance returns are attributable to the reversal of the stock price effects of pre-issuance earnings management (Teoh et al. 1998a, 1998b; DuCharme et al. 2004; Louis 2004; Gong et al. 2008a, 2008b). However, other studies maintain that stock prices are not generally affected by earnings management and/or that managers do not generally manage earnings prior to corporate events with the intent to mislead investors. Stein (1989), for instance, presents a signal-jamming model in which managers undertake wasteful actions to increase current earnings but a capital market composed of rational investors correctly anticipates this earnings inflation, reconstructs the unmanaged earnings series, and always prices the firm at fundamental value. Consistent with this view, Coles et al. (2006) find that abnormal accruals are low after firms announce out-of-the-money options will be cancelled and before replacement options are issued, but stock price is unaffected by these apparent manipulations. Erickson and Wang (1999) also argue that investors expect a firm to inflate earnings prior to a stock swap and, consequently, discount the firm s stock price at the stock swap announcement whether the firm actually manages earnings or not. Consistent with this view, Shivakumar (2000, p. 340) posits that, since issuers cannot credibly signal the absence of earnings management, investors treat all firms announcing an offering as having overstated prior earnings, and consequently discount their stock prices. Therefore, anticipating this market behavior, an issuer s best response is to manage earnings. Erickson and Wang (1999) and Shivakumar (2000) argue that managers of issuing firms manage earnings not because they are opportunistic but as a rational response to the anticipated market behavior. 7

9 Ball and Shivakumar (2008, p. 347) also raise doubt about the notion that managers report opportunistically prior to major corporate events because these settings also are characterized by higher than usual litigation and regulatory risk from inflating earnings, and higher than usual scrutiny by market monitors such as analysts, underwriters, auditors, boards, the press and other parties to the transaction, as well as by regulators. 7 Their reasoning is different from the argument formulated by Erickson and Wang (1999) and others. These studies argue that managers are likely to inflate earnings prior to the corporate events but not generally with the intent to mislead investors. In contrast, Ball and Shivakumar (2008) argue that managers are generally unlikely to inflate earnings prior to major corporate events because of the aforementioned countervailing factors. Nonetheless, both sets of studies question the notion that managers generally exhibit opportunistic reporting behavior prior to important corporate events. We examine the extent to which acquirers act opportunistically prior to stock swaps by analyzing the association between the timing of the swap announcements and completions and the acquirers financial reporting behaviors and stock returns. At earnings announcements, investors are generally uncertain about the nature of the reported earnings. Abnormal accruals can be associated with opportunistic behavior; however, they can also result from normal business activities. They can even reflect managers expectations of future performance. 8 Therefore, although investors are likely to discount abnormal accruals around earnings announcements, they are unlikely to fully do so unless they can associate them with specific managerial incentives. The uncertainty is likely to be resolved (at least partly) when the accruals are reversed or the managers incentives are later revealed through certain actions, such as a stock swap announcement. Subsequent announcements of certain events are likely to alert 7 Gong et al. (2008b, p. 64) note, however, that acquirers indeed face lawsuits after stock-for-stock mergers quite often and that the most common complaint in these lawsuits is that managers have misguided investors by issuing false and misleading statements. They also argue that the large incidence of lawsuits after stock-for-stock mergers is an indication that managers and directors might not be as cautious as they should. 8 See, e.g., Watts and Zimmerman (1986), Subramanyam (1996), Rees et al. (1996), Demski (1998), Guay et al. (1996), Arya et al. (2003), and Louis and Robinson (2005). 8

10 investors to the managers incentives and prior earnings manipulation activities (Erickson and Wang 1999). Extant studies suggest that attention affects investors information processing. 9 Hirshleifer et al. (2004), in particular, suggest that investor inattention causes abnormal accruals mispricing. Thus, it appears that, if managers report opportunistically prior to corporate events, they are more likely to strategically announce these events when they believe that investors are less attentive. This prediction would hold whether the managers inflate earnings to inflate their stock prices prior to the acquisitions (Louis 2004) or they inflate earnings and engage in stock swaps to hide overvaluation (Jensen 2005). Assuming that the fraction of inattentive investors increases on Fridays, DellaVigna and Pollet (2009) show that managers would tend to announce bad earnings news on Fridays. They associate the timing of the earnings news to managers attempts at maximizing short-term stock prices. However, because stock-for-stock acquirers use current prices to execute the mergers and because of the substantial impact that the initial market reaction to a stock swap announcement can have on the cost and the likelihood of successfully completing the merger, strategic timing can offer enormous (long-term) economic benefits to an overvalued acquirer. This notion is in line with Shleifer and Vishny s (2003) argument that stock-for-stock mergers create value for acquirers even when their long-term performance is negative. The timing of merger announcements could serve the acquirers interests at the expense of their targets. It might appear that the targets managers would not accede to the acquirers plans. However, while target managers can broadly influence the timing of merger announcements, the precise timing of the announcements is the prerogative of the acquirers managers. Furthermore, as Shleifer and Vishny (2003) explain, target managers often agree upon merger conditions even when they know that the conditions are not in the best interest of their 9 See, e.g., Patell and Wolfson (1982), Penman (1987), DeLong et al. (1990), Daniel, Hirshleifer, and Subrahmanyam (1998), Hong and Stein (1999), Shleifer (2000), Huberman and Regev (2001), Daniel, Hirshleifer, and Teoh (2002), Hirshleifer and Teoh (2003), Hirshleifer et al. (2004), Peng and Xiong (2006), Gabaix and Laibson (2006), Gabaix et al. (2006), Baker, Ruback, and Wurgler (2007), Bollerslev et al. (2007), Barber and Odean (2008), Cohen and Frazzini (2008), DellaVigna and Pollet (2009), Hirshleifer, Lim, and Teoh (2009), and Hou et al. (2009). 9

11 shareholders. Shleifer and Vishny (2003) suggest that acquirers can buy their agreement through the acceleration in the exercise of stock options or by granting them generous severance pay. Target managers can also agree to mergers for reasons of retirement or illiquid stock ownership. Cai and Vijh (2007) also suggest that target managers who face high illiquidity discounts tend to leave after the acquisitions are completed. Therefore, these managers are not very concerned about the acquirers long-term returns. II. Research design A. Measuring abnormal accruals Following the extant literature (e.g., Teoh et al. 1998a, 1998b), we proxy for earnings management using abnormal accruals, as measured by the residual from a modified version of the Jones (1991) model. Specifically, for each calendar quarter and two-digit SIC-code industry, we estimate the following model using all firms that have the necessary data on Compustat: 4 TA i = j-1 Q j,i + 4 ( SALES i - AR i ) + 5 PPE i + 6 LTA i + 7 ASSET i + i, (1) j=1 where TA is total accruals; Q j is a binary variable taking the value of one for fiscal quarter j and zero otherwise; ΔSALE is the quarterly change in sales; AR is the quarterly change in accounts receivable; PPE is property, plant, and equipment at the beginning of the quarter; LTA is the fourth lag of total accruals; and ASSET is total assets at the beginning of the quarter. Consistent with Dechow et al. (2003), Louis and White (2007), and Gong et al. (2008a, 2008b), we extend the extant discretionary accruals models by controlling for lagged accruals because some firms may inherently have higher accruals levels than others and accruals are likely to be correlated over time. The model controls for growth in cash sales, firm size, and overall accrual levels. We measure total accruals as earnings before extraordinary items and discontinued operations minus operating cash flows. All the variables, including the fiscal quarter indicators, are scaled by total assets at the beginning of the quarter. Deflating the model by assets transforms ASSET into a column of ones, allowing us to estimate the model with a standard 10

12 intercept. To mitigate the effect of outliers and errors in the data, within each calendar quarter and each fiscal quarter, we delete the top and bottom half-percentiles of the deflated TA, ( SALE- AR), PPE, and LTA. We also require at least 20 observations for each estimation. Following Kothari, Leone, and Wasley (2005), we adjust the estimated abnormal accruals for performance. Consistent with Louis (2004), among others, for each quarter and each industry (two-digit SIC code), we create five portfolios with at least four firms each by sorting the data into quintiles based on the return-on-assets from the same quarter of the previous year. The performance-matched abnormal accruals for a firm are the firm-specific abnormal accruals minus the mean abnormal accruals for its respective industry-performance-matched portfolio. B. Measuring long-term abnormal returns We first compute abnormal returns using Daniel et al. s (1997) portfolio-benchmark return adjustment procedure. 10 Fama (1998) argues that buy-and-hold approaches of computing post-event abnormal returns aggravate bad model problems through compounding and do not account for potential cross-sectional correlation problems. He advocates using the Fama- MacBeth approach. We first implement his proposed procedure using Carhart s (1997) fourfactor model. However, Louis and White (2007) observe that, when using the calendar-time approach, the mean abnormal returns of two subsamples can both be lower than the mean abnormal return of the full sample, which occurs when one subsample has more monthly observations and the average abnormal return for these monthly observations is lower than the average abnormal return for the other subsample. 11 This problem could be quite pronounced in 10 To construct the benchmark portfolios, we first assign each stock that has book value of equity on Compustat and price and shares outstanding on CRSP to a size quintile (using NYSE size quintile breakpoints). Within each size quintile, we rank the stocks based on their industry adjusted book-to-market ratios, and assign them to their book-tomarket quintiles, yielding a total of 25 size and book-to-market sorted fractiles. We then sort them into quintiles within each of the 25 size and book-to-market fractiles, based on the prior 12-month stock return. This results in a total of 125 fractiles. Then, we compute a monthly value-weighted return for each of the 125 fractile portfolios. The benchmark portfolios are reconstructed annually at the end of June. 11 Consider a simple case with three months. In month 1, there is a Friday announcer with a return of 15% and no return for the non-friday announcers. In month 2, there is a Friday announcer with a return of 20% and a non-friday announcer with a return of 13%. In month 3, there is a Friday announcer with a return of 25% and a non-friday 11

13 our setting given the small number of Friday announcements in a given year. Therefore, we also apply the Fama and MacBeth (1973) approach using a slightly different procedure. More specifically, to measure a firm s abnormal return over the year after the merger announcement, we compute 12 monthly abnormal returns for that firm, one for each of the 12 months immediately after the merger announcement. Then, for each of the 12 months, we compute the cross-sectional average abnormal return for the sample, which yields a time-series of 12 monthly average abnormal returns. Following Fama (1998), we base our inferences on the mean and the standard deviation of the time-series of the cross-sectional average monthly abnormal returns. 12 This procedure maintains the number of observations used to compute the time-series average abnormal returns somewhat constant while controlling for potential cross-sectional correlation problems (Fama and MacBeth 1973). Because it averages the returns instead of compounding them, the procedure also avoids the potential problems associated with compounding. C. Regression model We model the association between the pre-merger announcement abnormal accruals and the post-merger announcement abnormal returns as follows: 1 ABRET i = ( f,0 + f,1 ABAC f,i + f,2 UNLISTED f,i + f,3 POOL f,i + f,4 SAME_IND f,i f=0 + f,5 LBIDDER f,i + f,6 LSIZE f,i + f,7 RSIZE f,i + f,8 BM f,i + f,9 LCOVERAGE f,i ) + i, (2) where ABRET is the acquirer s post-merger announcement abnormal returns; ABAC is the acquirer s pre-merger announcement abnormal accruals; UNLISTED is a binary variable taking announcer with a return of 11%. The time-series of the average returns are then 15% (month 1), 20% (month 2), and 25% (month 3) for the Friday subsample, 13% (month 2) and 11% (month 3) for the non-friday subsample, and 15% (month 1), 16.5% (month 2), and 18% (month 3) for the full sample. The means of the time-series averages are 6.5% for the full sample, but 10% and 12% for the Friday and the non-friday subsamples, respectively. The mean for the full sample is lower than the means for both the Friday and the non-friday subsamples. 12 Fama (1998) suggests that, when using the Fama-MacBeth approach, one can estimate the monthly abnormal returns in any reasonable way, for example, with a matching firm or matching portfolio approach, or with a formal asset pricing model (p. 295). Accordingly, we compute monthly abnormal returns of individual firms using Daniel et al. s (1997) portfolio-benchmark return adjustment procedure. 12

14 the value one if the target is not a publicly traded company and zero if it is publicly traded; POOL is a binary variable taking the value one if the merger is accounted for by the pooling-ofinterest method and zero if it is accounted for by the purchase method; SAME_IND is a binary variable taking the value one if the two merging partners are in the same two-digit SIC code and zero otherwise; LBIDDER is the log of the number of bidders competing to acquire the target; LSIZE is the log of the acquirer s market value of equity at the end of the quarter prior to the merger announcement; RSIZE, relative size, is the ratio of the transaction s total value to the acquirer s market value of equity at the end of the quarter prior to the merger announcement; BM is the acquirer s book-to-market value of equity at the end of the quarter prior to the merger announcement; LCOVERAGE is the log of the number of analysts included in the last analyst consensus annual earnings forecast for the acquirer prior to the beginning of the abnormal return measurement window; f is a binary variable taking the value one for Friday announcers and zero for non-friday announcers; 0 is an intercept term; and is the residual. Prior studies identify only two factors that are associated with long-term returns after stock swaps: pre-merger abnormal accruals (Louis 2004) and book-to-market (Rau and Vermaelen 1998). Rau and Vermaelen (1998) argue that high book-to-market acquirers tend to be more cautious before engaging in major transactions and that their acquisitions are less likely to be motivated by hubris. Accordingly, they suggest that acquirers long-term returns are positively associated with book-to-market. However, Louis (2004) finds no evidence that the long-term market performance is associated with book-to-market ratios. Therefore, thus far, with the exception of abnormal accruals, there is no ex-ante evidence that any other factor is reliably associated with the long-term market performance. Nonetheless, we include the control variables in the model to ensure that our inferences are not due to potential omitted correlated variables. If acquirers that manage earnings strategically time their merger announcements and if they are successful in their strategies, then the difference in the post-merger announcement abnormal returns across the Friday and the non-friday announcers should be related to the pre- 13

15 merger announcement abnormal accruals. Therefore, we expect both the coefficient on ABAC 01 and the coefficient on ABAC 11 to be negative, but the coefficient on ABAC 11 to be more negative than the coefficient on ABAC 01. In notational terms, we expect: 11 < We estimate Model (2) using both buy-and-hold long-term returns in a pooled regression and the Fama-MacBeth procedure with monthly returns. To apply the Fama-MacBeth procedure, we estimate the model each month after the merger announcement month, using the abnormal returns for each specific month while leaving the explanatory variables unchanged. This process yields a series of coefficient estimates for each of the explanatory variables. Then, following Fama and MacBeth (1973) and Fama (1998), for each explanatory variable, we use the timeseries of the mean and the standard deviation of the coefficient estimates to make inferences. III. Sample description A. Sample selection The study covers announcements of completed stock swaps between U.S. companies made between January 1994 and December 2009, inclusively. We conduct the study over a period when it was relatively easy for investors to access financial and other corporate information. Starting in 1994, firms proxy statements became electronically available through the historical Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system. Therefore, we control for limitations on investors access to publicly available information, which might be more severe over the weekend. Investors could appear to underreact to Friday announcements simply because they do not have access to financial information over the weekend. However, information on EDGAR can be accessed anytime with relative ease, which mitigates this potential problem. The sample is obtained from the Security Data Company s (SDC) acquisition database. We include an observation in the sample if the form of the transaction (as defined on SDC) is a merger, the method of payment is reported on SDC, and the value of the transaction is disclosed 14

16 on SDC. The acquirer must also have Compustat data on equity book value, price, number of shares outstanding, and the variables to compute the abnormal accruals, and data from the Center for Research in Security Prices to compute the abnormal returns. We require that the transaction value and the acquirer market value be over two million dollars. The primary sample is limited to stock swaps, although for comparison purposes, we also analyze a sample of cash transactions. A transaction is deemed a stock swap if more than 50% of its value is financed with stock. 13 The average stock financing for our sample is about 94% of the total financing. We exclude banks because Compustat does not have the data necessary to compute abnormal accruals for banks. Our main sample includes 2,442 completed stock swap announcements that satisfy the selection criteria, with 306 Friday announcements and 2,136 non-friday announcements (those made between Monday and Thursday). 14 The sample includes only completed transactions; however, for the sake of completeness, we also analyze a sample of failed bids. B. Descriptive statistics Panel A of Table I presents the daily distribution of the sample. Managers show a preference for making announcements early in the week. The number of announcements declines steadily during the week, with 28.71% of the announcements made on Mondays and 12.53% on Fridays. While the percentages of Monday and Friday announcements depart substantially from the 20% that we would expect if the announcements were randomly distributed across the weekdays, the percentages for the other days are quite close to the 20%. The daily pattern of the announcements suggests that managers generally avoid making merger announcements on Fridays. This observation is consistent with the notion that managers are generally overconfident and that they typically view their mergers as value-enhancing (Roll 1986; Rau and Vermaelen 13 We use this definition of a stock swap to maximize the sample size; many stock-for-stock acquirers add some small amounts of cash to their offers to facilitate the transactions. However, our results are qualitatively similar if we limit the sample to pure (100%) stock transactions. 14 There are five announcements that were made on a Thursday and one that was made on a Wednesday that are mistakenly assigned Friday dates on SDC. We reclassify these announcements as non-friday announcements. However, the reclassification does not qualitatively change the results. 15

17 1998; Malmendier and Tate 2008). If managers believe that investors are less attentive on Fridays and they also believe that their mergers are value-enhancing, then they are less likely to announce them on Fridays, unless the acquisitions are driven by overvaluation. Another reason for announcing mergers at the beginning of the week is that it is generally easier to negotiate deals and have them approved over the weekend. However, this reason does not explain why there are more merger announcements on Wednesdays and on Thursdays than on Fridays. Panel B of Table I reports characteristics of the sample. The industry-relatedness of the merging partners is different across the Friday and the non-friday announcers, which can be seen as consistent with the notion that Friday announcers tend to be more overvalued than non-friday announcers. To the extent that managers are more likely to uncover overvaluation in their industries, it could be more difficult for an acquirer to buy a target in its industry with overvalued stocks. The relative size of the transactions is also marginally different. There is no evidence that the decision to account for a transaction by the pooling accounting method, the number of bidders, the market values of the acquirers, the acquirers book-to-market values, the acquirers analyst following, or the public status of the targets are correlated with Friday announcements. Prior studies find that stock-for-stock acquirers announcement returns are on average negative when the targets are listed and positive when they are unlisted. It could appear then that, if managers act strategically, they would tend to not announce acquisitions of unlisted targets on Fridays. However, like proxy fights (Ikenberry and Lakonishok 1993) and new exchange listings (Dharan and Ikenberry 1995), stock-for-stock acquisitions of unlisted targets are associated with average announcement returns and post-announcement returns of opposite signs. Louis (2012) also shows, through an analysis of acquirers returns, abnormal accruals, net operating assets, and insider trading, that stock-for-stock acquirers are likely more overvalued when the targets are unlisted than when they are listed. 15 Louis (2012) provides many reasons why managers could 15 For our sample, the Spearman correlation (p-value) between target unlisted status and the abnormal returns over the year after the merger announcement is (0.000). The correlation is no longer significant after controlling for pre-merger announcement abnormal accruals. 16

18 have stronger incentives to acquire unlisted targets with inflated stocks. Hence, if managers strategically time announcements of overvaluation-driven acquisitions, it is not surprising that they would also announce acquisitions of unlisted targets on Fridays. Furthermore, while some factors are associated with the market reaction to merger announcements, none is a perfect predictor of the market reaction. A deal characteristic could be negatively (positively) associated with merger announcement returns; however, managers would not necessarily announce every merger that displays this characteristic on a Friday (non-friday). They could be aware of other factors associated with the deal and/or the acquirer that could induce the opposite market reaction. It is also important to note that, while Louis and Sun (2010) focus on the general market perception of the information content of merger announcements, our study focuses on managers assessments of their firms valuations. IV. Results A. Pre-merger announcement abnormal accruals and merger announcement timing Table II reports pre-merger announcement abnormal accruals for Friday and non-friday announcers. The results in Panel A of Table II show that the average abnormal accruals (one-tail p-value) for the quarter (90 days) prior to the merger announcement (Quarter-1) are 0.60% (0.000) and 1.30% (0.000) of total assets for the non-friday announcers and the Friday announcers, respectively. The difference is statistically significant with a one-tail p-value of The medians (one-tail p-value) are 0.42% (0.000) and 0.37% (0.002). The difference is statistically insignificant, with a one-tail p-value of Erickson and Wang (1999, p. 162) find that stock-for-stock acquirers start reporting high unexpected accruals as early as three quarters before the news of a merger is first revealed in the financial press. Accordingly, we also analyze the abnormal accruals over the three quarters prior to the merger announcement. We find average cumulative abnormal accruals (p-value) for the three quarters (Quarter-3 to Quarter-1) of 0.93% (0.000) and 3.04% (0.000) of total assets for 17

19 the non-friday announcers and the Friday announcers, respectively. The difference is statistically significant, with a one-tail p-value of The median cumulative abnormal accruals (one-tail p-value) are 1.05% (0.000) and 1.89% (0.000). The difference is also statistically significant, with a one-tail p-value of The results in Panel A of Table II have at least two important implications. First, they suggest that acquirers that inflate earnings prior to stock swaps strategically make the merger announcements on Fridays, when investors are presumably less likely to be paying attention. It could be that some acquirers announce their transactions on Friday to allow the market enough time to digest the information and avoid potential overreaction to the news. However, it is not obvious why those acquirers with the most positive abnormal accruals would want the market to take more time to process the information. In addition, as we will later show, the Friday announcers tend to have worse long-term market performance. Those acquirers would have faced much higher acquisition costs if the negative stock performance were reflected in their stock prices around the merger announcement. The second implication is that, although acquirers report significantly positive abnormal accruals in the quarter immediately prior to the merger announcement (Quarter-1), the differential abnormal accrual across the Friday and the non- Friday announcers is particularly pronounced in Quarter-2 and Quarter-3. This observation suggests that acquirers that manage earnings increase their likelihood of misleading investors by distancing most of their earnings management activities from the merger announcement date. Consistent with Erickson and Wang (1999), we find no evidence of substantial positive abnormal accruals prior to the third quarter. The average cumulative abnormal accruals are only 1.11% and 2.79% for the non-friday announcers and the Friday announcers, respectively, when we extend the accrual accumulation period to Quarter-4. Therefore, although our inferences are qualitatively similar if we use the abnormal accruals over the first two or the first four quarters prior to the merger announcements, all our subsequent analyses will be based on the abnormal accruals over the first three quarters. 18

20 Panel B of Table II reports the cumulative abnormal accruals for the three quarters immediately prior to the merger announcement for each of the individual business days. The abnormal accruals are larger for Friday than for each of the other days. We are interested in the total amount of abnormal accruals reported prior to the merger announcement. Therefore, the sum of the abnormal accruals over the three quarters is our preferred measure. However, for some firms, we do not have abnormal accrual estimates for all three quarters, which can create artificial variations in our estimate of total abnormal accruals for the three quarters. To ensure that this potential source of measurement error does not drive our inferences, we also use the average (instead of the sum) of the abnormal accruals over the three quarters since the average is not affected by missing observations. The results are reported in Panel C of Table II. Again, the abnormal accruals are significantly larger for the Friday announcers than for the non-friday announcers, with Friday dominating each of the other days. B. Abnormal accruals and the lag between the announcement and the completion dates As mentioned earlier, there is generally a delay between the announcement and the closing dates of a merger. Subsequent news, disclosure, and earnings announcements can bring renewed attention to an acquirer and dilute the effect of the merger announcement timing. However, managers acting opportunistically can time the announcements to be as close to the completion dates as possible. Managers can also anticipate delays in a deal s completion and act accordingly. We therefore analyze the association between the timing of the merger completion and managerial reporting behaviors. We present results for the first week after the merger announcement and for each month afterwards. We follow this pattern because there is a large concentration of mergers that are completed within one week, but there are not enough weekly observations to conduct statistical analyses beyond the first week. The results reported in Table III show that the Friday effect is driven mainly by those acquisitions that are completed within 10 weeks (or 70 days) of the announcement date. The differential abnormal accrual over the three quarters prior to the merger announcement is 3.83% 19

21 of total assets for acquisitions with close completion dates (4.70% for the Friday announcers and 0.87% for the non-friday announcers) but only 1.00% for those with distant completion dates (1.91% for the Friday announcers and 0.91% for the non-friday announcers). One likely explanation for the timing effect is that, as time progresses, firms are likely to come back to the spotlight, and hence a firm that inflates its pre-merger announcement earnings is less likely to benefit from the strategic timing of the merger announcement. Hence, opportunistic managers time the announcements and completions of their transactions to reduce the likelihood of renewed attention to their firms before the completion dates. The 10-week cut-off ensures that the merger is closed before the subsequent earnings announcement, a period around which firms are likely to attract a lot of attention. 16 Closing before the earnings announcement also enables the managers to hide potential reversals of the pre-acquisition earnings management through consolidation and acquisition related costs. C. Are the observed abnormal accruals realistic? The average abnormal accrual for the Friday announcers with close completion dates is very large. To facilitate a deeper analysis of the data, we list the abnormal accruals for each acquirer in this subgroup in Appendix A. A quick inspection of this appendix reveals that some of the abnormal accruals are extremely large. 17 Survey results reported by Dichev et al. (2012) indicate that chief financial officers (CFOs) believe that, among firms that manage earnings within GAAP, the amount of earnings management can be substantial. The CFOs estimates of earnings management range from 1% to 65.5% of earnings. Nevertheless, some of the large abnormal accruals that we observe in Appendix A can be puzzling. For instance, Stratesec Inc (# 104 in Appendix A) has positive abnormal accruals of 51% of total assets over the three fiscal quarters prior to the announcement of its merger with Security Systems Integration Inc. Could 16 We also analyze the association between the timing of the merger completion and the pre-completion abnormal accruals. The results are similar to those reported in Table III. Friday announcers report substantially more abnormal accruals prior to the merger completion date than non-friday announcers when the merger is completed within the next 10 weeks. The effect essentially disappears for mergers completed more than 10 weeks after the announcement. 17 We discuss the other elements of Appendix A in Section IV-H. 20

22 such huge abnormal accruals result from earnings management? Is it plausible that firms could inflate earnings by that much and receive clean audit opinions? Or, do the abnormal accruals simply result from the inherent noise in the abnormal accrual estimation process and/or errors in the data? To assess the extent to which the observed abnormal accruals are realistic, we analyze Stratesec s filings with the Securities and Exchange Commission (SEC) around the merger announcement. We focus on Stratesec because it has by far the largest pre-merger abnormal accruals and has one of the largest post-merger announcement market losses (-82.8% abnormal returns over one year). It is therefore the case that seems most implausible. Our analysis, which is discussed in Appendix B, indicates that Stratesec s abnormal accruals are not the result of data or estimation errors. Nor are they the results of growth. They are instead due to earnings management. The analysis in the appendix, coupled with Dichev et al. s (2012) survey results, increases our confidence that our findings are reasonable. Stratesec is a very small firm and is therefore probably more likely to mislead investors by using earnings inflation and acquisitions to project the appearance of growth. However, pre-merger accrual inflation is not limited to small firms like Stratesec (see Appendix A). Untabulated results show that the average cumulative abnormal accruals over the three quarters prior to the merger announcement are 5.23% for the Friday announcers with close completion dates that fall in the bottom two size quintiles and 3.81% for those that fall in the top two size quintiles. D. Merger announcement timing and post-announcement market underperformance The post-merger announcement abnormal returns are presented in Table IV. Panel A reports results for the year immediately after the merger announcement. Consistent with the notion that managers strategically time their merger announcements, we find that the abnormal returns are substantially more negative for Friday announcers than for non-friday announcers. The difference is striking. The average (median) buy-and-hold abnormal returns over the year after the announcement are 3.85% (-12.06%) for the non-friday announcers and 14.30% (- 21