The effect of financial misreporting on corporate mergers and acquisitions

Transcription

1 The effect of financial misreporting on corporate mergers and acquisitions Merle Erickson Booth School of Business - University of Chicago Shane Heitzman Simon School of Business University of Rochester X. Frank Zhang Yale School of Management November 20, 2012 Abstract In this paper, we investigate how misreporting influences merger and acquisition decisions. We analyze a sample of 283 firms accused of committing accounting fraud by the SEC between 1985 and During the alleged fraud period, these firms pursued over 300 acquisitions valued at $305 billion in the aggregate. Managers that pursue acquisitions increase the risk that the fraud will be discovered during negotiations. On the other hand, successful transactions conceal misreporting by complicating the firm s accounting information after the transaction closes. Our results suggest that these concealment benefits outweigh the incremental detection costs: Fraud firms are more likely than non-fraud firms to acquire another company and are also more likely to acquire firms that have less public information, are harder to value, and have less similar operations. In addition, we find that fraud firms negotiate provisions that increase the likelihood of deal success and accelerate the closing date to speed up the consolidation of the target s and fraud firm s financial statements. Taken as a whole, our results suggest that managers engaged in accounting manipulation use acquisitions in an ex ante attempt to conceal misreporting and that these concealment benefits outweigh the increase in detection risk. We appreciate comments from Jerry Zimmerman and workshop participants at National University of Singapore, Penn State University, SUNY Buffalo, Yale University, University of Alabama, and the Information, Markets and Organizations conference at Harvard University. Mengjie Huang provided valuable research assistance.

2 The effect of financial misreporting on corporate mergers and acquisitions 1. Introduction When managers misrepresent the firm s performance, they face strong private incentives to conceal their misreporting (Karpoff et al. 2008). These managers also make real hiring and investment decisions, including acquisitions, while misreporting (McNichols and Stubben, 2008; Kedia and Phillipon, 2009). This results in a fundamental conflict between the cost and benefit of acquisitions for firms engaged in intentional misreporting. Under the concealment benefits hypothesis, an acquisition has the potential to support managerial attempts to conceal the accounting misstatement. Misreporting firms are therefore more likely to make acquisitions, choose targets that reduce the comparability and consistency of the financial statements, and close the deal just before the fiscal quarter end. Under the detection cost hypothesis, the negotiation process can increase scrutiny of the buyer s confidential accounting records and increase the likelihood of fraud detection. Thus, the manager should be less willing to engage in an acquisition and, if they choose to acquire, they should structure the deal to minimize the firm s exposure to enhanced scrutiny of its accounting information. The concealment benefit and detection cost hypotheses have competing views on the frequency and characteristics of acquisition activities. In this paper, we empirically test these two competing hypotheses. We analyze the acquisition decisions of a sample of 283 firms accused by the SEC of engaging in accounting fraud (fraud firms) between 1985 and These firms pursued 307 target firms during the period of the alleged fraud; the aggregate value of all acquisitions by fraud firms during this period is approximately $305 billion. 1 During the period of the alleged fraud, the fraud firms in our sample are 37% more likely than non-fraud firms to announce an 1 This is a lower bound estimate of the activity by misreporting firms, as we cannot identify firms that committed fraud but were not caught. 1

3 acquisition and shift total investment expenditures to acquisitions of existing businesses. Among those firms that enter the acquisition market in a given year, fraud firms purchase more entities and spend more as a fraction of total assets relative to other firms. For the 307 acquisition attempts during the alleged fraud period, the average target is valued at about $1 billion, has EBITDA of around $150 million, and annual sales that approach 19% of the acquiring firm s sales. This evidence suggests that the concealment benefits that arise after the transaction is completed are quite large. Given the decision to pursue an acquisition, we also examine how managers trade off these considerations in negotiations. From the concealment benefit perspective, the misreporting manager has incentives to select a target that makes it difficult for external parties to analyze the target s assets and operations and their impact on the subsequent financial results reported by the combined firm. Consistent with this argument, we predict that misreporting firms will be more likely to acquire privately held firms, subsidiaries rather than stand-alone firms, firms located abroad, as well as firms in different industries. 2 Based on the sample of completed deals, we find that misreporting firms on average are about 18% more likely to acquire a subsidiary than nonfraud firms and 34% more likely to acquire a firm in a different industry, evidence consistent with the concealment benefits hypothesis. The manager s decision to misstate accounting information should also have an impact on bargaining strategy. If detection costs are relevant, the manager will reduce the incentives for sellers and capital providers to scrutinize the firm and will retain options that allow the acquirer to withdraw from the deal at low cost. This should manifest in a lower rate of target termination 2 The detection cost hypothesis is silent on target type; we have no a priori reason to believe holding payment method, size, and structure constant why these target characteristics should influence the likelihood the fraud will be detected during negotiations. 2

4 fees and a higher likelihood of deal failure. 3 On the other hand, if the manager s incentive to do the deal is driven by post-closing fraud concealment benefits, we predict that misreporting managers will demand that the target agree to a termination fee provision and will be more likely to close the deal. While deal failure rates are no different after controlling for termination fees, we do find that misreporting managers are 23% more likely to demand that the target agree to a termination fee provision, suggesting that concealment benefits from completing the deal are important. Finally, the concealment benefits hypothesis should affect when the deal closes. Delaying the close by just a few days can push back the consolidation of the target and acquiring firms financial statements and the resulting expected concealment benefits for months. From the time a transaction is first announced, we find that misreporting firms complete the acquisition about 19% faster than other firms. Moreover, they are nearly three times more likely to close a large deal in the final week of the fiscal quarter. [Do we want to do some analysis of these large acquisitions and report how much higher quarterly earnings are due to the early closing in those large acquisitions ] This latter result is perhaps the most direct evidence that misreporting managers expect to obtain concealment benefits from consolidating the financial results of the newly-acquired target firm with the misreporting acquirer. Our work adds to an expanding literature on the relation between accounting misstatements and investment decisions (including acquisitions) with an objective to better understand the causal links between the incentives to misreport accounting information and the frequency and characteristics of acquisitions undertaken by the firm. 4 Our investigation focuses on how the 3 Target termination fees award a payment to the acquiring firm if the target breaks off the deal. 4 This approach is distinct from the existing literature that focuses on why managers manipulate financial information (Burns and Kedia 2006; Goldman and Slezak 2006; Armstrong et al. 2010; Wang 2011), how it is 3

5 decision to misreport affects subsequent investment decisions in the spirit of Kedia and Phillipon (2009) who argue and provide evidence that managers that misstate accounting information must then invest more than optimal in order to maintain investors optimistic perception about the firm s growth opportunities. We provide an alternative explanation that exploits the fact that managers face substantial private costs if the misreporting is detected and thus their acquisition decisions likely reflect a trade off of post-closing concealment benefits and detection costs. Our results also have implications for understanding how agency conflicts influence the decisions of corporate acquirers (Amihud and Lev 1981; Jensen 1986; Morck et al. 1990; Harford 1999). Given the expected private costs if the fraud is detected, the manager has strong incentives to pursue acquisitions that conceal the manipulation and these tend to compound the effect on shareholder value. Taken together, our findings contribute to the growing stream of literature on the role of agency costs, information asymmetry, and financial reporting on investments and acquisitions. In the next section, we review the prior literature. Section 3 discusses the data sources and sample attributes. Section 4 discusses specific predictions and presents empirical evidence. We discuss alternative explanations in Section 5 and conclude in Section Prior literature Our paper is related to the growing literature on the connection between misreporting and investment (including acquisitions), with an objective to better understand the causal links between the incentives to misreport accounting information and the frequency and characteristics of acquisitions undertaken by the firm. One potential association between misreporting and detected (Dyck et al. 2010; Dechow et al. 2011; Wang 2011), and the impact of fraud detection on managers, directors, shareholders, and creditors (Karpoff et al. 2008a and b; Fich and Shivdasani 2007; Graham et al. 2008). 4

6 acquisitions arises if managers misstate accounting information with the primary purpose of acquiring a target on better terms. Erickson and Wang (1999) find that acquirers in stock for stock mergers manipulate earnings in the periods leading up to a merger announcement to inflate the value of shares used to acquire the target s stock. Louis (2004) and Gong, Louis and Sun (2008) provide evidence that for these acquirers, there is a stock price reversal both before and after the deal announcement that is a function of the pre-acquisition managed earnings. Kravet, Myers, Sanchez and Scholz (2012) extend these studies and find that firms that restate accounting earnings are more likely to have completed stock for stock mergers during the periods for which their financial statements contained accounting misstatements. Bens, Goodman and Neamtiu (2012) offer a second reason for the link between misreporting and acquisitions: misstatements are driven by bad acquisition decisions in the past. If managers are more likely to be replaced following poor performance, they have incentives to bias reported earnings upward. The novel findings in Bens et al. (2012) suggest that managers concerned about losing their job following a pessimistic market reaction to an acquisition announcement are more likely to inflate subsequent performance reports to allay investor concerns and retain their position. The third set of studies reverses the direction of causality and focuses on how the decision to misreport affects subsequent investment decisions. Kedia and Phillipon (2009) argue that managers that misstate accounting information must then invest more than optimal in order to maintain investors optimistic perception about the firm s growth opportunities. McNichols and Stubben (2008) provide an alternative explanation that allows those responsible for misreporting to be distinct from those responsible for initiating and approving investment projects. They argue 5

7 that overinvestment occurring after misreporting begins is a consequence of managers using optimistic forecasts, perhaps unknowingly, derived from the misreported information. Our paper largely follows this third path. Complementing existing studies, we examine whether misreporting firms use acquisitions to conceal accounting fraud. Specifically, we examine how acquisition activities, target types, and acquisition structures are affected by the manager s incentives to conceal fraud. Under the detection cost hypothesis, acquisitions impose a cost on the manager by increasing the access to and scrutiny of the financial statements by outside parties, including sellers, financing providers, and investors. Thus, the manager should be less willing to engage in an acquisition and should structure deals to minimize the firm s exposure to enhanced scrutiny of its accounting information. Under the concealment benefits hypothesis, acquisitions benefit the manager by exploiting the financial reporting effects of a business combination to cover up misstatements. 5 Thus, misreporting buyers undertake more successful deals, choose targets that reduce the comparability and consistency of the financial statements, and speed up the closing process. To keep the analysis clear and organized, we discuss specific predictions before each set of tests in Section Data and sample 5 There is anecdotal evidence that fraud impacts merger decisions. For example, in the Department of Justice s fraud indictment of Richard Scrushy at HealthSouth, it was alleged that, It was further part of the conspiracy that defendant RICHARD M. SCRUSHY and co-conspirators would and did cover up, conceal, and keep secret the fraud, by: (a) controlling and limiting access to HealthSouth s financial information; (b) controlling the internal distribution of financial results; (c) providing fraudulent documentation and false information to its auditors; (d) providing false information to Federal and State taxing authorities; and (e) fraudulently using the acquisition of other companies to conceal fraudulent assets on HealthSouth s books and in its reports. (par. 36. Italics added) Similarly, in Biovail Corporation Securities Litigation, Case No. 03-CV-8917 (RO): According to the complaint, defendants knew throughout the Class Period that its earnings forecasts for Biovail could not be met as evidenced by the fact that the Company was considering emergency plans for bolstering its business through essential acquisitions which were designed to replace foreseeable declining revenues. 6

8 Our basic research question is straightforward: Do the acquisition preferences of misreporting managers reflect a tradeoff of fraud concealment benefits and detection costs? To address this question, we are interested in identifying cases of relatively aggressive accounting manipulation in which managers have strong incentives to conceal their actions. This requires an objective measure of accounting manipulation. Ideally, this measure will reliably capture the accounting manipulation of firms whether or not the misreporting is actually detected, but such a measure is difficult to construct and validate. Ex ante measures such as accounting accruals can reflect accounting manipulation (Bergstresser and Philippon 2006), but also capture real changes in the fundamentals of the firm. Ex post measures such as restatements are potentially more useful (for example, Efendi et al. 2007; Graham et al. 2008), but often arise for reasons that have little to do with accounting manipulation. 6 Karpoff et al. (2012) document that the primary restatement database, Audit Analytics, substantially overstates cases of aggressive misreporting. Our primary measure of accounting manipulation is whether or not the SEC alleges that the firm engaged in fraudulent financial reporting as evidenced by an AAER covering fiscal years between 1982 and 2003 as identified by Dechow et al. (2011). 7 The AAER are an important subclassification of enforcement actions analyzed by Karpoff et al. (2008b). We recognize that using a sample of detected fraud cases to identify accounting manipulation has drawbacks, particularly because we are interested in understanding the potential role of acquisitions in hiding the misreporting. In other words, our sample is comprised of firms that engaged in fraud and were actually caught, so firms that engaged in fraud but successfully concealed it are ultimately not treated as accounting manipulators. Separating the effects of fraud commission from fraud 6 For example, firms can restate insignificant amounts or restate in response to a misinterpretation of accounting rules. Class-action lawsuits have been used more recently to generate a sample of manipulating firms (Armstrong et al. 2010; Dyck et al. 2010). 7 We are grateful to Patty Dechow, Weili Ge, Chad Larson, and Richard Sloan for making this data available. 7

9 detection is not straightforward and this could lead to a potential interpretation problem. We handle this challenge in two ways. First, we explicitly consider the role of fraud detection risk in influencing mergers and acquisitions through the detection cost hypothesis. As it turns out, the predictions under a detection cost explanation either stand alone or run in the opposite direction of the concealment benefits hypothesis. Thus, our evidence allows us to draw inferences on which of these two effects dominate. Second, we consider an alternative measure of fraud based on an expectations model which is estimated using, but does not directly rely on, alleged cases of fraud. 8 Our analysis occurs at both the firm-year and transaction levels. Among the 283 firms identified in the AAER sample with sufficient data from Compustat and CRSP (i.e., fraud firms), 181 pursue at least one acquisition during the 1982 through 2005 sample period (as identified by SDC), and of those, 94 make at least one acquisition during the alleged fraud period. We identify 644 firm-years of data for fraud firms during the alleged frauds with sufficient accounting and stock market information. For the set of non-fraud firms that we use for comparisons, we include all years and deals between 1982 and 2005 with sufficient data. In Table 1, we summarize a number of characteristics for fraud and non-fraud firms. On average, the SEC accuses managers of engaging in fraud over two fiscal years. However, it is widely believed that the accounting manipulation underlying the fraud begins long before the fraud s alleged start date. Old or missing records and the federal statute of limitations can prevent the SEC from formally alleging fraud in too early a period. Because of this, the period directly preceding the alleged fraud is unlikely to serve as an effective control for expected non- 8 It is worth noting that nearly all studies of fraud determinants face a similar problem (e.g. Erickson et al. 2006; Armstrong et al. 2010; Dechow et al. 2011; and Schrand and Zechman 2011), and dealing with this problem is not straightforward. 8

10 fraud acquisition behavior by fraud firms. It is also possible to benchmark to post-fraud years, but such an analysis is hampered by sample attrition (often because of bankruptcy) or significant changes to the firm s management and control environment. Thus, we draw our main inferences from comparisons of acquisitions by fraud firms during the fraud years to a sample of firms that were never accused of fraud through an AAER during our sample period. We report distributions for the alleged fraud firms during the fiscal years of the fraud period and three years immediately before (pre-fraud) and after (post-fraud), as well as for control firms not subject to an AAER during the sample period (non-fraud firms). For each firm in the sample, we identify all acquisition attempts during the sample period. Deals treated as occurring during the fraud period are those announced during the fiscal period in which the SEC alleges the fraud occurred. For each acquisition attempt, we require the acquirer to seek 100% of the target firm stock, not own more than 50% prior to the acquisition, and that the acquisition value of the target s stock be at least $10 million (in 2000 dollars). An acquisition attempt is successful if the firm completes the deal within two years of the public announcement, otherwise we treat it as unsuccessful. In each year of the alleged fraud, 21% of firms make at least one acquisition attempt. By comparison, only 11% of non-fraud firms attempt an acquisition in a given year, and the frequency for fraud firms is significantly greater than for non-fraud firms at the 0.01 level. Relative to non-fraud firms, fraud firms are significantly larger, have higher market-to-book ratios and higher leverage. In Table 1, Panel B we summarize the characteristics of the attempted acquisitions during the period of the alleged fraud. The average target purchased by fraud firms is valued at about $1 billion and has sales of $660 million and EBITDA of about $150 million. Non-fraud firms purchase targets with an average value of about $380 million that have average EBITDA of 9

11 about $90 million and for which average sales are about $540 million. On a relative basis, fraud firms acquire targets with equity values averaging 13% of their own value (2% at the median). On the other hand, non-fraud firms purchase firms whose value averages 35% of the buyer s preacquisition value (9% at the median). Both the mean and median relative deal sizes are significantly higher for the non-fraud sample at the 1% level, a result driven by the fact that fraud firms are systematically larger than non-fraud firms. Fraud firm shareholders react positively to the deal announcement throughout the sample, with 3-day excess returns averaging 0.20%, but less so relative to non-fraud shareholders (3-day excess return=0.79%). The average positive reaction in both samples is driven by the large proportion of takeovers of private firms, consistent with the findings in Fuller et al. (2002). At the median, fraud firms close a deal in 59 days (87 at the mean) during the fraud period, significantly faster than the 81 days (110 at the mean) taken by non-fraud firms. Finally, we also provide evidence on target type and deal structure. 71% of fraud firm targets are private (vs. 74% for non-fraud firms), while 36% of all targets are subsidiaries rather than stand-alone firms (vs. 34% for non-fraud firms). Fraud firms purchase foreign firms in 27% of their deals compared to 18% for non-fraud firms and this difference is statistically significant at the 0.01 level. Fraud firms are more likely to acquire a target with a different 2-digit SIC code (60% vs. 43% of deals). When fraud firms acquire, they are significantly more likely to obtain target termination fee agreements (20% vs. 12%). Fraud firms appear to have a significantly higher success rate for deals announced during the fraud, closing 97% of announced deals compared to 94% for non-fraud firms (p < 0.1). 4. Evidence on accounting manipulation and corporate acquisitions 10

12 4.1. Acquisition activity We first address the link between financial misreporting and acquisition activity. An acquisition can provide a number of benefits to the manager manipulating financial information. The right target can generate financial slack for a manager that has exhausted the firm s existing capacity for earnings manipulation. The required consolidation of the acquirer s and target s financial statements introduces complexity into the financial reporting outputs used by auditors and shareholders and can make post-deal financial statement analysis substantially more difficult. 9 Moreover, the announcement of an acquisition is usually a material information event. Substantial director, investor, and analyst effort is devoted toward understanding the impact of the deal on firm value, particularly when there is less information about the target. This can shift information gathering activities toward understanding the strategic decisions of the firm and away from the financial reporting integrity of the acquirer. Under the concealment benefits hypothesis, managers use takeovers to both obscure the firm s true performance and shift attention to other activities of the firm, implying that misreporting firms should be more active in the acquisition market. But acquisitions are costly if the opportunity costs of both the capital required to complete the transaction and the manager s time are material. Moreover, acquisitions also subject the firm to greater scrutiny from outsiders. Banks and underwriters provide additional scrutiny, and target 9 A recent accounting standard, Statement of Financial Accounting Standards No Business Combinations, explicitly discusses in the introduction to the standard the complexity arising from acquisition accounting, as follows. Under Opinion 16 [APB No. 16], business combinations were accounted for using one of two methods, the pooling-of-interests method (pooling method) or the purchase method. Use of the pooling method was required whenever 12 criteria were met; otherwise, the purchase method was to be used. Because those 12 criteria did not distinguish economically dissimilar transactions, similar business combinations were accounted for using different methods that produced dramatically different financial statement results. Consequently: Analysts and other users of financial statements indicated that it was difficult to compare the financial results of entities because different methods of accounting for business combinations were used. 11

13 directors have a legal obligation to assess the intrinsic value of the acquirer s stock consideration in many circumstances. Under the detection cost hypothesis, managers avoid pursuing acquisitions as this increased scrutiny significantly increases the probability that the manipulation will be detected. Thus, we expect misreporting firms to be less active in the acquisition market under the detection cost hypothesis. In the first column of Table 2, we report the results of a logistic regression explaining the probability of announcing at least one offer during the year: Prob( 1 successful offer) = f(β 1 Fraud year + β 2 Controls + β T Year effects + β K Industry effects) (1) We measure merger activity as a binary variable equal to 1 if the firm announces at least one ultimately successful acquisition attempt during the year. 10 The variable of interest is an indicator variable equal to one for firm-years in which the SEC alleges that fraud occurred. 11 The set of control variables includes the log of equity value, the market-to-book asset ratio, profitability, cash holdings, and book leverage, all measured at the beginning of the fiscal year. Year and industry fixed effects are also included. Consistent with accounting manipulation having an impact on acquisition behavior, we find that fraud firms are significantly more likely to complete acquisitions during periods of alleged fraud (p < 0.001). The average marginal effect reported indicates that a firm later accused of engaging in accounting fraud is 3.7 percentage points more likely to announce a successful 10 Given the high rate of completing announced offers, the results are almost identical if we include unsuccessful deals. 11 We do not benchmark the fraud year effects against pre-fraud year effects because it is not clear when the fraud actually begins. The SEC can be constrained by the statute of limitations when pursuing accounting fraud actions. Moreover, the farther back in time the SEC must go, the more likely the evidence necessary to prevail is incomplete or missing. 12

14 acquisition in a given fraud year. With non-fraud firms announcing at least one successful deal in 10% of sample years, misreporting firms are thus 37% more likely to announce an acquisition attempt in a given year (3.7%/10% 37%). The coefficients on the control variables are largely as expected. For example, large firms and those with more cash on hand are more likely to make a deal. We also examine whether the level of deal activity is different for misreporting firms. In columns 2 and 3 of Table 2 we focus on years when firms enter the acquisition market by announcing at least one acquisition. The level of activity is defined as the number of successful deals announced during the year or the total amount paid in acquisitions as a fraction of the firm s beginning total assets. The coefficient of on the fraud firm indicator reported in column 2 suggests that among all firms that enter the acquisition market in a given year, a firm engaged in fraud will make about more acquisitions per year. 12 The level of increased activity translates into on average about $486 million of additional revenue from the target (Panel B of Table 1). Using the same sample of firms that complete at least one deal, and summing the value of all acquisitions for each firm-year, the results in column 3 show that deals by fraud firms add 7 percentage points more to the acquirer s asset base than deals by non-fraud firms (t = 1.84). Overall, the results support the concealment benefits hypothesis that misreporting firms tend to do more acquisitions. A related but separate question is whether these acquisitions actually add to the total investment of the firm or represent substitution from other forms of investment like direct capital investment or research and development. We measure the composition of investment as the fraction of total investment expenditures (acquisitions, R&D, and capital 12 Obviously, the acquirers in our sample buy entire firms, not fractions (we restrict ownership to be 100% for completed deals). However, to provide economic interpretation of our results, we multiply the 0.74 by the various mean and median values for fraud firms as reported in Table 1. 13

15 expenditures) allocated to acquisitions. Panel A of Table 1 is consistent with the findings in prior studies that firms invest significantly during fraud years. The average fraud firm adds 9% to the total assets through capital investment and 10% through mergers and acquisitions during the misreporting period. Non-fraud firms add a similar amount in capital expenditures, but less than half the amount (4%) in acquisitions. As a fraction of total investment (capital expenditures, R&D, and announced acquisitions), however, acquisitions represent 14% of total investment for fraud firms and just 8% for non-fraud firms. Controlling for industry, year, and firm characteristics, this difference remains significant in the regression results reported in column 4 of Table 2. Among all firms with positive investment expenditures in a given year, the fraction of investment allocated to acquisitions is 3.9 percentage points, or 50% larger (3.9%/8% 50%) for fraud firms (t = 4.26). In other words, given the decision to invest, misreporting firms allocate more capital to acquisitions (right word is acquisitions, not investments as before right?) thus adding substantial complexity to the post-investment financial reporting environment further supporting the concealment benefits hypothesis. Acquisition activities serve as the first step to distinguish between the concealment and detection cost hypotheses. We recognize the possibility that our control variables may not fully address the endogeneity issue between misreporting and acquisition activities. 13 Rather than using more sophisticated models, we next turn to specific attributes of acquisitions, such as the target types and deal closing times, to test the concealment benefits and detection cost hypotheses Accounting manipulation and target type 13 One possible alternative explanation is that the SEC identifies fraud on the basis of acquisition activity. However, prior research (Dyck, Morse and Zingales (2010)) does not suggest that the SEC identifies most frauds, and does not suggest that frauds in general are discovered due to acquisition activity. 14

16 Under the concealment benefits hypothesis, managers identify those firms that if acquired make the firm s post-closing financial reporting information more complex for insiders and outsiders to process. Increasing opacity and uncertainty about the true impact of the acquisition on the acquirer s financial statements reduces the utility of the financial statements for detecting earnings manipulation. Taking the decision to make an acquisition as given, the misreporting manager will prefer targets that have less public information before the deal, assets that are harder for outsiders to identify and analyze, and operations that have less in common with the acquirer. We consider a number of target types. First, we consider private companies. Historical and forecasted financial data are readily available for most public targets, but unless they have publicly traded debt, private targets will have minimal public information. And when an acquirer purchases a private target, the required level of disclosure regarding the target s assets and historical performance is generally minimal. Among acquisitions of private targets in our sample, SDC provides historical accounting data for only 27% of these deals. The SEC s disclosure requirements look to the relative size of the target based on total assets or income. If the target s assets or income is less than 20% of the acquiring company s total assets, the acquirer has no obligation to disclose the target s pre-acquisition historical information. 14 Second, we consider acquisitions of subsidiaries. A subsidiary s financial information is less transparent and acquiring-firm managers are likely to have more discretion over the ultimate 14 When the target makes up between 20% and 40% of the acquirer s assets or income, only the most recent fiscal year audited results must be disclosed. Between 40% and 50%, two years of audited results must be disclosed, and above 50%, three years. This threshold was 10% before 1996, but still only required a single year of results for acquisitions between 10% and 20% of the combined company. Beginning in 2000, no disclosure is required if the payment to target shareholders is entirely in cash and the acquiring firm shareholders do not vote. When the private target is a subsidiary of another corporation, the consideration is almost always in cash. When the unlisted target is a standalone entity, stock is used more frequently. See Officer (2007) for further analysis of purchase prices of private targets. See 17 C.F.R. 210 generally for current requirements. Rodrigues and Stegemoller (2007) provide a useful discussion of the requirements for disclosure of target financial information, and argue that many material acquisitions do not require disclosure of target financial statements. 15

17 financial statement impact in this type of deal than when acquiring a free-standing private or public firm. Third, we examine whether misreporting firms are more likely to acquire foreign firms. Foreign firm takeovers lead to similar limitations in pre-acquisition information as their financial statements will not generally be constructed following U.S. GAAP even if they are publicly traded. Finally, we examine whether misreporting firms tend to acquire targets from different industries. A diversifying acquisition reduces the correlation in economic performance across units and arguably impedes efforts to detect misreporting in the parent firm. In Table 3, we analyze the effect of acquiring firm accounting manipulation on the type of target acquired. The dependent variable takes a value of 1 when the acquirer purchases a target of a given type (for example, a subsidiary). We control for acquiring firm size, market-to-book, profitability, cash holdings, leverage, and announcement year and industry effects. As the first column of Table 3 indicates, fraud acquirers are not more likely to purchase private targets (the coefficient on the fraud year indicator is insignificant). In the second column we compare acquisitions of subsidiary entities to those of stand-alone public and private firms. Misreporting firms show a strong preference for subsidiaries and are 5.4 percentage points more likely to acquire one (p = 0.03), suggesting that subsidiary targets provide more concealment benefits for fraud firm managers. Given that subsidiary acquisitions make up about 34% of the private firms acquired by non-fraud firms, this implies that fraud firms are about 16% more likely to acquire a subsidiary (computed as 5.4%/34% 16%). In the third column of Table 3, we focus on the propensity to acquire a foreign target. The results from that analysis indicate that misreporting acquirers are about 2.8 percentage points more likely to acquire a foreign firm than non-fraud acquirers, but the effect is not significant (p = 0.12). 16

18 In the final column of Table 3, we investigate whether misreporting firms are more likely to acquire a firm in a different industry. We define industries at the 2-digit SIC level and find that misreporting firms are 14.5 percentage points more likely to acquire a target in a different industry (p < 0.01) than non-fraud firms. Since 43% of the acquisition targets of non-fraud firms are in a different two-digit industry, these estimates suggest that conditional on a deal, misreporting firms are 34% more likely to purchase a firm in a different industry (14.5%/43% 34%). Overall, we interpret our evidence in this section as somewhat consistent with the prediction that firms engaged in accounting manipulation use acquisitions to provide concealment benefits by selecting target firms that impede the ability of investors, creditors, and regulators to understand what the firm acquired and how the acquisition affects post-transaction financial information Deal failure and termination fees If fraud detection risk is a material consideration during negotiations, the misreporting manager will value the option to withdraw from negotiations. This effect should manifest in a lower rate of target termination fees (a payment to the acquiring firm if the target breaks off the deal) and a higher likelihood of deal failure. If concealment benefits dominate detection costs, we expect managers will demand target termination fee agreements to raise the target s cost of breaking off deal negotiations and increase the likelihood of deal completion (Bates and Lemmon 2003; Officer 2003). In Table 4, we find that fraud firms have a 2.9 percentage point higher probability of including a target termination fee than non-fraud firms (p = 0.025). Because 17

19 non-fraud firms employ these agreements in 12% of deals, this implies that deals by fraud firms are about 24% more likely to contain a termination fee provision (2.9%/12% 24%). We also examine whether misreporting affects the probability of deal failure independent of the termination fee agreement and find no evidence that it does. However, a termination fee agreement is associated with a 6.1 percentage point decline in the likelihood an announced deal will fail (p < 0.01). One implication is that fraud firm managers impact deal outcomes through their use of provisions such as termination fees. However, a well-known limitation of this analysis is that deal failure rates are based on announced offers and thus understate the true failure rate that includes unsuccessful offers that were conveyed privately but not announced publicly. Thus, observed deal completions rates provide a relatively weak test of the impact of misreporting on deal success or failure. [This paragraph is very confusing and I don t know what we are concluding I think it needs a rewrite. We say no evidence in the first sentence, then report a significant difference, then at the end say that this is a weak test. Please think about how to explain this test more clearly] 4.4. Closing speed Finally, we examine the impact of misreporting on the time it takes to complete a deal once it is announced. We focus on the log of the number of days from announcement to completion so that we can interpret the coefficient as the percentage impact on closing speed in the logarithm format. After excluding deals that are announced and completed on the same day (usually small deals and comprising about 25% of the sample), the results in Panel A of Table 5 imply that once the deal is announced, fraud firms complete a deal 19% (=ln(nonfraud closing days) ln(fraud closing days)) faster than non-fraud firms (p < 0.01). Given that the average completion time is 18

20 110 days for non-fraud firms, our results imply that fraud firm managers take actions that cut the time it takes to complete the transaction by 19 days. 15 Delaying the close of the acquisition by just a few days can postpone consolidation with the target s financial statements by months. Thus, a more direct test of the concealment hypothesis is based on the timing of deal closings. We expect misreporting firms will attempt to accelerate deal closings to occur before the end of a fiscal quarter. 16 For each successful deal, we identify the fiscal quarter end closest to the deal close date. Deals that close the last seven days of the fiscal quarter (ending on the quarter end date) represent week 0. Deals that close the first week of the next fiscal quarter represent week 1, and so on. For all deals, the distribution of deal close dates around the nearest quarter end dates is represented in Figure 1a and suggests that fraud firms do close a higher percentage of all deals (16.2%) just before the fiscal quarter close (week 0) relative to non-fraud firms (11.3%). The difference between these frequencies, as reported in panel B of Table 5, is significant (p < 0.01). If the manager s desire to accelerate the close is a function of the materiality of the deal, then we expect to find stronger results for large acquisitions. In figure 1b, we focus on targets valued at least 10% of the acquirer, and find that fraud firms close 28.8% of their large deals in the final week of the quarter while non-fraud firms close 11.3% of their large deals during the same week. 17 The difference is significant (p < 0.01) and provides further support for the conjecture 15 Interestingly, Grinstein and Hribar (2004) find that acquiring CEOs that take longer to complete a large deal have higher future compensation. They interpret this result as longer completion times requiring more effort which the CEO is compensated for. In this light, our results suggest that fraud firm managers perceive greater benefits from speeding up the close and place a lower value on the potential wage increase from extending the closing period. 16 We thank Maureen McNichols for suggesting this test. 17 The apparent acceleration of closing dates to meet the close of the fiscal period is also reflected in the times to completion. In untabulated tests, we find the median deal closing in the last week of the fiscal quarter take 87 days to close for non-fraud firms, and only 30 days to close for fraud firms (p-value of difference < 0.02). In contrast, the median deal closing in the first week of the fiscal quarter takes 76 days to close for non-fraud firms and 71 days to close for fraud firms (p-value of difference = 0.79). 19

21 that firms making deals to conceal misreporting will accelerate the closing dates to get the target s financial information on the combined entity s books as quickly as possible. 18 In panel C, we estimate logistic regressions in which the dependent variable is a binary variable equal to one if the deal closes in the final week of the closest fiscal quarter, and zero otherwise. We control for firm size and year effects, and the results confirm our univariate findings fraud firms are more likely to close deals during the last week of the quarter and this propensity is an increasing function of target size. These findings are robust to controlling for a number of other attributes of the deal. Overall, the evidence is consistent with the concealment benefit hypothesis. Not only do misreporting firms complete more acquisitions, they increase the fraction of total investment allocated to acquisitions. Misreporting firms also tend to acquire subsidiaries and firms in different industries. Finally, misreporting firms are more likely to demand termination fee agreements and to close the deal just before the fiscal quarter end. 5. Additional analyses and alternative explanations 5.1. Does the acquisition delay the fraud detection? To this point, the empirical analyses have centered on a comparison of fraud firms to nonfraud firms. In this section, we try an alternative approach that compares fraud firms that make acquisitions during the first year of the fraud period to those that do not. If the primary role of accounting manipulation is to mislead investors about the fundamental value of the firm, then actions that conceal that manipulation should lead to slower fraud detection and price correction. We partition fraud firms based on whether or not they announced a successful acquisition in the first year of the fraud. This approach avoids a potential mechanical relation between deal 18 During the years before the fraud, there is no apparent clustering of deal closings around fiscal period end. 20

22 activity and the duration of the fraud that would arise if we looked to multiple years. In Table 6, Panel A we report the mean and median firm characteristics, measured at the end of the last prefraud year, for firms that announce a deal in the first year (19% of firms) and those that do not (81% of firms). Based on a comparison of the medians, the firms are largely similar in terms of cash holdings and leverage, while acquirers are larger, have higher market to book, and are slightly more profitable than non-acquirers. As argued earlier, actions that conceal the fraud should lead to a slower detection and price adjustment. Thus, if mergers and acquisitions have the effect of concealing underlying financial manipulation, we would expect overvaluations to persist longer in fraud firms completing acquisitions than in fraud firms not completing acquisitions. In Table 6, Panel A, we report descriptive statistics on the cumulative raw returns over various windows before and after the start of the alleged misreporting. At the univariate level, only long-run returns before and after the fraud start date appear significantly different between the two groups. A concealment explanation for acquisitions implies that to the extent the misreporting is impounded in stock prices and results in overvaluation, acquisitions intended to conceal the fraud will delay the recognition of misreporting and the stock price correction. To test this notion formally, we adopt a model of return reversal. The dependent variable is the cumulative stock return that begins in the first month of the fiscal year the fraud is alleged to have occurred and ends 3, 6, 12, 24, or 48 months out. The primary independent variable that serves as a proxy for overvaluation is the prior stock returns over an equally long window that ends in the last month before the first year of the alleged fraud. To test whether firms that do deals appear better able to conceal the fraud, we interact prior returns with an indicator for M&A activity in the first year of fraud. We also control for firm size and market-to-book. 21

23 The results, reported in Table 6, Panel B, imply that the reversal of stock returns is significantly delayed for firms doing deals in year 1, based on return windows between 6 and 24 months. For example, the results suggest that firms with strong returns leading up the fraud period are less likely to experience a reversal in the months following the fraud if they make an acquisition. When the return period equals 24 months, the coefficient on the interaction between prior returns and acquisition activity is 0.24 (p = 0.02), and the interpretation of the coefficient is. In our study, we do not incorporate exactly when the fraud was detected as measuring that construct is not straightforward. However, the result that non-acquiring fraud firms have faster reversals of stock overvaluation is consistent with acquisitions having the effect of temporarily concealing the misreporting from shareholders An alternative measure of misreporting Following recent studies on accounting manipulation, we rely on enforcement actions taken by the SEC against firms with alleged material misstatements to objectively identify a set of firms with a demand for transactions that conceal accounting manipulation. While this approach is unlikely to include firms that did not misreport, it likely misses a substantial number of firms that did misreport but successfully concealed their actions from outsiders. 19 Dealing with this issue empirically is not straightforward and an accepted method has not yet emerged. Ideally, we would like to use the available data to identify manipulating firms without reference to whether or not they were actually caught. But conceptually, such a metric is unlikely to be a particularly powerful way to identify manipulating firms. Manipulation demands opacity such that 19 While our interest lies in understanding the demand for acquisitions as a way to reduce the ex ante likelihood the firm s misreporting will be detected, an alternative interpretation of the results is simply that it reflects differences in the ex post likelihood of getting caught. This same criticism applies to nearly all studies that rely on samples with clear selection issues. This criticism, however, leads to the opposite prediction in nearly all of our tests, and as a result biases us against finding results consistent with the concealment motive. 22