Financial Misreporting and Corporate Acquisitions

Transcription

1 Financial Misreporting and Corporate 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 September 24, 2013 Abstract Managers who perpetrate corporate fraud have strong private incentives to conceal it. An acquisition offers an opportunity to distract auditors and regulators, reduce transparency through consolidated financial reports, and bundle concealed losses with merger-related write-offs. However, an acquisition also increases the risk that fraud will be discovered during negotiations. In this paper, we investigate how misreporting decisions influence subsequent 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. Our results suggest that concealment motives explain acquisition activity for firms engaged in fraud. Fraud firms are more likely than non-fraud firms to acquire another company, are more intensive acquirers and speed up closing dates to accelerate consolidation. They also choose targets that have less public information, are harder for outsiders to value, operate in different industries and negotiate provisions that increase the likelihood of deal success. Taken together, our results suggest that acquisitions are used by managers in an ex ante attempt to conceal misreporting. 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 Financial Misreporting and Corporate Acquisitions 1. Introduction Prior research suggests that managers who misrepresent firm performance exploit the firm s overvaluation to make new investment, including acquisitions (McNichols and Stubben 2008; Kedia and Philippon 2009). But these managers also face strong private incentives to conceal their misreporting (Karpoff, Lee and Martin 2008b), and there is growing anecdotal evidence that acquisitions played a key role in attempts to cover up fraud at Olympus, Tyco and Cendant, among others. Despite potentially large economic effects, we have little systematic evidence that managers who commit fraud view the acquisition market as an avenue to conceal it. In this paper, we test a fraud concealment hypothesis for mergers and acquisitions. The core presumption of this hypothesis is that the manager views an acquisition as a tool to delay or prevent discovery of an ongoing fraud. An acquisition serves this purpose on many dimensions. First, it distracts resource-constrained regulators and auditors who review and approve the transaction and related disclosures. Second, the consolidated financial statements that now include the acquired firm are less transparent about the economic performance of the fraud firm; disentangling the impact of the acquisition is difficult because information about the target is often opaque. Third, target firms can collude with buyers to defer earnings until after the deal closes. Finally, acquisition accounting permits firms to immediately write off substantial amounts for restructuring charges or (at the time) acquired research and development, thus providing cover for recording pre-deal losses. A fraud concealment motive for acquisitions predicts that misreporting firms are more intensive acquirers, select targets that reduce financial statement transparency and allow for accounting flexibility, and speed up the deal closing to accelerate financial statement consolidation. 1

3 A contrasting view of the impact of an acquisition is that the negotiation process actually increases fraud detection risk by triggering enhanced scrutiny of the firm s financial statements. If this detection risk hypothesis is descriptive, we expect managers to be less willing to engage in an acquisition, but if they choose to, should structure the deal to minimize scrutiny of the firm s accounting information. The fraud concealment and detection risk hypotheses have unique views on the frequency and characteristics of acquisitions for firms engaged in financial misreporting. We analyze acquisitions by 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 alleged fraud period at an aggregate value of approximately $305 billion. 1 During the alleged fraud, fraud firms in our sample are 37% more likely than non-fraud firms to announce an acquisition. Among those firms that enter the acquisition market, fraud firms purchase more entities, show stronger growth in their asset base through acquisitions, and shift aggregate investment expenditures toward acquisitions, all relative to non-fraud firms. Further, the target firms acquired are significant in relation to various acquirer metrics. For the 307 acquisition attempts during the alleged fraud period, the average target is valued at $1 billion, has EBITDA of around $150 million, and annual sales that approach 19% of the acquiring firm s sales. The fraud concealment hypothesis predicts that the misreporting manager needs the target s financial results consolidated with the fraud firm s as soon as possible; pulling the close of the acquisition ahead by just a few days can get the financial reports consolidated months earlier. From the time a transaction is first announced, we find that misreporting firms complete the acquisition about 19% faster than non-fraud firms. Moreover, fraud firms are nearly three times more likely to close a large deal in the final week of the fiscal quarter. This latter result is 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. 2

4 perhaps the most direct evidence that misreporting managers anticipate benefits from consolidating their financial information with a newly-acquired target. Given the decision to pursue an acquisition, we also examine how managers trade off these considerations in selecting and negotiating with a target. These tests are less direct than those based on closing dates but are not conditional on the timing of the close. Fraud concealment motives predict that the misreporting manager selects targets that make it more difficult for external parties to analyze the target s prior financial position and performance and thus its impact on the financial results reported by the combined firm. Consistent with this argument, misreporting firms should be more likely to acquire privately held firms, subsidiaries (rather than stand-alone firms), foreign firms and 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 other firms and 34% more likely to acquire a firm in a different industry, evidence consistent with the fraud concealment hypothesis. The manager s decision to misstate accounting information should also have an impact on bargaining strategy. If fraud concealment motives drive the acquisition, 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. 3 We 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. 4 2 The detection risk hypothesis is silent on target type; we have no a priori reason to believe holding payment method, size, and structure constant that these target characteristics should influence the likelihood that the fraud will be detected during negotiations. 3 We also allow the possibility for detection risk to cause managers to reduce incentives for sellers and capital providers to scrutinize the firm and retain options that allow the acquirer to withdraw from the deal at a low cost. This should manifest in a lower rate of target termination fees and a higher likelihood of deal failure. 4 To supplement the evidence on behavior during the bargaining process, we are currently gathering data on the timing of negotiations, fraud firm manager s propensity to initiate the deal and their involvement in merger negotiations. 3

5 Our work adds to an expanding literature on the relation between accounting choice and investment 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. 5 Our investigation focuses on how the decision to misreport affects subsequent investment decisions and complements recent evidence that suggests managers who 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 the managers who face substantial private costs if the misreporting is detected also have decision rights over the firm s investment policy. Thus, our results also have implications for understanding how agency conflicts influence corporate acquisitions (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 at the expense of shareholders. 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 related literature and provide motivation for the prediction that the manager s decision to engage in fraud affects subsequent acquisition behavior. 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 6. 5 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 2013), how it is detected (Dyck et al. 2010; Dechow et al. 2011; Wang 2013), and the impact of fraud detection on managers, directors, shareholders, and creditors (Karpoff et al. 2008a and 2008b; Fich and Shivdasani 2007; Graham et al. 2008). 4

6 2. Related literature and motivation Our paper is related to the growing literature on the connection between misreporting and investment. The objective of this paper is 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 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 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 a 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 jobs following a pessimistic market reaction to an acquisition announcement are more likely to inflate subsequent performance reports. A third set of studies reverses the direction of causality and focuses on how the decision to misreport in one period affects investment decisions in a subsequent period. Kedia and Philippon (2009) argue that managers that misstate accounting information must then overinvest to 5

7 maintain investors optimistic expectations 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 that overinvestment occurring after misreporting begins is a consequence of managers using optimistic forecasts derived from the misreported information. Our investigation follows this third path and complements existing studies by examining whether misreporting managers turn to the acquisition market to conceal accounting fraud. Specifically, we examine how acquisition intensity, target types, and acquisition structures are affected by the manager s incentives to conceal fraud. A key assumption is that a manager who misreports has strong private incentives to conceal the misreporting from others. Under the fraud concealment hypothesis, the acquisition market offers the manager a number of devices to delay or prevent outsiders from discovering the misreporting. The first of these is to divert attention. The announcement of an acquisition is usually a material 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. Regulators and auditors are involved in reviewing the transaction and related disclosures and understanding the impact of the acquisition on the parent s financial reports. The existence of binding resource constraints implies that allocating effort to analyzing the impact of the acquisition reduces the effort allocated to analyzing the financial reporting integrity of the acquirer. According to a February 1, 2009 article in Mergers and Acquisitions by Ken MacFayden, companies will use M&A to create enough noise that the din distracts regulators from seeing through to the swindle. 6

8 A second and related effect occurs because post-close consolidated financial statements are less transparent and make it tougher to disentangle the performance of the old fraud firm from the newly acquired firm. The required consolidation of the acquirer s and target s financial statements after the close reduces the transparency of financial reporting outputs used by auditors and shareholders and can make post-deal financial statement analysis substantially more difficult. 6 Again, MacFayden writes, Malfeasance in M&A is not confined to target companies either. Deloitte's [Yogesh] Bahl notes, You also see it happen from the acquirer's perspective. Once these companies start mixing up the financials it clouds the level of transparency that once existed. Consistent with this statement, the Department of Justice indictment of HealthSouth executives alleges 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: (e) fraudulently using the acquisition of other companies to conceal fraudulent assets on HealthSouth s books and in its reports. 7 Indeed, academic research in accounting often excludes firm-years with acquisitions because of the lack of consistency in the financial statements, particularly with respect to comparisons of successive balance sheets (Collins and Hribar 2002). 6 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. 7 In Biovail Corporation Securities Litigation, Case No. 03-CV-8917 (RO), the SEC argued that 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. 7

9 A third device to conceal fraud occurs if the target colludes with the buyer and manipulates their own earnings before the deal closes. The buyer achieves this earnings slack by having the target delay revenue recognition until after the deal close or accelerates loss recognition to before the close. Research by Chen, Thomas and Zhang (2013) documents this spring loading of earnings in acquisitions: target firms understate earnings between the deal announcement and deal close, the understatement is driven by overstating expenses, and investors in the acquiring firm do not anticipate it. In a complaint against a senior manager of BYSIS, the SEC contends that the firm improperly recorded as revenue bonus commission income already earned but not recorded by a company BISYS had acquired (SEC vs. Wevodau 2008). A more well-known case involves Tyco s acquisition of Raychem where it was claimed, after the deal was announced in May but before it was completed in August [Raychem employees] were asked by Tyco officials to do such things as accelerate the payment of expenses, hold back the posting of payments received until after the acquisition date which they refused to do and overstate reserves. The implied purpose, they say, was to help boost Tyco's post-acquisition cash flow. (Greenberg 2002). Finally, misreporting buyers can exploit acquisition accounting to recognize hidden losses by relabeling them as merger-related restructuring charges or write-offs of acquired in-process research and development. Put differently, managers use the acquisition to justify one-time charges to the income statement that can be manipulated to include losses that fraud firm managers are concealing from investors. According to an SEC complaint against Peregrine Systems (emphasis added): In the last quarter of fiscal 2001 Peregrine exploited its acquisition of the company Extricity to conceal sham receivables from public view. That quarter Peregrine improperly wrote off $30 million in receivables to the "Acquisition Costs & Other" expense line item of Peregrine's income statement. Through its officers and employees, Peregrine knew at the time, that (a) a substantial portion of these receivables should not have been recorded as revenue in the first place, (b) the receivables were not impaired as a result of the Extricity acquisition, and therefore (c) it was inappropriate to make it appear to the investing public that the write-off related to a non-recurring event. 8

10 Similarly, a Department of Justice indictment against officials at CUC (predecessor of Cendant): defendants WALTER A. FORBES and E. KIRK SHELTON and their coconspirators caused CUC to establish a merger reserve of approximately $127 million to pay expenses arising from CUC's acquisition of Ideon, and additional merger reserves totaling approximately $53 million to pay expenses arising from CUC's acquisition of Davidson, Sierra, and other, smaller companies.the conspirators falsely represented, and caused CUC to falsely represent, to the SEC and the investing public that the size of the Ideon reserve was based on CUC's good-faith estimate of the probable amount of such acquisition-related costs. In fact, the conspirators knew that they had fraudulently included tens of millions of dollars in the Ideon reserve in excess of that amount. 8 During our sample period, managers also had discretion in allocating the purchase price to in-process research and development and immediately expensing it upon acquisition. Current rules preclude immediate expensing of the R&D assets acquired, but still require impairment testing. Based on these assertions about the role of acquisitions in attempts to conceal fraud, we make the following predictions. First, we expect misreporting buyers will be more intensive acquirers, choosing not only to acquire more often, but to acquire more entities and allocate more capital to those acquisitions. Second, we predict misreporting firms will choose targets that reduce the transparency of the firm s financial statements. Third, because fraud concealment benefits are driven by the impact on the reporting environment following the close, we predict that fraud firms will speed up the closing process, especially around fiscal period ends. Lastly, while we cannot observe firms that were not caught, we can observe how long the fraud lasted before discovery by the market. If mergers are successful in delaying fraud, we predict stock price reversals will be slower for those fraud firms that do enter the acquisition market. Acquisitions are a costly way to conceal fraud given the opportunity costs of the manager s time and the capital required to complete the transaction. More importantly, an acquisition can also subject the firm to even greater scrutiny from outsiders. Banks and underwriters provide 8 In a related case, executive at Japanese firm Olympus was recently found to have misled investors by passing millions in unrecognized investment losses through the income statement as merger-related expenses. 9

11 additional scrutiny of buyers seeking capital and target directors have a legal obligation to assess the intrinsic value of the acquirer s stock consideration in many circumstances. This increased scrutiny increases the probability that the manipulation will be detected; by how much is not clear. Thus, a contrasting view of the fraud concealment hypothesis is that the threat of increased detection risk causes managers to pass up acquisitions so as to avoid increased scrutiny of the financial statements by outside parties, including sellers, financing providers, investors, auditors and regulators. If this is descriptive, the manager should be less willing to engage in an acquisition and for deals it does do, should structure them to minimize scrutiny of its accounting information. We discuss specific predictions before each set of tests in Section Data and sample Our basic research question is straightforward: does ongoing financial reporting fraud influence the manager s acquisition decisions? To address this question, we are interested in identifying cases of fraudulent 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 within GAAP (Bergstresser and Philippon 2006), and also capture real changes in the fundamentals of the firm. Ex post measures such as restatements are used extensively (e.g., Efendi et al. 2007; Graham et al. 2008). However, restatements are not always a correction of fraudulent accounting. 10

12 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 and identified by Dechow et al. (2011). 9 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 detection is not straightforward and this could lead to a potential interpretation problem. We handle this two ways. First, we consider the role of fraud detection risk in influencing mergers and acquisitions. As it turns out, the predictions under a detection risk hypothesis either stand alone or run in the opposite direction of the fraud concealment hypothesis. Thus, our evidence allows us to draw inferences on which of these two effects, if any, dominate. Second, we consider an alternative measure of fraud based on an expectations model that is estimated using, but does not directly rely on, alleged cases of fraud. Finally, 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 2012), as they typically use fraud firms that were actually caught to define their sample. 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 9 We are grateful to Patty Dechow, Weili Ge, Chad Larson, and Richard Sloan for making this data available. 11

13 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 earlier periods. Because of this, the period directly preceding the alleged fraud is unlikely to serve as an effective control for expected nonfraud 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 structure. 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, own no more than 50% prior to the acquisition and value the target s stock at no less than $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, 20% of fraud firms make at least 12

14 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 about $90 million and for which sales approximate $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 typically larger. 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 nonfraud 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 13

15 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 < 0.05). 4. Evidence on accounting manipulation and corporate acquisitions 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. Under the fraud concealment 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. A detection risk explanation predicts the opposite. 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: 1 = (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 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. 14

16 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 < 0.01). 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 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 an acquisition. 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 0.74 more acquisitions per year. 12 The level of increased activity translates into an average of $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 Moreover, the farther back in time the SEC must go, the more likely the evidence necessary to prevail is incomplete or missing. 12 Obviously, the acquirers in our sample buy entire firms, not fractions (the unit of measurement is the whole firm, and 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. 15

17 value of all acquisitions by firm-year, the results in Column 3 show that deals by fraud firms add 7 percentage points more to the their asset base than deals by non-fraud firms (t = 1.84). Overall, the results support the fraud concealment prediction that misreporting firms tend to do more acquisitions. A related question is whether these acquisitions actually increase net investment by 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 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 their total assets through capital investment and 10% through mergers and acquisitions during the misreporting period. Non-fraud firms show similar growth 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 thus adding substantial complexity to the post-investment financial reporting environment further supporting the fraud concealment hypothesis. The evidence above suggests that concealment benefits matter. However, we recognize the possibility that our control variables may not fully address the endogeneity issue between 16

18 misreporting and acquisition activities. 13 In the next section, we focus on a set of tests that hinge on the prediction that fraud concealment benefits are derived from timely consolidation of target and acquirer Closing speed We consider the impact of misreporting on the time it takes to complete a deal once it is announced. We first 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. 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 3 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 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 about 19 days. 14 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 benefits hypothesis focuses on the timing of deal closings. If closing the deal before the fiscal period ends is important, we expect misreporting firms will accelerate the deal closing to just before the fiscal quarter end. 15 In other words, we expect a higher percentage of deal closings before the fiscal quarter end for fraud firms than for non-fraud firms. For each successful deal, we identify 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 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. 15 We thank Maureen McNichols for suggesting this test. 17

19 the fiscal quarter end date closest to the deal close date. Deals that close in 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 3, 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 deals. In Figure 1b, we focus on targets valued at least 10% of the acquirer. We 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, with the difference significant at the 1% level. 16 In both Figures 1a and 1b, the percentage of deal closings tend to be lower in the first few weeks after the quarter end for fraud firms than for nonfraud firms, with the exception of the first week. In an untabulated analysis, we examine the years before the fraud of the same firm and find no apparent clustering of deal closings around the fiscal quarter end. Taken together, these results provide strong support for the conjecture that misreporting firms decrease transparency by making deals and accelerate the closing dates to get the target s financial information on the combined entity s books as quickly as possible. 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 fiscal quarter, and zero otherwise. We control for firm size and year effects, and the results confirm our univariate findings fraud 16 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 takes 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). 18

20 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 and provide direct evidence that misreporting managers value the decrease in transparency from consolidating a newly-acquired entity Accounting manipulation and target type Under the fraud concealment hypothesis, managers identify those firms that if acquired make the firm s post-closing financial reports more complex for insiders and outsiders to process. Increasing opacity and uncertainty about the true impact of an acquisition on the acquirer s financial information reduces the precision of financial statement analysis for detecting earnings manipulation. Therefore, taking the decision to acquire 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. 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 little 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 (based on firm disclosures) 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 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. If it is between 40% and 50%, two years of audited results must be 19

21 Second, we consider subsidiaries as the target of acquisitions. Acquiring-firm managers have more discretion to negotiate the financial statement impact of a subsidiary acquisition relative to an acquisition of a free-standing private or public firm. Moreover, acquiring firm managers are likely to have relatively more bargaining power as many subsidiaries are sold to satisfy the selling parent s liquidity demands (Officer 2007). Third, we examine whether misreporting firms are more likely to acquire foreign firms. Foreign firm takeovers lead to similar limitations in preacquisition 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 impedes efforts to detect misreporting in the parent firm. In Table 4, 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 4 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 focus on acquisitions of subsidiary versus stand-alone public and private firms. We find that misreporting firms are 5.4 percentage points more likely to acquire a subsidiary (p = 0.03), suggesting that subsidiary targets provide fraud concealment benefits for disclosed, and if 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. 20

22 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 4, 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). In the final column of Table 4, we investigate whether misreporting firms are more likely to make a diversifying acquisition. 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 go outside their industry when pursuing a target (14.5%/43% 34%). Overall, we interpret the evidence in this section as consistent with the prediction that firms engaged in accounting manipulation use acquisitions to provide fraud concealment benefits. Selecting subsidiary firms and firms in other industries leads to a reduction in transparency that impede the ability of investors, creditors, and regulators to understand what the firm acquired and how the acquisition affects the post-transaction financial reports Deal failure and termination fees If there is a material risk that fraud will be detected during negotiations, the misreporting manager will value the option to withdraw from negotiations. This should manifest in a lower incidence of target termination fees (a payment to the acquiring firm if the target breaks off the 21

23 deal) and a higher likelihood of deal failure. But if fraud concealment benefits dominate detection risk considerations, 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 5, 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 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 that an announced deal will fail (p < 0.01). This implies that fraud firm managers do influence deal success through their more frequent use of provisions such as termination fees. 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 completion rates provide a relatively weak test of the impact of misreporting on deal success or failure. Overall, our evidence is consistent with fraud concealment motives having a measureable impact on the market for acquisitions by misreporting firms. Not only do misreporting firms increase acquisition activity, they push to close deals just ahead of a fiscal period end date. Misreporting firms also tend to acquire subsidiaries and firms in different industries and attempt to increase deal success through termination fee agreements. 22