Using Real Activities to Avoid Goodwill Impairment Losses: Evidence and Effect on Future Performance

Transcription

1 Journal of Business Finance & Accounting Journal of Business Finance & Accounting, 42(3) & (4), , April/May 2015, X doi: /jbfa Using Real Activities to Avoid Goodwill Impairment Losses: Evidence and Effect on Future Performance ANDREI FILIP, THOMAS JEANJEAN AND LUC PAUGAM Abstract: We examine whether managers postpone the recognition of goodwill impairment by manipulating cash flows and the consequences of such a strategy on future performance. According to SFAS 142, an impairment loss must be recognized if the reporting unit s total fair value to which goodwill has been allocated is less than its book value. A growing body of empirical evidence shows that managers delay the recognition of goodwill impairment in accounting books. However, past literature is silent on how managers convince various gatekeepers (e.g., auditors, financial analysts) that recognizing an impairment loss is unnecessary although it seems economically justified. SFAS 142 requires managers to forecast future cash flows to justify the decision to recognize, or not, an impairment loss. Therefore, we predict that managers manipulate upward current cash flows to support their choice to avoid reporting an impairment loss. We also test whether or not this real earnings management is detrimental to future performance. Based on a sample of US firms over the period , we document that firms suspected of postponing goodwill impairment losses exhibit significantly positive discretionary cash flows compared to various control groups. We also find that this real activities manipulation is detrimental to future performance. Keywords: goodwill impairment, earnings management, cash flow management, real activities. 1. INTRODUCTION Accounting for acquired goodwill has been subject to considerable debate. SFAS 142 (now codified in ASC 350), issued in 2001, abolished goodwill amortization but requires goodwill to be tested periodically for impairment using estimates of its current fair value. Estimates of the current fair value of goodwill rely on assumptions about management s future actions, including managers conceptualization and implementation of firm strategy. Such expectations are difficult to verify and audit. Standard setters expect that managers will, on average, use estimates of goodwill s fair value to convey private information about future cash flows. Yet, this view is challenged by agency theory which predicts managers will, on average, exploit the unverifiable in goodwill accounting rules to manage financial reports opportunistically in line with The Authors are from ESSEC Business School, Cergy-Pontoise 95021, France. Address for correspondence: Luc Paugam, ESSEC Business School, Accounting and Management Control, 1 Avenue Bernard Hirsch, Cergy-Pontoise 95021, France. paugam@essec.edu 515

2 516 FILIP, JEANJEAN AND PAUGAM their own private incentives. Ramanna and Watts (2012) analyze a sample of firms with market indications of goodwill impairment, and find a 69% frequency of goodwill nonimpairment. They find some evidence consistent with the prediction of agency theory (positive association between goodwill non-impairment and CEO compensation, CEO reputation concerns, and debt covenant violation concerns). In contrast, they find no support for predictions explaining non-impairment related to the existence of flexibility or with managers possessing positive private information about the firm. Li and Sloan (2011) also show that goodwill impairments lag behind deteriorating operating performance and negative stock returns by at least two years, and Hayn and Hughes (2006) and Jarva (2009) also document a significant delay between economic impairment and the recognition of an impairment loss. In this study, we focus on goodwill impairment for several reasons. First, it plays a critical role for investors ability to monitor managers capital allocation decisions. Goodwill emerges from past acquisitions and impairment reflects management inability to extract value from past acquisitions. Second, goodwill accounts for a significant amount of public firms balance sheet. 1 Third, under SFAS 142, if there is a decline in the value of a reporting unit, goodwill must be tested for impairment, making goodwill the most sensitive asset type to a decline in firm value. Fourth, goodwill, unlike tangible assets, is a specific non-listed asset that is not separable and for which the value is estimated with discounted projected cash flows. It is the most sensitive asset for which impairment tests rely on multiple fair value assessments allowing discretion in the choice to impair the asset or not. 2 Under US GAAP and IFRS, impairment tests are crucial to guarantee timely loss recognition and maintain conservatism of the financial report (Amiraslani et al., 2013; Kim et al., 2013; and Lawrence et al., 2013), as they ensure that assets are not carried at more than their economic value. Standard setters and market regulators have recently echoed concerns regarding untimely impairment in the US and Europe. The SEC expressed concerns as a staff member indicated in 2008 that it would not be reasonable for a registrant to simply ignore recent declines in their stock price, as the declines are likely indicative of factors the registrant should consider in their determination of fair value, such as a more-than-temporary repricing of the risk inherent in any company s equity that results in a higher required rate of return or a decline in the market s estimated future cash flows of the company (Fox, 2008). Hans Hoogervorst, Chairman of the IASB, also acknowledges his concerns about goodwill resulting from business combinations and admits that [g]iven its subjectivity, the treatment of goodwill is vulnerable to manipulation of the balance sheet and the P&L (Hoogervorst, 2012). The European Securities and Markets Authority (ESMA) found that significant impairment losses of goodwill recognized in 2011 were limited to a handful of issuers, particularly in the financial services and telecommunication industry (ESMA, 2013). 3 This remark illustrates regulators concerns about economic 1 Our sample covers all COMPUSTAT observations available from 2003 to Overall, approximately 60% of observations have goodwill. Goodwill accounts for a mean (median) of 16.7% (12.2%) of total assets for observations with goodwill. 2 To some extent, brands are also entity-specific and subject to manipulation. However, in our sample and in contrast with the amount of goodwill, other intangible assets account for a small fraction of total assets. The median value of other intangible assets is only 2.7% of total assets (and brands account for only a fraction of these other intangible assets). 3 The implementation of goodwill impairment testing under IFRS provides similar rules to SFAS 142 since the revision of IAS 36 in 2004.

3 USING REAL ACTIVITIES TO AVOID GOODWILL IMPAIRMENT LOSSES 517 impairments not always being booked in a timely manner. Finally, the press recently discussed insufficient and untimely recognition of economic impairment for goodwill as The Economist explained that bosses go to inordinate lengths to delay recognizing such supposedly irrelevant, non-cash losses (The Economist, 2013). In this paper, we investigate two related research questions. First, we examine how managers postpone goodwill impairment recognition. We investigate whether managers use real activities to improve current cash flows in order to delay goodwill impairment recognition. The reason for engaging in such a costly strategy is motivated by the audit process of goodwill impairment testing. Auditors, along with other gatekeepers (e.g., financial analysts) base their appreciation about the absence of impairment on business plans developed by management. Such business plans consist in projecting current cash flows over a finite horizon and a terminal value. The higher the level of current cash flows, the more reasonable high level of future cash flows and terminal value will appear to auditors. This creates an incentive for managers to take various actions in order to increase current cash flows such as cutting discretionary expenditures like research and development (R&D), advertising, or selling, general and administrative (SG&A) expenses. Other means can be used to increase operating cash flows, such as stretching suppliers payables, using products in inventories to meet demand, collecting account receivables faster, and cutting on various operating cash expense. Finally, investment in capital expenditures could also be affected by a willingness to avoid booking economic impairment, as cutting capital expenditures improves free cash flows used in valuation models. We contend that managers enhance the credibility of their business plans by managing upward current cash flows. Therefore, we hypothesize that firms that delay goodwill impairment tend to have abnormally high current cash flows as a result of these actions. Second, we investigate the future performance of firms avoiding impairment recognition. Past research has established that real earnings management is costly as it decreases future growth opportunities. Following this argument, postponing goodwill impairment by using real activities to increase current cash flows should be associated with lower future performance. An alternative view is that under the threat of a goodwill impairment, firms are more efficient and, therefore, better manage their operations and avoid making unnecessary expenditures and expenses. Under this perspective, the use of real activities to delay goodwill impairment should be associated with higher future performance. To discriminate between these two competing arguments, we examine the operating and market performance of firms up to two years after their decision to delay the recognition of an impairment loss. We rely on two identification strategies to detect firms that postponed goodwill impairment. Our first strategy consists in matching goodwill impairers with nonimpairers in the same industry, year and with the closest market-to-book (MTB) ratio of equity at the beginning of the year. The goodwill impairment test is conducted at the level of the reporting units to which goodwill has been allocated, for which limited information is available to outsiders. Nonetheless, a firm s outsiders, including investors, analysts and researchers, often rely on market value at the firm level, such as the market-to-book ratio (MTB), as a benchmark to assess whether book values are overstated (e.g., Beatty and Weber, 2006; Ramanna and Watts, 2012; Roychowdhury and Martin, 2013; Chen et al., 2014; and KPMG, 2014). Suspect firms, according to our first identification strategy, are matched non-impairers as they do not report impairment losses, whereas comparable firms decided to do so. As a robustness test,

4 518 FILIP, JEANJEAN AND PAUGAM we supplement our first strategy with a second, following Ramanna and Watts (2012), based on firms exhibiting a MTB below one. We identify as Suspect firms as all nonimpairers with booked goodwill and two consecutive MTB below one at the end of the fiscal year. When market values continue to be below book values over extended lengths of time, it is likely that managers delay the recognition of goodwill write-downs in order to boost earnings (Roychowdhury and Martin, 2013). Our test samples are drawn from a larger set of 38,667 US firm-year observations spanning from 2003 to Consistent with prior literature, we find that goodwill impairment losses are large and infrequent. On average, approximately 65% of impairers book one impairment loss over a three-year window and write off 55% of beginning of the year goodwill (9% of lagged total assets) in one time. This pattern suggests that, on average, the recognition of goodwill impairment is delayed until it becomes inevitable. As noted by Li and Sloan (2011), goodwill can be seen as the value of intangible assets resulting from external growth activities that take a long time to develop, such as reputation, brand name, good market position and human capital; so it is unlikely that economic goodwill impairment appears and disappears within one year. Roychowdhury and Martin (2013) also explain that untimely asset write-down represent cumulative losses that should have been recognized in prior periods and/or include the effect of earnings bath. They further point out that the timing and size of impairment decisions are related. Impairment recorded with a delay, for example when permanent impairments are more likely, are expected to be larger since such impairments are essentially stronger confirmation of negative eventualities. 4 Our main finding is that firms postponing goodwill impairment in their accounting books manage their current levels of cash flows upward compared to firms that recognize an impairment loss. We find evidence of upward cash flow management using three different proxies of cash flow management: (1) real activities affecting cash flows such as discretionary expenditures and production management, (2) operating cash flow management, and (3) free cash flow management. This pattern of unexpected positive cash flows is consistent with managers manipulating current cash flows to support their choice not to report impairment loss in financial statements. We also show that the importance of cash flow management is greater for firms with larger amount of goodwill in their balance sheet. Additionally, we find that the recognition of impairment loss is associated with big bath accounting among impairers that exhibit large and negative income-decreasing abnormal accruals (excluding the impairment loss) in the year of impairment. Our findings hold after controlling for factors affecting earnings management and cash flow management such as firm size, sales growth, financial leverage, change in the financial structure, audit quality, and analyst coverage. We also show that non-impairers that are likely to carry impaired goodwill exhibit a lower change in future operating performance, present lower future stock returns and cumulated abnormal returns than impairers over one-to-two years after impairment avoidance. These results are consistent with the argument that these unexpectedly high levels of current cash flows of firms that delayed impairment through 4 Some large negative shocks could trigger large goodwill impairment loss but the data reveals a systematic pattern in the recognition of very large goodwill impairment losses by a majority of impairers. Large economic shocks should be less frequent than small shocks. By contrast, the frequency of small goodwill impairment losses, which would be consistent with economic impairment, seems abnormally low.

5 USING REAL ACTIVITIES TO AVOID GOODWILL IMPAIRMENT LOSSES 519 manipulation of production, R&D expenses, advertising, SG&A, sales terms or capital expenditures are detrimental to future performance. Our paper makes several contributions to the literature. First, we contribute to the goodwill literature by examining how management achieves a delay in the recognition of goodwill impairment. Goodwill is usually the first asset in listed firms balance sheet (in weight), and goodwill impairment is usually large and has a significant impact on earnings. In the future, in developed and mature economies, goodwill is likely to increase as a percentage of total assets as firms are more likely to grow through mergers and acquisitions than through organic growth. The reliability of mark-to-model fair values required by the current impairment regime is challenged. Impairment of goodwill does not allow investors to monitor managers capital allocation decisions. We shed light on how managers may influence modeled values as we demonstrate that cash flows (real activities) management is likely to be a tool used to support the non-recognition of economic impairment. This result informs standard setters, auditors, financial analysts and regulators regarding the implementation and potential shortcomings of current goodwill impairment testing rules. Second, we also bring a methodological contribution. Past literature identified Suspect firms, i.e., non-impairers, likely to carry economically impaired goodwill, as firms with a MTB ratio below one over one or two years that do not impair goodwill. This strategy is relatively stringent and ensures that identified firms carry impaired goodwill (low type I error ) as the market is unlikely to be inefficient over an extended length of time. However, such an approach potentially overlooks firms with MTB above one but still carrying impaired goodwill (high type II error ). 5 As explained by Lawrence et al. (2013), due to (unconditionally) conservative accounting, some assets could require impairment even though the aggregate MTB is higher than one (see also Beaver and Ryan, 2005). Our approach relies on identifying Suspect firms by selecting non-impairers that can be matched with impairers that have similar characteristics (in particular in terms of year, industry and lagged MTB). Such a strategy allows larger samples to be obtained and thus stronger statistical inferences to be drawn. Finally, we also provide additional evidence on the future performance effect of real activity management. Prior literature is inconclusive. While Gunny (2010) finds that real activities manipulation to meet or beat earnings forecasts is associated with higher future performance, Bhojraj et al. (2009) find the opposite. In the context of seasoned equity offering (SEO), Cohen and Zarowin (2010) and Roychowdhury et al. (2012) find a negative association between real activity management and SEO underperformance. We find that manipulation of current cash flows motivated by goodwill impairment avoidance may be detrimental to future performance. The rest of this paper is organized as follows. In Section 2, we review accounting regulations and changes in accounting standards related to goodwill accounting and impairment tests. In Section 3, we present related literature and develop our hypotheses. In Section 4, our research methodology is presented. In Section 5, we present the sample and our findings. In Section 6, we discuss additional and robustness tests and we conclude in Section 7. 5 Indeed, goodwill is tested for impairment at the reporting unit level. A firm with a market-to-book ratio above one can also be a Suspect firm.

6 520 FILIP, JEANJEAN AND PAUGAM 2. IMPAIRMENT TESTS: BACKGROUND (i) Impairment Tests Before and After the Application of SFAS 142 SFAS 142, Goodwill and other Intangible Assets, went into effect for calendar-year companies on January 1, 2002 and supersedes APB Opinion 17, Intangible Assets (FASB, 2001). 6 Goodwill is no longer amortized and is instead tested for impairment at least annually. Like other assets in the balance sheet, goodwill has always been subject to a test for impairment under SFAS 121, Accounting for the Impairment of Long-Lived Assets and for Long-Lived Assets to be Disposed of, now superseded by SFAS 144, Accounting for the Impairment or Disposal of Long-Lived Assets. That is, if there were reasons to suspect that the amount recorded in the balance sheet was no longer recoverable from future operations, it was tested to see if it had, in fact, diminished in value. Aetna, Black & Decker, Disney, JDS Uniphase and Lucent are firms that impaired goodwill under the goodwill amortization regime, that is, prior to 1 January Nevertheless, SFAS 142 introduced two major differences regarding impairment testing related to the frequency of testing, and impairment testing procedures. SFAS 142 requires that a firm tests goodwill for impairment at least once a year. The test can be performed at any time during the year as long as it is done consistently. In addition, goodwill should be tested for impairment if an event or circumstances change that would more likely than not reduce the fair value [...] below its carrying value. Examples of such events include (SFAS ): A significant adverse change in legal factors or in the business climate; an adverse action or assessment by a regulator; unanticipated competition; a loss of key personnel; a more-likely-than-not expectation that a reporting unit or a significant portion of a reporting unit will be sold or otherwise disposed of; the testing for recoverability under Statement 121 of a significant asset group within a reporting unit; recognition of a goodwill impairment loss in the financial statements of a subsidiary that is a component of a reporting unit. Regarding impairment testing procedures, SFAS 121 relied on a first stage based on undiscounted cash flows. If the sum of all undiscounted cash flows expected to be generated by the assets with which goodwill was allocated were greater than the amount recorded in the balance sheet for those assets (goodwill included), there was no impairment. APB 17 suggested several methods for measuring impairment of enterprise-level goodwill. One was the market value method, which compared a company s net book value to its market capitalization. If book value exceeded the market capitalization, i.e., when the MTB ratio was below one, the excess carrying amount was written off. This approach is similar to our second identification strategy based on a MTB ratio less than one for two consecutive years. Other suggested methods included undiscounted cash flows and discounted cash flows. SFAS 142 is more specific and implies greater expertise in the field of valuation. It requires the use of a fair value or discounted cash flows test applied at the reporting unit level. A reporting unit is an operating segment or one level below the segment level currently disclosed in footnotes to the financial statements under SFAS 131, Disclosures about Segments of an Enterprise and Related Information. Different reporting units may be tested for impairment at different times. Consequently, we should see 6 SFAS 142 is now codified in ASC 350.

7 USING REAL ACTIVITIES TO AVOID GOODWILL IMPAIRMENT LOSSES 521 more impairments post SFAS 142, as well as more important charges since goodwill is no longer amortized. To implement impairment tests under SFAS 142, the standard requires that goodwill and all other assets and liabilities associated with an acquired business be allocated to one or more reporting units following the acquisition (SFAS 141, Business combinations). The purchase price of the target entity has to be allocated to the fair value of identifiable tangible and intangible assets. The identification and valuation process of acquired tangible and intangible assets, namely purchase price allocation, implies that goodwill has to be considered only as a residual. Goodwill is divided among the reporting units that are expected to benefit from the synergies of the business combination. Once the goodwill has been allocated to reporting units, issuers apply a two-step test. In a first step, the issuer compares the fair value of the reporting unit to its book value. The fair value of the reporting unit is defined as the amount at which the unit could be bought or sold in a current transaction between willing parties (SFAS 142). If the reporting unit is publicly traded, its market value would provide the best evidence of fair value. However, it is rare that firms can simply rely on a quoted market price. Managers use some combination of other methods to determine fair value, all involving a significant amount of judgment (Hilton and O Brien, 2009). Managers usually acknowledge that they use specific assumptions for impairment testing purposes in annual reports. The SFASB s preferred approach for estimating fair value is a discounted cash flow (DCF) model. The fair value estimate of the reporting unit is highly dependent upon the assumptions that management makes, in particular projected cash flows, growth assumption and the discount rate. Fair value can also be based on multiples of earnings, revenues or other performance measures such as earnings before interest and taxes (EBIT) or earnings before interest, taxes, depreciation, and amortization (EBITDA), when appropriate. Once the fair value of the reporting unit is computed, it is compared to the book value. If the fair value is greater than the book value, the impairment test ends and there is no charge to earnings. If the fair value is less than the book value, the impairment test continues to step two. Step two consists of comparing the implied fair value of goodwill to the recorded value. 7 The fair value of the reporting unit computed in step one is allocated to the individual assets and liabilities in the reporting unit, as if a new purchase price allocation was performed under SFAS 141 (note that it is only done to test goodwill for impairment; the actual assets and liabilities of the reporting unit are not revalued). The excess of the fair value of the reporting unit over the sum of the fair values of the net assets is the implied goodwill. The impairment charge is the excess of the goodwill recorded on the balance sheet over the implied value of goodwill. However, while the charge resulting from this test is limited to the recorded goodwill alone, there may be additional write-downs resulting from the application of SFAS 144 to the other assets of the reporting unit. Once written-down, goodwill cannot be written-up, unlike other assets. 7 Step two is one of the main differences between SFAS 142 and its IFRS counterpart: IAS 36. Under IAS 36, impairment charge is simply equal to the difference between the book value and the fair value of the reporting unit (called the cash generating unit).

8 522 FILIP, JEANJEAN AND PAUGAM (ii) Costs and Benefits of Untimely Impairments for Stakeholders External parties, such as investors, do not observe the fair value estimates and estimation procedure that underpin managers impairment decisions. This is particularly true for unlisted entity-specific assets such as goodwill (Roychowdhury and Martin, 2013). SFAS 142 provides managers with discretion regarding the choice to record asset write-downs. This discretion can either be exploited opportunistically in line with managers own private incentives or used to reflect the economic reality. The timeliness of asset impairment has important consequences for firms main stakeholders, i.e., managers, shareholders and debt holders. Such consequences are exacerbated in the case of goodwill that emerges from large and strategic capital allocation decisions. For managers, as Roychowdhury and Martin (2013) explain, goodwill impairment is essentially admission of failure to extract value out of past acquisitions as goodwill impairments are a signal of overpaid acquired firms. Managers are likely to benefit from delayed impairment and overstated earnings, at least in the short term, in the form of higher compensation. Muller et al. (2012) show that managers are more likely to trade shares of their own company before announcing goodwill impairments. Delayed impairments also allow managers to protect their reputation, avoid debt covenant violation and grant managers a flexibility option as they can wait to receive confirming signals that goodwill is permanently impaired. Nonetheless, delaying the recognition of economic impairment can also be costly for managers over the mid-term. Delaying goodwill impairment is likely to increase the amount of future impairment booked in a single time as goodwill economic impairment loss accumulates. However, this may also allow managers to use impairment as part of big bath accounting (Roychowdhury and Martin, 2013). Managers reputations may also be at greater risk if goodwill impairments are too ostensibly delayed. Overall, empirical evidence suggests that, on average, benefits from delaying goodwill impairment tend to be greater than costs associated with delayed impairment. For shareholders and potential investors, delayed goodwill impairments exacerbate information asymmetries. Outsiders are at an information disadvantage relative to insiders about the economic value of goodwill, a non-separable asset capturing expected benefits from the efficient and effective management of other assets. The Securities and Exchange Commission (SEC) required additional disclosures to be provided in management discussion and analysis about goodwill impairment (Carnall et al., 2009). As explained by Singh (2014): Goodwill write-offs, if done in a timely manner, are of interest in terms of the signal they send about the value of the company s intangible assets, the company s future earnings prospects, and an assessment of the amount paid for acquisitions. In a survey conducted by KPMG (2014), one respondent argues that Goodwill is relevant to assess the financial outcome of the original decision made and to hold the Board and senior management accountable for capital allocation decisions. Delayed impairments are harmful for investors as they lose a critical informative signal. However, existing shareholders may benefit from delayed impairments as they prevent the violation of debt covenant contracts. To the extent that the market does not realize that goodwill is economically impaired, in case of equity offering, delaying impairment also allows new shares to be issued at a higher price, reducing the dilution effect for existing shareholders.

9 USING REAL ACTIVITIES TO AVOID GOODWILL IMPAIRMENT LOSSES 523 Overall, debt holders appear to be net losers of untimely impairments. Delaying the recognition of goodwill write-downs reduces protection against inappropriate investment decisions, reduces the effectiveness of debt covenants, and increases the ability to distribute a firm s wealth to shareholders due to inflated earnings. (iii) The Role of Auditors There are limits to managers ability to delay goodwill impairments imposed by institutional and governance factors. Auditors have an important role as they have a fiduciary duty to limit accounting choices to make sure that they respect the main objective of accounting standards and that managers do not circumvent the spirit of GAAP. Roychowdhury and Martin (2013) explain that auditors assess the propriety of managers accounting choices including those pertaining to asset write-downs. International Auditing Standards (ISA) also prescribe auditors to verify the estimates of fair value of goodwill (see ISA 600). However, little is known about the exact nature of work that auditors undertake and the extent of the emphasis they put on reviewing goodwill impairment tests. In the US, auditors do not provide details to outsiders about the work they perform to audit firms financial statements. Conversely, in some European countries, such as France, audit reports contain a justification of assessment section, which provides (limited) information explaining to outsiders the main reasons for the opinion. In the justification of assessments, auditors generally include about half a page of discussion on the estimates and assumptions made by management. Much of that discussion is related to estimates and assumptions used by management in testing goodwill for impairment (as well as for other topics such as accounting for post-employment benefits and provision for risks). We include in an Appendix several examples of the statements that can be found in the justification of assessment of French auditors reports. There is no reason to think that French or other European auditors perform their work more diligently than their US counterparts, therefore, it is likely that auditors in the US also pay close attention to impairment tests. Assumptions made by management appear to be critical in the assessment of impairment tests by auditors. If auditors scrutinize assumptions chosen by management, an alternative or supplementary approach to delay the recognition of goodwill impairment is to improve the real level of current cash flows. 3. HYPOTHESES Prior studies which focus on goodwill impairment can be classified into two categories. The first category examines the determinants of goodwill impairment timeliness. Godfrey and Koh (2009) identify US listed companies reporting goodwill impairment losses from 2002 to 2004 and focus on firms motives for reporting goodwill impairment losses. They find that impairment write-offs are negatively associated with firms underlying investment opportunities and return on assets. They interpret their findings as an indication that the impairment regime reflects firms underlying economic attributes. Consistent with this conclusion, Ahmed and Guler (2007) provide evidence that goodwill write-offs and goodwill are more strongly associated

10 524 FILIP, JEANJEAN AND PAUGAM with stock returns and stock prices after the adoption of SFAS 142 than before. Chen et al. (2008) find that impairments under SFAS 142 are more timely than under prior rules. However, they also find that the market anticipates impairment losses subsequent to initial adoption. They argue that timeliness can be improved further after the adoption of SFAS 142. Indeed, multiple incentives exist for managers to avoid or delay impairment recognition including debt and compensation contracts (Watts and Zimmerman, 1986) or management reputation (Francis et al., 1996). As noted by Rockness et al. (2001): The effects [of goodwill impairment] on financial results and ratios will be very significant in years of impairment, and it is hard to see how fair values for goodwill will be objectively determined. The new impairment charges are prime candidates for movable expenses from one period to another to achieve desired earnings targets. Jarva (2009) examines a sample of non-impairment firms in which there are indications that goodwill is impaired. He fails to find convincing evidence that these firms are opportunistically avoiding impairments. By contrast, Ramanna and Watts (2012) provide evidence that non-impairment is increasing with financial characteristics that serve as proxies for greater incentives to avoid goodwill impairment and unverifiable fair value-based discretion (CEO compensation, CEO reputation, debt covenant violation concerns and managers flexibility under the SFAS 142 write-off rules). They fail to find evidence that non-impairment is due to managers possession of favorable private information about the firm. A second category of studies analyzes the stock market consequences of goodwill (non-) impairment. Impairments of assets are typically perceived as a negative asset pricing signal by market participants (Fields et al., 2001). However, Li and Sloan (2011) show that market participants tend to anticipate goodwill impairment as most of the (stock) under-performance of firms that recognize goodwill impairment occurs prior to the actual impairment charge, indicating that in general, investors are aware of the issues that may lead to a subsequent impairment long before the actual impairment is recognized. Hayn and Hughes (2006) also demonstrate that goodwill impairment lags behind the economic impairment of goodwill and estimate that the average lag ranges between three to four years. Jarva (2009) investigates a sample of firms reporting goodwill impairments in , and find that SFAS 142 goodwill writeoffs have significant predictive ability for one- and two-year-ahead cash flows. This is consistent with the notion that goodwill impairments are, on average, more closely related to economic factors than to opportunistic behavior, but does not imply that goodwill impairments are always timely. Jarva (2009) also finds indications that SFAS 142 goodwill write-offs lag behind the economic impairment of goodwill. Overall, this literature shows that impairment as prescribed by SFAS 142 provides useful indications to market participants and better reflects the economic nature of goodwill than the previous amortization regime. However, it seems that managers use the discretion to delay the accounting recognition of goodwill write-offs. In other words, goodwill impairments are subject to a high degree of manipulation by managers due to the subjectivity permissible by accounting standards and managers use this discretion to withhold the disclosing of bad news consistent with Kothari et al. (2009). If motives to use this discretion are, to some extent, documented by various studies, the means by which managers can convince auditors or financial analysts that the recognition of an accounting goodwill impairment is unnecessary are still unknown. We contribute by filling this gap in the current literature.

11 USING REAL ACTIVITIES TO AVOID GOODWILL IMPAIRMENT LOSSES 525 According to SFAS 142, impairment tests are typically based on the discounted cash flow (DCF) model, which is the preferred method. Therefore, the outcome of the test is affected by different parameters: the discount rate used, the growth assumption chosen over the business plan (usually a three- to five-year cash flow projection period), the duration of short-term growth during the business plan, the percentage of profit margins, the choice on whether or not to include an extrapolation period to go from short- to long-term growth, and the terminal growth assumption used to compute terminal value (Penman, 2006). As noted by Ramanna and Watts (2012), most of these parameters are subject to a degree of managerial discretion. Therefore, gatekeepers of the firm financial reporting processes (e.g., auditors, audit committee, financial analysts) may challenge forecasts on the starting point of the DCF model, i.e., the operating cash flow generated for the current year. 8 The current cash flow level can have a material impact on the amount of terminal value that is used to project cash flows after the explicit horizon of the business plan. 9 This starting point is all the more important in that the current year cash flow is usually considered as the starting point for any valuation (Penman, 2006). In addition, Tversky and Kahneman (1974) showed that economic agents (including managers and auditors) tend to use an anchoring-and-adjustment heuristic for estimating unknown quantities. In other words, forecasting starts with information one does know and then adjusts until an acceptable value is reached. This anchoring bias is widely documented (Epley and Gilovich, 2006). Auditors can accept a degree of optimism from managers choices related to these parameters but will ultimately limit managers discretion if forecasted values are inconsistent with the current level of cash flows. Since current cash flows affect the outcome of the impairment test, we hypothesize that firms will try to avoid or delay the recognition of impairment by improving current cash flows. Formally, we test the following hypothesis (stated in alternate form): H1: Firms facing pressure to recognize a goodwill impairment loss are more likely to manage upward current cash flows. Such real activities management (RM) may take various forms, such as a manipulation of production, a reduction of R&D, advertising, and SG&A expenses (Roychowdhury, 2006), or of capital expenditures. Cash flow manipulation through real activities management can also be achieved by stretching suppliers payables, collecting account receivables faster, selling off inventories, or cutting various operating cash expenses. The effect of real activities management on future performance is a priori unclear. Managers may use real activities as a signaling mechanism (Gunny, 2010). Gunny (2010) argues that if real activities are used to meet or beat benchmarks, then managers credibly signal their quality as real activities management is seen as costly (Graham et al., 2006). In addition, real activity management may provide benefits to the firm that enables better performance in the future. For example, Bartov 8 The same argument applies if management chooses another valuation method such as a multiple of EBITDA, as the higher the current level of EBITDA, the higher the value of the reporting unit by applying a multiple to current EBITDA. 9 Terminal values, which are usually computed based on the last cash flow forecast of the business plan, usually account for more than 60% of the total value of a firm/reporting unit.

12 526 FILIP, JEANJEAN AND PAUGAM (1993) provides evidence consistent with managers selling fixed assets to avoid debt covenant violations. Alternatively, managers have opportunistic motives in adopting real activities management strategies (Cohen and Zarowin, 2010). Cohen and Zarowin (2010) report that managing earnings at the time of the seasoned equity offering (SEO) via real activities is more costly in the long run than doing so via accruals. Roychowdhury et al. s (2012) findings reveal that both accruals manipulation and real activities manipulation are inducing over-valuation at the time of a SEO. However, earnings management is most closely and predictably linked with post-seo stock market under-performance when it is driven by real activities manipulation. Such an adverse consequence of real activities management on future performance is not limited to the SEO setting. For instance, by contrast to Gunny (2010), Bhojraj et al. (2009) show that firms that beat analyst forecasts by using real and accrual earnings management have worse operating performance and stock market performance in the subsequent three years than firms that miss analyst forecasts without earnings management. Potential effects of real activities management to avoid goodwill impairment are also two-sided. On the one hand, managing production levels to increase cash flows, delaying or cutting SG&A, advertising, R&D expenditures, or spending less on capital expenditures in order to improve current cash flows is likely to harm long-term performance by reducing investment opportunity set. Under the efficient-market hypothesis (EMH), stock prices of firms using sub-optimal investment or production levels to manage their current cash flows will be lower once it becomes known to market participants, in turn reducing stock returns once investors realize this sub-optimal strategy is adopted to increase current cash flows. This would imply a negative relationship between real activities management and future performance in our sample. On the other hand, the threat of having to recognize an impairment loss may foster a more efficient use of available resources. Under this approach, cutting discretionary expenses, capital expenditures and managing the level of output to increase current period cash flows could actually reflect effective control over investments and production. In that case, investors would react favorably to firms engaging in efficient resource allocation. Given the above arguments, we state H2 as follows: H2: Managing cash flows to avoid or delay the recognition of an accounting impairment has no consequence on the long-term performance of the firm. (i) Overall Research Design 4. RESEARCH DESIGN In this paper, we examine potential cash flows manipulation in order to avoid or delay the recognition of economic impairment. This requires identifying firms that should impair goodwill but decide to postpone or avoid impairment recognition in their books. In other words, we have to identify firms subject to an economic impairment but which decided to avoid the accounting impairment. The goodwill impairment test is conducted at the reporting unit level. However, reporting units fair values are usually not observable. The only possible option to assess if managers avoid the

13 USING REAL ACTIVITIES TO AVOID GOODWILL IMPAIRMENT LOSSES 527 recognition of goodwill impairment consists in using firm-level data. 10 The MTB allows comparison of the market expectation of future cash flows generated by the firm (numerator) to the carrying book value of net assets (denominator). Such properties of the MTB led researchers to use it to identify firms carrying impaired assets in many empirical studies (e.g., Beatty and Weber, 2006; Hilton and O Brien, 2009; Ramanna and Watts, 2012; Lawrence et al., 2013; Roychowdhury and Martin, 2013; Chen et al., 2014). The SEC also considers that MTB is a relevant indicator of potential goodwill impairment. In a survey about the usefulness of the goodwill impairment test, KPMG (2014) also explains that although interviewees agreed that a MTB less than one does not automatically warrant an impairment loss, it is an important indicator that requires further assessment of a potential goodwill impairment. Given that economic impairments are unobservable, we rely on two strategies to identify Suspect firms, i.e., firms carrying impaired goodwill but not recognizing an impairment. Since MTB below one is merely an indicator, to ensure the robustness of our results, we rely on a first strategy that allows the identification of firms with a MTB that could be above one but that are likely to carry impaired assets. Our first strategy consists in matching firms that booked an impairment (Control group) with firms carrying goodwill that did not book an impairment but belong to the same industry (two-digit SIC code) during the same year and have the closest lagged MTB. 11 Firms in a given industry-year are likely to be subject to the same current economic conditions. If, for a given valuation level, a firm books an impairment, it is likely that a competitor with the same valuation and no impairment loss carries impaired goodwill. Under this assumption, we label non-impairing firms that are matched with firms that book an impairment as Suspect firms. This approach allows us to obtain a relatively large number of Suspect firms from which we can draw statistical inferences. As a robustness test, we also rely on a second identification strategy similar to Ramanna and Watts (2012). It consists in identifying firms that (1) carry goodwill, (2) did not book an impairment and (3) exhibit a MTB lower than one for two consecutive years. A MTB below one over a short period could indicate that managers have positive private information about the true economic value of goodwill and that the market is potentially inefficient. However, over a longer length of time, market inefficiencies are unlikely, and a MTB below one during two years is likely to indicate that goodwill is permanently impaired. Firms with MTB below one for an extended period of time can reasonably be considered as carrying impaired goodwill. This approach is suggested in several empirical studies to identify impaired goodwill (Hayn and Hughes, 2006; Ramanna and Watts, 2012; Roychowdhury and Martin, 2013) and is suggested by market regulators such as the SEC in the US or the ESMA in Europe as a proxy for goodwill impairment. Under this identification strategy, the likelihood that identified Suspect firms are not carrying economically impaired goodwill (type I error) is low. However, this identification strategy is rather stringent as firms with MTB above one may also have impaired goodwill. The likelihood of failing to identify true Suspect firms (type II error) is relatively high. For instance, a firm with operations across two 10 In Section 6, we conduct a robustness test and focus on firms with one large reporting segment for which firm-level data are more relevant. 11 In our matched samples, we restrict the difference in market-to-book ratio between impairers and matched non-impairers to a maximum of 0.25 to allow meaningful comparisons between Suspect firms and Control firms.

14 528 FILIP, JEANJEAN AND PAUGAM lines of business may have economically impaired goodwill in one reporting segment although its firm-level MTB is above one. The operations in a profitable line of business could compensate the low performing operations in the other. 12 Figure 1 summarizes our empirical approach. In our first identification strategy, we first consider firms that booked an impairment (i.e., from Groups 1 and 2) and we find a match, i.e., a firm comparable in terms of industry, year and lagged MTB from Groups 3 and 4. Any matched firm is considered as Suspect and is compared to its control. In our second identification strategy, we identify Suspect firms from Group 3 as firms carrying goodwill, without impairment and a MTB below one during two years. These firms are then compared to firms drawn from Group 1 (firms with similar MTB but that booked an impairment). YES Economic Impairment Accounting Impairment YES Group 1 Group 2 NO Group 3 Group 4 NO To test H1, we examine the level of real activity management strategies used to increase current cash flows in the year of identified economic impairment. We also study the level of accruals management since prior literature indicates that delayed asset impairment can be associated with big bath accounting (Li and Sloan, 2011; Roychowdhury and Martin, 2013). For this sub-sample of firms, we examine whether they exhibit (1) higher abnormal level of current cash flows, and (2) higher incomeincreasing abnormal accruals (excluding the impairment loss) than the Control group firms after controlling for various factors affecting accruals and real activities management. Finally, to examine whether cash flow management harms or enhances future performance (H2), we compute and compare for Suspect firms and Control firms change in adjusted return on assets (ROA), raw and cumulative abnormal returns over one and two years following the suspected impairment avoidance. (ii) Measuring Real Activities Management to Improve Current Cash Flows We measure the use of real activities by management to increase current cash flows with three different proxies: (1) real activities affecting cash flows identified in the real earnings management literature (RM) (Roychowdhury, 2006; Cohen et al., 2008; Cohen and Zarowin, 2010; and Zang, 2012), (2) operating cash flow management (CFM), and (3) free cash flow management (FCFM). We note that these three measures somehow overlap but they allow cash flow management to be captured across several dimensions. First, following Roychowdhury (2006), we examine real activities management intended to increase cash flows across two dimensions: (1) cutting discretionary 12 We examine this possibility in Section 6, where we focus on firms with a large business segment to which goodwill is most likely allocated.