Goodwill Impairments after the Implementation ... - University of Tilburg [PDF]

Goodwill Impairments after the Implementation of IFRS 3 A Tool for Earnings Management?

M. Stumpel August, 2012

Goodwill Impairments after the Implementation of IFRS 3 A Tool for Earnings Management?

Master Thesis Author:

Michelle Stumpel

ANR:

60.87.89

University:

Tilburg University

Faculty:

Tilburg School of Economics and Management

Department:

Accountancy

Supervisor Tilburg University:

S. van der Meulen (08-03-2012 - 31-05-2012)

Supervisor Tilburg University:

A.J.M. Verriest (31-05-2012 – 30-08-2012)

Second reader Tilburg University:

J.P.M. Suijs

Company:

KPMG

Supervisor KPMG:

W.M. Rijnen

Date of completion:

17-08-2012

ACKNOWLEDGMENTS On August 30, 2005, I started as a student at Tilburg University and today, exactly 7 years later, I am leaving with a Master of Science degree. When someone says being a student means so much more than just studying, I could not agree more. During my time at Tilburg University I got to know all the aspects of a student’s life. From living with roommates and partying 4 to 5 nights a week, to being active at a student’s association and organizing one of Tilburg University’s largest events, the Economic Business weeks Tilburg. It all contributed to who I am today. Despite the 7 years it took for me to graduate, I would not want to change a thing and I will carry the numerous good memories with me when taking the next steps into life. There are several people I would like to thank for making my graduation possible. First of all, thanks and appreciation go to Sofie van der Meulen and Arnt Verriest. Privileged by having 2 supervisors, their suggestions and feedback encouraged me to think critically and improve my thesis which resulted in a work I am very proud off. Next, I would like to show gratitude to my supervisor from KPMG, Willem Rijnen. As he was always available for discussions and help, his notes and guidance were of great support during my internship at KPMG. Furthermore, a special thanks goes to my sweet and supporting friends who made the 7 years in Tilburg both fantastic as well as unforgettable, and it would have not been the same without them. I am positive that the future will bring us many more great and memorable moments. Gratitude goes to Suzan Pepers for being my statistics hero and to Josine Jansen for her critical notes and feedback. Last but certainly not least, thanks and great appreciation go to my parents and sister. Their unconditional support and love means the world to me, and it is great knowing they will always stand by my side.

Michelle Stumpel Amstelveen, August 2012

ABSTRACT This study examines the influence of reporting incentives in determining the timing and the amount of an IFRS 3 goodwill impairment in the period 2006-2009. The substitution of historical cost accounting with a fair value-based approach increased managers’ discretion in deciding on the fair value of goodwill and the amount written off. Empirical results reveal that executive turnover affects the impairment decision, as firms with a recently installed CEO are more likely to report a (larger) goodwill impairment. Results are also consistent with big bath accounting behavior. Firms with an unexpected negative earnings surprise seem to be accelerating the impairment recognition. The inferences hold after controlling for economic factors and other determinants of goodwill writeoffs. Overall, the results imply that reporting incentives affect the accounting choices relating to the timing and amount of goodwill written off, and it can therefore be concluded that goodwill impairments are used as a tool to manage a firms’ earnings. The study contributes to the accounting literature by expanding the knowledge on the accounting consequences of IFRS 3 and the impairment testing approach. Additionally, an alternative and indirect method for investigating earnings management is explored.

TABLE OF CONTENTS 1.

INTRODUCTION .............................................................................................................................................................. 7

2.

LITERATURE REVIEW .....................................................................................................................................................10 2.1 2.2 2.3 2.4 2.5

3.

ACCOUNTING FOR INTANGIBLE ASSETS .................................................................................................................................. 10 SFAS 141 AND 142 ......................................................................................................................................................... 11 IFRS 3 ........................................................................................................................................................................... 12 EARNINGS MANAGEMENT................................................................................................................................................... 14 RESEARCH QUESTION ......................................................................................................................................................... 15

HYPOTHESES DEVELOPMENT ........................................................................................................................................17 3.1 MANAGERIAL REPORTING INCENTIVES ................................................................................................................................... 17 3.1.1 Change in senior management ............................................................................................................................... 18 3.1.2 Debt-covenant hypothesis ...................................................................................................................................... 19 3.2 EARNINGS MANAGEMENT................................................................................................................................................... 19 3.2.1 Big bath accounting ................................................................................................................................................ 20 3.2.2 Income smoothing .................................................................................................................................................. 21

4.

DATA AND METHODOLOGY...........................................................................................................................................22 4.1 SAMPLE SELECTION ........................................................................................................................................................... 22 4.2 RESEARCH MODEL ............................................................................................................................................................. 24 4.2.1 Dependent variables ............................................................................................................................................... 26 4.2.2 Reporting incentives ............................................................................................................................................... 26 4.2.3 Control variables ..................................................................................................................................................... 27

5.

RESULTS ........................................................................................................................................................................30 5.1 5.2 5.3 5.4

6.

SENSITIVITY ANALYSES ..................................................................................................................................................44 6.1 6.2 6.3 6.4

7.

DESCRIPTIVE STATISTICS..................................................................................................................................................... 30 CORRELATIONS ................................................................................................................................................................ 34 LOGISTIC REGRESSION ....................................................................................................................................................... 38 MULTIPLE LINEAR REGRESSION ........................................................................................................................................... 40

DEBT TO EQUITY RATIO ...................................................................................................................................................... 44 TOTAL ASSETS AS MEASURE OF FIRM SIZE ............................................................................................................................... 46 CONTROL FOR COUNTRY EFFECTS.......................................................................................................................................... 46 PROXY BIG BATH ACCOUNTING ............................................................................................................................................ 47

CONCLUSION, LIMITATIONS, AND FUTURE RESEARCH ...................................................................................................49 7.1 7.2 7.3

CONCLUSIONS .................................................................................................................................................................. 49 LIMITATIONS ................................................................................................................................................................... 50 FUTURE RESEARCH ............................................................................................................................................................ 50

REFERENCES ...........................................................................................................................................................................52 APPENDIX I .............................................................................................................................................................................55 APPENDIX II ............................................................................................................................................................................56 APPENDIX III ...........................................................................................................................................................................57 APPENDIX IV ..........................................................................................................................................................................60

1.

INTRODUCTION

Together with the growing significance of intangible assets and goodwill for companies worldwide, the debate surrounding the recognition and valuation of these assets also intensified. In response to the critics, the Financial Accounting Standards Board (FASB) issued Statement of Financial Accounting Standards (SFAS) No. 142 – Goodwill and Other Intangible Assets, which fundamentally changed the accounting for goodwill. The new standard has eliminated goodwill amortization and requires firms to test their goodwill balance for impairment. In 2004, the International Accounting Standards Board followed the FASB by introducing International Financial Reporting Standard (IFRS) 3 Business Combinations, which contains similar goodwill reporting requirements. By substituting historical cost-based measures by a fair value approach, the standard setters aimed at improving comparability, transparency and the decision usefulness of accounting information (Jerman and Manzin, 2008; Hamberg et al., 2011). As the standards also provide managers with more flexibility in determining the value of goodwill and the existence and amount of an impairment, opportunities arise for managers to engage in some sort of earnings management (Massoud and Raiborn, 2003; Van de Poel et al., 2009). Motivated by the possible effects of the increased flexibility and judgment due to the adoption of the new goodwill accounting standards, the current study examines factors determining the timing and the amount of an IFRS 3 goodwill impairment. Specifically, after controlling for economic determinants of the impairment loss, the current study looks at both executive turnover and a firm’s debt contracting affecting the decisions regarding goodwill expense recognition. Additionally, the current study investigates whether the increased managerial discretion in deciding on goodwill impairments results in the presence of two well-known earnings management techniques: big bath accounting and income smoothing. In order to study the effect of reporting incentives on the goodwill impairment decisions, a sample of 487

1

companies listed on the pan-European electronic stock exchange ‘Euronext’ in the period 2006-2009 is identified. The transition year 2005 is deliberately excluded from the current research, as the transitional impairment test in the adoption period may provide additional incentives to accelerate goodwill write-offs. The period 2006-2009 is believed to provide cleaner evidence on the factors affecting the impairment choices. To examine the dichotomous decision whether or not a goodwill impairment charge is reported, a logistic regression model is used. In order to examine the amount of a goodwill impairment and the factors influencing the amount, a multiple linear regression model is used. Both models contain several control variables, such as firm size and industry classification, and also control for economic factors which are expected to affect the impairment decisions.

1

The sample of 487 companies over the period 2006-2009 results in 1748 firm-year observations being used in the current research.

7

The results of the logistic regression indicate that firms are more likely to report a goodwill impairment when they have recently experienced a change in the chief executive officer (CEO) position. This is consistent with the conjecture that newly installed CEOs may exercise greater scrutiny over the value of assets or change the firm’s strategy (Francis et al., 1996; Riedl, 2004). Further results on the logistic regression model show that the probability of taking a write-off is larger for firms with an unexpected negative earnings surprise. Consistent with big bath accounting behavior, firms accelerate the goodwill impairment since they already experience a loss in earnings. The results of the analysis on the determinants of the amount of goodwill written off are consistent with those of the logistic regression. New CEOs report larger impairment losses and the reported impairments are also larger when the firms’ earnings turn out to be below expectations. Additionally, the multiple linear regression indicates that firms with a positive earnings surprise also write off a larger amount of goodwill. This result is consistent with the desire to show smooth earnings patterns and thus with income smoothing behavior. By means of several sensitivity analyses, the findings on goodwill impairments being used to engage in big bath accounting behavior are confirmed. Besides, when using dummy variables to control for country effects instead of a macroeconomic proxy, a significant positive association between a firm’s debt to equity ratio and the magnitude of the impairment loss has been found. Although this result is contrary to the expectation that highly indebted firms delay impairment charges, it could signal demand for conservative reporting by the firm’s debtholders. Overall, the results suggest that managerial reporting incentives affect the firms’ accounting choices relating to the timing and amount of an IFRS 3 goodwill impairment. Hence, goodwill impairments are at least to some extent used as a tool for earnings management. This study contributes to the existing accounting literature in the following ways. First of all, the consequences of moving to a fair value approach, providing managers with a significant amount of discretion, are documented. While the literature on the implementation of SFAS 142 is quite extensive, the accounting consequences of IFRS 3 and the switch to an impairment-only approach are less researched, especially in the period subsequent to the transition year. Next, the current study provides an alternative approach to identifying earnings management behavior. The most common technique for measuring earnings management is trying to isolate the discretionary part of accruals. This research uses a more indirect approach as it examines the increased managerial discretion arising from the provisions in IFRS 3, and the use of the goodwill impairment test as a tool to manage a firms’ earnings. The current research is relevant to society, especially to regulators, standard setters and oversight bodies, as it provides insight on the impact of fair value accounting and the quality of financial statements after the implementation of IFRS. One of the main objectives of the IASB and the IFRSs is to increase the comparability of financial statements, but the increased managerial discretion and judgment and the opportunities to manage earnings do not seem to contribute to achieving this objective. Policy makers as well as investors have to be aware of these implications of the application of IFRS. Additionally, the importance of high quality corporate governance practices needs to be emphasized, in order to obtain high quality and more comparable statements.

8

The remainder of the study is organized as follows. The next section provides theoretical background on the financial accounting standards for goodwill and discusses related research. Section 3 develops the hypotheses and section 4 discusses the sample selection procedure and the research method. Section 5 presents the descriptive statistics and the main empirical results. Section 6 presents several sensitivity analyses and section 7 concludes the current study.

9

2.

LITERATURE REVIEW

This section provides an overview of the existing literature on goodwill and the impairment test in order to substantiate the current research with theoretical background. First, the concept of goodwill and the changes in the financial reporting standards for goodwill are described. Next, earnings management and the methods available to use goodwill impairments as a tool to manage earnings will be discussed. 2.1 Accounting for Intangible Assets In recent years, a new economic era with a growing significance of intangible assets has been entered. Today intangibles such as brand names, licenses, patents and customer relations have become more and more valuable. Since acquisitions remain an extremely popular method to achieve corporate growth (Datta and Grant, 1990; Sirower and O’Byrne, 1998), it is important to adequately measure the value of intangible assets purchased and the goodwill that arises from the acquisition. The problem however is that intangibles in general are difficult to price and particularly goodwill is considered one of the most complex intangible assets (Lhaopadchan, 2010). Goodwill is measured as the excess of the business acquisition price over the fair market value of a target firm’s identifiable net assets (Gara, 2007; Jerman and Manzin, 2008; Hamberg, Paananen and Novak, 2011). According to Baldi (2009), goodwill arises in either of two ways: internally generated goodwill or acquired goodwill (as part of the acquisition of another firm). Due to the fact that goodwill from acquisitions can be derived from a purchase valuation, only purchased goodwill is recorded on the balance sheet. Hence, as only this type of goodwill can be reliably measured, it will be the focus of the current study. Goodwill has been a controversial subject for decades, with the debate focused around the recognition of goodwill as an asset, the treatment of goodwill after its initial recognition, and its link to the income statement (Jahmani et al., 2010). Prior to 2001, Accounting Principles Board Opinion No. 17 (APB 17) Intangible Assets in the United States, and International Accounting Standard 22 (IAS 22) Business Combinations in Europe, both required goodwill to be recognized as an asset and amortized on a straight-line basis over its useful life (IAS 22.44, 1983; IAS 22.50, 1983; APB No. 17, 1970). The useful life of goodwill according to APB 17 covered a period not exceeding 40 2

years, while IAS 22 considered the useful life to be only 20 years. During the 1990s, the number of corporate acquisitions was soaring and investors, regulators and executives desired more adequate information and proper ways to identify, measure and manage intangible assets and goodwill (Jerman and Manzin, 2008). At the same time, the amortization method for goodwill under APB 17 and IAS 22 became criticized since the balance sheet amounts did not adequately reflect the value of goodwill and provided investors and analysts with misleading information about the true value of the firm (Lhaopadchan, 2010).

2

For more specific aspects and attributes of Accounting Principles Board Opinion No. 17 (APB 17) and International Accounting Standard 22 (IAS 22), refer to Table 1.

10

In response to the request for improved standards to account for business combinations, intangible assets, and goodwill, the Financial Accounting Standards Board (FASB) placed the issue on its agenda in August 1996 (Sevin et al., 2007). Three primary goals of the Board were ‘to improve the consistency of the procedures used in accounting for an acquisition of a business’; ‘to improve the relevancy and transparency of information provided to investors, creditors and other users of financial statements’ and ‘to improve international comparability’ (FASB, 2004). The FASB recognized that externally purchased goodwill may have an indefinite useful life, and thus is not a systematically wasting asset that decreases in value on an uniform basis (Massoud and Raiborn, 2003; Gara, 2007). Together, these goals and insights led to a thorough alteration of the reporting treatment for intangible assets and goodwill. 2.2 SFAS 141 and 142 st

On July 1 2001, the FASB released the Statement of Financial Accounting Standards (SFAS) No. 141: Business Combinations and No. 142: Goodwill and Other Intangible Assets, which respectively replaced APB No. 16 Business Combinations and APB No. 17 Intangible Assets, both dating from the early 1970s. SFAS 141 and 142 fundamentally changed the accounting for goodwill, moving from a historical cost approach to a fair value-based accounting method (Lhaopadchan, 2010). The two most significant changes were the elimination of the pooling of interests method of accounting for a business combination, and the shift in accounting for goodwill from the amortization approach to an impairment testing approach (Massoud and Raiborn, 2003). Prior to SFAS 141, firms were able to choose between two methods when accounting for an acquisition: the ‘pooling of interests method’ or the ‘purchase method’. The main difference is that the purchase method recognizes all additional intangible assets, including any goodwill acquired, while the pooling of interests method does not recognize any intangible assets or goodwill besides those that were already recorded by the acquired firm (Jerman and Manzin, 2008). As a consequence, almost identical transactions were accounted for using different methods, with managers showing a clear preference for the pooling of interests method (Van de Poel et al., 2009). Hence, the financial results of firms were difficult to compare because of the different methods available to record a business combination. The elimination of the pooling of interests method by SFAS 141 resulted in all transactions being accounted for using the purchase method. This improved the comparability of information on business combinations provided in financial reports (SFAS 141, 2001). The second major change of SFAS 142, as compared to APB 17, is that the standard takes a different approach on how to account for intangible assets and goodwill subsequent to their initial recognition. The goodwill amortization approach was eliminated and instead firms are required to, at least annually, undertake a two-stage impairment test to evaluate their goodwill balance. The first step is to determine whether an impairment exists, by comparing the fair value of a reporting unit with its book value. Next, when the value of goodwill is considered to be below its current book value, an impairment loss has to be recognized in income from continuing operations, 11

and goodwill has to be written down. The FASB states that these changes in accounting for goodwill improve the financial reporting quality, because the financial statements of acquiring entities will better reflect the underlying economic value of the intangible assets and goodwill. Financial statement users will be better able to understand the investments in these assets and the subsequent performance of investments (SFAS 142, 2001). These suggested improvements are supported by studies of Massoud and Raiborn, (2003), Sevin et al. (2007), and Baldi (2009), however some critical notes can be made. As recognized by the FASB itself, there may be more volatility in reported income than under the previous standards, because impairment losses are likely to occur irregularly and in varying amounts (SFAS 142, 2001). Managers often do not appreciate earnings volatility and therefore may have an incentive to choose a convenient time to recognize impairment losses (Jahmani et al., 2010). In addition, SFAS 142 provides managers with more flexibility in determining the fair value of goodwill, the existence, and the amount of a goodwill impairment (Beatty and Weber, 2006; Lhaopadchan, 2010). 2.3 IFRS 3 In 2001, the International Accounting Standards Board (IASB) was formed to succeed the International Accounting Standards Committee (IASC). Due to an increasing demand for harmonized global financial accounting and the request for similar changes in accounting for intangibles and goodwill (IASB, 2008), the IASB quickly followed the FASB by releasing International Financial Reporting Standard (IFRS) 3 Business Combinations in 2004, which superseded IAS 22 Business Combinations. Moreover, IAS 36 Impairment of Assets and IAS 38 Intangible Assets were revised to adopt an impairment-only approach (Schultze, 2005). Similar to SFAS 141, IFRS 3 prohibits the use of the pooling of interests method and allows business combinations only to be accounted for using the full-purchase method (Watrin et al., 2006). In addition, IFRS 3 also abandons the amortization of goodwill. Instead, the book value of goodwill has to be tested for impairment on a regular basis, but at least once a year, in accordance with IAS 36 Impairment of Assets (Hamberg and Beisland, 2010). This is done by comparing the carrying amount with the recoverable amount. When the carrying amount exceeds the recoverable amount, an impairment loss has to be recognized (IAS 36.90, 2004). The changes in IFRS 3, IAS 36 and IAS 38 resulted in carrying intangible assets and goodwill at their fair value, which increased the relevance and usability of financial accounts for financial statement users (Lhaopadchan, 2010). Simultaneously, the standards of the FASB and the IASB moved towards each other which increased the international comparability and global harmonization of accounting standards. Of course, differences between SFAS 142 and IFRS 3 still remain. For example, the impairment test of goodwill differs between the two accounting standards (Jerman and Manzin, 2008) Other examples are technical differences, such as the requirements for a business combination and the definition of fair value (IASB, 2008).

12

TABLE 1 Overview and comparison of the different accounting standards: Accounting Principles Board (APB) Opinion 17 Intangible Assets, Statement of Financial Accounting Standard (SFAS) 142 Goodwill and Other Intangible Assets, International Accounting Standard (IAS) 22 Business Combinations and International Financial Reporting Standard (IFRS) 3 Business Combinations. Accounting Principles Board Opinion 17

Statement of Financial Accounting Standard 142

International Accounting Standard 22

International Financial 3 Reporting Standard 3

Issued by AICPA (1970)

Issued by FASB (2001)

Issued by IASC (1983)

Issued by IASB (2004)

4

Accounting for business combinations at date of acquisition

Purchase method or pooling of 5 interests method.

Only purchase method is allowed.

Purchase method or pooling of interests method.

All business combinations must be accounted for using the purchase method.

Goodwill at date of acquisition

Only when the purchase method has been used: difference between the cost of acquisition and the fair value of the net assets of the acquired unit.

Difference between the cost of acquisition and the fair value of the net assets of the acquired unit.

Only when the purchase method has been used: difference between the cost of acquisition and the fair value of the net assets of the acquired unit.

The excess of the cost of the business combination over the acquirer's share of the net fair values of the identifiable assets, liabilities and contingent liabilities.

Treatment of goodwill after acquisition

Amortization on a straight-line basis over a period not to exceed 40 years.

Tested for impairment at least annually, and more often if certain indicators are present. An impairment loss must be recognized on the income statement if goodwill has declined below its book value.

Amortization on a straight-line basis over a period not to exceed 20 years.

Tested for impairment at least annually in accordance with IAS 36 Impairment of Assets.

3

Late 2007 and early 2008, both SFAS 141 and IFRS 3 have been revised, which resulted in SFAS 141(R) and IFRS 3R (Woodlock and Peng, 2008). The revisions made are beyond the scope of the current study and therefore the initial versions of the statements will be used. 4 The purchase method or the acquisition method requires the acquirer to recognize and measure the identifiable assets acquired, the liabilities assumed, and any noncontrolling interest. Also, goodwill arising on the acquisition should be recognized and measured (SFAS 141). 5 Under the pooling of interests method, the assets and liabilities of the acquirer and the acquiree are combined; no other assets or liabilities are recognized as a result of the transaction. The pooling method therefore only recognizes the intangible assets that were already recorded by the acquiree. Next to that, when using the pooling of interests method, no goodwill is generated on the acquisition.

13

With the move to goodwill impairment testing under SFAS 142 and IFRS 3, the FASB and the IASB turn to fair value accounting of the acquired goodwill shown on the firms’ consolidated balance sheets (Lhaopadchan, 2010). According to the FASB, the primary advantage of fair value accounting is the increased relevance of the financial statements since assets and liabilities will be measured at their current values. Besides that, the fair value approach is supposed to increase the comparability and consistency of financial reports (Shortridge et al., 2006). Additionally, the IASB aims at increasing the decision usefulness of accounting information by using fair valuebased measures instead of historical cost measures (Hamberg et al., 2011). 2.4 Earnings management As pointed out by Beatty and Weber (2006), the introduction of SFAS 142 and IFRS 3 provides managers with more flexibility in determining the fair value of goodwill, the existence, and the amount of a goodwill impairment. An unintended consequence of these new standards is that it introduces significant room for management judgment and bias (Guler, 2006). Consequently, managers may use the increased flexibility to engage in some sort of earnings management (Massoud and Raiborn, 2003). Following Healy and Wahlen (1999), earnings management can be defined as: “Managers using judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company, or to influence contractual outcomes that depend on reported accounting numbers.” When studying earnings management, a distinction can be made between income increasing earnings management and income decreasing earnings management. The term income increasing earnings management refers to ‘accounting procedures that shift reported earnings from future periods to the current period’ (Sweeney, 1994). An example of an incentive for income increasing earnings management are the debt contracts of a firm. In order to avoid costly covenant violations, Sweeney (1994) finds that managers are more likely to make incomeincreasing accounting changes. Contrarily, income decreasing earnings management is consistent with executives depressing the earnings temporarily in order to increase the likelihood of a desired outcome (Beneish, 2001). As Peasnell et al. (2005) find, managers appear to manipulate reported earnings downward when pre-managed earnings exceed a certain threshold. Reasons may be to transfer the excess amount to future periods when economic conditions are poor, or to comply with market demands for a smooth and steady earnings growth pattern (Zucca and Campbell, 1992; Chung et al., 2002). Healy and Wahlen (1999) identify several motivations for earnings management, including increasing executive compensation, reducing the likelihood of violating lending agreements and influencing the stock market. Van de Poel et al. (2009) affirm this by noticing that asset impairment decisions may be influenced by personal reporting incentives instead of purely economic factors. Executives can take advantage of the flexibility introduced by SFAS 142 and IFRS 3 by either not recognizing an impairment when they need to, or by recognizing an impairment only because it is beneficial for them to do so. 14

Besides the personal reporting incentives managers may encounter when making impairment decisions, the external demand for firms to meet earnings forecasts and to increase the share price may also be a motivation for managers to manipulate earnings (Jordan and Clark, 2011). In order to meet markets’ earnings expectations and create a smooth earnings path, a variety of techniques are available for managers. Two of the most well-known are big bath accounting and income smoothing. Big bath accounting is a form of income increasing earnings management and involves taking a one-time overstatement of charges against income to reduce assets, in order to avoid future write-down expenses (Zucca and Campbell, 1992; Sevin and Schroeder, 2005). This overstatement is usually recorded in the same period as the period in which the firm already experiences earnings below expectations. The reasoning behind it is that the market punishes a firm less for the additional loss since earnings are already bad, and future income statement benefits arise because of the lack of losses to be taken in the future (Massoud and Raiborn, 2003; Jordan and Clark, 2011). Alternatively, large losses on assets may be taken in a period when earnings exceed analysts’ forecasts. This practice, known as income smoothing, has been described earlier. Both of the earnings management techniques are expected to arise because of the subjectivity inherent in the impairment testing approach under SFAS 142 and IFRS 3 (Massoud and Raiborn, 2003; Guler, 2006; Jordan et al., 2007). Since managers have some discretion in deciding when to take an impairment loss, they can accelerate taking the loss when earnings are below expectations or when earnings are unexpectedly high in a particular year. 2.5 Research question The method of accounting for goodwill has changed significantly the past decade with the adoption and implementation of Statement of Financial Accounting Standards (SFAS) No. 141: Business Combinations and No. 142: Goodwill and Other Intangible Assets in the United States and International Financial Reporting Standard (IFRS) 3 Business Combinations in the European Union. The move from historical cost accounting to a fair value approach has provided executives with more discretion and flexibility in making estimations and assumptions with respect to, for example, the allocation of goodwill and the fair value measurement. With this, management is given the possibility to choose and manage earnings through the selective application of provisions in IFRS 3 and SFAS 142, either to achieve personal reporting goals or meet the external markets’ demand for smooth earnings growth (Sevin and Schroeder, 2005). The accounting literature contains several studies that examine the implementation of SFAS 142 and whether this new standard allows for the use of earnings management techniques. Since literature on the adoption of IFRS 3 and the role of financial reporting incentives in goodwill impairment decisions is quite limited, the current study tries to fill part of this gap in the literature. There are two main reasons of why different results are expected to be found in an European setting. First of all, the legal system of most European countries is quite different from the legal system in the US. As recognized by Leuz et al. (2003), managers and controlling owners may have incentives to manage earnings in order to conceal their private control benefits from outsiders and to reduce the likelihood of outside intervention. An example of a private control benefit insiders have is the possibility to transfer firm assets to other firms owned by family. Part of 15

a country’s legal system concerns the protection of outside investors from the effects of private control benefits and mitigation of insiders’ incentives to manage accounting earnings. Therefore, earnings management appears to be lower in economies with strong investor rights and strong legal enforcement (Leuz et al., 2003). The US is usually classified as being a ‘common law’ country which is characterized by dispersed ownership, strong investor rights and strong legal enforcement. The legal system of most European countries is, on the other hand, described as ‘code law’. These economies typically have low investor protection rights, which can result in more earnings management (Ball et al., 2000; Leuz et al., 2003; Van Tendeloo and Vanstraelen, 2005). The second reason why different results may be found in Europe, is the effect of the abolishment of goodwill amortization on a firms’ earnings. European firms amortized their acquired goodwill over a relatively shorter period of time (10 or 20 years instead of a maximum of 40 years), which leads to a larger effect on the reported earnings when the goodwill amortizations were abolished (Hamberg et al., 2011). Since the abolishment has a larger effect on European firms’ earnings, the opportunistic behavior and reporting incentives found in US studies should at least be equally important in Europe. Taken together, this makes the role of managerial reporting incentives and earnings management after the adoption of IFRS 3 an interesting topic to study. The overall research question of this study is the following: To what extent are goodwill impairment decisions, following the implementation of IFRS 3, used as a tool for earnings management?

16

3.

HYPOTHESES DEVELOPMENT

The introduction of the impairment testing approach by IFRS 3, has provided managers with an increased discretion in making estimates and assumptions regarding the fair value of goodwill (Sevin and Schroeder, 2005; Beatty and Weber, 2006), and consequently also regarding the amount of goodwill written off. Due to this increased discretion, a door has opened for executives to creatively manage earnings, either to achieve personal goals or to meet market demands. Section 3.1 describes a set of hypotheses that involve two reporting incentives that may affect the goodwill impairment decision. In practice, the impairment decision is composed of two accounting choices: the decision whether or not to report an impairment, and the decision that captures the size of the impairment (Van de Poel et al., 2009). Both of these choices will be addressed when developing the hypotheses. Besides managerial reporting incentives, studies that have investigated goodwill impairments in line with earnings management theories big bath accounting and income smoothing will be reviewed in section 3.2. The two techniques may be used by management to meet markets’ demand for a smooth and steady earnings growth pattern and goodwill impairments seem to be a tool that can help to accomplish that. 3.1 Managerial reporting incentives When returning to the definition of earnings management as given by Healy and Wahlen (1999), the first two sets of hypotheses will be focused on the second part of that definition: “(…) to influence contractual outcomes that depend on reported accounting numbers.” Increasing management compensation or reducing the likelihood that lending agreements are violated, are examples of motivations for earnings management (Healy and Wahlen, 1999). As previously mentioned, the two goodwill impairment decisions can be used to influence reported earnings in a year, and thus also the outcome of contractual agreements. Since goodwill impairment losses are a subset of asset write-offs in general, two separate streams of literature can be identified: studies that have examined the link between reporting incentives and the asset impairment decisions, and research that focuses on goodwill impairments and the accompanying reporting standards SFAS 142 and IFRS 3. Francis et al. (1996) studied discretionary asset write-offs in a period when there was still little authoritative guidance on accounting for most types of asset impairments. They found that managerial incentives to manipulate earnings play no significant role in write-off decisions for inventory and property, plant and equipment (PP&E), but do play a substantial part in explaining goodwill write-offs. Riedl (2004) looks at write-offs of long-lived assets prior and subsequent to the issuance of SFAS 121 Accounting for the Impairment of Long-Lived Assets in 1995. His results show that after the implementation of SFAS 121, write-offs have significantly greater association with reporting incentives than those reported prior to the standard. This suggests that managers apply more discretion in reporting write-offs and are more likely to act opportunistically subsequent to the adoption of the standard.

17

Besides studies on general asset impairments, considerable research has addressed goodwill impairments, the implementation of SFAS 142, and management’s discretion in the goodwill accounting choices. Beatty and Weber (2006) studied the factors affecting managers’ impairment decisions in the SFAS 142 transition period. Their evidence suggests that firms’ debt contracting, CEO turnover, bonus plans and exchange delisting incentives can influence the amount and timing of goodwill write-offs. A similar study has been done by Guler (2006), but he chose a period subsequent to the adoption of SFAS 142 for his research. He finds that executives’ reporting incentives affect the decision to recognize a SFAS 142 write-off. Specifically, a negative association between a firm having a bonus-based compensation plan and the likelihood of recognizing a goodwill impairment charge has been found. A study that examines IFRS 3 and IAS 36 instead of SFAS 142 is conducted by Van de Poel, Maijoor and Vanstraelen (2009). They explore the influence of reporting incentives on the goodwill impairment test in the two years after the mandatory implementation of IFRS in the European Union. Results show that the impairment decision is significantly associated with several financial reporting incentives. Based on prior literature, this research focuses on two managerial reporting incentives that can play a role in making the goodwill impairment decisions: the appointment of a new CEO and the debt contracts of a firm. Zucca and Campbell (1992) recognize that it is difficult to assess all the factors that might motivate a manager to record a goodwill impairment, and conclusions on motivating factors have to be drawn from available data. On one hand, an impairment decision may be driven by economic factors such as poor firm performance and increased competition, which can result in a decline in the value of goodwill. On the other hand, the impairment decisions might be driven by personal reporting incentives of management (Van de Poel et al., 2009). In the latter, management may use their discretion to decide on the amount and the timing of the impairment charge. 3.1.1

Change in senior management

The first reporting incentive is related to the tenure of management en especially the tenure of the chief executive officer (CEO). Existing literature finds a positive association between the decision to take a goodwill impairment and a recent change in CEO or senior management. Masters-Stout et al. (2008), who explore goodwill impairments after the implementation of SFAS 142, reason that more goodwill is impaired in the earlier years of the CEO’s tenure since previous management can be blamed for (poor) acquisition decisions. Also, expensing goodwill early on will lead to more chance of realizing improved future earnings under the supervision of new management. Both Francis et al. (1996) and Riedl (2004) find that new senior management may exercise greater scrutiny over the value of existing assets or may change the firm’s strategic focus, which both are likely to result in an impairment. Based on these findings, it is expected that the positive relation between a change in executive position and the decision to record a goodwill impairment charge will also hold in an IFRS environment. Additionally, it is expected that the impairment amount will be relatively larger when a new CEO has been installed. Therefore, the first hypotheses will be:

18

H1a: Firms with a recent change in the CEO position are more likely to record a goodwill impairment charge, ceteris paribus. H1b: Firms with a recent change in the CEO position will record relatively larger goodwill impairment charges, ceteris paribus 3.1.2

Debt-covenant hypothesis

Accounting numbers are often used in order to regulate and monitor the contractual agreements between a firm and its stakeholders. Two common forms of contractual agreements are compensation contracts and lending contracts. Lending contracts are used to ensure that managers do not take actions that benefit the stockholders at the expense of the firms’ creditors. (Healy and Wahlen, 1999). An example might be firms taking on risky projects, which increases the risk of default. Creditors suffer when the project fails, because the firm may be unable to make debt payments. Debt covenants place contractual restrictions on actions of the firm to make sure they are able to 6

fulfill their financial obligations. According to the debt-covenant hypothesis , the closer a firm is to violating one of its accounting-based debt covenants, the stronger managements’ incentives to manipulate the financial reporting process in order to avoid costly violation of the covenant (Dichev and Skinner, 2002). Support for the debtcovenant hypothesis is found in the existing accounting literature. Rield (2004) finds that managers are likely to make income-increasing accounting decisions in order to avoid debt covenant violations. As such, managers have an incentive to delay a goodwill impairment charge since this expense will lower the firms’ earnings. Beatty and Weber (2006) also find evidence of managers avoiding covenant violation by delaying expense recognition. Based on the previous reasoning and on prior research, the following is hypothesized: H2a: Firms with accounting-based debt covenants are less likely to record a goodwill impairment charge, ceteris paribus. H2b: Firms with accounting-based debt covenants will record relatively smaller goodwill impairment charges, ceteris paribus. 3.2 Earnings management Besides the reporting incentives managers may be exposed to, the current study looks into the decision of management to engage in either of two well-known earnings management patterns: big bath accounting or income smoothing. Firms may have various reasons of why they make use of these methods, such as showing smooth earnings patterns and a steady growth rate, meeting analysts’ expectations and influencing the share price 6

The debt-covenant hypothesis stems from the well-known Positive Accounting Theory (PAT) developed by Ross Watts and Jerold Zimmerman in the late 1970s. In short, this theory is concerned with the explanation and prediction of accounting practices. It provides reasons for observed practices, such as the strong incentive for managers to manipulate earnings when almost reaching the debt covenants (Whitley, 1988).

19

(Zucca and Campbell, 1992; Massoud and Raiborn, 2003; Jordan and Clark, 2011). Several studies have examined asset write-downs in the early 90s when the number of events was low and specific regulatory guidance was absent. The results seem to be mixed. Zucca and Campbell (1992) provide support for asset write-downs being used to manage earnings, but Francis et al. (1996) find that write-offs are actually decreasing when the firm has unusual poor or unusual good performance. This is contrary to the earnings management literature, which suggests that write-offs are increasing when the firm has unexpectedly high or low earnings. When looking at more recent studies that focus specifically on goodwill impairment charges being used to manage earnings, evidence seems to be unambiguous. Van de Poel et al. (2009) find that firms in the two years after the implementation of IFRS 3 are more likely to take goodwill impairments when earnings are unexpectedly low or unexpectedly high. The same results are reported by Guler (2006) who studied goodwill impairment charges subsequent to the adoption of SFAS 142. Additional evidence on big bath earnings management is delivered by Jordan et al. (2007) and Jordan and Clark (2011), although one has to be cautious when interpreting their results since they do not control for economic factors having an effect on the timing and amount of the impairment decision. 3.2.1

Big bath accounting

The big bath accounting theory suggests that firms postpone asset write-downs and record a larger loss in a period in which the firm experiences below expected earnings (Zucca and Campbell, 1992). Reasoning for such behavior is that the market perceives the extra loss as less significant, since earnings are already below what was expected. Accelerating the recording of an impairment provides an opportunity to realize increased future earnings or reduce the variability of future earnings, since goodwill impairment charges are likely to be lower in the future (Massoud and Raiborn, 2003; Jordan and Clark, 2011). As mentioned earlier, the big bath accounting theory is a widely investigated subject and ‘taking a bath’ has often been mentioned as a motivation for asset write-downs (Zucca and Campbell, 1992). However, most research on this theory stems from the US and concerns the year of transition to a new standard. Jordan et al. (2007) notice that a transition year provides an additional incentive to recognize a large goodwill impairment, since it can be reported below-the-line as the result from changing an accounting principle. Studies investigating big bath accounting in the period subsequent to the changes in accounting for goodwill are scarce. Guler (2006) finds a negative association between earnings and the likelihood of an SFAS 142 write-off, however his evidence is not statistically significant. Reviewing prior literature, the current study hypothesizes that, also in years without a change in accounting standards, managers are more likely to record a larger impairment loss when earnings are below expectations. H3a: Firms with unexpectedly low earnings are more likely to record a goodwill impairment charge, ceteris paribus. H3b: Firms with unexpectedly low earnings will record relatively higher goodwill impairment charges, ceteris paribus. 20

3.2.2

Income smoothing

Income smoothing has not been studied as widely as the big bath accounting theory, but it remains a popular earnings management pattern through which management aspires to achieve a steady rate of earnings growth (Zucca and Campbell, 1992). As the market tends to reward firms that meet expectations and show smooth earnings patterns, managers have an incentive to take a large impairment write-off when earnings are above expectations (Massoud and Raiborn, 2003; Zucca and Campbell, 1992). Prior studies on the income smoothing theory however do not seem to be consistent. For example, whereas Riedl (2004) finds a positive association between earnings and asset impairment charges, Francis et al. (1996) find a negative association when examining general asset write-offs. Guler (2006) and Van de Poel et al. (2009) both find a positive relation between a period with higher than expected earnings and goodwill impairment charges under respectively SFAS 142 and IFRS 3. The study of Guler (2006) considers income smoothing in the period after the adoption of SFAS 142, but Van de Poel et al. (2009) only consider the years 2005 and 2006 which comprise the transition period of IFRS 3. By means of the following hypotheses, the current study tries to contribute to the earnings management literature, and find some evidence on the presence of income smoothing in the period subsequent to the implementation of IFRS 3. H4a: Firms with unexpectedly high earnings are more likely to record a goodwill impairment charge, ceteris paribus. H4b: Firms with unexpectedly high earnings will record relatively higher goodwill impairment charges, ceteris paribus.

21

4.

DATA AND METHODOLOGY

Section 4 commences by outlining the empirical part of the study. Building on existing relevant literature, four hypotheses were developed, as described in the previous section. In order to test these hypotheses, a sample of listed companies in Europe has been composed. An extensive description of the sample and its construction is provided in section 4.1. Next, a separate regression model will be build for each of the two goodwill impairment decisions in section 4.2. This section also gives a throughout discussion of the dependent, independent and control variables. 4.1 Sample selection 7

The initial sample consists of all firms listed on the pan-European electronic stock exchange “Euronext” , 8

during the sample period 2006-2009 . The main reason for choosing the Euronext as initial company sample is the general comparability of the firms. All firms listed on the Euronext have to comply with the same listing requirements and use the same reporting and accounting standards. Listing requirements consist of both financial criteria and requirements concerning the firms’ stocks and securities. The initial sample results 1.926 potential firm observations. Firms that fall within the banking or financial services industries are excluded from the sample, as financial institutions have specific accounting requirements that substantially differ from those of industrial and commercial companies (Van Tendeloo and Verstraelen, 2005). Excluding financial firms with two-digit SIC (Standard Industrial Classification) codes 60-67, reduces the sample to 1.499 firms. Subsequently, all inactive and delisted companies are removed from the sample, as well as companies not reporting in compliance with the IFRS accounting standards, reducing the sample to 579 firms. As of 2005, all listed companies in the European Union are obliged to prepare their financial statements conform IFRS, but some firms have the possibility to file for deferment (EC Decision 1606/2002). Additionally, firm observations for which not all the required data is available are deleted from the sample. Especially important here is that all firms in the sample should have a goodwill opening balance in either one or all of the years 2006 to 2009. The result is a final sample of 489 firms, representing 1.748 firm-year observations. In Table 2 a summary of the sample selection process can be found. The time span of the current research consists of the years 2006 to 2009, deliberately excluding 2005, the year of mandatory adoption of IFRS by EU listed firms. A transition year provides additional incentives and opportunities to record impairment losses since in that year the loss can be reported ‘below-the-line’ and thus does not affect net income (Guler, 2006; Jordan et al., 2007). Under normal circumstances the loss has to be included as an expense in income from continuing operations, which reduces net income. In order to exclude any 7

In order to take advantage of the harmonization of the EU financial markets, the Euronext was formed in 2000 following a merger of the Amsterdam Stock Exchange, Brussels Stock Exchange and Paris Bourse. Euronext N.V. is based in the Netherlands and has subsidiaries in Belgium, France, Portugal and the United Kingdom. 8

The initial sample of firms has been retrieved from the ORBIS database.

22

additional incentive to record an impairment provided by the transition year, the time span is limited to the years 2006 to 2009. Additionally, it could be interesting to see whether or not the impairment decisions are influenced by managerial reporting incentives, instead of solely economic incentives, in regular reporting years. TABLE 2 – Sample Selection

Firms Initial sample Observations deleted:  Firms with two-digit SIC codes 60-67 (‘Finance, Insurance, and Real Estate’)  Inactive or delisted  Not reporting according to the IFRS accounting standards in the period 2006-2009 a  Missing data b Final sample Write-off observations Non-write-off observations

1,926

Firmyears 7,704

(427)

(1,708)

(441) (479)

(1,764) (1,916)

(92) 487 218 269

(568) 1748 391 1357

a

The main reason for firm observations to be excluded from the sample is because they did not have a separate goodwill account on their opening balance sheet in either one of the years 2006-2009. b

A more expanded table of the sample selection can be found in APPENDIX III, Table III, including the number of firm-year observations per year (2006, 2007, 2008 and 2009).

Graph 1 shows the total sample, subdivided into the number of firm-year observations per country. As can be seen, the sample is dominated by firm observations from France (68%). As has been touched upon earlier, European countries have a different legal system compared to the US. Most countries in Europe are classified as ‘code law’ and have relatively weaker investor protection rights, which can result in more pervasive earnings management (Leuz et al., 2003). LaPorta et al. (1998) document that especially French civil law countries offer the worst legal protection to investors. Since the sample used for the current research consists primarily of French companies, it is even more likely that indications for the presence of earnings management will appear.

23

GRAPH 1 – Company Composition

1400

Number of firmyear observations

1200 1195

1000 800 600 400 200

210

2

4

1 209 117 3

4

3

0

4.2 Research model In order to test the four hypotheses sets as described in section 3, two separate regression models will be developed. Hypothesis set 1 and 2 focus on two managerial reporting incentives that could affect the timing and the amount of the goodwill impairment charge. Hypotheses 1a and 1b examine the incentive a recently installed CEO might have to accelerate the recording of an impairment loss or to take a relatively larger impairment respectively. Possible reasons could be that new senior management may exercise greater scrutiny over the value of assets, may change the strategic focus of the firm, or wants to place the blame for poor acquisitions on previous management (Francis et al., 1996; Riedl, 2004; Masters-Stout et al., 2008). Hypotheses 2a and 2b argue that executives may want to delay the recording of impairment write-offs in order to avoid breaching the covenants written on debt contracts. Since covenant violations are costly, managers have strong incentives to manipulate the reporting process and thus delaying the recognition or recording a relatively smaller impairment loss (Dichev and Skinner, 2002). Next, hypotheses sets 3 and 4 are build to examine whether management makes use of either one of two earnings management techniques: big bath accounting or income smoothing. Hypotheses 3a and 3b state that when a firm experiences a negative earnings surprise, management is more likely to report an impairment loss and the loss is expected to be relatively higher, which is consistent with the big bath accounting theory. Hypotheses 4a and 4b examine an opposite statement as it is hypothesized that firms with unexpectedly high earnings are also more likely to take a (higher) goodwill write-off. The capital market tends to reward firms with smooth earnings patterns, so by recording an impairment the variation in earnings is tempered. The following two regression models are built to test the 4 hypotheses sets. Regression Model 1 focuses on the decision of firms to record a goodwill impairment charge. Since this decision is a dichotomous choice, a logistic regression model will be used similar to Beatty and Weber (2006) and Hamberg et al. (2011). Regression Model 2

24

assesses the influence of the determinants on the amount of the goodwill written off by means of a multiple linear regression model. Model 1:

IMPAIRMENTit = α + β1CHANGE_CEOit + β2DEBTit + β3BATHit + β4SMOOTHit + β5GDP_GROWTHit + β6∆SALESit + β7∆OCFit + β8∆ROAit + β9MTBVit + β10GOODWILLit + β11SIZEit + β12CAP_INTENSITYit + β13DMind1i + β14DMind2i + β15DMind3i + β16DMind4i + β17DMind5i + β18DMind6i + β19DMind7i + εit

Model 2: IMPAIRMENT_AMOUNTit α + β1CHANGE_CEOit + β2DEBTit + β3BATHit + β4SMOOTHit + β5GDP_GROWTHit + β6∆SALESit + β7∆OCFit + β8∆ROAit + β9MTBVit + β10GOODWILLit + β11SIZEit + β12CAP_INTENSITYit + β13DMind1i + β14DMind2i + β15DMind3i + β16DMind4i + β17DMind5i + β18DMind6i + β19DMind7i + εit

Where: Expected sign IMPAIRMENTit

=

IMPAIRMENT_AMOUNTit

=

CHANGE_CEOit

=

DEBTit BATHit

= =

SMOOTHit

=

indicator variable taking value of 1 if a firm reports an impairment loss at the end of year t, and 0 otherwise. the reported amount of the write-off deflated by total assets t-1 for write-off firms, and 0 for non-write-off firms. indicator variable equal to 1 if the firm experienced a change in the CEO position in years t-1 or t, and 0 otherwise. the debt to equity ratio of firm i in year t. equals the change in firm i’s pre-impaired earnings before taxes from t-1 to t divided by total assets t-1, when this change is below the industry median of non-zero negative values, and 0 otherwise. equals the change in firm i’s pre-impaired earnings before taxes from t-1 to t divided by total assets t-1, when this change is above the industry median of non-zero positive values, and 0 otherwise.

+

-

+

25

GDP_GROWTHit

=

∆SALESit

=

∆OCFit

=

∆ROAit

=

MTBVit CAP_INTENSITYit GOODWILLit

= = =

SIZEit

=

the percentage change in Gross Domestic Product (GDP) from year t-1 to t in the country in which firm i is established. the change in firm i’s sales from year t-1 to year t, divided by total assets at the end of year t-1. the change in firm i’s operating cash flow from year t1 to year t, divided by total assets at the end of year t1. the percentage change in firm i’s return on assets (ROA) from year t-1 to t. the market-to-book value of firm i in year t. the ratio of long-term assets over total assets. ratio of firm i’s opening balance of goodwill on total assets at t-1. the natural logarithm of firm i’s total assets in year t.

-

-

+ +/-

The variable DMind1i takes the value of 1 when firm i is classified in industry 1, and 0 otherwise. The variables DMind2i to DMind7i are interpreted in a similar manner. Goal of including these variables is controlling for industry-fixed effects. Worldscope (via Datastream) is used to acquire most of the data necessary for the current research. 4.2.1

9

Dependent variables

The research focuses on both of the goodwill impairment choices: the decision whether or not to record an impairment charge in a particular year, and the decision that determines the magnitude of the goodwill impairment. The first decision, taking a write-off or not, is a dichotomous choice. Therefore, the dependent variable (IMPAIRMENTit) is an indicator variable that takes the value of 1 if a firm records a goodwill impairment in year t, and 0 otherwise (Beatty and Weber, 2006; Van de Poel et al., 2009). Next, the dependent variable in Regression Model 2 (IMPAIRMENT_AMOUNTit), is measured as the magnitude of the reported goodwill write-off for firm i in year t, divided by lagged total assets (Rield, 2004). An important note here is that the impairment amount is reflected as a positive number and cannot take a value below zero. 4.2.2

Reporting incentives

Hypothesis 1a and 1b examine the tenure of the CEO of a firm and his or her goodwill impairment choices. The independent variable CHANGE_CEOit is an indicator variable which is coded 1 if a firm experienced a change in the CEO position in years t-1 or t, and 0 otherwise. For this variable, a dummy has to be created in SPSS (D_CHANGE_CEOit). Consistent with prior research, a positive association between the appointment of a new CEO 9

An overview of the different proxies created and the Worldscope codes used to collect the data can be found in APPENDIX II.

26

and the timing and magnitude of a goodwill impairment loss is expected. Information for this variable has been hand-collected by comparing the board composition of firms in annual reports. Hypothesis set 2 looks into the debt contracts of a firm. Most private debt agreements include restrictive covenants to limit potential conflicts of interest between the firms’ debtholders and shareholders (Begley, 1990). Dichev and Skinner (2002) notice that the presence of covenants is likely to affect managers’ reporting decisions in order to avoid costly covenant violations. Managers are more likely to use income-increasing and asset-increasing accounting procedures (Begley, 10

1990). The independent variable DEBTit measures the debt to equity ratio of a firm in year t. A higher debt ratio increases the probability of violating debt covenants, so highly indebted firms are expected to avoid goodwill impairment charges or record relatively low charges (Hamberg et al., 2011). The third and fourth hypotheses sets focus on the presence of the earnings management patterns big bath accounting and income smoothing. To see whether the flexibility of managers with regard to the goodwill impairment choices is exploited by managing the firms’ earnings, two proxies are added to the regression model. As a proxy for big bath reporting behavior, BATH it is included. Consistent with the studies of Riedl (2004) and Guler (2006), BATHit equals the change in firm i’s pre-impaired earnings before taxes from t-1 to t, divided by total assets t-1, when this change is below the industry median of non-zero negative values, and 0 otherwise. When the change in earnings is below the industry median, one can say that earnings are ‘unexpectedly’ low. Since earnings are already below expectations, the opportunity arises to accelerate the impairment loss or take relatively large losses, which may result in increased future earnings (Massoud and Raiborn, 2003). Therefore, a negative relation between BATHit and the impairment write-off decision is predicted. Similarly, SMOOTHit equals the change in firm i’s pre-impaired earnings before taxes from t-1 to t, divided by total assets t-1, when this change is above the industry median of non-zero positive values, and 0 otherwise (Riedl, 2004; Guler, 2006). When earnings are exceeding the expectations, managers may have an incentive to record a larger goodwill write-off since this decreases the surprisingly high earnings and contributes to a smooth earnings pattern (Zucca and Campbell, 1992; Massoud and Raiborn, 2003). Hence, a positive association between SMOOTHit and the impairment decision is expected. 4.2.3

Control variables

The regression model includes several control variables in order to capture factors other than the reporting incentives that could influence the goodwill impairment decisions. For example, managers have to be cautious for value declines of assets, attributable to poor firm and industry performance, and changes in the economic climate

10

The debt to equity ratio as a proxy for closeness to debt covenants has been criticized by, among others, Dichev and Skinner (2002) and Riedl (2004). Alternatives proposed are the inclusion of an indicator variable, taking the value of 1 if the firm has private debt (Riedl, 2004) or identifying debt covenant information using the DealScan database (Dichev and Skinner, 2002). Due to absence of access to DealScan, and the lack of a consistent definition and reporting method among firms of their private debt agreements, the debt to equity ratio is used in this research.

27

(Van de Poel et al., 2009). The regression model includes a number of proxies to reflect the economic factors that have an effect on the two impairment decisions. Based on the regression models of Riedl (2004) and Van de Poel et al. (2009), the variable GDP_GROWTHit is included to capture the effect of changes in the overall economic environment of the country in which firm i is based. A negative change suggests economic decline, which negatively affects the fair value of a firm’s assets. A decline in value increases the chance of firms taking an impairment and the amount of the impairment is likely to be relatively higher than when there is no significant decline in the fair value. Additionally, four firm-specific proxies are included, which are associated with the economic performance and condition of a firm. These variables are constructed based on the regression models used by Riedl (2004), Beatty and Weber (2006), Guler (2006), and Van de Poel et al. (2009). The first, ∆SALESit, is measured as the change in a firm’s net sales revenue from year t-1 to year t, deflated by the total assets at the end of year t-1. The second variable is ∆OCFit, which is the change in a firm’s operating cash flow from period t-1 to t, divided by total assets at the end of period t-1. Both of these variables are deflated by total assets since Rield (2004) recognizes that scale could significantly affect the regression results. As a third proxy for the economic condition of a firm, change in return on assets (∆ROAit) is included in the regression models, which is measured as the percentage change in return on assets for a firm from period t-1 to t. All three variables are predicted to have a negative sign, since a decrease in either of these measures signals a decrease in the economic condition of the firm, which increases the chance of reporting an impairment and the amount of the impairment loss. The fourth and final proxy associated with the economic performance of a firm is the market to book value (MTBV it). A high market to book value may imply that a firm has lots of growth options and will therefore have no need to impair goodwill. Hence, a negative sign for this variable is expected. Besides the control variables included to cover the economic factors influencing the goodwill impairment decision, the model also controls for the size of a company, the relative level of the goodwill opening balance, the capital intensity of a firm, and industry fixed effects. Following Van Tendeloo and Verstraelen (2005), Guler (2006), Lapointe-Antunes et al. (2008) and Van de Poel et al. (2009), SIZEit is measured as the natural logarithm of the firms’ total assets

11

at the end of year t. No sign is predicted for the variable SIZE it, since prior evidence on this

variable seems to be mixed (Sevin and Schroeder, 2005). To control for the effect of the size of the firms’ goodwill balance on the impairment decision, the proxy GOODWILLit is included in the model. This variable is measured as the ratio of firm i’s opening balance of goodwill on total assets at t-1 (Van de Poel et al., 2009). The reasoning here is that firms having a larger amount of goodwill on their balance sheet, may be more likely to incur an impairment since the amount of goodwill exposed to the impairment test is greater. Hence, a positive sign is predicted for this variable. Next, the model controls for the differences in capital intensity across firms and countries. The proxy CAP_INTENSITYit is measured as the ratio of long-term assets

12

over total assets for firm i in year t (Leuz, et al.,

11

A natural logarithm has to be used here in order to include the relative effect of size on the impairment decision instead of the absolute effect. 12

Long-term assets are calculated by subtracting Total Current Assets (WC02201) from Total Assets (WC02999).

28

2003). A negative association between the capital intensity proxy and the dependent variables is expected. Finally, to control for industry effects that might influence the impairment decision, the sample is divided into industries 13

based on the first two digits of the firm’s SIC code. For each of the resulting 8 categories , 7 dummy variables are created and included in the regression model. An overview of the total company sample, subdivided into the 8 different industry categories, can be found in Table 3 on pg 30.

13

The category ‘Finance, Insurance and Real Estate’ (SIC 60-67) has been excluded from the company sample, as has already been explained. The category ‘Nonclassifiable Establishments’ (SIC 99) does not contain any companies, so is therefore excluded as well. This leaves 8 different industry categories.

29

5.

RESULTS

Section 5 presents the results of the data analyses in SPSS. First, section 5.1 provides the descriptive statistics for the entire sample as well as the distribution among the different industries. Additionally, the descriptive statistics concerning the variables incorporated in the two Regression Models are discussed. Section 5.2 gives a correlation analysis by means of a Pearson correlation matrix and a Spearman correlation matrix. Together with the descriptives, initial observations can be made about the sample and the variables. Subsequently, section 5.3 and 5.4 discuss the logistic regression and the multiple linear regression, respectively. The results arising from these regressions can be used to draw conclusions on the four hypotheses sets. 5.1 Descriptive Statistics Table 3 provides an overview of the company sample, divided into different industries, and some descriptive statistics for the sample. Firms are assigned to an industry based on the first two digits of their Standard Industrial 14

Classification (SIC) code . As has been explained before, one important industry group has been excluded from the sample: banking or financial services firms with SIC code 60-67. The first and second column in Table 3 show the 15

name of the industry category and the accompanying US SIC codes . The remainder of the table gives an overview of the number of firm-year observations within each industry category, as well as the total number of firm-year observations in the sample. Additionally, the number of write-off observations within each category is provided in the fifth column. The number of firm-year observations in each industry category and the number of impairment observations in the different categories are converted into percentages in order to give more insight into the distribution of observations among the industry categories. TABLE 3 – Industry Composition US SIC Codes 01 – 09 10 – 14 15 – 17 20 – 39 40 – 49 50 – 59 70 – 89 91 – 97 Total

Industry category Agriculture, Forestry and Fishing Mining Construction Manufacturing Transportation, Communications, Electric, Gas and Sanitary Services Wholesale Trade Services Public Administration

Number of firm-years 18 33 89 771 191

Percentage (%) of total 1,0% 1,9% 5,1% 44,1% 10,9%

Impairment

Percentage (%) of total

4 7 24 179 62

22,2% 21,2% 26,9% 23,2% 32,5%

177 461 8 1748

10,1% 26,4% 0,5% 100%

34 81 0 391

19,2% 17,6% 0,00% 22,4%

14

‘Euronext’ uses the Industry Classification Benchmark (ICB) system to categorize companies into industries. Neither the ORBIS database nor Datastream contains ICB codes so therefore the current study classifies companies based on their SIC code. 15 Categories and SIC code ranges are based on the overview provided on the NAICS (North American Industry Classification System) website. Retrieved from: http://www.naics.com/. 30

From Table 3 can be seen that the majority of firm observations occur within two industries: ‘Manufacturing’ (44,1%) and ‘Services’ (26,4%). The other industry categories contain significantly less observations. Important to notice is that when looking at the percentage of impairment observations for each category, the differences between industries are much smaller. The industries ‘Construction’ (26,9%) and ‘Transportation, Communications, Electric, Gas and Sanitary Services’ (32,5%) have a relatively high frequency of impairment observations, compared to the percentage of impairments for the total sample. However, one has to be cautious with interpreting these results since both categories comprise only a small fraction of the total sample. The industry ‘Services’ has, despite its high number of firm observations, a relatively low frequency of reported goodwill impairments. This seems remarkable since the ‘Services’ industry is expected to be the most intangible intensive, which makes the chance of reporting an impairment more likely (Bens et al., 2011). An explanation could be that the economic performance in the ‘Services’ industry is less volatile than in other industries, meaning that the value of goodwill stays relatively stable (Duff & Phelps, 2010). 16

Table 4 presents the descriptive statistics for the total company sample used in the current research . The variables and proxies from Model 1 as well as from Model 2 are included in Table 4. The descriptive statistics are used to explore the collected data and to make some general observations. Table 4 shows the number of observations available for a particular variable (N), as well as the mean, the median, the standard deviation, and the minimum and maximum value a variable can take. Also, descriptives on the first and third quartile are provided in Table 4, which gives insight into the dispersion of the data. Several observations can be made from the descriptive statistics overview. First of all, 22% of the firms has recorded a goodwill impairment in either one or more of the years 2006-2009. This is in line with results found in prior research (Riedl, 2004; Van de Poel et al., 2009) although the percentage of impairments found in the current research is somewhat higher. Next, it is important to notice the number of observations that provide conditions for either big bath accounting or income smoothing. For 354 firm-year observations (20,3%), a condition is found that could provide a manager with the incentive to engage in big bath accounting by accelerating the reporting of a goodwill impairment. Conditions for 17

income smoothing are present in 507 (29%) of the observations. Descriptive results for the variables BATHit and SMOOTHit are in line with prior research (Van de Poel et al., 2009). The final point to make on the descriptive statistics of the total company sample is that in approximately 21% of the observations, a change in the CEO position in either year t-1 or in year t takes place.

16

Consistent with prior research (Beatty and Weber, 2006), the continuous variables have been winsorized at the top 1% and the bottom 1%. This method eliminates the effects of extreme observations, without losing any of the observations from the sample. The results are not sensitive to the variables being winsorized. 17

These results are not tabulated.

31

TABLE 4 18 Descriptive Statistics Total Sample (N = 1748)

Dependent variable IMPAIRMENTit IMPAIRMENT_AMOUNTit Reporting incentives CHANGE_CEOit DEBTit BATHit SMOOTHit b Control variables GDP_GROWTHit ∆SALESit ∆OCFit ∆ROAit MTBVit CAP_INTENSITYit GOODWILLit SIZEit

N

Mean

Median

Standard Deviation

Q1

Q3

Minimum

Maximum

1748 1748

0,2237 0,0029

0 0

0,4168 0,0139

0 0

0 0

0 0

1 0,1179

1748 1748 1748 1748

0.2054 0,8736 -0,0169 0,0233

0 0,5708 0 0

0,40410 1,1122 0,0428 0,0553

0 0,2131 0 0

0 1,1013 0 0,0279

0 -0,3276 -0,2358 0

1 6,9842 0 0,3526

1748 1748 1748 1748 1748 1748 1748 1748

0,0254 0,0706 0,0100 -0,2990 1,9554 0,4709 0,1553 19,9192

0,0391 0,0417 0,0086 -0,0932 1,5050 0,4590 0,1141 19,5738

0,0313 0,2291 0,0754 2,5527 1,6919 0,2025 0,1408 2,1152

-0,0173 -0,0288 -0,0232 -0,5061 0,9375 0,3166 0,0403 18,2240

0,0488 0,1476 0,0429 0,1618 2,4 0,6236 0,2386 21,3149

-0,0390 -0,5793 -0,2386 -16,6204 -0,4 0,0709 0,0003 16,0878

0,2857 0,9925 0,2773 9,5714 10,49 0,9167 0,5843 25,0688

b

The model also controls for industry fixed effects by including 7 dummy variables. The descriptives for the industry dummies are not provided since an overview of the company sample subdivided into industries has already been discussed.

In order to see whether there are differences between the group of firm-years reporting an impairment and the firm-years not taking an impairment, the sample is divided into two subsamples. Table 5 provides the descriptive statistics for the impairment sample and the no impairment sample. Independent t-tests are performed to compare the means of the two subsamples, and to test for significant differences between the sample medians, a Wilcoxon ranksum test has been performed. By means of the subdivision of the total sample, several observations can be made. First, the mean goodwill impairment amount is only 1,31% of the beginning of the period total assets. On average, goodwill constitutes 17% of total assets for the firm-observations with an impairment loss, which is significantly higher than the amount of goodwill on total assets for the non-write-off sample (15%). This seems to be valid, since with a larger amount of goodwill on the balance sheet, the chance of having to report an impairment becomes higher.

18

An overview of the descriptive statistics for each of the years 2006, 2007, 2008 and 2009 can be found in Table IV, Appendix III.

32

TABLE 5 Descriptive Statistics Impairment Sample and No Impairment Sample Impairment Sample

No Impairment Sample

N

Mean

Median

Std. Dev.

391 391

1 0,0131

1 0,0024

0 0,0270

1357 1357

0 0

0 0

0 0

391

0,3120

0

0,4639

1357

0,1746

0

0,3798

DEBTit

391

0,9966

0,7409

1,0745

1357

0,8382

0,5240

1,1207

BATHit

391

-0,024

0

0,0489

1357

-0,0148

0

0,0405

SMOOTHit

391

0,0188

0

0,0503

1357

0,0246

0

0,0566

Control variables GDP_GROWTHit

391

0,0218

0,0234

0,0311

1357

0,0265

0,0428

0,0314

∆SALESit

391

0,0591

0,0210

0,2379

1357

0,0740

0,0463

0,2265

∆OCFit

391

0,0076

0,0057

0,0664

1357

0,0107

0,0094

0,0778

∆ROAit

391

-0,502

-0,1618

2,8388

1357

-0,2405

-0,084

2,4621

MTBVit

391

1,7787

1,3800

1,5651

1357

2,0063

1,5500

1,7239

CAP_INTENSITYit

391

0,4857

0,4721

0,1869

1357

0,4666

0,4565

0,2066

GOODWILLit

391

0,1697

0,1520

0,1309

1357

0,1511

0,1034

0,1434

SIZEit

391

20,735

20,5253

2,1892

1357

19,6840

19,328

2,0343

Dependent variable IMPAIRMENTit IMPAIRMENT_AMOUNTit Reporting incentives CHANGE_CEOit

N

Mean

Median

Std. Dev.

t-test (p-value)

***

z-statistic (p-value)

***

5,360 (0,000) ** 2,486 (0,013) *** -3,255 (0,001) * -1,948 (0,052)

-5,923 (0,000) *** -5,007 (0,000) *** -3,965 (0,000) *** -2,673 (0,008)

***

-2,880 (0,004) ** -2,447 (0,014) -1,023 (0,306) ** -2,027 (0,043) *** -2,630 (0,009) * -1,761 (0,078) *** -3,692 (0,000) *** -8,362 (0,000)

-2,607 (0,009) -1,131 (0,258) -0,775 (0,439) * -1,651 (0,099) ** -2,475 (0,014) * 1,731 (0,084) ** 2,308 (0,021) *** 8,849 (0,000)

***

*** ** *

, , , denotes significance at the 0,01, 0,05, or 0,10 level, respectively.

Next, the financial performance variables ∆SALESit, ∆OCFit, ∆ROAit and MTBVit are all lower for the impairment sample compared to the no impairment sample, with ∆ROAit and MTBVit being significantly lower. These results signal worse financial performance for the firms-years in the impairment sample, which is in line with prior expectations. A decline in firm performance negatively affects the fair value of assets, and thus increases the likelihood of having to take an impairment (Riedl, 2004). Consistent with prior research, impairment firms have a significantly higher rate of turnover in the CEO position (Riedl, 2004; Guler, 2006). Additionally, compared to the no impairment sample, observations in the impairment sample have a significantly stronger decrease in their earnings, as can be seen by a lower BATHit for these firms. Remarkable is the significant difference between the impairment sample and the no impairment sample for the variable SMOOTH it. The mean value for SMOOTH it is lower for the impairment sample, which is contrary to what has been expected. According to the theory on income smoothing behavior, firms experiencing higher than expected earnings are more likely to report an impairment

33

and the impairment loss is expected to be bigger (Massoud and Raiborn, 2003; Zucca and Campbell, 1992). The same unexpected result has been found by Riedl (2004), although he does not provide an explanation for the finding. A final remark that can be made is the significant difference in firm size between the impairment sample and the no impairment sample. A possible explanation may be that larger firms usually undertake more acquisitions which results in a higher goodwill balance, and therefore the chance of an impairment loss increases. As can be seen from Table 5, the results from the Wilcoxon ranksum tests support the results found with the independent t-tests, providing additional evidence on the significant differences between the impairment sample and the no impairment sample. APPENDIX III, Table IV, Panel A presents an overview of the descriptive statistics for each of the years 2006 to 2009. Panel B focuses on the years 2006 and 2009 and tests whether there are significant differences between those years. In November 2008 the financial crisis, which began in the sub-prime sector of the securitized US mortgage market, blew over to Europe (Carmassi et al., 2009). The economic decline resulting from the crisis has an effect on both European banks as well as on firm performance. This becomes visible in Table IV, APPENDIX III, as almost every economic control variable in 2009 has significantly decreased compared to 2006. Also, firms reported more goodwill impairments in 2009, which signals a decline in fair value. The effect of the financial crisis on fair value accounting and goodwill impairments is beyond the scope of the current study but could be an interesting topic for future research. 5.2 Correlations Table 6 presents the Pearson correlation matrix. From this matrix, general observations can be made on the correlation between a set of variables. First of all, it is important to check whether multicollinearity exists between two variables. The term multicollinearity is used when two independent variables are highly correlated (>.75). When this is the case, it may be wise to delete one of the two variables from the regression model since otherwise biases can arise. Variables that are highly correlated essentially measure the same thing, and together they contribute a lot in explaining the dependent variable, even though individually they do not contribute significantly to the model. As can be seen from Table 6, there are no signs for multicollinearity between any set of independent variables, so there is no need to further assess or resolve multicollinearity . Next, the correlation matrix shows that two out of the four reporting incentive-related variables (CHANGE_CEOit and BATHit) are significantly correlated with both the IMPAIRMENTit variable and the IMPAIRMENT_AMOUNTit variable. This provides initial support for the entire first and third hypothesis sets. The independent variables DEBTit and SMOOTHit only shows a significant correlation with the dependent variable IMPAIRMENTit. Remarkable is that, contrary to the predictions made in hypothesis set 2, the correlation sign between DEBTit and IMPAIRMENTit is positive instead of negative. Also contrary to the expectations is the negative relation between SMOOTH it and IMPAIRMENTit. Except for ∆SALESit, the economic control variables all show the expected (negative) correlation with IMPAIRMENTit and IMPAIRMENT_AMOUNTit, with ∆ROAit and MTBVit being significantly correlated with both dependent variables. 34

The independent variables BATHit and SMOOTHit are both significantly correlated with the proxies reflecting the economic condition and performance of a firm: GDP_GROWTHit, ∆SALESit, ∆OCFit, ∆ROAit and MTBVit. This seems reasonable since the economic performance of a firm is also reflected in the change in earnings, which is used to construct the proxies for big bath accounting and income smoothing. Table 7 presents the Spearman correlation matrix. The Spearman correlation matrix controls for any outliers, unequal variances or non-normality that could be present and affect the Pearson correlations. The Spearman correlation coefficients are in line with those found in the Pearson correlation matrix (Table 6). It can therefore be concluded that there are no significant differences between the rank correlations and the ordinary correlations and thus there are no outliers that could affect the results. Since all the continuous variables have been winsorized at the top 1% and bottom 1%, it was not expected that significant outliers would still be present.

35

TABLE 6 Pearson Correlation Matrix

36

TABLE 7 Spearman Correlation Matrix

37

5.3 Logistic Regression In order to test hypothesis H1a, hypothesis H2a, hypothesis H3a and hypothesis H4a, a logistic regression analysis has been performed. The 4 hypotheses are focused around the decision whether or not to report an impairment loss in a certain year, and the factors that may influence that decision. As noted earlier, the decision to take an impairment is a dichotomous choice, which makes the use of a binary logistic regression the most suitable. To test the 4 hypotheses, and to examine whether or not an independent variable is useful and has explanatory power, 6 reduced versions of Regression Model 1 are created in Table 8. Reduced Model 1-I in Table 8 performs the regression with only the three general control variables GOODWILLit, SIZEit, and CAP_INTENSITYit. Next, Model 1-II adds the five proxies that control for economic factors influencing the impairment decision: GDP_GROWTHit, ∆SALESit, ∆OCFit, ∆ROAit and MTBVit. As can be seen by the Model χ² and the significant change in the χ² (compared to Model 1-I), the model including the five economic variables is a good fitting model and the independent control variables have significant explanatory power in predicting the outcome of IMPAIRMENTit. Additionally, the Nagelkerke R Square shows that 10,4% of the variation in IMPAIRMENTit is explained by the economic factors and the general control variables. As expected, firms with poorer financial performance and less growth options are more likely to take a write-off, as indicated by the significantly negative coefficients on ∆ROAit and MTBVit. Furthermore, a decrease in a country’s Gross Domestic Product (GDP) increases the chance of firms situated in that country to record an impairment loss. These results are in line with those found by Guler (2006), Beatty and Weber (2006) and Van de Poel et al. (2009). Model 1-III is used to examine whether a recent change in a firm’s CEO position leads to a greater likelihood of that firm deciding to report an impairment loss. The significant and positive coefficient on CHANGE_CEOit supports 19

hypothesis H1a. The results for H2a are reported in Model 1-IV, column 5. The significance of the change in the χ² is not very strong, indicating that the variable DEBTit does not have as much additional explanatory power as the variable CHANGE_CEOit and the economic control variables. Contrary to the expectations, DEBT it shows a statistically positive coefficient at the 10% level. This would indicate that highly indebted firms are more likely to take an impairment instead of delaying the impairment. All together, there is no support found for hypothesis H2a.

19

The coefficients for a logistic regression have to be interpreted differently than those for a linear regression. B = 0,701 for CHANGE_CEOit signals that an increase in this independent variable is associated with a greater likelihood of the dependent variable event (in this case deciding to record an impairment loss) occurring. Additional explanation is given by the Exp(B) value, which is not presented in Table 8. Exp(B) for 0,701 is 2,0158 which means that when a firm experiences a change in the CEO position, this firm is approximately 2 times more likely to record an impairment loss.

38

TABLE 8 Regression Model 1: Dichotomous Logistic Regression Model Ia Prediction INTERCEPT

?

II b

III c

IV d

Ve

VI f

VII g Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

(p-value)

(p-value)

(p-value)

(p-value)

(p-value)

(p-value)

-6,014***

-5,947***

-6,011***

-5,863***

-6,557***

-5,960***

(p-value) -6,525***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

Reporting Incentives CHANGE_CEO it

+

0,690***

0,701***

(0,000)

(0,000) DEBTit

-

0,109*

0,100*

(0,058)

(0,077) BATHit

-

-5,818***

-6,165***

(0,000)

(0,000) SMOOTHit

+

0,146

0,259

(0,913)

(0,848)

Economic Factors -4,506**

-3,944*

-4,311**

-4,264**

-4,496**

-3,453

(0,032)

(0,064)

(0,041)

(0,044)

(0,033)

(0,108)

0,300

0,322

0,294

0,536*

0,291

0,521

(0,331)

(0,299)

(0,341)

(0,083)

(0,362)

(0,105)

-0,758

-0,751

-0,705

0,104

-0,777

0,045

(0,394)

(0,396)

(0,429)

(0,909)

(0,391)

(0,961)

-0,042*

-0,041*

-0,041*

-0,017

-0,041*

-0,018

(0,072)

(0,081)

(0,076)

(0,481)

(0,073)

(0,462)

-0,091**

-0,097**

-0,107**

-0,087**

-0,092**

-0,110**

(0,034)

(0,023)

(0,014)

(0,042)

(0,034)

(0,011)

1,631***

1,487***

1,466***

1,549***

1,418***

1,488***

1,475***

(0,001)

(0,003)

(0,004)

(0,002)

(0,005)

(0,003)

(0,004)

0,281***

0,293***

0,287***

0,288***

0,312***

0,293***

0,302***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

-1,536***

-1,543***

-1,567***

-1,679***

-1,421***

-1,544***

-1,618***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

No. of Observations Chi-square χ² (model)

1748

1748

1748

1748

1748

1748

1748

107,407***

122,956***

147,992***

125,982***

140,517***

122,968***

167,289***

Chi-square χ² (change)

-

15,549***

25,036***

3,026*

17,561***

0,012

44,333***

Nagelkerke R Square

0,091

0,104

0,124

0,106

0,118

0,104

0,139

GDP_GROWTHit

-

∆SALESit

-

∆OCFit

-

∆ROAit MTBVit

-

Control Variables h GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY . it it it b Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it c Reduced model with predictors: (Constant), CHANGE_CEO , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it it d Reduced model with predictors: (Constant), DEBT , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it it e Reduced model with predictors: (Constant), BATH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it it f Reduced model with predictors: (Constant), SMOOTHit, GOODWILLit, SIZEit, CAP_INTENSITYit, GDP_GROWTHit, ∆SALESit, ∆OCF it, ∆ROAit, MTBVit. g Full model with predictors: (Constant), CHANGE_CEO , DEBT , BATH , SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , it it it it it it it it it ∆OCF it, ∆ROAit, MTBVit. h The regression model also controls for industry-fixed effects.

39

The results for H3a and H4a, reported in model 1-V and 1-VI, are used to investigate whether executives make use of the earnings management techniques big bath accounting or income smoothing. Strong support is provided for hypothesis H3a with a highly significant positive coefficient on the big bath behavior proxy BATHit. The importance of this variable in predicting the outcome of dependent variable IMPAIRMENT it is also shown by a significant χ² change of 17,561. It can therefore be confirmed that firms experiencing unexpectedly low earnings are more likely to record a goodwill impairment charge. The results on this proxy are in line with the studies of Riedl (2004) and Francis et al. (1996). Consistent with the fourth hypothesis, a positive coefficient is found on SMOOTHit. However, the result is not statistically significant so the notion that managers are more likely to report an impairment when earnings are unexpectedly high cannot be supported. The full model 1-VII combines all the reporting incentive-related variables and control variables. Consistent with the predictions made in hypothesis H1a and H3a, CHANGE_CEOit and BATHit are significantly related to the decision to report a goodwill impairment. Further, the economic control proxy MTBVit as well as the control variables GOODWILL it, SIZEit, and CAP_INTENSITYit show a significant coefficient throughout the 7 models in Table 8. 5.4 Multiple Linear Regression Table 9 shows the regression results for Regression Model 2, which is designed to test hypotheses H1b, H2b, H3b and H4b. The four hypotheses involve the factors influencing the magnitude of the impairment loss that is recorded by a firm. Based on prior literature (Riedl, 2004; Beatty and Weber, 2006; Van de Poel et al., 2009), and since the data for the variable IMPAIRMENT_AMOUNTit is censored at zero, a Tobit regression seems to be the most appropriate. As the current study uses SPSS to analyze data, a Tobit regression cannot be performed since this specific type of regression is not available in SPSS. Therefore, a multiple linear regression will be used for analyzing Model 2, which is the most suitable alternative available. Similarly to Regression Model 1, six reduced versions of Regression Model 2 are created in order to test the hypotheses and to examine whether or not the independent variables contribute to the prediction of the size of the impairment loss written off. Model 2-I only includes the three general control variables GOODWILL it, SIZEit, and CAP_INTENSITYit, whereas Model 2-II adds the five proxies that control for the influence of the economic condition and performance on the 2

amount of goodwill written off. As can be seen by the significant change in R (0,010, p = 0,003), and the F-test for Model 2-II, the overall model is significant in explaining the dependent variable and the five economic variables significantly improve the prediction of the magnitude of goodwill impairment. Except for ∆SALESit, all the economic factors have the expected negative coefficient, with ∆OCFit, ∆ROAit, and MTBVit having a significant influence on the impairment amount reported. When examining the economic factors and their relation with the dependent variable IMPAIRMENT_AMOUNTit in more detail, ∆SALESit seems to have a positive influence on the amount of the impairment loss instead of the expected negative influence. This result indicates that a negative change in a firms’ sales does not provide an incentive for managers to report a larger impairment loss, and thus may not be a primary indicator of the firm’s performance and economic condition. Additionally to the economic control factors, 40

TABLE 9 Regression Model 2: Multiple Linear Regression Model Ia Prediction INTERCEPT

?

II b

III c

IV d

Ve

VI f

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

VII g Coefficient

(t-statistic)

(t-statistic)

(t-statistic)

(t-statistic)

(t-statistic)

(t-statistic)

(t-statistic)

0,006*

0,007*

0,007*

0,008**

-0,003

0,006

-0,005

(1,778)

(1,927)

(1,945)

(2,065)

(-0,904)

(1,543)

(-1,371)

Reporting Incentives CHANGE_CEO it

+

DEBTit

-

BATHit

-

0,103***

0,085***

(4,324)

(3,753) 0,035

0,036

(1,411)

(1,385)

-0,351***

-0,348***

(-14,285)

(-14,144) SMOOTHit

+

0,047*

0,080***

(1,778)

(3,201)

Economic Factors -0,037

-0,030

-0,035

-0,027

-0,036

-0,017

(-1,453)

(-1,179)

(-1,363)

(-1,118)

(-1,408)

(-0,702)

0,059**

0,060**

0,059**

0,125***

0,047*

0,105***

(2,212)

(2,276)

(2,204)

(4,864)

(1,691)

(4,019)

-0,065***

-0,066***

-0,065***

0,002

-0,074***

-0,012

(-2,677)

(-2,704)

(-2,644)

(0,092)

(-2,987)

(-0,510)

-0,043*

-0,042*

-0,042*

0,025

-0,041*

0,030

(-1,786)

(-1,761)

(-1,771)

(1,095)

(-1,722)

(1,291)

-0,052**

-0,056**

-0,061**

-0,050**

-0,058**

-0,072***

(-2,131)

(-2,294)

(-2,415)

(-2,175)

(-2,364)

(-3,028)

0,152***

0,143***

0,142***

0,146***

0,130***

0,145***

0,134***

(5,462)

(5,137)

(5,124)

(5,218)

(4,925)

(5,199)

(5,113)

-0,023

-0,019

-0,026

-0,024

0,033

-0,010

0,039

(-0,809)

(-0,688)

(0,348)

(-0,844)

(1,221)

(-0,355)

(1,419)

-0,034

-0,029

-0,030

-0,037

-0,002

-0,030

-0,014

(-1,166)

(-0,974)

(-1,019)

(-1,223)

(-0,075)

(-1,038)

(-0,483)

No. of Observations

1748

1748

1748

1748

1748

1748

1748

Model R Square

0,022

0,032

0,042

0,033

0,132

0,034

0,146

R Square change

0,022***

0,010***

0,010***

0,001

0,100***

0,002*

0,114***

Model F-test

3,826***

3,801***

4,768***

3,685***

16,476***

3,765***

15,519***

GDP_GROWTHit ∆SALESit ∆OCFit ∆ROAit MTBVit

-

Control Variables h GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a Reduced model with predictors: (Constant), GOODWILLit, SIZEit, CAP_INTENSITYit. b Reduced model with predictors: (Constant), GOODWILLit, SIZEit, CAP_INTENSITYit, GDP_GROWTH it, ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. c Reduced model with predictors: (Constant), CHANGE_CEO , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it it d Reduced model with predictors: (Constant), DEBT , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it it e Reduced model with predictors: (Constant), BATH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it it f Reduced model with predictors: (Constant), SMOOTHit, GOODWILLit, SIZEit, CAP_INTENSITYit, GDP_GROWTHit, ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. g Full model with predictors: (Constant), CHANGE_CEO , DEBT , BATH , SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , it it it it it it it it it ∆OCF it, ∆ROAit, MTBVit. h The regression model also controls for industry-fixed effects.

41

the amount of goodwill reported on a firm’s opening balance sheet has a highly significant influence on the amount of goodwill written off. Consistent with prior research (Hamberg et al., 2011), these results imply that larger impairments are reported when the goodwill opening balance is higher. Interesting to notice is the control variable SIZEit. The size of a firm did have a significant and positive influence on the decision to record an impairment loss (Table 8), but the magnitude of the loss is not necessarily larger than for smaller firms. Prior researchers (Beatty and Weber, 2006; Hamberg et al., 2011) did not express a specific expectation on the association between the size of a firm and the impairment amount reported. Their results are mixed, but never statistically significant, so it can therefore be concluded that the size of a firm has no pronounced influence on the magnitude of the impairment loss. The results for hypothesis H1b are reported in Model 2-III. The inclusion of the independent variable CHANGE_CEOit increases the R Square of the model to 4,2%, which provides evidence for the significant explanatory power of the variable in explaining IMPAIRMENT_AMOUNTit. Model 2-III shows a significant positive coefficient for the variable CHANGE_CEOit (T = 4,324, p = 0,000). In line with prior research, the hypothesis that newly installed CEOs will record larger goodwill impairment charges can therefore be supported. With this result, confirmation for the entire first hypothesis set has been found since earlier it has been supported that firms with new CEOs are more likely to decide to write-off their goodwill. Model 2-IV in column 5 provides the results for hypothesis H2b. The variable DEBTit has a positive coefficient, which is contrary to the expectations. Additionally, the result on DEBTit is insignificant which implies the variable has no influence on the size of the impairment loss recorded. As has been discussed in section 4, using the debt to equity ratio as a measure for closeness to debt covenant violation has been criticized by several researchers (Dichev and Skinner, 2002; Riedl, 2004). They are concerned the ratio is not a valid measure for closeness to covenant violation and may contain measurement errors. This could provide an explanation for the insignificant result found for the variable DEBTit. The results for hypotheses H3b and H4b, which focus on big bath accounting behavior and income smoothing behavior, are presented in Model 2-V and 2-VI. Similar to the results of the logistic regression, including the proxy BATHit in the regression model has a very large effect on the R Square, as it increases to 13,2%. The highly significant positive coefficient found for BATHit, indicates the variable is also extremely important in explaining the amount of impairment loss reported. Model 2-V provides evidence consistent with hypothesis H3b, indicating that an unexpected negative change in earnings leads to a larger goodwill impairment loss reported. Together with the results found in Table 8, the entire third hypothesis set can be supported. Similar results and a highly significant coefficient for the big bath proxy are also found by Riedl (2004). Evidence consistent with hypothesis H4b can be found in Model 2-VI, which shows a positive coefficient for SMOOTH it (T = 1,778, p = 0,076). Even though SMOOTHit is only significant at the 10% level in model 2-VI, the full model presented in column 8 of Table 9 provides additional support for the significant explanatory power of SMOOTH it. Together with the results from the logistic regression it can be concluded that firms with unexpectedly high earnings do not necessarily decide to 42

report an impairment, but when they do the loss is larger than the loss of firms with a normal earnings level. Therefore, there seems to be some evidence for firms engaging in income smoothing behavior. The full model IV, which includes all the reporting incentive-related independent variables, explains 14,6% of the variation in the dependent variable IMPAIRMENT_AMOUNTit. Additionally, the full model shows a significant relation between three of the four reporting incentive-related independent variables (CHANGE_CEOit, BATHit, and SMOOTHit) and the dependent variable predicting the size of the goodwill impairment loss.

43

6.

SENSITIVITY ANALYSES

Section 6 performs a number of sensitivity analyses in order to check the robustness of the results found in section 5. Since the debt to equity ratio as a measure for closeness to debt covenant violation is a widely discussed topic, section 6.1 provides an alternative measure to test hypotheses H2a and H2b. Next, section 6.2 examines whether the results are influenced by the proxy used for measuring the size of a firm. Further, instead of including the variable GPD_GROWTHit to control for the effect of differences between countries and their economic conditions, section 6.3 explores the regression results using dummy variables to control for country effects. Finally, as the proxy for big bath accounting behavior potentially contains a lot of noise, two alternatives are used to validate the BATHit variable and the results on H3a and H3b. 6.1 Debt to equity ratio The first sensitivity analysis performed will examine whether the results reported in section 5 are sensitive to the choice of proxy for the firm’s debt level. The independent variable DEBTit is used to see whether highly indebted firms are likely to delay goodwill impairment charges in order to avoid debt covenant violation, and has been measured using the firm’s debt to equity ratio (Hamberg et al., 2011). As has been noticed earlier, the debt to equity ratio as a measure for the firm’s closeness to debt covenant violation is a criticized subject (Dichev and Skinner, 2002; Riedl, 2004). An alternative measure used in the studies of Beatty and Weber (2006) and Guler (2006) is the ratio of total debt to total assets. Remarkable is that both studies do not have a priori prediction on the coefficient of the debt variable. As Guler (2006) hypothesizes, debt covenants may also lead to higher scrutiny on the financial reporting process, including the accounting discretion with respect to the impairment testing. The current study retains the prediction of a negative association between the firm’s debt level and the two impairment-related dependent variables (IMPAIRMENTit and IMPAIRMENT_AMOUNTit). To capture the alternative proxy, the variable DEBTit is replaced by the variable DEBT2it in both Regression Model 1 and Regression Model 2. The results for the adjusted regression models are presented in Table 10 and Table 11. The results for the binary logistic regression model, evaluating the decision of firms to report an impairment loss or not, are consistent with those reported in Table 8. The full Model 1-VII in Table 10 actually loses some of its explanatory power and significance as the Model χ² and the Nagelkerke R Square are somewhat smaller. Additionally, the model including DEBT2it loses the earlier reported weak significance on the debt variable, even though these results were contrary to the expectations. Similarly, including the variable DEBT2it into the multiple linear regression model does not lead to different results as those found in Table 9. CHANGE_CEO it, BATHit, and SMOOTHit all have a significant influence on the prediction of the amount of the write-off recorded by a firm, whereas DEBT2it seems to have no influence on the outcome of the dependent variable.

44

TABLE 10

TABLE 11

Dichotomous Logistic Regression Model: Sensitivity Analysis Incorporating Alternative Measure DEBT2 it

Multiple Linear Regression Model: Sensitivity Analysis Incorporating DEBT2 it

Ia II b III c IV d Ve VI f VII g Prediction Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient INTERCEPT

?

(p-value)

(p-value)

(p-value)

(p-value)

(p-value)

(p-value)

-6,014***

-5,947***

-6,011***

-5,956***

-6,557***

-5,960***

(p-value) -6,637***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

Reporting Incentives CHANGE_CEO it DEBT2 it BATHit SMOOTHit

∆SALESit

0,701***

0,675***

(0,000)

(0,000)

-

0,282

0,220

(0,377)

(0,497)

-

-6,165***

-5,864***

(0,000)

(0,000)

+

?

0,006*

0,007*

0,007*

0,007*

-0,003

0,006

-0,005

(1,778)

(1,927)

(1,945)

(1,956)

(-0,904)

(1,543)

(-1,509)

0,146

0,358

(0,913)

(0,792)

CHANGE_CEO it

+

0,083***

0,103***

(3,686)

(4,324) DEBT2 it

-

0,020

0,029

(0,405)

(1,148) BATHit

-

-0,351***

-0,348***

(-14,281)

(-14,144) SMOOTHit

+

0,047*

0,080***

(1,778)

(3,223)

Economic Factors -

∆OCFit

-

∆ROAit

-

MTBVit

INTERCEPT

Reporting Incentives +

Economic Factors GDP_GROWTHit

Ia II b III c IV d Ve VI f VII g Prediction Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic)

-

-4,506**

-3,944*

-4,481**

-4,264**

-4,496**

-3,660*

(0,032)

(0,064)

(0,033)

(0,044)

(0,033)

(0,087)

0,300

0,322

0,304

0,536*

0,291

0,527

(0,331)

(0,299)

(0,325)

(0,083)

(0,362)

(0,101)

-0,758

-0,751

-0,745

0,104

-0,777

-0,006

(0,394)

(0,396)

(0,403)

(0,909)

(0,391)

(0,995)

-0,042*

-0,041*

-0,041*

-0,017

-0,041*

-0,018

(0,072)

(0,081)

(0,073)

(0,481)

(0,073)

(0,460)

-0,091**

-0,097**

-0,089**

-0,087**

-0,092**

-0,091**

(0,034)

(0,023)

(0,038)

(0,042)

(0,034)

(0,033)

Control Variables h

GDP_GROWTHit ∆SALESit

-

-0,037

-0,030

-0,036

-0,027

-0,036

-0,019

(-1,453)

(-1,179)

(-1,415)

(-1,118)

(-1,408)

(-0,768)

0,059**

0,060**

0,059**

0,125***

0,047*

0,106***

(2,212)

(2,276)

(2,228)

(4,864)

(1,691)

(4,030)

(-2,677)

(-2,704)

(-2,657)

(0,092)

-0,074*** -0,013 (-0,532) (-2,987)

-0,043*

-0,042*

-0,043*

0,025

-0,041*

0,029

(-1,786)

(-1,761)

(-1,792)

(1,095)

(-1,722)

(1,271)

-0,052**

-0,056**

-0,051**

-0,050**

-0,058**

-0,063

(-2,131)

(-2,294)

(-2,105)

(-2,175)

(-2,364)

(-2,740)

0,152***

0,143***

0,142***

0,144***

0,130***

0,145***

0,132***

(5,462)

(5,137)

(5,124)

(5,160)

(4,925)

(5,199)

(5,046)

-0,023

-0,019

-0,026

-0,022

0,033

-0,010

0,041

(-0,809)

(-0,688)

(0,348)

(-0,788)

(1,221)

(-0,355)

(1,518)

-0,034

-0,029

-0,030

-0,037

-0,002

-0,030

-0,012

(-1,166)

(-0,974)

(-1,019)

(-1,224)

(-0,075)

(-1,038)

(-0,412)

No. of Observations

1748

1748

1748

1748

1748

1748

1748

∆OCFit

-

∆ROAit

-

MTBVit

-0,065*** -0,066*** -0,065*** 0,002

-

Control Variables h 1,631***

1,487***

1,466***

1,490***

1,418***

1,488***

1,414***

(0,001)

(0,003)

(0,004)

(0,003)

(0,005)

(0,003)

(0,000)

0,281***

0,293***

0,287***

0,291***

0,312***

0,293***

0,307***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

-1,536***

-1,543***

-1,567***

-1,612***

-1,421***

-1,544***

-1,516***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

No. of Observations Chi-square χ² (model)

1748

1748

1748

1748

1748

1748

1748

107,407*** 122,956*** 147,992*** 123,735*** 140,517*** 122,968*** 164,248***

Model R Square

0,022

0,032

0,042

0,033

0,132

0,034

0,145

Chi-square χ² (change)

-

15,549***

25,036***

0,779

17,561***

0,012

41,292***

R Square change

0,022***

0,010***

0,010***

0,001

0,100***

0,002*

0,113***

Nagelkerke R Square

0,091

0,104

0,124

0,104

0,118

0,104

0,137

Model F-test

3,826***

3,801***

4,768***

3,646***

16,476*** 3,765***

GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY . it it it b Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it c Reduced model with predictors: (Constant), CHANGE_CEO , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , it it it it it it it ∆ROAit, MTBVit. d Reduced model with predictors: (Constant), DEBT2 , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , it it it it it it it it MTBVit. e Reduced model with predictors: (Constant), BATH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , it it it it it it it it MTBVit. f Reduced model with predictors: (Constant), SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , it it it it it it it it MTBVit. g Full model with predictors: (Constant), CHANGE_CEO it, DEBT2it, BATH it, SMOOTH it, GOODWILLit, SIZEit, CAP_INTENSITYit, GDP_GROWTH it, ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. h The regression model also controls for industry-fixed effects.

GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

15,439***

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY . it it it b Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , ∆ROA , it it it it it it it MTBVit. c Reduced model with predictors: (Constant), CHANGE_CEO , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , it it it it it it ∆OCF it, ∆ROAit, MTBVit. d Reduced model with predictors: (Constant), DEBT2it, GOODWILLit, SIZEit, CAP_INTENSITYit, GDP_GROWTHit, ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. e Reduced model with predictors: (Constant), BATH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , it it it it it it it ∆ROAit, MTBVit. f Reduced model with predictors: (Constant), SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , GDP_GROWTH , ∆SALES , ∆OCF , it it it it it it it ∆ROAit, MTBVit. g Full model with predictors: (Constant), CHANGE_CEO , DEBT2 , BATH , SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , it it it it it it it GDP_GROWTH it, ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. h The regression model also controls for industry-fixed effects.

45

Taken together, using the ratio of total debt to total assets as a measure for closeness to covenant violation does not provide evidence consistent with hypothesis H2a or H2b and is therefore not superior to using the firm’s debt to equity ratio. 6.2 Total assets as measure of firm size The second robustness check investigates whether or not the results are sensitive to the decision to measure firm size as the book value of assets instead of measuring size as the market capitalization. Both methods are used in prior literature, which may suggest the measures are equally useful. Similar to the main regression models, there is no a priori prediction on the association of the firm size with the two goodwill impairment decisions. To see if different results arise when the natural logarithm of market capitalization is being used instead of the logarithm of total assets, the proxy SIZE2it replaces SIZEit in Regression Model 1 and 2. The untabulated results for this alternative measure are not significantly different from the main results presented in Table 8 and 9. The logistic regression shows a highly significant positive coefficient on SIZE2 it, indicating that larger firms are more likely to report an impairment loss. Also similar to earlier results, the alternative measure for firm size does not have a significant influence on the magnitude of the impairment charge. To conclude, the results of the main regressions and the regressions performed including SIZE2 it demonstrate that there are no differences in using either the book value of assets or the market capitalization as measure for control for firm size. 6.3 Control for country effects Based on the studies of Riedl (2004) and Van de Poel et al. (2009), Regression Model 1 and Regression Model 2 include the proxy GDP_GROWTHit to capture macroeconomic effects and to control for the overall economic climate of a country. From the results in Table 8 can be observed that the change in GDP level of a specific country has a significant influence on the decision whether or not to recognize a goodwill impairment. As the proxy seems to be valuable in determining the outcome of the dependent variable, it could be interesting to explore an alternative measure that controls for country effects. As can be seen from Graph 1 on page 24, the total sample contains firm-year observations from 10 different countries. Similar to LaPorta et al. (1998), a dummy variable for each of the countries, with one so-called omitted dummy, is created. Table V and VI in APPENDIX IV present the regression results for the models controlling for country effects by means of dummy variables instead of the GDP_GROWTHit proxy. Several noticeable differences appear when comparing the main regression results and the alternative regression results. First of all, the Nagelkerke R Square for the logistic regression model controlling for country effects by means of dummy variables is 0,161, compared to 0,139 for the regression model in Table 8. This result indicates that the inclusion of country dummies increases the explanatory power of the entire model in predicting the outcome of the dependent variable IMPAIRMENTit. Basically, the model becomes better when controlling for country effects with dummy variables than when using the GDP_GROWTHit proxy. The larger Model χ² and χ² change in Table V support the notion that including the country dummies results in a better fitting model. 46

The second remarkable difference between the regression results is the statistical significance of the DEBTit coefficient that arises on both the logistic regression and the linear regression in Table V and VI (APPENDIX IV). Even though the results are contrary to the expectations in hypotheses H2a and H2b, the level of a firm’s debt to equity ratio seems to be positively related to the decision whether or not to record an impairment loss and to the magnitude of the loss written off. These alternative results could provide evidence for the previously mentioned hypothesis of Guler (2006), stating that debt covenants introduce higher scrutiny on the exercise of accounting discretion with respect to the impairment testing method. Related to this is that high leverage and thus the presence of many debtholders, could result in more conservative financial reporting (Ball et al., 2008). As debt covenants are often written in terms of financial statement variables, debtholders demand timely information on these variables. The recording of impairment losses is a possible way of reporting conservative earnings and therefore provides an explanation for the positive coefficient on DEBT it. A final remark that can be made on the results presented in Table V and VI is that the previously found significant coefficients on CHANGE_CEOit, BATHit, SMOOTHit, MTBVit and GOODWILLit do not change when country dummies are included instead of the proxy GDP_GROWTHit. 6.4 Proxy big bath accounting As pointed out by Rield (2004), the proxy used to capture big bath accounting behavior may alternatively reflect the economic performance of a firm. When a firm experiences a large negative change in earnings, this may either way lead to a write-off charge on goodwill. It may therefore be unclear whether the impairment is driven by the incentive to engage in big bath reporting behavior or merely by the negative economic result. In order to validate the presented results on the BATHit proxy, and to provide additional assurance that BATH it is indeed capturing the intended reporting behavior, two alternative definitions are explored. Based on the research of Francis et al. (1996) and Riedl (2004), BATH2it is equal to any negative change in firm i’s pre-impaired earnings before taxes, and 0 otherwise. BATH3it equals the negative change in a firm’s pre-impaired earnings when the level of pre-write-off earnings in year t is also negative, and 0 otherwise. Both these alternative measures are scaled by lagged total assets. The logistic regression and linear regression results are presented in Table 12 and 13. As has been done prior to all of the sensitivity analyses above, the correlation matrices including the alternative variables BATH2it and BATH3it are checked for signs of multicollinearity. None of the correlation coefficients point towards multicollinearity being present, so further investigation does not seem to be necessary. Both the inferences on BATH2it and BATH3it, as well as the other variables in Table 12 and 13 are consistent with those in the primary analyses. Moreover, the Model χ², the model F-tests and the R Squares are similar to those of the main regressions (Table 8 and 9). These results provide additional evidence on managers engaging in big bath accounting behavior by accelerating the goodwill impairment charge when earnings are below expectations.

47

TABLE 12

TABLE 13

Dichotomous Logistic Regression Model: Sensitivity Analysis Incorporating Alternative Measures BATH2 it and BATH3it

Multiple Linear Regression Model: Sensitivity Analysis Incorporating Alternative Measures BATH2 it and BATH3it

BATHit BATH2 it BATH3 it Prediction Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient (p-value) INTERCEPT

?

(p-value) -6,525***

(p-value)

(p-value)

(p-value)

(p-value)

-6,557***

-6,615***

-6,609***

-6,366***

-6,270***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

Reporting Incentives CHANGE_CEO it DEBTit

-

SMOOTHit

+

0,690***

0,690***

0,680***

(0,000)

(0,000)

(0,000)

0,109*

0,109*

0,108*

(0,058)

(0,058)

(0,059)

-6,165***

-5,818***

-6,542***

-6,261***

-5,457***

-4,679***

(0,000)

(0,000)

(0,000)

(0,000)

(0,001)

(0,003)

0,259

0,454

-0,067

(0,848)

(0,737)

(0,961)

Economic Factors

∆SALESit ∆OCFit ∆ROAit MTBVit

INTERCEPT

?

(t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic) -0,005 -0,003 -0,004 -0,006 -0,002 -0,003 (-1,371) (-0,904) (-1,000) (-1,588) (-0,490) (-0,825)

Reporting Incentives +

BATHit

GDP_GROWTHit

BATHit BATH2 it BATH3 it Prediction Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient

CHANGE_CEO it DEBTit

+ -

BATHit

-

SMOOTHit

+

0,085***

0,086***

0,073***

(3,753)

(3,811)

(3,246)

0,035

0,034

0,026

(1,411)

(1,383)

(1,068) -0,348*** -0,351*** -0,349*** -0,357*** -0,344*** -0,340*** (-14,144) (-14,285) (-14,044) (-14,324) (-14,516) (-14,331) 0,080*** 0,023*** 0,064*** (3,201) (3,634) (2,563)

Economic Factors -

-4,264**

-3,453

-4,017*

-3,200

-4,538**

-3,729*

(0,044)

(0,108)

(0,058)

(0,137)

(0,031)

(0,082)

0,536*

0,521

0,567*

0,542*

0,458

0,456

(0,083)

(0,105)

(0,068)

(0,092)

(0,140)

(0,158)

0,104

0,045

0,182

0,100

-0,300

-0,325

(0,909)

(0,961)

(0,841)

(0,913)

(0,738)

(0,721)

-0,017

-0,018

-0,016

-0,016

-0,023

-0,025

(0,481)

(0,462)

(0,510)

(0,498)

(0,336)

(0,296)

-0,087**

-0,110**

-0,085**

-0,108**

-0,089**

-0,111***

(0,042)

(0,011)

(0,047)

(0,012)

(0,037)

(0,010)

Control Variables a

GDP_GROWTHit ∆SALESit ∆OCFit ∆ROAit MTBVit

-

-0,027

-0,017

-0,018

-0,007

-0,041*

-0,033

(-1,118)

(-0,702)

(-0,748)

(-0,305)

(-1,706)

(-1,360)

0,125***

0,105***

0,131***

0,109***

0,118***

0,102***

(4,864)

(4,019)

(5,076)

(4,166)

(4,635)

(3,880)

0,002

-0,012

0,004

-0,011

-0,028

-0,040*

(0,092)

(-0,510)

(0,178)

(-0,477)

(-1,188)

(-1,697)

0,025

0,030

0,026

0,031

0,016

0,018

(1,095)

(1,291)

(1,110)

(1,359)

(0,701)

(0,804)

-0,050**

-0,071*** -0,054**

-0,072***

(-2,175)

-0,072*** -0,047** (-3,028) (-2,053)

(-2,966)

(-2,361)

(-3,008)

0,130***

0,134***

0,131***

0,135***

0,116***

0,120***

(4,925)

(5,113)

(4,951)

(5,153)

(4,392)

(4,542)

0,033

0,039

0,031

0,040

0,030

0,034

(1,221)

(1,419)

(1,158)

(1,461)

(1,136)

(1,246)

-0,002

-0,014

(-0,075)

(-0,483)

-0,004 (-0,141)

-0,015 (-0,549)

-0,030 (1,136)

-0,010 (-0,366)

Control Variables a 1,418***

1,475***

1,427***

1,484***

1,365***

1,430***

(0,005)

(0,004)

(0,005)

(0,004)

(0,007)

(0,005)

0,312***

0,302***

0,313***

0,304***

0,308***

0,295***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

-1,421***

-1,618*** (0,000)

-1,624*** (0,000)

-1,449*** (0,000)

-1,641*** (0,000)

CAP_INTENSITYit

(0,000)

-1,429*** (0,000)

No. of Observations Chi-square χ² (model)

1748

1748

1748

1748

1748

1748

No. of Observations

1748

1748

1748

1748

1748

1748

140,517*** 167,289*** 141,999*** 168,990*** 134,261*** 160,032***

Model R Square

0,132

0,146

0,131

0,146

0,137

0,146

Chi-square χ² (change)

17,561***

44,333***

19,038***

46,029***

11,299***

37,070***

R Square change

0,100***

0,114***

0,099***

0,114***

0,105***

0,114***

Nagelkerke R Square

0,118

0,139

0,119

0,141

0,113

0,134

Model F-test

16,476*** 15,519*** 16,295

GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a The regression model also controls for industry-fixed effects.

GOODWILLit SIZEit

+ +/-

15,582*** 17,167*** 15,593***

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a The regression model also controls for industry-fixed effects.

48

7.

CONCLUSION, LIMITATIONS, AND FUTURE RESEARCH

This section provides a summary of the research conducted and an overview of the main empirical results. Additionally, some limitations of the current study are discussed, as well as several suggestions for future research. 7.1 Conclusions The current study examines factors that influence managers’ decisions regarding the timing and the amount of goodwill impairment charges. In 2005, the accounting for goodwill fundamentally changed when the IASB followed the FASB and moved to a fair value-based accounting approach. With the introduction of IFRS 3 and the alteration of IAS 36 and IAS 38, the amortization of goodwill method was abandoned and henceforward, goodwill has to be tested for impairment at least annually. Besides the increased relevance and transparency of the financial statement accounts as well as the improved comparability, IFRS 3 provides managers with more discretion in determining the fair value of goodwill and the amount of goodwill written off. The flexibility surrounding the goodwill impairment choices leaves significant room for management judgment and bias. This study investigates several determinants of the two impairment decisions in order to see whether earnings are managed through the selective application of the provisions in IFRS 3. Specifically, the influences of executive turnover and debt agreements are investigated, together with the earnings management patterns big bath accounting and income smoothing. In order to examine the effect of reporting incentives on the goodwill impairment decisions, a logistic regression and a multiple linear regression have been performed. The empirical analyses reveal that firms with a recently installed CEO are more likely to report a goodwill impairment charge. Additionally, these firms also record higher charges compared to firms not experiencing a change in the chief executive position. Further, the analyses indicate the presence of big bath accounting behavior in the period after the implementation of IFRS 3 and the introduction of the impairment testing approach. The results suggest that firms experiencing unexpectedly low earnings are more likely to record a (larger) goodwill impairment. Results for firms with large positive earnings surprises are not as unanimous. The presence of higher than expected earnings alone does not provide managers with the incentive to record a goodwill impairment charge. But when an impairment is reported, the loss is likely to be larger than the loss of a firm with a normal earnings level. The provisions in IFRS 3 and IAS 36 thus provide firms with the opportunity to engage in a form of earnings smoothing although the incentive is not very strong. All of the inferences above hold after controlling for economic factors influencing the impairment decisions, and other determinants of goodwill write-offs such as firm size and the amount of the goodwill opening balance. Taken together, the results imply that after the implementation of IFRS 3, reporting incentives affect the accounting choices relating to the timing and amount of goodwill written off, and it can therefore be concluded that goodwill impairments are used as a tool to manage a firms’ earnings.

49

7.2 Limitations The study is subject to a number of limitations. The first limitation is due to the sample selection procedure. Even though several advantages were mentioned for choosing the Euronext as initial company sample, there are also disadvantages to this approach. The main disadvantage is that German and UK firms are not listed on the 20

Euronext. With Germany and the United Kingdom being two of the most important economies in Europe , excluding these firms puts a constraint on the generalization of the results. A second limitation is due to the previously mentioned difficulty surrounding the proxy measuring closeness to debt covenant violation. According to Dichev and Skinner (2002), debt covenants are usually written on private debt agreements, but these are often unobservable. This is the reason why researchers use proxies, such as a firm’s debt to equity ratio, for closeness to covenant violation. However, these proxies contain measurement error and may be difficult to interpret. The current study also deals with these problems, since the results on the DEBTit variable are mixed and inconsistent with the predictions. Using a debt to total assets ratio (DEBT2it) instead of the debt to equity ratio does not resolve the issues as results on that variable are also inconclusive. Therefore, the proxies used for measuring closeness to covenant violation pose a serious limitation to the current research. It would be interesting to see the results on hypotheses H2a and H2b when using more specific loan agreement information from the DealScan database. A final limitation is due to the variable capturing big bath accounting behavior. Although the results on the BATH it variable have been partially verified by using the alternative measures BATH2 it and BATH3it, it still contains noise and could also capture a residual economic effect rather than big bath accounting behavior. Additionally, as recognized by Riedl (2004), there are competing interpretations regarding big bath behavior. Managers could, instead of reporting opportunistically, use their discretion to signal information on the underlying firm performance to the market. The current study does not take this second possibility into account. 7.3 Future research Besides addressing the limitations identified in the previous section, several recommendations to extent the research were encountered. First, as has been noticed by reviewing the additional descriptive statistics in APPENDIX III Table IV, there seem to be significant differences between the mean values of some of the variables in the years 2006 and 2009. The financial crisis, which began in Europe in 2008, appears to be the main explanation for the significant decreases in economic control variables such as ∆SALESit and ∆ROAit. Additionally, Table IV shows differences in the two impairment variables between the years 2006 and 2009. This could provide an initial signal for the effects of the financial crisis on the decisions regarding IFRS 3 goodwill impairments. Expanding this notion could be worth investigating, especially since prior literature suggests that fair value accounting contributed to the financial crisis (Laux and Leuz, 2009). A second recommendation could be to examine an additional reporting incentive that may affect the decision to take a goodwill write-off, consistent with Beatty and Weber (2006) and Guler (2006): the compensation plan of an executive. Managers with a bonus plan directly linked to the 20

Information retrieved from: http://www.ec.europa.eu/eurostat

50

firms’ earnings, may have the incentive to delay or reduce goodwill impairment charges in order to increase their personal wealth. Because of time constraints and data collection problems, the current study did not incorporate a proxy related to the incentives arising due to compensation plans. A third possibility for extending the current research could be to include the effect of stock price performance on the goodwill impairment decisions. Francis et al. (1996) and Riedl (2004) both include market-based measures as explanatory variable for impairment write-offs. A primary advantage of including a variable measuring a firm’s stock returns is that it also captures expectations of future performance. A negative association between the market-based measure and the amount of goodwill written off would be expected, as a declining stock price signals worse financial performance and thus a higher likelihood of reporting an impairment charge.

51

REFERENCES Accounting Principles Board Opinion No. 17 (1970). Issued by Accounting Principles Board of the American Institute of Certified Public Accountants (AICPA). Retrieved from: http://clio.lib.olemiss.edu/u?/aicpa,322 Baldi, F. (2009). Assessing the Value of Growth Option Synergies from Business Combinations and Testing for Goodwill Impairment: A Real Options Perspective. Journal of Applied Corporate Finance, Vol. 21, No. 4, pp. 115-124 Ball, R., Kothari, S.P., and Robin, A. (2000). The effect of international institutional factors on properties of accounting earnings. Journal of Accounting and Economics, Vol. 29, pp. 1-51 Ball, R., Robin, A., and Sadka, G. (2008). Is financial reporting shaped by equity markets or by debt markets? An international study of timeliness and conservatism. Review of Accounting Studies, Vol. 13, pp. 168-205 Beatty, A. and Weber, J. (2006). Accounting Discretion in Fair Value Estimates: An examination of SFAS 142 Goodwill Impairments. Journal of Accounting Research, Vol. 44, No. 2, pp. 257-288 Begley, J. (1990). Debt Covenants and Accounting Choice. Journal of Accounting and Economics, Vol. 12, pp. 125139 Beneish, M.D. (2001). Earnings Management: A Perspective. Managerial Finance, Vol. 27, No. 12, pp. 3-17 Bens, D.A., Heltzer, W., and Segal, B. (2011). The Information Content of Goodwill Impairments and SFAS 142. Journal of Accounting, Auditing and Finance, Vol. 26, No. 3, pp. 527-555 Carmassi, J., Gros, D., and Micossi, S. (2009). The Global Financial Crisis: Causes and Cures. Journal of Common Market Studies, Vol. 47, No. 5, pp. 977-996 Chung, R., Firth, M., and Kim, J-B. (2002). Institutional monitoring and opportunistic earnings management. Journal of Corporate Finance, Vol. 8, pp. 29-48 Datta, D.K. and Grant, J.H. (1990). Relationships Between Type of Acquisition, The Autonomy Given to the Acquired Firm, and Acquisition Success: An Empirical Analysis. Journal of Management, Vol. 16, No. 1, pp. 29-44 Dichev, I.D. and Skinner, D.J. (2002). Large-sample Evidence on the Debt Covenant Hypothesis. Journal of Accounting Research, Vol. 40, No. 4, pp. 1091-1123 Duff & Phelps (2010). 2010 Goodwill Impairment Study. Retrieved from: http://www.duffandphelps.com Financial Accounting Standards Board (2004). Summary of FASB Tentative Decisions on Business Combinations. Retrieved from: http://www.fasb.org Francis, J., Hanna, J.D. and Vincent, L. (1996). Causes and Effects of Discretionary Asset Write-Offs. Journal of Accounting Research, Vol 34, pp. 117-134 Gara, S.C. (2007). Market reaction to Accounting Regulatory Changes: Adoption of SFAS 142. Journal of American Academy of Business, Vol. 10, No. 2, pp. 289

52

Guler, L. (2006). Goodwill Impairment Charges under SFAS No. 142: Role of Executives’ Incentives and Corporate Governance. Working paper. Retrieved from: http://proquest.umi.com/pqdlink?did=1394656061&Fmt=7&clientI d=79356&RQT=309&VName=PQD Hamberg, M. and Beisland, L. (2010). Changed Methods to Account for Goodwill – Did it really make a difference? Working Paper. Retrieved from: http://www.snee.org/filer/papers/506.pdf Hamberg, M., Paananen, M., and Novak, J. (2011). The Adoption of IFRS 3: The Effects of Managerial Discretion and Stock Market Reactions. European Accounting Review, Vol. 20, No. 2, pp. 263-288 Hayn, C. and Hughes, P. (2005). Leading indicators of goodwill impairment. Journal of Accounting, Auditing and Finance, 21, pp. 223-265 Healy, P.M. and Wahlen, J.M. (1999). A Review of the Earnings Management Literature and Its Implications for Standard Setting. Accounting Horizons, Vol. 13, No. 4, pp. 365-383 International Accounting Standard 22 (1983). Issued by International Accounting Standards Committee (IASC). Retrieved from: http://www.iasplus.com International Accounting Standard 36 Impairment of Assets (2004). Issued by International Accounting Standards Board (IASB). Retrieved from: http://www.iasplus.com International Accounting Standards Board (2008). Business Combinations Phase II. Retrieved from: http://www.iasb.org International Financial Reporting Standard 3 Business Combinations (2004). Issued by International Accounting Standards Board (IASB). Retrieved from: http://www.iasplus.com Jerman, M. and Manzin, M. (2008). Accounting Treatment of Goodwill in IFRS and US GAAP. Organizacija, Vol. 41, No. 6, pp. 218-225 LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., and Vishny, R. (1998). Law and Finance. Journal of Political Economy, Vol. 106, pp. 1113-1155 Leuz, C., Nanda, D., and Wysocki, P.D. (2003). Earnings Management and Investor Protection: An International Comparison. Journal of Financial Economics, Vol. 69, No. 3, pp. 505-527 Lhaopadchan, S. (2010). Fair value accounting and intangible assets: Goodwill impairment and managerial choice. Journal of Financial Regulation and Compliance, Vol. 18, No. 2, pp. 120-130 Jordan, C.E. and Clark, S.J. (2011). Big Bath Earnings Management: The Case of Goodwill Impairment Under SFAS No. 142. Journal of Applied Business Research, Vol. 20, No. 2, pp. 63-70 Jordan, C.E., Clark, S.J. and Vann, C.E. (2007). Using Goodwill Impairment to Effect Earnings Management during SFAS No. 142’s Year of Adoption and Later. Journal of Business & Economic Research, Vol. 5, No. 1, pp. 23-30 Massoud, M.F. and Raiborn, C.A. (2003). Accounting for Goodwill: Are We Better Off? Review of Business, Vol. 24, No. 2, pp. 26-32

53

Masters-Stout, B., Costigan, M.L., and Lovata, L.M. (2008). Goodwill impairments and chief executive officer tenure. Critical Perspectives on Accounting, Vol. 19, pp. 1370-1383 Peasnell, K.V., Pope, P.F., and Young, S. (2005). Board Monitoring and Earnings Management: Do Outside Directors Influence Abnormal Accruals? Journal of Business Finance & Accounting, Vol. 32, No. 7-8, pp. 1311-1346 Riedl, E.J. (2004). An Examination of Long-Lived Asset Impairments. The Accounting Review, Vol. 79, No. 3, pp. 823852 Schultze, W. (2005). The Information Content of Goodwill-Impairments under FAS 142: Implications for External Analysis and Internal Control. Schmalenbach Business Review, Vol. 57, No. 3, pp. 276-297 Sevin, S. and Schroeder, R. (2005). Earnings management: evidence from SFAS No. 142 reporting. Managerial Auditing Journal, Vol. 20, No. 1, pp. 47-54 Sevin, S., Schroeder, R. and Bhamornsiri, S. (2007), Transparent financial disclosure and SFAS No. 142. Managerial Auditing Journal, Vol. 22, No. 7, pp. 671-687 Sirower, M.L. and O’Byrne, S.F. (1998). The Measurement of Post-Acquisition Performance: Toward a Value-Based Benchmarking Methodology. Journal of Applied Corporate Finance, Vol. 11, No. 2, pp. 107-121 Statement of Financial Accounting Standards No. 141 Business Combinations (2001). Issued by Financial Accounting Standards Board (FASB). Retrieved from: http://www.fasb.org Statement of Financial Accounting Standards No. 142 Goodwill and Other Intangible Assets (2001). Issued by Financial Accounting Standards Board (FASB). Retrieved from: http://www.fasb.org Van de Poel, K., Maijoor, S. and Vanstraelen, A. (2009). IFRS goodwill impairment test and earnings management: the influence of audit quality and the institutional environment. Working paper. Retrieved from: http://www.fdewb.unimaas.nl/isar2009/02_17_van_de_poel_maijoor_vanstraelen.pdf Van Tendeloo, B. and Vanstraelen, A. (2005). Earnings Management under German GAAP versus IFRS. European Accounting Review, Vol. 14, No. 1, pp. 155-180 Watrin, C., Strohm, C., and Struffert, R. (2006). The Joint Business Combinations Project. The CPA Journal, January 2006 Whitley, R.D. (1988). The Possibility and Utility of Positive Accounting Theory. Accounting, Organizations and Society, Vol. 13, No. 6, pp. 631-645 Woodlock, P. and Peng, G. (2009). How Will Valuation Changes Affect M&A Deals? The Journal of Corporate Accounting & Finance, Vol. 20, No. 4, pp. 49-61 Zucca, L.J. and Campbell, D.R. (1992). A Closer Look at Discretionary Writedowns of Impaired Assets. Accounting Horizons, Vol. 6, No. 3, pp. 30-41

54

APPENDIX I TABLE I - Overview of the variables in the regression model and their measurement DEPENDENT VARIABLES IMPAIRMENTit

Indicator variable taking value of 1 if a firm reports an impairment loss at the end of year t, and 0 otherwise.

IMPAIRMENT_AMOUNTit

The reported amount of the write-off deflated by total assets t-1 for write-off firms, and 0 for non-write-off firms. REPORTING INCENTIVES

CHANGE_CEOit

Indicator variable equal to 1 if the firm experienced a change in the CEO position in years t-1 or t, and 0 otherwise.

DEBTit

The debt to equity ratio of firm i in year t.

DEBT2it

The debt to total assets ratio of firm i in year t.

BATHit

Equals the change in firm i’s pre-impaired earnings before taxes from t-1 to t divided by total assets t-1, when this change is below the industry median of non-zero negative values, and 0 otherwise.

SMOOTHit

Equals the change in firm i’s pre-impaired earnings before taxes from t-1 to t divided by total assets t-1, when this change is above the industry median of non-zero positive values, and 0 otherwise. ECONOMIC FACTORS

∆SALESit

The change in firm i’s sales from year t-1 to year t, divided by total assets at the end of year t-1.

∆OCFit

The change in firm i’s operating cash flow from year t-1 to year t, divided by total assets at the end of year t-1.

∆ROAit

The percentage change in firm i’s return on assets (ROA) from year t-1 to t.

MTBVit

The market-to-book value of firm i in year t.

GDP_GROWTHit

The percentage change in Gross Domestic Product (GDP) from year t-1 to t in the country in which firm i is established. CONTROL VARIABLES

GOODWILLit

Ratio of firm i’s opening balance of goodwill on total assets at t-1.

SIZEit

The natural logarithm of firm i’s total assets in year t.

SIZE2it

The natural logarithm of firm i’s market capitalization in year t.

CAP_INTENSITYit

The ratio of long-term assets over total assets.

DMindi

The model controls for industry-fixed effects by including k-1 dummy variables, representing the 8 industry categories in the company sample. The variable DMind1i takes the value of 1 when firm i is classified in industry 1, and 0 otherwise.

55

APPENDIX II 21

TABLE II - Overview of the used data and codes in Datastream and Worldscope NAME

CODE

VARIABLE(S)

DESCRIPTION

Goodwill

WC18280

GOODWILL

Goodwill Impairment

WC18225

Total Assets

WC02999

IMPAIRMENT, IMPAIRMENT_AM OUNT IMPAIRMENT_AM OUNT, BATH, BATH2, BATH3, SMOOTH, ∆SALES, ∆OCF, CAP_INTENSITY, GOODWILL, SIZE

Goodwill represents the excess cost over the fair market value of the net assets purchased. Loss due to the impairment of goodwill.

Total Sales

WC01001

∆SALES

Earnings before Interest, Taxes, & Depreciation (EBITDA)

WC18198

Return on Assets (ROA)

WC08326

Debt-to-Equity

WC08231

Net Cash Flow – Operating Activities

WC04860

Market Capitalization

WC08001

BATH, BATH2, EBITDA represent the earnings of a company before interest BATH3, SMOOTH expense, income taxes and depreciation. It is calculated by taking the pretax income and adding back interest expense on debt and depreciation, depletion and amortization and subtracting interest capitalized. ∆ROA (Net income before preferred dividends + ((interest expense on debt-Interest capitalized) * (1-tax rate))) / last year's total assets * 100 DEBT (Long term debt + short term debt & current portion of long term debt) / common equity * 100 ∆OCF Net Cash Flow – Operating Activities represent the net cash receipts and disbursements resulting from the operations of the company. It is the sum of funds from operations, funds from/used for other operating activities and extraordinary items. SIZE2 Market Price-Year End * Common Shares Outstanding

Market to Book Value

MTBV

MTBV

Total Current Assets

WC02201

CAP_IN TENSITY

Debt to Total Assets

WC08236

DEBT2

Total assets represents the sum of total current assets, long term receivables, investment in unconsolidated subsidiaries, other investments, net property plant and equipment and other assets.

Net sales or revenues represent gross sales and other operating revenue less discounts, returns and allowances.

Refers to the share price of a company divided by the net book value Represents cash and other assets that are reasonably expected to be realized in cash, sold or consumed within one year or one operating cycle. Generally, it is the sum of cash and equivalents, receivables, inventories, prepaid expenses and other current assets. (Short Term Debt & Current Portion of Long Term Debt + Long Term Debt) / Total Assets * 100

21

All names, codes and descriptions are retrieved from the Worldscope Database – Datatype Definitions Guide by Thomson Financial (2003).

56

APPENDIX III TABLE III Sample Selection per Year (Total N = 1748)

Initial sample Observations deleted: Firms with two-digit SIC codes 60-67 (‘Finance, Insurance, and Real Estate’) Inactive or delisted Not reporting according to the IFRS accounting standards in period 20062009 a Missing data Final sample Write-off observations Non-write-off observations

2006 1,926

2007 1,926

2008 1,926

2009 1,926

(427)

(427)

(427)

(427)

(441) (479)

(441) (479)

(441) (479)

(441) (479)

(153) 426 87 339

(145) 434 81 353

(138) 441 108 333

(132) 447 115 332

a

The main reason for firm observations to be excluded from the sample is because they did not have a separate goodwill account on their opening balance sheet in a specific year or because CEO information for one of the years was unavailable.

57

TABLE IV a PANEL A: Descriptive Statistics Per Year 2006-2009 (Total N = 1748 ) 2006 Mean

Std. Dev

N

426 426

0,2042 0,0020

0,4039 0,0112

426 426 426 426

0,1643 0,8253 -0,0128 0,0308

426 426 426 426 426 426 426 426

0,0491 0,1376 0,0094 -0,0971 2,3481 0,4536 0,1397 19,7921

N Dependent variable IMPAIRMENTit IMPAIRMENT_AMOUNTit Reporting incentives CHANGE_CEOit DEBTit BATHit SMOOTHit Control variables GDP_GROWTHit ∆SALESit ∆OCFit ∆ROAit MTBVit CAP_INTENSITYit GOODWILLit SIZEit a

2007 Mean

2008 Mean

Std. Dev

N

434 434

0,1866 0,0018

0,3901 0,0108

0,3710 1,0700 0,0408 0,0618

434 434 434 434

0,1843 0,8205 -0,0089 0,0278

0,0025 0,2468 0,0728 2,7013 1,7843 0,2044 0,1359 2,1530

434 434 434 434 434 434 434 434

0,0512 0,1234 0,0063 -0,0656 2,4819 0,4677 0,1512 19,9328

N

2009 Mean

Std. Dev

Std. Dev

441 441

0,2449 0,0046

0,4305 0,0186

447 447

0,2573 0,0033

0,4376 0,0133

0,3882 1,0267 0,0308 0,0619

441 441 441 441

0,2177 0,9680 -0,0224 0,0172

0,4131 1,2238 0,0485 0,0468

447 447 447 447

0,2528 0,8781 -0,0231 0,0178

0,4351 1,1132 0,0469 0,0480

0,0119 0,2175 0,0741 2,1011 1,8759 0,2054 0,1382 2,1071

441 441 441 441 441 441 441 441

0,0262 0,0913 0,0086 -0,3933 1,3232 0,4732 0,1607 19,9955

0,0110 0,2118 0,0759 2,6323 1,3358 0,2009 0,1425 2,0916

447 447 447 447 447 447 447 447

-0,0229 -0,0649 0,0157 -0,6249 1,6939 0,4882 0,1687 19,9517

0,0079 0,1775 0,0785 2,6910 1,4565 0,1985 0,1451 2,1117

Data have been winsorized at the top 1% and bottom 1%.

58

TABLE IV PANEL B: Descriptive Statistics 2006 and 2009

Dependent variable IMPAIRMENTit IMPAIRMENT_AMOUNTit Reporting incentives CHANGE_CEOit DEBTit BATHit SMOOTHit Control variables GDP_GROWTHit ∆SALESit ∆OCFit ∆ROAit MTBVit CAP_INTENSITYit GOODWILLit SIZEit

N

2006 Mean

N

2009 Mean

Std. Dev

Std. Dev

426

0,2042

0,4039

447

0,2573

0,4376

426

0,0020

0,0112

447

0,0033

0,0133

426

0,1643

0,3710

447

0,2528

0,4351

426

0,8253

1,0700

447

0,8781

1,1132

426

-0,0128

0,0408

447

-0,0231

0,0469

426

0,0308

0,0618

447

0,0178

0,0480

426

0,0491

0,0025

447

-0,0229

0,0079

426

0,1376

0,2468

447

-0,0649

0,1775

426

0,0094

0,0728

447

0,0157

0,0785

426

-0,0971

2,7013

447

-0,6249

2,6910

426

2,3481

1,7843

447

1,6939

1,4565

426

0,4536

0,2044

447

0,4882

0,1985

426

0,1397

0,1359

447

0,1687

0,1451

426

19,7921

2,1530

447

19,9517

2,1117

t-test (p-value) -1,863* (0,063) -1,645 (0,100) -3,238*** (0,001) -0,714 (0,475) 3,474*** (0,001) 3,450*** (0,001) 183,688*** (0,000)

13,856*** (0,000) -1,222 (0,222) 2,891*** (0,004) 5,946*** (0,000) -2,537** (0,011) -3,043*** (0,002) -1,106 (0,269)

*** ** *

, , , denotes significance at 0,01, 0,05 or 0,10 levels respectively.

59

APPENDIX IV TABLE V

TABLE VI

Dichotomous Logistic Regression Model: Sensitivity Analysis Controlling for Country Effects

Multiple Linear Regression Model: Sensitivity Analysis Controlling for Country Effects

Ia II b III c IV d Ve VI f VII g Prediction Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient INTERCEPT

?

(p-value)

(p-value)

(p-value)

(p-value)

(p-value)

(p-value)

-6,263***

-6,281***

-6,343***

-6,130***

-6,957***

-6,296***

(p-value) -6,816***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

Reporting Incentives CHANGE_CEO it DEBTit BATHit SMOOTHit

0,757***

0,744***

(0,000)

(0,000)

-

0,127**

0,135**

(0,034)

(0,026)

-

-6,365***

-5,814***

(0,000)

(0,000)

+

?

0,007**

0,008**

0,008**

0,009**

-0,003

0,007*

-0,004

(2,016)

(2,083)

(2,110)

(2,335)

(-0,731)

(1,718)

(-1,049)

0,170

0,180

(0,899)

(0,895)

CHANGE_CEO it

+

0,089***

0,108***

(3,914)

(4,542) DEBTit

-

0,051**

0,057**

(2,013)

(2,125) BATHit

-

-0,347***

-0,345***

(-13,996)

(-13,924) SMOOTHit

+

0,046*

0,078***

(1,735)

(3,104)

Economic Factors -

∆OCFit

-

∆ROAit

-

MTBVit

INTERCEPT

Reporting Incentives +

Economic Factors ∆SALESit

Ia II b III c IV d Ve VI f VII g Prediction Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic) (t-statistic)

-

0,089

0,144

0,091

0,358

0,080

0,381

(0,760)

(0,623)

(0,754)

(0,223)

(0,791)

(0,213)

-0,587

-0,595

-0,532

0,291

-0,610

0,185

(0,515)

(0,506)

(0,556)

(0,751)

(0,507)

(0,842)

-0,045*

-0,043*

-0,044*

-0,020

-0,045*

-0,020

(0,051)

(0,065)

(0,057)

(0,402)

(0,052)

(0,407)

-0,097**

-0,103**

-0,115***

-0,090**

-0,097**

-0,115***

(0,026)

(0,018)

(0,009)

(0,035)

(0,025)

(0,008)

Control Variables h

0,044

0,048*

0,045*

0,114***

0,033

0,099***

(-1,500)

(1,936)

(1,793)

(4,682)

(1,254)

(3,970)

-0,061**

-0,062*** -0,061**

0,004

(-2,523)

(-2,580)

(-2,494)

(0,162)

-0,070*** -0,011 (-0,475) (-2,830)

-0,042*

-0,041*

-0,041*

0,025

-0,040*

0,030

(-1,757)

(-1,711)

(-1,707)

(1,088)

(-1,692)

(1,319)

-0,057**

-0,061**

-0,071*** -0,052**

(-2,358)

(-2,504)

(-2,818)

(-2,257)

-0,063*** -0,077*** (-3,223) (-2,580)

0,149***

0,143***

0,141***

0,146***

0,129***

0,144***

0,133***

(5,303)

(5,077)

(5,046)

(5,174)

(4,825)

(5,136)

(5,001)

-0,032

-0,028

-0,035

-0,036

0,026

-0,019

0,028

(-1,124)

(-0,991)

(0,211)

(-1,261)

(0,942)

(-0,663)

(1,011)

-0,019

-0,015

-0,016

-0,024

0,010

-0,018

-0,002

(-0,621)

(-0,499)

(0,606)

(-0,787)

(0,338)

(-0,570)

(-0,073)

No. of Observations

1748

1748

1748

1748

1748

1748

∆SALESit

-

∆OCFit

-

∆ROAit

-

MTBVit

-

Control Variables h 1,526***

1,436***

1,411***

1,487***

1,352***

1,438***

1,393***

(0,003)

(0,005)

(0,006)

(0,004)

(0,008)

(0,005)

(0,007)

0,274***

0,285***

0,277***

0,276***

0,307***

0,285***

0,290***

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

(0,000)

-1,230***

-1,242***

-1,242***

-1,362***

-1,122***

-1,244***

-1,285***

(0,003)

(0,002)

(0,003)

(0,001)

(0,006)

(0,002)

(0,002)

No. of Observations Chi-square χ² (model)

1748

1748

1748

1748

1748

1748

1748

135,386*** 145,678*** 174,406*** 150,049*** 163,988*** 145,694*** 195,094***

Model R Square

0,028

0,037

0,048

0,039

0,134

0,038

1748 0,149

Chi-square χ² (change)

-

10,291**

28,729***

4,372**

18,321***

0,016

49,416***

R Square change

-

0,009***

0,011***

0,003**

0,097***

0,002*

0,112***

Nagelkerke R Square

0,114

0,122

0,145

0,126

0,137

0,122

0,161

Model F-test

2,583***

2,848***

3,620***

2,923***

11,113*** 2,858***

GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY . it it it b Reduced model with predictors: (Constant), GOODWILLit, SIZEit, CAP_INTENSITYit, ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. c Reduced model with predictors: (Constant), CHANGE_CEO , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it d Reduced model with predictors: (Constant), DEBT , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it e Reduced model with predictors: (Constant), BATH it, GOODWILLit, SIZEit, CAP_INTENSITYit , ∆SALES it , ∆OCF it, ∆ROAit, MTBVit. f Reduced model with predictors: (Constant), SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it g Full model with predictors: (Constant), CHANGE_CEO , DEBT , BATH , SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , it it it it it it it it ∆OCF it, ∆ROAit, MTBVit. h The regression model also controls for industry-fixed effects and for country effects.

GOODWILLit SIZEit CAP_INTENSITYit

+ +/-

11,158

***, **, *, denotes significance at 0,01, 0,05 and 0,10 levels, respectively. a Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY . it it it b Reduced model with predictors: (Constant), GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it c Reduced model with predictors: (Constant), CHANGE_CEO , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , it it it it it it it MTBVit. d Reduced model with predictors: (Constant), DEBT , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it e Reduced model with predictors: (Constant), BATH , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it f Reduced model with predictors: (Constant), SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , ∆SALES , ∆OCF , ∆ROA , MTBV . it it it it it it it it g Full model with predictors: (Constant), CHANGE_CEO , DEBT , BATH , SMOOTH , GOODWILL , SIZE , CAP_INTENSITY , it it it it it it it ∆SALES it, ∆OCF it, ∆ROAit, MTBVit. h The regression model also controls for industry-fixed effects and for country effects.

60