Examining the patterns of goodwill impairments in Europe and the US

Transcription

1 Examining the patterns of goodwill impairments in Europe and the US Paul André, HEC Lausanne, Andrei Filip, ESSEC Business School, France, Luc Paugam, HEC Paris, France, Forthcoming in Accounting in Europe Abstract: We examine the patterns of goodwill impairments in Europe and in the US over the period from 2006 to 2015, for a sample of more than 35,000 firm-year observations. We define the timeliness of goodwill impairments as the frequency of accounting impairments conditional to indications of economic impairments. We measure indications of economic impairment with three metrics: equity market value minus equity book value less than goodwill, market-to-book smaller than one, and negative EBITDA. Our research strategy leads us to draw very different conclusions than those in the recent EFRAG (2016) study. While median levels of goodwill on the books between US and European firms are relatively similar, we find several indications that US firms recognize timelier impairments, at least during 2008 and 2009, i.e., the early years of the financial crisis. We further document that US impairers write down a much greater percentage of their beginning balance of goodwill than European impairers. During the financial crisis, the median level of impairment by US firms was 63% of opening goodwill in 2008 and 40% in 2009, whereas median European write-downs were only 6% and 7% of goodwill, respectively. Even though European firms are more likely to impair over multiple years, the cumulative impairments never come close to the level of US firms, be it in a single year or cumulative over multiple years. We also find that the frequency of accounting impairment is small compared to the number of firms presenting evidence of economic impairment: only 20 to 25% of firms recognize impairments depending on the measure of economic impairment. This has often been interpreted by academics as a sign of untimely writeoffs. Accounting differences between US GAAP and IFRS are unlikely to explain our results. One caveat of our analysis is that it does not allow us to draw conclusions on whether the observed differences between US and European firms are driven by differences in conditional conservatism and/or big bath accounting practices. Key words: Goodwill, Impairment, IFRS 3, IAS 36, Europe, US Acknowledgements: We thank Anastasia Borisova, Leonidas Doukakis, Francesco Mazzi, Alain Schatt and Ioannis Tsalavoutas for very helpful comments.

2 1. Introduction Accounting for goodwill has been the source of never ending debates in the academic and practitioner literature going as far back to William Harris paper The law and practice in relation to goodwill delivered to the Manchester Accountants Student Society in , if not before. The recent Post-Implementation Review (PIR) into International Financial Reporting Standards (IFRS) 3 Business Combinations has offered another opportunity to open up the discussion. Recall that the International Accounting Standards Board (IASB) rejected the amortisation and impairment approach, primarily because it is not possible to reliably determine the useful life and the pattern of consumption of goodwill, so that the amortisation charge over any given period is only an arbitrary estimate. Also, the Financial Accounting Standards Board (FASB) when publishing Statement of Financial Accounting Standards (SFAS) 142 (2001) argued that (ASBJ, EFRAG and the OIC (July 2014, p. 11)): non-amortisation of goodwill with adequate impairment testing and appropriate disclosure promotes transparency in financial reporting, and thus provides useful information to those who rely on financial statements. A Research Group formed of members from Accounting Standard Board of Japan (ASBJ), European Financial Reporting Advisory Group (EFRAG) and Organismo Italiano di Contabilità (OIC) (July 2014, p. 4) note that: many constituents questioned the usefulness of the information resulting from the impairment-only approach. It also found that preparers and auditors are concerned about the cost and subjectivity of the impairment testing in accordance with International Accounting Standards (IAS) 36 Impairment of Assets and whether impairment losses are recognised in a timely manner. Furthermore, many have indicated that the impairment-only approach may have played a role in the financial crisis. That is, since the current approach does not allow the depiction of yearly consumption of acquired goodwill, impairment losses often come too late. This effect has been made evident in recent years in which many entities recognised impairment losses of goodwill years after the financial crisis, when financial markets, possibly, had already taken them into account. In response to the PIR of IFRS 3, the IASB and FASB have initiated research projects on the topic. These have been discussed at board meetings over the spring of Also, EFRAG (2016) examines goodwill and impairment levels for non-financial constituents of the S&P Europe 350. The study provides some benchmarking with constituents of the US S&P 500. The study does not highlight major differences between the US and European firms with regard to impairment recognition patterns. Impairment recognition is a key accounting mechanism ensuring conditional conservatism of financial reports (Amiraslani, Iatridis, and Pope 2013; Roychowdhury and Martin 2013; André, Filip and Paugam 2015). Conditional conservatism, or timely loss recognition, is defined as the greater aggressiveness in the recognition of bad news relative to the recognition of good news (Basu 1997). Impairments are instrumental in achieving conditional conservatism because they ensure that assets are not carried on the balance sheet at a value greater than their economic ( recoverable ) value and that a loss be recognized on the income statement when this is the case. Yet, several studies investigating US data document that goodwill accounting impairments tend to be untimely (e.g., Hayn and Hughes 2006; Ramanna and Watts 2012; Li and Sloan 2015; Filip, Jeanjean and Paugam 2015). André, Filip and Paugam (2015) also document evidence consistent with untimely goodwill impairment in 1 See reference in Brief (1969) 1

3 the post-ifrs adoption period for European firms. However, to the best of our knowledge, we are unaware of a comprehensive study exploring differences in the timeliness of goodwill impairment recognition across US and European firms. We look at the frequency and magnitude of goodwill impairments in Europe and in the US over the period from 2006 to 2015, i.e., the post-ifrs adoption period for European listed firms. Our sample includes more than 35,000 firm-year observations. We define timeliness as the association between economic indicators suggesting that goodwill is impaired and actual accounting impairments. We measure economic impairment by three alternative metrics used in the literature: equity market value minus equity book value less than goodwill, market-tobook smaller than one, and negative EBITDA. We examine and compare the frequency of accounting impairments to indications of economic impairments across US and European firms. First, we document that US firms report larger but less frequent impairments than European firms. The differences between US impairers and European impairers are large: the median impairment for US impairers accounts for 33% of beginning of the year goodwill while it is only 5% for European impairers. Next, our analysis indicates that, relative to European firms, US firms are more likely to impair goodwill when there are indications that goodwill is economically impaired. The differences between US and European firms are most striking during the years 2008 and 2009, when US firms were much more aggressive in the recognition of goodwill impairment than their European counterparts. We find consistent evidence of greater timeliness for US firms relative to European firms using our three measures of economic impairments in both univariate and multivariate analyses. This study contributes to the literature by exploring the timeliness of goodwill impairments. We find that although rules with regard to goodwill impairments are relatively similar in the US and in Europe, firms exhibit very different patterns of goodwill impairment recognition. We discuss several factors that may explain such differences. In addition, we believe that our results are important to standard setters and regulators considering the adequacy of accounting standards and the enforcement of compliance with impairment tests. Our results seem to suggest that a margin for improvement may exist with regard to the timeliness of goodwill impairments in Europe. The remainder of this paper is organized as follow. We present prior research in section 2, describe our sample in section 3, present our results in section 4, discuss avenues for future research in section 5 and conclude in section Prior research The timeliness and informativeness of goodwill impairments has been the object of many studies since the change to the impairment approach. For a recent literature review of European data, see Schatt, Doukakis, Bessieux-Ollier and Walliser (2016) in this issue. Boennen and Glaum (2014), D Arcy and Tarca (2016) and Wen and Moehrle (2015) also discuss US findings. These reviews conclude that there is significant evidence of potential untimeliness of goodwill write offs. Identifying firms with economic impairment is critical for our research objective. Three alternative metrics of possible economic impairment at the firm level have been used in the literature. First, Beatty and Weber (2006) and Verriest and Gaeremynck (2009) consider that an impairment is likely if a firm s market value minus its book value of equity is smaller than the amount of goodwill reported on the balance sheet. Second, Ramana and Watts (2012) argue that there is potentially such indication if a firm s market-to-book is lower than one for a certain 2

4 period. Finally, firms exhibiting a pre-impairment operating loss could also be candidates for an impairment (EFRAG 2016). In 2013, the European Securities and markets Authority (ESMA) published a report on its investigation of accounting practices related to impairment testing of goodwill and other intangibles. They looked at 235 issuers with significant goodwill in Impairments were limited to a handful of firms mostly in telecommunications and financial services. They also note that of the 43% of the sample with a market-to-book lower than one, only 47% recognized an impairment loss. Verriest and Gaeremynck (2009) examine FTSE 300 firms in 2005 and Out of the 47 firms with likely impairments, i.e., market value minus book value of equity smaller than goodwill, only 53% actually impair. The ESMA report also criticises major disclosures as being boiler plate and not sufficiently entity-specific. Major concerns include: 1) limited discussion of key management assumptions; 2) lack of consistency in cross-firm sensitivity analysis; 3) little use of external information sources when determining fair value less cost to sell using discounted cash flows; 4) very optimistic projection of future growth rates and 5) missing information on discount rate used at the cash-generating unit. Tsalavoutas, André and Dionysiou (2014) find similar issues with disclosures when looking at a worldwide sample. Nevertheless, the issue is still of interest in this post IFRS 3 PIR period. Both the IASB and EFRAG have recently worked on this topic. Philippe Danjou, IASB Board member, presented some trend statistics at the Accounting in Europe/Financial Reporting Standards Committee of the EAA Symposium at the EAA Conference in Maastricht in May 2016 for a sample of non-financial members of the CAC 40. Filippo Poli from EFRAG, also at the Accounting in Europe/Financial Reporting Standards Committee of the EAA Symposium at the EAA Conference in Maastricht in May 2016, presented preliminary results that can now be found in EFRAG (2016). EFRAG s study is based on a sample of non-financial members of the S&P Euro 350, also compared with US S&P 500 firms Sample Table 1, presents our sample selection process. We start with all publicly listed firms in Europe and in the US as of December 31, 2015 that is some 8,500 firms in Europe and 13,000 in the US as per the Thomson EIKON database. After eliminating financials and firms without industry identifiers, we are left with 4,295 European firms and 6,873 US firms. We then collect data for the ten year period from , last year of available data at the time of the download. We choose to start in 2006, one year after the adoption of IFRS by EU listed firms to avoid the noise from the adoption year. After deleting observations with missing data, negative MTB firms and firms not using IFRS or US GAAP for at least two consecutive years, we are left with 27,172 and 28,897 European and American firm-year observations, respectively. Some 20% of the European and 18% of the US observations do not report any goodwill at the beginning of the year and are dropped from the analysis. Our final sample consists of 18,538 and 16,525 European and US firm-year observations, respectively. 3 (Table 1 here) 2 Both set of slides are available for EAA members at 3 We note that our sample suffers from survivorship bias but are unable to conjecture on how this might affect our results. 3

5 4. Results 4.1 Descriptive statistics Table 2 presents descriptive statistics of our sample. Our European and US firms are of similar average size, just over 4 billion euros. US sample firms seem to have higher Market-to-Book (MTB) ratios, 3.50 vs 2.34 but European firms slightly perform better; they exhibit a Returnon-Assets (ROA) of 8.8% vs. 7.7% for US firms. We focus on median levels of goodwill and goodwill impairments because their distributions are highly skewed. The level of goodwill on the books is almost similar, average (median) Goodwill/Total Assets is 16.8% (12.2%) in the US and 16.7% (11.8%) in Europe. The means are not statistically different between US and European firms. We know from prior literature that goodwill levels on the books are important. The EFRAG (2016) study looking at the constituents of the S&P Europe 350 finds a similar ratio of 17% when excluding the financial sector (3.5% when included). Our reported percentage of goodwill is higher than the 9% reported by Bens, Heltzer and Segal (2011) that includes the Compustat universe of 92,390 US firms from because their sample also includes firms with zero goodwill. It is also slightly higher than the Glaum, Schmidt, Street and Vogel (2013) 2005 sample of 357 European firms with an average of 11%. Goodwill to Equity is also quite similar between Europe and the US with mean (median) 47.4% (30.00%) and 50.6% (27.2%), respectively. Mazi, André, Dionysiou and Tsalavoutas (2016) find similar levels of goodwill to net book value. 4.2 Levels of goodwill over time (Table 2 here) Table 3 presents some statistics on the importance of goodwill over time. The median level of goodwill to total assets for the European sample ranges between 9.5% in 2006 to 12.5% in Figure 1a plots median levels over time. Median goodwill levels were at their highest in 2008, coming down during the financial crisis and remaining relatively stable thereafter. A similar pattern is observable from the median goodwill to equity levels (see Figure 1b). The median goodwill to equity for European firms during the period ranges between 26.3% in 2006 and 31.4% in In contrast, the median level of goodwill to total assets for the US sample ranges between 10.9% in 2009 to 13.6% in The difference between the US median is statistically higher when compared to the European median. Again, a similar pattern is observable for the median goodwill to equity levels. The median goodwill to equity for US firms during the period ranges between 24.3% in 2009 and 31.6% in Comparing the two blocs, we notice a different pattern: US levels take a significant drop in 2008 whereas there is a drop in Europe from 2009 to 2011 but not as sharp as that of US 4

6 firms. By 2015, US levels are comparable to European ones. Table 3 shows that US levels are statistically lower in both 2008 and 2009 during the financial crisis Timeliness of goodwill impairments (Table 3 and Figures 1a and b) We compare the levels of accounting impairments with various firm-level measures of economic impairment. As discussed above, we use three measures to identify firms carrying economically impaired goodwill: 1) market value of equity minus book value of equity less than book goodwill (MV BV Equity < GDWL); 2) market-to-book less than one (MTB < 1); and 3) negative earnings before interest, tax, depreciation and amortization (EBITDA < 0). We then assess the timeliness of accounting impairments by observing the conditional frequency of accounting impairment to each of our three measures of economic impairments. When using our first measure of economic impairment MV BV Equity < GDWL, Panel a of Table 4 reports the following over the period: - European firms were more likely to exhibit economic impairment than US firms (42.4% vs. 27.9%); - European firms were more likely to have an accounting impairment (unconditionally to evidence of economic impairment) than US firms (15.3% vs. 10.3%); - Both European firms and US firms are more likely to impair when there is evidence of economic impairment (in about 20% of cases), than when there is no evidence of economic impairment; - Yet, there is an important difference in 2008 between US and European firms: US firms took an impairment when there were indication of economic impairment in 34.9% of cases whereas it was only in 23.6% of cases for European firms. Moving to the next measure of economic impairment MTB < 1, in Panel b of Table 4, we find similar results: - European firms are more likely to exhibit economic impairment signs (27.2% for European firms vs. 15.4% for US firms) - European firms more likely to impair unconditionally to indications of economic impairments (15.3% vs. 10.3% US). - However, US firms are more likely to take an accounting impairment when MTB < 1 (25.9% vs. 19.7% for Europeans), for each year of our sample period. - Further, again in 2008, US firms were much more likely to have impaired if MTB < 1 (more than 40%) compared to Europeans (23.3%). Finally, Panel C of Table 4 presents results for our third metric of likely economic impairment: negative EBITDA. Differences for this measure are again somewhat similar. Slightly more US firms had negative EBITDA than European firms during the period covered, 18.0% and 15.2% respectively. Out of these, 25.8% of US firms take an impairment whereas 24% of Europeans do so over the ten years. However, the difference in remains quite striking: US firms in both years were much more likely to impair is they exhibit negative 4 We acknowledge that the two groups have different starting points. US firms stop amortizing goodwill as early as 2001 whereas most European firms stopped amortizing goodwill when they adopted IFRS in Table 3 does indicate that US firms had significantly higher GDLW/TA in 2006 (11.9% vs. 9.5%) but there is no difference when looking at GDWL/EQ (around 26%). 5

7 EBITDA: 50.7% for US firms vs 31.3% for European firms in 2008 and 37.5% for US firms and 26.2% for European firms in Level of impairments over time (Table 4 and figures 2a, 2b and 2c here) We now turn to levels of impairment and present what we believe to be the most striking difference between US and European firms. Table 5 and Figure 3 first present median levels of impairment when a firm does impair (we exclude firms that did not impair) independent of indications of economic impairment. The level of impairment over the 10-year-period, i.e., impairment to opening goodwill level IMPt/GDWLt-1, for US firms is 33.1% of opening goodwill whereas it is a mere 5.3% for European firms. While the relative level of US impairments is consistently far greater every year of the period examined, the most significant difference occurs in 2008: US firms wrote off 62.6% of goodwill that year whereas European firms wrote off only 6.6% of goodwill. We further examine differences in the level of impairment when there is a sign of economic impairment. Results can be found in Table 5 and Figures 4a, b and c. Again we focus on our same three measures of economic impairment: 1) MV BV Equity < GDWL; 2) MTB < 1; and 3) EBITDA < 0. Over the 10 year window we examine, we discover even sharper differences when conditioned on signs of economic impairment. Under our first measure, MV BV Equity < GW, US firms impair 39.8% of opening goodwill whereas Europeans firms only impair 6.4% of opening goodwill. Looking at impairments when MTB < 1, we find US firms write off 61.9% of opening goodwill whereas European firms only impair 9.2% of goodwill. When negative EBITDA is used as evidence of economic impairment, we find that US firms impair as much as 81.6% of opening goodwill whereas Europeans write down only 32.7%. Figures 4a, b and c confirm this trends in all years (except for 2006) with the most striking differences occurring in 2008 and When US firms exhibit negative EBITDA in 2008 or 2009, they impair 94.5% and 80.6% respectively when Europeans only write off 40.5% and 28.7% of opening goodwill. (Table 5 and Figures 4a, 4a and 4c here) We also investigate differences in relative goodwill impairments over multiple years in Table 6. It could have been argued that European firms take smaller impairments in a single year but make more frequent impairments in many years so that the cumulative write-off would be similar to the write-off observed for US firms in a single year. Table 6 presents descriptive statistics for the sub-sample of firms that impair goodwill in year t and shows the number and magnitude of impairments for impairers over a three-year window. Results confirm that US firms are more likely to take a write-off in a single year (i.e., year t) over any 3-year overlapping period (61.9% of cases for US firms vs. 44.1% for European firms) whereas European firms are more likely to book impairments over two and three year windows (31.9% of cases with two years of write downs and 24.0% of cases with three subsequent years of write-offs) than US firms (28.1% and 10.1%, respectively). Nevertheless, cumulative impairments to opening goodwill for US firms are 50.4% if in one year, 59.3% if in two years and 57.3% if in three years. European firms never reach these levels: 9.9% if in one year, 12.9% if in two years and 11.0% if in three years. 6

8 (Table 6 here) In an additional analysis, we examine a multivariate logistic model to see if indications of economic impairment in the year of impairment (in t) or two previous years (in t-1 in t-2) explain the occurrence of an accounting impairment in year (t). We estimate the following probit model: Pr(DIMPt = 1) = b0 + b1decimpt + b2decimpt-1 + b3decimpt-2 + b4dus + b5dus*ecimpt + b6dus * decimpt-1 + b7dus * decimpt-2 + b8gdwl/tat-1 + b9size + Industry Fixed Effects + Year Fixed Effects + ε where: DIMP = 1 if the firm impairs goodwill, and 0 otherwise; decimp = 1 for one of our three measures of economic impairment: 1 if MV BV Equity < GDWL; MTB < 1; and 3) EBITDA < 0, and 0 otherwise. We use these three measures alternatively in three different regressions. dus = 1 for US firms, and 0 otherwise; GDWL/TA = opening amount of goodwill expressed in percentage of total assets SIZE = natural logarithm of total assets. Following Basu (1997) and the univariate analysis presented above, we reason that the association between dummies for current year economic impairment indicators decimpt (economic impairment in t) and the likelihood of an accounting goodwill impairment DIMPt (accounting impairment in t) measures the timeliness of goodwill impairment, i.e., impairment timeliness is measured with coefficient b1 for European firms. We focus on differences between US and European firms in the timeliness of impairment. The main coefficient of interest is b5 that captures the incremental timeliness of impairments booked by US firms relative to impairments booked by European firms. The total timeliness of impairments for US firms is measures by (b1 + b5). The association between lagged economic impairment indicators decimpt-1 (economic impairment in t-1) and decimpt-2 (economic impairment in t-2) and the likelihood of accounting impairment DIMPt (accounting impairment in t) measures the delayed response to economic goodwill impairments. In other words, delayed response to economic impairment for European firms is measured by coefficient b2 (one year delay) and b3 (two years delay). The incremental response to delayed impairment for US firms is captured by b6 (one year delay) and b7 (two year delay). Coefficient b0 captures the likelihood of an impairment independent from poor economic performance for European firms. It can be interpreted as a measure of unconditional conservatism. Coefficient b4 measures the incremental unconditional conservatism by US firms. We also control for opening goodwill balance and size that are likely to affect the likelihood of an impairment. We also include industry and year fixed effects. Table 7 reports estimation results of the model. Estimation results confirm our prior analysis. Coefficient b1 (decimpt in year t) that captures the timeliness of goodwill impairment for European firms is positive, whatever the measure of economic impairment used (significant at less than 1%, two sided tests). This implies that impairments in Europe exhibit positive timeliness. However, we are unable to assess whether it is the adequate level of timeliness. Signs of economic impairment in the two years before (decimpt-1 and decimpt-2) also explain 7

9 impairments in year t. This indicates a delayed response to economic impairment in the part of European firms. Our model allows to compare unconditional and conditional conservatism between US firms and European firms. First, we confirm that in any given year, US firms are less likely to take an impairment as compared to Europeans: coefficient b4 is always negative (significant at less than 1%, two sided tests). This is evidence that US firms are less unconditionally conservative than European firms with regard to goodwill impairment recognition. Second, indications of economic impairment in year t for US firms are significantly more likely to lead to an accounting impairment in year t than for European firms across our three measures of economic impairment: coefficient b5 on dus*decimpt is positive (significant at less than 1%, two sided tests). This is evidence of greater conditional conservatism with regard to impairment tests in the US relative to Europe. Further, we generally do not find conclusive evidence of a stronger response from US firms to lagged indication of economic impairments relative to European firms (see coefficients on dus*decimpt-1 or dus*decimpt-2). Overall, US firms appear to be timelier than Europeans in recognizing economic goodwill impairments on their books while European firms are unconditionally more conservative. (Table 7 here) 5. Potential Explanations and Avenues for Future Research Further research could investigate several potential explanations of the reported differences between European and US firms. We briefly discuss in this section the role of differences in accounting standards and the impact on conditional conservatism and big-bath behaviour. Accounting differences While IFRS and US GAAP are quite similar with respect to business combinations, there nevertheless some differences that could potentially affect the timing and size of goodwill impairments. First, since the 2007 revised version of the US SFAS 141 Business Combinations, effective in 2009, the US requires an acquirer to recognize the assets acquired, the liabilities assumed, and any non-controlling interest in the acquiree at the acquisition date, measured at their fair values as of that date, even for acquisition less than 100%. This is sometimes called the full goodwill approach or measuring the non-controlling interest at the fair value approach rather than the proportionate share method previously used in IFRS. The IASB also revised IFRS 3 at that time to allow a choice of either method. Measuring the non-controlling interest at the fair value approach leads to greater levels of goodwill booked. We know of no study having examined in detail the choices made by European firms since 2009, however, our results do not document any significant increase in goodwill levels at least up to Second, in the US, goodwill is allocated to reporting units (RU) that are expected to benefit from the synergies. A RU is an operating segment or one level below an 5 Tsalavoutas, André and Dionysiou (2014) note the following: Out of the 76 companies for which acquisitions involve between 50% and 99% of the acquiree s assets, 33 remain silent on how the non-controlling interest is measured. Hence, users do not receive full information as IFRS 3 now offers two potential ways of measuring non-controlling interest. Additionally, only 11 companies (14.4%) explicitly state that they measure their non-controlling interest at fair value (full goodwill approach) 8

10 operating segment. Under IFRS, goodwill is allocated to cash-generating units (CGU). A CGU is defined as the smallest identifiable group of assets that generates cash inflows that are largely independent of the cash inflows from other assets or groups of assets. Further a CGU should represent the lowest level within the entity at which the goodwill is monitored for internal management purposes; and should not be larger than an operating segment as defined by IFRS 8. Since a CGU is likely smaller than a RU, testing the recoverable amount of a smaller unit to its carrying value could potentially lead to a greater incidence of impairments. Managers may have fewer options to delay goodwill impairment if goodwill is allocated to smaller units in which the value of assets are less likely to compensate one another. We find a greater incidence for European firms but also lower levels of impairments. Third, US GAAP requires a two-step approach whereas IFRS has a one-step approach. Note that in the US, an entity may first assess qualitative factors to determine whether the twostep goodwill impairment test is necessary. If the entity determines, based on the qualitative assessment, that it is more likely than not that the fair value of a reporting unit is below its carrying amount, the two-step impairment test is performed. Under the US approach, Step 1 involves comparing the fair value and the carrying amount of the reporting unit, including goodwill. If the fair value of the reporting unit is less than the carrying amount, Step 2 is completed to determine the amount of the goodwill impairment loss, if any. Step 2 involves assessing the goodwill impairment which is measured as the excess of the carrying amount of goodwill over its implied fair value. The implied fair value of goodwill, which is calculated in the same manner that goodwill is determined in a business combination, is the difference between the fair value of the reporting unit and the fair value of the various assets and liabilities included in the reporting unit. Any loss recognized is not permitted to exceed the carrying amount of goodwill. The impairment charge is included in operating income. Under IFRS, the recoverable amount of the CGU or group of CGUs (i.e., the higher of its fair value minus costs to sell and its value in use) is compared with its carrying amount. Any impairment loss is recognized in operating results as the excess of the carrying amount over the recoverable amount. The impairment loss is allocated first to goodwill and then on a pro rata basis to the other assets of the CGU or group of CGUs to the extent that the impairment loss exceeds the book value of goodwill. The first stage qualitative assessment in the US could suggest less occurrence of impairments, it is not clear which overall effect of these methods can have on timeliness and level of impairment. Conditional conservatism Impairment testing is a classic example of conditional conservatism. Conditional conservatism is the greater aggressiveness in the recognition of bad news than in the recognition of good news and is considered a key qualitative characteristic of financial reporting (Watts 2003; Francis, LaFond, Olsson and Schipper 2004; Ecker et al. 2006; Ball, Kothari, and Robin 2000; Dechow, Ge and Schrand 2010; and Kothari, Ramanna and Skinner 2010). This form of newsdependent prudence ensures that potential economic losses are reported in earnings in a timely fashion, whereas the recognition of potential economic gains is delayed. Conditional conservatism is distinguished from unconditional conservatism, also known as ex-ante or newsindependent prudence, consisting in systematically understating the book value of net assets relative to their economic value, independent from any news, e.g., automatic expensing of R&D costs (Pope and Walker 2003; Beaver and Ryan 2005). 9

11 There is a large academic literature that suggests that US firms are more conditionally conservative than European firms, at least the code law countries (Ball et al. 2000, Kothari, Ramanna and Skinner 2010). André, Filip and Paugam (2015) suggest that the decrease in conditional conservatism they document in Europe in the post IFRS period is linked to untimely goodwill impairment decisions. Amiraslani, Iatridis and Pope (2013) indicate that the level of timeliness is related to country specific factors such as the level of stock market development, ownership concentration, investor protection and enforcement. More work is probably needed to understand such cross-country differences and whether the affect goodwill impairment patterns. Our results suggest that US firms are being more conditionally conservative than European firms as a whole with regard to goodwill impairments. Big bath behaviour The big bath behaviour has also been documented for a long time. Under certain circumstances, managers may have incentives to clean the books, i.e., write off as much assets as possible. One such circumstance is suggested early on by Healy (1985) in his examination of bonus schemes. He cites (p. 86): If earnings are so low that no matter which accounting procedures are selected target earnings will not be met, managers have incentives to further reduce current earnings by deferring revenues or accelerating write-offs, a strategy known as 'taking a bath'. This strategy does not affect current bonus awards and increases the probability of meeting future earnings' targets. Another circumstance where big bath behaviour has been documented is when there is a change in CEO or CFO. New management can clean the slate and blame the old management for the poor results this period, this again increasing the probability of showing better results in the future. Early evidence appeared in Murphy and Zimmerman (1993). While we do not have specific data on CEO/CFO turnover for our sample, CEO turnover data from Strategy& of PWC (formerly Booz & Company) for the top 2,500 International companies since 2000 ( does not suggest US CEO turnover could explain differences. For example, in both 2008 and 2009, forced CEO turnovers were more important in Western Europe than the US/Canada. Our results indicate that having negative EBITDA leads to a greater frequency and level of goodwill impairment. While being in a loss position can suggest some economic difficulties, we cannot rule out stronger big bath behaviour in the US relative to Europe. 6. Conclusion We look at the patterns of goodwill impairments in Europe and the US over the period from 2006 to 2015, i.e., the post-ifrs period for European listed firms. Our sample includes more than 35,000 firm-year observations. Using three definitions of economic impairment, we examine the level of accounting impairments to economic impairments as a measure of timeliness. We examine median levels of goodwill and goodwill impairments since their distributions are highly skewed. While median levels of goodwill on the books are relatively similar between US and European firms, there are indications that US firms were more likely to impair when there is economic indicators of potential impairment, in particular in the early years of the financial crisis in We further document that US firms when they impair, impair a much 10

12 greater percentage of their goodwill (median impairments to opening goodwill for US firms is 33% whereas it is only 5% for European firms). During , US firms that impaired wrote off 63% and 40% of goodwill, respectively, whereas European impairers wrote-down only 6 to 7% of goodwill. European firms are more likely to impair on multiple years, however, even the cumulative impairments never approach the level US firms take, be it in one or multiple years. The percentage of write-downs are at their highest when there are any signs of economic impairment for the US firms (40% when market value of equity minus book value of equity is smaller than goodwill; 62% when market-to-book is smaller than one or 82% when EBITDA is negative) but mostly only when EBITDA is negative for European firms (33% write-downs). This type of behaviour when EBITDA is negative may also be viewed as big bath behaviour. We also document that the actual number of impairments are small compared to the number of firms exhibiting economic impairment: only 20 to 25% of firms impair depending on the measure of economic impairment chosen. While this has often been interpreted by academics as a sign of untimely write-offs, we also show that in any given year there is at least as many firms booking an impairment while not exhibiting at least one of our measures of economic impairment. This could be a sign of either delayed recognition by many firms or that our measures to capture economic impairment are inaccurate. We leave it to future research to distinguish between these two explanations. There are clear evidence that US firms are timelier if one considers that our measures of economic impairment are good proxies. Our research strategy leads us to draw very different conclusions than those in the recent EFRAG (2016) study. One, we have a larger sample of firms since their study limits itself to Europe S&P 350 companies and US S&P 500 firms. Two, we focus on the median and examine not the full sample (where many firms do not impair) but only those that have impaired or that have impaired when showing signs of potential economic impairment. Taking the average when many firms do not impair gives very low levels of average impairment losses. Further, we do not compare aggregate or absolute levels of impairment which can be difficult to analyse. Three, we exclude financial firms. These firms further skew the distributions of the level of goodwill or impairments. More work would be needed to reconcile both sets of results. Finally, we also leave to future researcher the task of explaining whether the larger frequency and level of impairments by US firms is driven by greater conditional conservatism (and if so understanding the demand for this) and/or by big bath behaviour. Further research could also further examine differential perception by CFOs and management teams on the perceived complexities about implementing impairment testing such as undertaken by Mazzi, Liberatore and Tsalavoutas (2016), in this issue. References Accounting Standard Board of Japan (ASBJ), European Financial Reporting Advisory Group (EFRAG) and Organismo Italiano di Contabilità (OIC), Should goodwill still not be amortised? Available from: %20and%20Amortisation/140725_Should_goodwill_still_not_be_amortised_Researc h_group_paper.pdf Amiraslani, H., Iatridis, G., and Pope, P.F. (2013). Accounting for asset impairment: a test for IFRS compliance accross Europe, Working Paper, Center for Financial Analysis and Reporting Research (CFARR), Cass Business School. 11

13 André, P., Filip, A., and Paugam, L. (2015). The effect of Mandatory IFRS Adoption on Conditional Conservatism in Europe. Journal of Business Finance & Accounting, 42(3-4), Ball, R., Kothari, S.P., and Robin, A. (2000). The Effect of International Institutional Factors on Properties of Accounting Earnings, Journal of Accounting & Economics, 29(1), Basu, S. (1997). The Conservatism Principle and the Asymmetric Timeliness of Earnings, Journal of Accounting & Economics, 24(1), Beatty, A., and Weber, J. (2006). Accounting Discretion in Fair Value Estimates: An Examination of SFAS 142 Goodwill Impairments. Journal of Accounting Research, 44(2), Beaver, W.H., and Ryan, S.G. (2005). Conditional and unconditional conservatism:concepts and modeling, Review of Accounting Studies, 10, Bens, D.A., Heltzer, W., and Segal, B. (2011). The Information Content of Goodwill Impairments and SFAS 142, Journal of Accounting, Auditing & Finance, 26(3), Boennen, S. and Glaum, M. (2014). Goodwill accounting: A review of the literature. Available at SSRN: Brief, R.P. (1969). An Econometric Analysis of Goodwill: Some Findings in a Search for Valuation Rules. Accounting Review, 44(1), D Arcy, A. and Tarca, A. (2016). Reviewing goodwill accounting research: What do we really know about IFRS 3 and IAS 36 implementation effects? Working paper, Vienna University of Economics and Business. Dechow, P., Ge, W. and Schrand, C. (2010). Understanding earnings quality: A review of the proxies, their determinants and their consequences, Journal of Accounting & Economics, 50, Ecker, F., Francis, J., Kim, I., Olsson, P.M. and Schipper, K. (2006). A returns-based representation of earnings quality, Accounting Review, 81, European Financial Reporting Advisory Group (EFRAG). (2016). What do we really know about goodwill impairment? A quantitative study. Available at European Security and Markets Authority (ESMA). (2013). European enforcers review of impairment of goodwill and other intangible assets in the IFRS financial statements. Available from Filip, A., Jeanjean, T., and Paugam, L. (2015). Using Real Activities to Avoid Goodwill Impairment Losses: Evidence and Effect on Future Performance. Journal of Business Finance & Accounting, 42(3-4), Francis, J., LaFond, R., Olsson, P.M. and Schipper, K. (2004). Costs of equity and earnings attributes, Accounting Review, 79, Glaum, M., Schmidt, P., Street, D. L., and Vogel, S. (2013). Compliance with IFRS 3 and IAS 36 required disclosures across 17 European countries: company-level and country-level determinants. Accounting & Business Research, 43(3), Hayn and P. J. Hughes (2006). Leading Indicators of Goodwill Impairment, Journal of Accounting, Auditing & Finance, 21(3), Healy, P.M. (1985). The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics, 7,

14 Kothari, S. P., Ramanna, K., and Skinner, D.J. (2010). Implications for GAAP from an Analysis of Positive Research in Accounting, Journal of Accounting & Economics, 50(2&3), Li, K.K. and Sloan, R.G., (2015). Has Goodwill Accounting Gone Bad? Working paper available at SSRN: Mazzi, F., André, P., Dionysiou, D., and Tsalavoutas, I. (2016) Compliance with goodwill related mandatory disclosure requirements and the cost of equity capital. Accounting & Business Research forthcoming. Available at SSRN: Mazzi, F., Liberatore, G., and Tsalavoutas, I. (2016). Insights on CFOs Perceptions about Impairment Testing Under IAS 36. Accounting in Europe, this issue. Murphy, K., and Zimmerman, J. (1993). Financial Performance Surrounding CEO Turnover. Journal of Accounting and Economics, 16(1-3), Pope, P.F., and Walker M. (1999). International Differences in the Timeliness, Conservatism, and Classification of Earnings, Journal of Accounting Research, 37(Suppl.), Ramanna, K., and Watts, R. L. (2012). Evidence on the use of unverifiable estimates in required goodwill impairment. Review of Accounting Studies, 17(4), Roychowdhury, S. and Martin, X. (2013). Understanding Discretion in Conservatism: An Alternative Viewpoint, Journal of Accounting & Economics, 56(2&3), Verriest, A., and Gaeremynck, A. (2009). What determines goodwill Impairment? Review of Business and Economics, 29(2), Schatt, A., Doukakis, L., Bessieux-Ollier, C., and Walliser, E. (2016). Do goodwill impairments by European firms provide useful information to investors? Accounting in Europe, this issue. Tsalavoutas, I., André, P., and Dionysiou, D., Worldwide application of IFRS 3, IAS 36 and IAS 38, related disclosures, and determinants of non-compliance. ACCA research report No Available from Watts, R.L. (2003). Conservatism in accounting part i: Explanations and implications, Accounting Horizons, 17, Wen, H.J., and Moehrle, S.R. (2015). Accounting for Goodwill: A Literature Review and Analysis. Available at SSRN: 13

15 14.0% 13.5% 13.0% 12.5% 12.0% 11.5% 11.0% 10.5% 10.0% 9.5% 9.0% Median level of goodwill (Europe) Pct of TA Median level of goodwill (US) Pct of TA Figure 1a. Median level of goodwill in percentage of total assets: Europe vs. US 35.0% 33.0% 31.0% 29.0% 27.0% 25.0% 23.0% Median level of goodwill (Europe) Pct of Equity Median level of goodwill (US) Pct of Equity Figure 1b. Median level of goodwill in percentage of common equity: Europe vs. US 14

16 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Europe: pct of obs. acc. impairment when econ. impairment US: pct of obs. acc. impairment when econ. impairment Figure 2a. Percentage of observation with an accounting impairment when also indications of economic impairment (Economic impairment defined as MV - Equity < GDWL) 45.0% 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% Europe: pct of obs. acc. impairment when econ. impairment US: pct of obs. acc. impairment when econ. impairment Figure 2b. Percentage of observation with an accounting impairment when also indications of economic impairment (Economic impairment defined as MTB < 1) 15

17 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Europe: pct of obs. acc. impairment when econ. impairment US: pct of obs. acc. impairment when econ. impairment Figure 2c. Percentage of observation with an accounting impairment when also indications of economic impairment (Economic impairment defined as EBITDA < 0) 16

18 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Europe: median level of impairment IMP/GDWLt-1 US: median level of impairment IMP/GDWLt-1 Figure 3 Median level of impairment when firms impair 17

19 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Europe: median level of impairment IMP/GDWLt-1 US: median level of impairment IMP/GDWLt-1 Figure 4a. Median level of impairment when also indications of economic impairment (Economic impairment defined as MV BV Equity < GDWL) 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Europe: median level of impairment IMP/GDWLt-1 US: median level of impairment IMP/GDWLt-1 Figure 4b. Median level of impairment when also indications of economic impairment (Economic impairment defined as MTB < 1) 18

20 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Europe: median level of impairment IMP/GDWLt-1 US: median level of impairment IMP/GDWLt-1 Figure 4c. Median level of impairment when also indications of economic impairment (Economic impairment defined as EBITDA < 1) 19

21 Table 1 Sample Observations Europe US # Pct # Pct # of firms in EIKON 8, % 13, % (-) Firms with unavailable SIC code -2, % -4, % (-) Financial institutions (SIC=6xxx) -1, % -1, % (=) # of firms 4, % 6, % (x10) # of observations , % 68, % (-) # of observations without IFRS/ US GAAP for two consecutive years -12, % -24, % (-) # of observations with unavailable data or negative MTB -3, % -15, % (=) # of firms year 27, % 28, % (-) # of observations with no lagged goodwill -8, % -12, % (=) Final sample 18, % 16, % MTB: Market-to-book 20

22 Table 2 Descriptive statistics ROA MTB GDWL/TA GDWL/EQ ASSETS Europe (N=18,538) Mean ,151 Std. Dev ,862 Min P Median P ,830 Max ,632 US (N= 16,525) Mean ,235 Std. Dev ,660 Min P Median P ,919 Max ,395 Difference in means Difference t value p value *** *** *** ROA: net income/total assets; MTB: market-to-book; GDWL/TA: goodwill/total assets; GDWL/EQ: goodwill/total equity; ASSETS: total assets (millions of euros) 21