How Do Firms Implement Impairment Tests of Goodwill?

In March 2004 the revised IAS 36 was approved by the IASB (2004a). According to it, firms must carry out an impairment test of goodwill on at least a yearly basis. If, as a result of this, the recoverable amount of goodwill is less than its carrying amount, the carrying amount shall be reduced to its recoverable amount (e.g., an impairment loss is recognized). IAS 36 defines the recoverable amount as the higher of an asset's ‘value in use’ and ‘fair value less costs to sell’. This is a radical change in accounting for goodwill. Previously, international accounting standards required recognition of goodwill subject to amortization over its useful lifetime.

We conduct a descriptive study which examines how Danish companies implement the revised IAS 36. We identify what firms do in cases where IAS 36¹ provides little or no specific guidance. The work focuses on two issues in IAS 36 that have received attention in the literature: (a) how firms define a cash-generating unit (CGU), and (b) how those firms measure the recoverable amount of a CGU. Technical aspects related to carrying out impairment tests of goodwill are examined: for instance, how firms calculate cost of capital and how terminal value is estimated in calculating the recoverable amount. However, we do not address if estimates seem fair.²

Based on the feedback from the survey, we further examine factors which may explain why some firms seem to be in better compliance with IAS 36 than other firms. This should provide useful help in how to improve compliance.

An empirical examination of how firms implement impairment tests of goodwill seems warranted. First, while prior research (see, e.g., Henning et al. 2000; Jennings et al. 2001; Bugeja and Gallery, 2006) examines the information content of goodwill and goodwill amortization (the correlation between accounting numbers for goodwill and market data [stock prices and stock returns]), we examine how firms implement impairment tests of goodwill. Anderson (2004) argues that the new and revised standards on business combinations and intangible assets have been well documented but there has been little discussion of how these requirements will be followed in practice.

Second, IAS 36 is a standard that involves substantial judgment. An example is the identification of a cash generating unit (IAS 36, 68). Further, while value in use is one of the measurement concepts in IAS 36, the standard is silent on how various value drivers should be estimated. For example, IAS 36 does not provide guidance about how terminal value should be measured. In addition, IAS 36 deviates from traditional valuation practice on certain issues. A prime example is para. 55 that requires discount rates to be pre-tax rates.³ In fact, this policy was questioned by those who responded to the exposure draft on IAS 36 (BC91). Petersen et al. (2006) provide evidence that discount rates in practice are estimated on an after-tax basis. This approach is also supported by a number of popular valuation textbooks such as Koller et al. (2005) and Damodaran (2002).

Finally, IAS 36 is a complicated standard which requires specific knowledge of valuation techniques. Producers of accounting information (firms) and their financial advisers must possess the necessary technical skills (e.g., be able to apply valuation models) and economic expertise (e.g., be able to make reasonable budget assumptions) to carry out impairment tests. These skills are needed, as auditors of a firm may not assist in such valuation work under strict ethical and independence rules (Anderson, 2004).

Many argue that an impairment test only approach seems a logical step in the development of accounting for goodwill. First, the underlying logic for removing the traditional amortization methodology is that the amortization on a straight-line basis over a number of years contains no information value for those using financial statement (Jennings et al., 2001). Moreover, IFRS 3 (IASB, 2004b) no longer requires that companies perform the almost impossible task of estimating the useful life of goodwill (Jansson et al., 2004). Second, the impairment approach should provide users of financial statements with better information, as goodwill is not automatically amortized (Colquitt and Wilson, 2002; Bens and Heltzer, 2005). Finally, goodwill impairment tests would be operational and capture a decline in the value of goodwill (Donnelly and Keys, 2002).

On the other hand, the new approach in accounting for goodwill may be questioned on several grounds. First, while amortizing of goodwill is considered to be arbitrary, it is easy to apply in comparison with the impairment approach. The major benefit of amortization, on a straight line basis, is that it is possible to predict its impacts on earnings with greater accuracy (Stevenson and McPhee, 2005). This is in line with Lachnit and Müller (2003), who argue that to achieve comparability among firms, adjustments for different accounting practices are necessary. Among the different kinds of treatment of goodwill, they regard amortization over a predetermined time horizon as the most useful practice. Only by amortizing on a systematic basis is it possible to determine ‘normalized’ or ‘permanent’ income as a measure of earnings power. Second, conducting a detailed test for impairment on every asset and associated goodwill from initial acquisition at the end of each reporting period may be time consuming and costly (McGreachin, 1997; Rockness et al., 2001). Third, the European Financial Reporting Advisory Group (EFRAG) in their comments to IFRS 3 expresses ‘major concerns’ over IFRS 3. EFRAG argues that the standard would introduce unreliable measurements of the recoverable amount of goodwill.

In light of the above, it is relevant to examine how Danish firms implement impairment tests of goodwill. To do so, questionnaires were sent to all Danish listed companies that recognize goodwill.⁴ Thereby, we obtain a unique dataset that provides in-depth information concerning how impairment tests of goodwill are carried out. The results indicate that practice varies considerably among firms. Further, some firms do not define a CGU and hence do not comply with IAS 36. We also find inconsistencies in the way firms estimate recoverable amounts. Areas of concern include calculating the pre-tax discount rate, adjusting for risk and estimating the cash flow in the terminal period. Finally, our multivariate analysis reveals that the inconsistencies detected in our study are less likely to occur in firms that systematize their impairment testing procedures and involve persons with considerable valuation experience.

Since international accounting standards are mandatory for all listed firms in the EU and have been adopted by listed firms in several countries—for example, Australia and New Zealand⁵ adopted the international accounting standards on 1 January 2005 and 1 July 2007, respectively—our results should be of interest to firms, auditors, standard setters and users of financial statements outside a Danish setting.⁶ Among other things, our findings can be used to improve current practice, especially in areas where firms seem to have difficulties in carrying out impairment tests in compliance with IAS 36.

ACCOUNTING FOR GOODWILL UNDER IFRS 3 AND IAS 36

IFRS 3, 36 requires that acquired assets, liabilities and contingent liabilities are recognized at fair value in the annual accounts of the acquiring firm and if they comply with recognition criteria, regardless of whether they have been recognized previously. It requires that, to the extent the purchase price exceeds the total fair value of the identifiable net assets, the difference is recognized as goodwill. It does not permit amortization of goodwill, but requires goodwill to be tested for impairment as set forth in IAS 36.

IAS 36 demands that each separate asset is tested for impairment, if there are any indications of impairment. Assets with an indeterminable lifetime, which are not amortized, must be tested for impairment at least on a yearly basis. Goodwill cannot be tested separately, as it represents resources that cannot be identified or quantified reliably. Goodwill is therefore allocated to the individual CGU or group of CGUs that benefit from the acquired goodwill (IAS 36, 80). If the recoverable amount of the CGU, to which goodwill is allocated, is lower than the carrying value of that CGU, the difference is recognized as an impairment loss in the income statement. If this difference is greater than the carrying amount of goodwill, the total goodwill amount is written off, while the remaining difference is allocated to (subtracted from) the remaining assets on a pro rata basis (IAS 36).⁷

The recoverable amount for an asset, CGU or group of CGUs is the higher of its fair value less costs to sell and its value in use. Applying value in use of a CGU is problematic, as it requires forecasting of future earnings and estimating an appropriate discount factor. Discounted cash flow calculations have previously been applied to financial assets and liabilities, where the expected future cash flows to a much larger extent are a good proxy for the future realized cash flows than will be the case for discounted cash flow calculations used to test for impairment.

RESEARCH DESIGN

A survey-based analysis is adopted to examine how Danish firms apply IAS 36 in practice. Given the descriptive nature of our study, surveys seem superior to other research methods (Yin, 1994). Surveys complement other research methods based on either large samples or case studies. As pointed out by Graham and Harvey (2001), large studies are the most common type of empirical analysis offering statistical power and cross-sectional variation. On the other hand, large sample studies have the weakness related to variable specification and the inability to ask qualitative questions. Case studies offer details and unique information regarding corporate behaviour; however such samples are small and results are often sample-specific. Surveys as adopted here offer a balance between large sample studies and case studies. We obtain a larger sample than in case studies, improving the validity. Further, we are able to ask detailed questions which improve the quality of the data. However, surveys (questionnaires) are not without potential problems. There are risks that the respondents are not representative of the population and that the questions are misunderstood. Each of these issues is addressed below.

Based on a careful review of IAS 36 and related literature, we developed a draft questionnaire that focuses on (a) identifying a CGU and (b) measuring its recoverable amount. The draft pilot questionnaire was circulated to auditors from leading auditing firms and CFOs employed at large Danish firms. The revised questionnaire incorporated their suggestions including a (few) new questions as well as a refinement of some of the existing questions. Further, we sought advice from colleagues who specialize in surveys and received valuable feedback on design issues and ideas as how to increase the response rate. As a final test the questionnaire was mailed to four firms, which led to additional modifications of the questionnaire.

Based on the feedback from the firms, five areas with apparent non-compliance with IAS 36 (labelled ‘inconsistencies’) were identified. In order to further explore those inconsistencies, we identified seven firm characteristics which may explain non-compliance. Multiple regression analyses were run to examine which of the seven factors explain non-compliance.

SAMPLE AND DESCRIPTIVE STATISTICS

To identify firms that carry goodwill on the balance sheet, data were collected from Account Data, a large database that records accounting data from Danish listed firms. Eighty out of 165 firms that were listed on the Copenhagen Stock Exchange as of 1 March 2006 recognized goodwill as an asset.⁸

Each of those 80 firms was contacted by phone in order to identify the person in charge of impairment testing.⁹ These personal contacts served two purposes. We wanted to make sure that the questionnaire was mailed to the right person (i.e., the person deeply involved in impairment testing), and that the personal contacts would enable us to make follow-up phone calls either to remind respondents to fill in the questionnaire or for clarification purposes. Four of the 80 firms did not wish to participate in the survey. An internet-based questionnaire was submitted to the remaining 76 firms. The link was forwarded directly to the person in charge of impairment tests. The respondents also received a letter outlining the purpose of the survey. After approximately three (six) weeks a reminder (second reminder) was e-mailed to the persons who had not yet filled in the questionnaire. Finally, after another couple of weeks, we called the firms that had still not completed the questionnaire and encouraged them to do so.

Eventually, 58 firms responded to each question in the questionnaire. An additional four firms submitted usable feedback that was sufficient to be included in the analysis, although a few questions were left unanswered, leaving a final sample of 62 observations. Launsø and Rieper (2000) argue that feedback from at least 70% is sufficient to provide a picture of the population. Our questionnaire was filled in by 72.5%, while additionally 5% of the respondents provided feedback on the major part of the questionnaire.¹⁰ A feedback rate of over 70% is quite unusual for questionnaires. For example, Graham and Harvey (2001) dispatched 4,440 questionnaires, of which just 392 were returned, an 8.8% response rate. Mukherjee et al. (2004) examined merger motives and obtained 75 usable responses: 11.8% of the 636 delivered responses.

In order to check the validity of the data (completed questionnaire) we received, respondents were contacted, and asked to elaborate on their data and provide any further information they might have. This additional check confirmed the general results as reported in the following sections.

CASH GENERATING UNITS

METHODS FOR ESTIMATING THE RECOVERABLE AMOUNT

Risk

In accordance with IAS 36, 55 and 56 and appendix A15, risk specific to the asset must be accounted for by either adjusting the cash flow or the discount rate. This is in line with recommendations in prevalent corporate finance literature (e.g., Brealey and Myers, 2003), who name the two methods ‘the certainty equivalent’ and the ‘risk-adjusted discount rate’. Theoretically, the two methods yield identical results and choosing between them shouldn't matter. Previous studies have shown that practitioners almost exclusively use the risk-adjusted discount rate method (Petersen et al., 2006). It is also the method that is best described in the popular valuation literature (e.g., Koller et al., 2005; Damodaran, 2002). In addition beta values, which are used to estimate the risk adjusted discount rate, are reported in financial databases like Bloomberg. Beta (β) is measured as

where Cov, r_ass, r_m and Var represent covariance, return on assets, market return and variance, respectively. The certainty equivalent method estimates the cash flow that makes an investor indifferent between a certain cash flow and an uncertain cash flow at a given point of time. The certainty equivalent method requires that the cash flows are adjusted for risk. This adjustment can be shown as CE(CF) = (Expected cash flow – b[u_m-r_f]). u_m is the expected rate of return on the market portfolio and b is measured as

In the following we examine how firms adjust for risk.

Eight respondents adjust for risk in the cash flows and 31 respondents adjust for risk in the discount rate (Table 9). IAS 36, 55–56 speaks of ‘adjustments to the cash flows’ without specifying how these adjustments should be made. Since the standard is silent as how to risk adjust cash flows, we want to examine current practice. The results are reported in Table 10.

Where do you adjust for the risk, when you apply the DCF-model?	N
The cash flows (nominator)	8
Discount rate (denominator)	31
Don't know / do not wish to answer	16
Total	55

How do you adjust cash flows for risk, when you apply the DCF-model?	N
By making a conservative estimate of the free cash flows	4
By weighting the likelihood of possible future cash flows	1
They expected free cash flows are reduced by multiplying them with a factor calculated via the risk adjusted required rate of return (the factor can for explicit budget years be estimated as follows:((1+rf)/(1+r))n)	0
They expected free cash flows reduced with a fixed percentage	1
Don't know / do not wish to answer	2
Other (please specify in the textbox below)	0
Total	8

Four respondents adjust for risk by measuring cash flows conservatively. While this procedure adjusts for risk, it will be unbiased only by chance. Furthermore, in practice it is unclear what ‘measuring conservatively’ means. What some might perceive as conservative, others might perceive as the most likely outcome. One respondent adjusts for the risk by probability weighting future cash flows. This corresponds to what IAS 36, appendix A defines as expected cash flow. This method simply weights the different likely outcomes (cash flows) and adds them. The method is proposed as an alternative to estimation of the expected cash flows rather than the most likely cash flow. A probability weighting of the different cash flows, therefore, is not the correct way to adjust for risk. One respondent reduces the expected cash flows with a fixed percentage. There is no way to tell how this percentage has been calculated, but if the same fixed percentage is used across business units that carry different risks, the adjustment for the risk is not correct. Two respondents do not inform how they adjust for risk in the cash flows. At least six firms do not properly adjust for risk.

IAS 36, 56 requires that the discount rate reflects the risk of the assets; that is, the risk of each CGU. If the discount rate is adjusted for risk, cash flows should be discounted with the weighted average cost of capital (WACC) or a variant thereof (e.g., implicit discount factor for comparable firms) that reflects the risk of each CGU.²⁰

As reported in Table 11 approximately half (13) of those who adjust for risk in the denominator use WACC, which is correct if the WACC reflects the risk to the specific CGU. Of the fifteen respondents who use WACC, twelve use different WACCs across CGUs (not reported). They argue that this is due to the fact that business activities carry different risks, for example, due to CGUs that are situated in different parts of the world (country risk) and differences in the financial leverage in the individual CGUs. Two respondents, who argue that risk is adjusted in the discount factor, use the risk free interest rate. Clearly, the risk free rate does not take risk into account. Of seven respondents that use the equity holders' required rate of return, three are financial institutions. The remaining four use interest bearing debt as well as equity in their capital structure and should use WACC or a variant hereof as the discount rate. The two respondents who respond ‘other’ use a ‘combination of WACC and average rate of return on excess cash’ and an ‘estimated discount factor’. It is difficult to evaluate an ‘estimated discount factor’, while a ‘combination of WACC and average rate of return on excess cash’ is incorrect (one respondent).

Your responses to the previous questions have shown that in applying the DCF model you discount after-tax cash flows with an after-tax discount rate. Which of the following after-tax discount rates do you apply?	N
Implicit discount factor for comparable firms	0
WACC	13
The risk free rate	2
The average borrowing rate	0
The marginal borrowing rate	0
Equity holders required rate of return	6
Don't know / do not wish to answer	1
Other (please specify in the textbox below)	2
Total	24

In conclusion, of the 27 who relate how they adjust for risk, at least six (seven) firms adjust for risk in the cash flow (choice of discount rate) in a way not supported by theory, and are thus considered not to be in compliance with IAS 36. Considering how well this area is portrayed in the finance literature, these results are surprising.

WHAT EXPLAINS INCONSISTENCIES IN IMPAIRMENT TESTS?

The descriptive analysis reveals that some of the methods used when defining a CGU are not in compliance with IAS 36. We also find inconsistencies in the way that firms estimate the recoverable amount. These findings naturally beg the question of what explains these inconsistencies. We therefore examine firm characteristics that may account for these discrepancies. Such an analysis may help firms to improve their impairment testing, assist auditors to focus on areas that need more attention and enable standard setters to consider how to improve the standard.

Based on previous research and certain relevant firm characteristics, a number of factors are identified which we believe explain inconsistencies in carrying out impairment tests in practice. Below, we list five factors which could affect how impairment tests are applied:²¹

A multiple regression analysis is applied to test to what extent the total number of inconsistencies (that firms commit) might be explained by the five firm characteristics just described. This leaves the following model:

()

For all five firm characteristics the coefficient (β₁–β₅) is expected to be negative. ‘Inconsistencies’, which is our response variable in equation (1), is measured as the number of inconsistencies divided by total responses (maximum five responses).²² If for example one out of four responses from a firm is categorized as an inconsistency that firm receives a score of 1/4.

Our analysis of impairment practice among Danish firms reveals five areas of inconsistency:

In Table 13 we summarize these inconsistencies.

It is important to stress that the reported statistics in Table 13 are affected by missing observations of the explanatory variables (firm characteristics). This implies that fewer inconsistencies are included in the analysis than are found in the descriptive study. For example, seventeen firms do not estimate the pre-tax discount rate according to theory (and IAS 36), but due to missing observations of certain explanatory variables only fourteen of those firms are included. While some of the inconsistencies only contain a limited number of observations or have a limited variability, Table 13 also shows that our response variable in equation (1), inconsistencies, contains 49 firm observations. Eighty per cent (39 firms) of these firms commit at least one inconsistency.

Table 14 provides descriptive statistics on inconsistencies and the firm characteristics (explanatory variables).

The descriptive statistics show that 50% (median) of the sample firms obtain an inconsistency score of 0.5. This indicates that half the responses for those firms are not in compliance with IAS 36. As is apparent in Table 1, size is not normally distributed. We therefore use the natural logarithm of turnover. The descriptive statistics on size reported in Table 14 show that the transformed size variable is normally distributed. On average, goodwill makes up 6% of total assets. In the most extreme case goodwill makes up 31% of total assets. Thus, for most firms the magnitude of goodwill seems modest. This supports the view that these firms may not offer as much attention to impairment tests, as the effect on earnings on possible impairment losses may be negligible. Most firms apply the same valuation (impairment) model across CGUs. Approximately 50% of the persons involved with impairment testing have substantial experience with valuing firms in general. Finally, a manual for impairment testing has been developed by 50% of firms.

In Table 15 we report preliminary evidence on equation (1).

As expected, inconsistencies are negatively correlated with all five firm characteristics. In fact the correlations between inconsistencies and ‘other experience with valuation’ and ‘manual’ are negative and significant at the 1% level. Thus, ‘other experience with valuation’ of firms has a positive effect on the number of correctly implemented impairment tests. Furthermore, firms which use manuals for impairment testing purposes experience fewer inconsistencies than firms which refrain from making manuals.

There is a tendency that large firms make an impairment test manual (significant at the 3% level). This is also evident for firms where the magnitude of goodwill is substantial (significant at the 5% level). Firms which use the same valuation model across CGUs also produce manuals more frequently (significant at the 1% level). These results indicate that large firms, firms with a substantial amount of goodwill and firms that use the same valuation model across CGUs are more systematic in carrying out impairment tests. The correlation matrix reported in Table 15 suggests that multicollinearity is not an issue in the subsequent multiple regression analysis.²³

Table 16 reports the results from the multiple regression analysis. In line with the correlation coefficients reported in Table 15, the coefficients on ‘manual’ and ‘other experience with valuation’ are negative and significant at the 1% level. Thus, after controlling for the other variables both ‘manual’ and ‘other experience with valuation’ remain significant. None of the other firm characteristics explains variations in inconsistencies.

In order to examine the robustness of the reported results, additional tests are conducted. First, we replace (log) revenue with (log) total assets as a proxy for size. The reported results remain robust if this alternative measure of size is applied. Also examined is the sensitivity of the coding of our response variable (inconsistencies). For example, one may argue that firms which use multiples when estimating the terminal value violate the internal coherence of the discounted cash flow model. Using these alternative specifications for inconsistencies (not reported) did not affect the reported results.

Various test diagnostics (not reported) further reveal that the standard assumptions in linear regression analysis are not violated. As mentioned above, multicollinearity is not an issue. Likewise, diagrams for the standardized residuals support the assumption that the residuals are normally distributed. These further analyses support the conclusions in our research.

CONCLUSIONS AND PERSPECTIVES

We examine how firms define a CGU and how those firms measure the recoverable amount of a CGU adopting a survey approach. A risk inherent in surveys is that the respondents are not representative of the population and that the questions may be misunderstood. We put considerable effort into reducing those risks. As a result we obtained a high response rate (more than 70%) and received feedback from respondents who are deeply involved in carrying out impairment tests. Our survey generally supports the proposition that a common practice has not yet been established. Based on our analysis it is difficult to say whether this simply reflects that firms adopt an approach suited to their organizational and economic structures or that firms are uncertain as to how to apply the standard. We also find that some firms do not define a CGU in compliance with IAS 36. Twenty firms do not include corporate assets in a CGU for impairment testing purposes, which is a violation of IAS 36, 102. Further, we find inconsistencies in the way that firms estimate the recoverable amount. None of the firms used the iteration method to transform an after-tax discount rate to a pre-tax discount rate. Firms also experience difficulties in risk adjusting cash flows and discount rates. The estimation of the free cash flow in the terminal period is another area of concern.

A multivariate analysis reveals that the inconsistencies detected in our study are less likely when the impairment tests are carried out by persons with considerable valuation experience and when manuals are prepared. This indicates that firms which systematize their impairment testing encounter fewer inconsistencies.

Since IAS 36 was adopted in March 2004 the survey responses are obtained from respondents that (in many cases) have limited experience with impairment tests of goodwill. It may therefore be premature to draw any strong conclusions based on our results. However, firms can use our findings to improve the way that they carry out impairment tests. Auditors can use our results to focus on areas that deserve further attention. Finally, standard setters are better able to evaluate the current standard and consider how improve it. For example, a follow through example that clearly explains how the value of a CGU is estimated based on a pre-tax/after-tax basis and how to adjust for risk in the cash flow and discount rate, respectively, would be useful.

This study permits the following extensions. Research could be extended to other accounting items with a similar complexity. Provisions, stock options and financial instruments are some of the accounting items which deserve to be further researched. It also seems appropriate to examine how users of financial statements read, interpret and use accounting information, including challenging accounting items such as write-offs of goodwill. This is also supported by the fact that new and more complex accounting standards are emerging. A recent research project in Norway (Kinserdal, 2006) documented that financial analysts do not believe that pension liabilities in Norwegian firms are valued differently (based on different assumptions). Such results may question, whether financial analysts use all value-relevant accounting information in valuation of firms.