Market value impact of goodwill
impairment: latest evidence from the UK
Master Thesis
Master in International Finance, Amsterdam Business School Prepared by: Jaemi Lou Samonte
Supervisor: drs. J.F. (Dennis) Jullens Date: August 2015
Abstract
This paper extends prior research by examining the potential market value impact of goodwill impairment announcements, utilizing event study methodology applied in the context of the UK market and under the current impairment-only regime which replaced the previous standards of periodic amortization and/or optional impairment.
Overall results for the full sample reveal positive but low cumulative abnormal returns (CAR) within the short event window. For the subsamples, it was found that impairment announcements made alongside positive operating earnings disclosures are linked to positive short window CARs, in contrast to simultaneous negative earnings or other disclosures that are associated with negative CARs. Moreover, more adverse share price reaction around the event date was found for firms that incurred greater goodwill impairment in relation to total assets (higher goodwill impairment materiality) and those that have lower analyst coverage (higher information asymmetry). Lastly, positive CARs were observed in the one-year period following announcement date, which are more pronounced for firms announcing simultaneous negative earnings, those that incurred more material goodwill impairment charge and those belonging to the high information asymmetry group.
Although the study did not result in strong evidence regarding the impairment-returns relation on the basis of the full sample, it does support previous research that found links between the share price impact of goodwill impairment and simultaneous announcements (in particular of operating earnings), the materiality of goodwill impairment sustained and the information environment that a firm is in. Moreover, it provides additional evidence indicating recovery of share prices post-announcement, contrary to the notion of market underreaction to goodwill impairment announcements, and opens a new dimension to such debate given the difference in magnitude of recovery observed within the various subsamples examined.
Table of Contents
1. Rationale and aim of research ... 1
1.1 Main research question ... 1
1.2 Choice of methodology ... 3
1.3 Choice of sample – geography and timeframe ... 4
1.4 Impact of firm-specific attributes... 6
1.5 Post-announcement effects ... 6
2. Regulatory framework ... 7
3. Literature review... 8
4. Methodology and hypothesis ... 12
5. Data ... 15
6. Results ... 18
7. Robustness ... 23
8. Conclusion ... 25
9. References ... 28
1
1. Rationale and aim of research
1.1 Main research question
Under the current set of International Financial Reporting Standards (IFRS), a firm is obliged to record goodwill in its financial statements whenever it acquires another firm for a price that is greater than the fair value of the acquiree’s assets net of liabilities. Subsequent to acquisition, such recorded goodwill must be subject to an impairment test at least annually. If there is evidence that the recoverable amount of the business segment (known as “cash-generating unit”) is less than its carrying value or the amount by which its net assets are reported in the balance sheet, goodwill has to be fully or partially written off. This entails a reduction in goodwill balance on the asset side of the balance sheet as well as an above-the-line expense item in the income statement that negatively impacts income from continuing operations and ultimately net profits.
According to Duff & Phelps’ 2014 European Goodwill Impairment Study, about 87% of companies in Europe (as proxied by the Stoxx® Europe 600 Index) has reported goodwill in their balance sheets, while each year 23%-35% of such firms has booked a goodwill impairment charge during the sample period 2010-2013. This is in line with the findings of the European Securities and Markets Authority (ESMA) in its study entitled “Review on the application of accounting requirements for business combinations in IFRS financial statements” (2014) which reports that about 86% of the firms reviewed has recognized goodwill in their financial statements, with goodwill representing an average of 54% of the acquisition price. Clearly, goodwill and goodwill impairment significantly impact firms’ financial statements, but should investors also be concerned about their potential influence on firms’ equity value and share price especially when goodwill is subsequently reviewed and found to be impaired?
In theory, share prices should react to new information that reaches the public given a semi-strong efficient equity market. The key question is whether goodwill impairment charges as disclosed in financial statements or announced by management indeed represent newly-available information. On one hand, it can be thought that unfavorable business and economic conditions
2
which underlie goodwill impairment are already known to and assimilated by investors prior to actual recording of any impairment charge, thus there should be limited value impact upon goodwill impairment announcement. For instance, in a report released by KPMG entitled “Who cares about goodwill impairment? A collection of stakeholder views” (2014), majority of the interviewees believe that goodwill impairment has a “confirming rather than predicting value” since value is assessed based on cash flows, multiples or residual income and that goodwill impairment charges generally lag behind actual economic developments. Moreover, goodwill impairment testing and measurement is thought to be a highly complex exercise involving a great deal of subjectivity, which limits its reliability and usefulness in assessing value. On the other hand, goodwill impairment disclosures may be deemed as conveying new, valuable information about a firm and thus should influence its stock price upon announcement. As Hirschey and Richardson (2003) puts it, goodwill impairment as with other types of asset write-offs indeed do not coincide with changes in assets or cash flows. However, the informational value of goodwill impairment lies in the fact that it provides a signal about changes in a firm’s future profit-generating potential which cause investors to revise their expectations about the firm and thus their assessment of value. In other words, goodwill impairment may “foreshadow more bad news” in that it can serve as an indication that financial troubles, which may have already been reflected in historical profits and cash flows, might continue on in the future. Similarly, Li et al. (2010) explain that goodwill impairment testing entails a comparison between current forward-looking assessments of profit-making potential versus previous forecasts, and any impairment represents disclosure of a change in management’s forecast of future earnings from the relevant assets. Such updated assessments and forecasts are otherwise not previously available to the market. The main aim of this study is thus to seek evidence in support of one of these
alternative schools of thought – that is, to investigate whether or not goodwill impairment indeed has an impact on firm’s equity value and share price.
Aside from informing the fundamental debate above, this paper specifically aims to contribute to existing literature by examining a key economy and M&A market, the United Kingdom, with the latest data (2010-2013) unaffected by major crisis periods and occurring under the current impairment-only accounting standard, using an event study framework which seeks to measure the causal link, if any, between goodwill impairments and any share price impact. In addition, firm and announcement-specific attributes were considered to provide richer insight into the
3
drivers of the impairment-returns relation, as well as long-window returns to assess any possible post-announcement effects. The following paragraphs provide further explanation of these dimensions.
1.2 Choice of methodology
There have been a few studies that investigated this topic by utilizing various statistical/econometric methods in the context of different geographies and time periods. The most common methodologies used are accounting-based market valuation models and event study models. The former group follows the widely-used Ohlson (1995) model wherein market value of equity is regressed against various financial statement items such as book value of equity, earnings and other specific items like goodwill impairment in order to test the significance of these variables on market value. In contrast, the latter framework looks at possible share price impact of specific announcements, such as goodwill impairment, by estimating an expected “normal” return from historical data and comparing it with actual return around the announcement date, in line with the seminal work of Ball and Brown (1968) and Fama et al. (1969) who examined stock price effects of earnings announcements and stock splits, respectively. Comparing the two approaches, the key advantage of the event study methodology is that it allows a more direct testing of causality than market valuation models. Kothari and Warner (2006) highlight that the usefulness of event studies lie in the fact that they enable measurement of the impact of certain events on the wealth of firm’s shareholders. In a paper by Morck and Yeung (2011), the authors remarked that “Event studies are perhaps the most direct test for causality available to economists”. Furthermore, model misspecification is a potential risk in utilizing a market valuation-based model. Klimczak (2009) points out that the advantages of market valuation models are their relative flexibility in incorporating different variables as well as ease of implementation as compared to event studies, which may be the reason for their popularity. It is also interesting to note that, in the study by Li et al. (2010), the only existing study to the author’s knowledge which examined the value relevance of goodwill impairment using both methods, evidence supporting value relevance of goodwill was found using both approaches but was relatively weaker for the market valuation model than the event study model. Lastly, event studies enable testing of share price reaction pre- and post-announcement dates which is useful for assessing any partial anticipation or underreaction among market players. In
4
light of the above, this study was conducted using an event study framework to investigate the research question at hand.
1.3 Choice of sample – geography and timeframe
Most of the existing literature on this topic have focused on the US and UK markets with sample periods in the late 1990s to early 2000s during which goodwill impairment accounting was still either discretionary or in its early adoption or transitional phase. For this study, the UK market
was revisited, however with latest data spanning the years 2010-2013 which falls under the post-IFRS adoption era in Europe wherein goodwill impairment accounting has become mandatory for public firms.
The UK remains an interesting subject due to its position as a key economy and financial market, with the London Stock Exchange ranking 4th in the world and 1st in Europe in terms of market capitalizationi. In terms of M&A activity, the UK has been consistently placed 1st or 2nd in terms of value and deal volume in Europeii. In line with this, aggregate goodwill impairment charges among UK firms has also been highest or second highest in the region in terms of euros and number of events (Duff & Phelps (2014)). Lastly, practicality considerations such as availability of information, language considerations and size and manageability of data within the given time frame are additional motivations for choosing the UK market as the subject of this study.
The period 2010 to 2013 is an interesting timeframe to investigate the association of goodwill impairment with share value in the UK for three key reasons. First and foremost, it represents the latest data which has not been covered by any existing study. Secondly, this period represents a relatively stable period for financial markets, even with some slowdown in 2011 due to the European debt crisis which was less severe than the market crashes in early 2000s and 2008. This is important since inclusion of crisis periods tend to distort expected or “normal” return measurements especially with relatively short estimation periods (e.g. one year) for event studies (See Chen (2014) and Jovanovic and Fox (2010)). For instance, Laghi et al. (2013) found that the value relevance of goodwill impairment increases during crisis periods such as in 2008 pointing to possible increased attention among investors. Furthermore, since goodwill impairment measurement involves estimating fair value or value in use (discounted expected future cash flows), for which general macroeconomic conditions was cited as the most important
5
driver, periods of economic slowdown and uncertainty generally lead to more impairment being recorded by firms (Duff & Phelps (2014)). Li et al. (2010) who had a sample period spanning 1997-2002 highlighted the clustering of impairment charges around 2002 following the “dotcom crash” and the resulting higher significance of results for that year and warned about the possible consequences in interpreting their results. As shown in the latest Duff and Phelps study, aggregate goodwill balance (additions net of impairments) among the sample European firms has been relatively stable from 2010 to 2013. Thus, examining such period make for a more homogenous setting for studying the association between goodwill impairment and share prices and should avoid the challenges as described above.
Lastly, the period 2010 to 2013 comes years after the adoption of IFRS in Europe in 2005 which made goodwill impairment accounting mandatory for publicly-listed companies. A few studies have looked into the impact on value relevance of the adoption to the current impairment-only rule versus earlier standards of amortization and/or discretionary impairment but the results are not unanimous. On one hand, discretionary goodwill impairment may render the information delivered more novel and hence enhance its information value and market impact, but may also allow more opportunities for using this accounting disclosure for earnings manipulation thus affecting the credibility of the information. For instance, Francis et al. (1996), who investigated this topic during the discretionary impairment era, wrote “There exists a notion that management takes advantage of the discretion afforded by accounting rules to manipulate earnings either by not recognizing impairment when it has occurred, or by recognizing it only when it is advantageous (to them) to do so.” But the same reasoning may also be applied to the impairment-only rule since management discretion is still involved, for instance in determining business segments (cash-generating units), allocating of goodwill among them, estimation of fair values and cash flows and so on. In addition, mandatory goodwill impairment could also lead to unreliable disclosures given that there are firms with insufficient capacity to comply with the mandatory testing and measurement requirements, or it could lead to more frequent impairment charges which may lessen the value impact of each charge. Unlike several prior studies, this paper is not intended to be a comparative study of pre- and post-adoption eras; however, it still aims to indirectly contribute to the discussion by examining the latest available data several years after the impairment-only standard has been put in place.
6
1.4 Impact of firm-specific attributes
This study has also looked into certain firm- and announcement-specific attributes which may provide richer insight into the relation between goodwill impairment and share price performance, in particular the presence of simultaneous announcements, goodwill materiality and information asymmetry. First, following the work of Hirschey and Richardson
(2003), the impact of other disclosures made at the same time as the announcement of goodwill impairment was investigated. The authors found that simultaneous disclosures, particularly of positive operating earnings, negative operating earnings and other miscellaneous information, can either have a diluting or amplifying influence on the relation between goodwill impairment and market value. Second, in line with Li et al. (2010), materiality of goodwill as measured by goodwill impairment amount over total assets was considered. It is logical to expect that the more material goodwill impairment is for a certain firm, the greater is its share price impact, although there are some disparities in the existing literature as regards this matter. Lastly, information asymmetry as quantified by number of analysts following the firm was also examined, similar to the work of Bens et al. (2007). The thinking behind this is that the more analysts following a certain firm, the better the information environment meaning that analysts may be able to anticipate any goodwill impairment by monitoring firm fundamentals and management actions, thus potentially reducing the value impact of actual impairment announcements.
1.5 Post-announcement effects
Finally, this study also investigated the possible underreaction of investors to goodwill impairment announcements. This was done by examining cumulative abnormal returns during
the post-announcement period using the event study framework. Hirschey and Richardson (2003) found significant additional negative abnormal returns one year after goodwill impairment announcement, the size of which is linked to the magnitude of the share price impact around announcement date suggesting possible underreaction in the market, whereas Li et al. (2010) found positive abnormal returns post-announcement. This is an interesting question to address as it could provide evidence on how investors assimilate impairment information and whether or not they should be concerned about further share price deterioration in the period following the announcement.
7
2. Regulatory framework
Accounting for business combinations and purchased goodwill has been subject to much debate among market participants and standard-setters, as reflected in varying accounting standards implemented within different geographies through the years. This situation has largely changed with the adoption of IFRS in Europe in 2005, which harmonized business combination and goodwill accounting in the United States, Europe and other jurisdictions implementing IFRS. From that point until present, the common acceptable accounting treatment for goodwill has been balance sheet recognition upon acquisition followed by periodic impairment review and consequent write-off if necessary.
The US was the first to formally implement the current standard with the adoption of Statement of Financial Accounting Standards (SFAS) 141 (Business Combinations) and SFAS 142 (Goodwill and Other Intangible Assets) in 2001. Such standards require recognition of goodwill in the balance sheet (under the “purchase” method) and periodic goodwill impairment testing based on a fair value or discounted cash flow approach, with recognition of goodwill impairment loss in the income statement if necessary without possibility of reversal in future periods. Prior to that, both “pooling” and “purchase” methods were allowed under Accounting Principles Board (APB) Opinion 16, under which goodwill was not recognized at all in the financial statements using the pooling method. In case the purchase method is implemented, goodwill was treated as a wasting asset with finite life and therefore subject to annual amortization up to a maximum of 40 years (following APB Opinion 17).iii
Within the European Union, local generally accepted accounting principles (GAAP) in each country were followed prior to the mandatory adoption of IFRS among its members in 2005. In the United Kingdom (see AbuGhazaleh et al. (2013), the 1980s saw a vast divergence of accounting treatment of goodwill in practice ranging from permanent recording of goodwill in the books (no amortization or impairment), periodic amortization, to immediate write-off. In 1984, Statement of Standard Accounting Practice (SSAP) 22 was implemented which prescribed the immediate write-off of goodwill against reserves but otherwise also allowed it to be capitalized and subsequently amortized. Due to criticisms of such approach, Financial Reporting Standards (FRS) 10 and 11 were adopted which forbid immediate write-off and instead required capitalization followed by amortization for up to 20 years, annual impairment reviews or both,
8
with the first approach being widely preferred by firms in practice. The adoption of IFRS in Europe in 2005 lead to the abolishment of such alternative accounting treatments for goodwill in place of the current impairment-only approach.
With the implementation of IFRS 3 (Business Combinations) and the revised International Accounting Standards or IAS 36 (Impairment of Assets) and IAS 38 (Intangible Assets) by the International Accounting Standards Board (IASB) in 2004, which were adopted in Europe in 2005, the previous pooling method for business combinations and annual amortization of goodwill were prohibited and replaced with recognition of goodwill in the balance sheet and impairment testing at least annually and whenever there is indication of impairment. This impairment-only concept is fundamentally the same as the current US standards under SFAS 141 and 142 except for some differences. In essence, goodwill is regarded as impaired if the “recoverable amount” of a specific business segment (referred to as “cash-generating unit”) is less than its carrying amount, which is the amount by which it is reported in the financial statements. The recoverable amount is either the fair value less disposal costs (a market-based measure) or value in use (an internal measure based on expected cash flows), whichever is higher. This study provides a good opportunity to examine the relation of goodwill impairment with market values under such current accounting regime, specifically by looking into the UK market during the most recent period 2010 to 2013.
3. Literature review
Most of the existing literature that examined the association between goodwill impairment and market value have focused on the US and UK markets with sample periods in the late 1990s to early 2000s during which goodwill impairment accounting was still either discretionary or in its early adoption or transitional phase. This study closely follows the work of Hirschey and Richardson (2003) and Li et al. (2010) which covered the US and UK markets, respectively, during earlier time periods and under previous accounting regimes.
One of the earliest studies on this topic was by Francis et al. (1996) which looked at the US market during the period 1989-1992. During such time when asset write-offs were yet to be
9
subject to clear authoritative standards and were therefore discretionary (except for inventory), the authors found no significant market reaction to announcements of goodwill impairment. Using an event study framework and US data from 1992-1996, Hirschey and Richardson (2003) found a significant negative share price effect of goodwill impairment charges around announcement date (Days -1 to 0). They observed a similar adverse impact both within the one-year periods prior to and after announcement which the authors attributed to possible partial anticipation of and underreaction to goodwill impairment announcements, respectively. Furthermore, they found that the size of the negative valuation effects in the post-announcement period is linked to the size of the share price impact around announcement date (short window), supporting their underreaction conclusion. Li et al. (2010) on the other hand, utilized both market value and event study models for UK data in 1997-2002, and found evidence supporting the negative association between goodwill impairment and the market value of firms, albeit weaker using the first model. Also, the authors observed positive abnormal returns in the pre-announcement period as well as recovery of share price post-pre-announcement date, in contrast to the findings of Hirschey and Richardson (2003), although the event period used was much shorter at 65 days and the whole sample period was said to be generally characterized by rising cumulative abnormal returns. In terms of approach, Hirschey and Richardson (2003) segregated the sample observations according to “messiness” of announcement (i.e. if they are purely goodwill impairment announcements or accompanied by other announcements such as operating earnings or other disclosures) and performed testing on each resulting subsample. In contrast, Li et al. (2010) only selected observations which represent a single goodwill impairment during the sample period which were not confounded by other announcements.
Several studies have also examined the impact of specific attributes related to the firm or announcement on the goodwill impairment to market value relation. Please refer to Table 1 for the summary of key findings. As mentioned, Hirschey and Richardson (2003) considered the type of announcement according to “messiness” or the existence of other simultaneous
disclosures (such as operating earnings). Whereas the authors observed an overall significant
adverse share price impact upon goodwill impairment announcement using the full sample, they found that those accompanied by positive earnings disclosures resulted in positive but immaterial cumulative abnormal returns (CARs), while those that were announced together with negative
10
earnings and other miscellaneous disclosures (such as corporate restructuring, asset sales etc.) exhibited negative and statistically significant CARs, concluding that the market interprets company announcements in the context of other company information. Related to this, Li et al. (2010) also considered the operating earnings backdrop of each firm upon goodwill impairment announcement (whether they reported positive or negative EBITDA in the year that the goodwill impairment announcement was made) and found that those with negative EBITDA experienced a sharper drop in share price upon impairment announcement compared to positive EBITDA firms. In addition, the authors found that firms belonging to the highest quartile in terms of
materiality of goodwill impairment (as measured by goodwill impairment over total assets) and
those that did not release any prior information on the possible goodwill impairment, sustained greater share price declines upon announcement compared to the lowest quartile based on impairment materiality and those who did not release any prior information, respectively. Related to the materiality of goodwill impairment, the size of impairment was considered in several studies (although not conceptually identical since the former is a relative measure and the latter is absolute). The event studies conducted by Z. Li et al. (2011) and Escaffre & Sefsaf (2010) found evidence that the size or magnitude of the goodwill impairment is related to the size of the share price reaction, whereas Francis et al. (1996) and Hirschey and Richardson (2003) found that size is not related, rather it is the fact that goodwill impairment was announced that appear to influence stock price. Laghi et al. (2013) also looked at goodwill impairment materiality but found results in the opposite direction as goodwill impairment was found to be more value-relevant for firms with smaller goodwill impairment over goodwill balance ratio, remarking that investors may exert more caution when low impairment losses are recognized. Another paper which investigated the impact of several firm-specific attributes is by Bens et al. (2007). Using cross-sectional tests on US data over the period 1996-2003, the authors found that higher information asymmetry (less analyst coverage) and greater ability to comply with new
standards (larger firms) are associated with higher value impact of goodwill impairment
announcements, whereas firm complexity (greater number of business segments) is not. The second attribute was linked by the authors to the credibility of the announcement, explaining that goodwill impairment announcements made by smaller firms may be deemed less informative by the market resulting in attenuated market reaction since these firms tend to be less able to effectively implement mandatory impairment requirements. Other credibility-related attributes
11
were also looked into in the study by Knauer and Wöhrmann (2012). They found evidence that the lower the level of legal protection in a jurisdiction, which the authors linked to greater degree of management discretion in financial reporting, and the less verifiable explanation provided by
management regarding the impairment charge, the more negative is the reaction to goodwill
impairment announcements. The authors explained that in those cases, “investors take into account that the true goodwill impairment might be higher than the announced write-off and therefore react more negatively”, thus linking lower credibility with greater share price impact, in sharp contrast with Bens et al. (2007) who linked lower credibility (smaller firms with poorer ability to comply with standards) with smaller share price impact.
A group of related papers from the US that examined goodwill numbers reported under SFAS 142 and its predecessor SFAS 121iv, in particular those by Z. Li et al. (2011), Bens et al. (2007) and Chen et al. (2004), also found a significant and negative relation between goodwill impairment and stock price. Comparing the two standards, which differ primarily in that the trigger for goodwill impairment is based on fair value or discounted cash flows in the former versus undiscounted cash flows in the latter, the authors found evidence that the value relevance of goodwill impairment has actually decreased following adoption of the new standard. Z. Li et al. (2011) explained that their findings might be rooted to the more regular recording and smaller magnitude of impairment losses (leading to a lower “news value” of impairments) under the new regime, while Bens et al. (2007) pointed to a possible reduction in the reliability of information due to the greater difficulty and level of management discretion involved in implementing the current fair value-based standard. Chen et al. (2004), on the other hand, found improved valuation effects of goodwill impairment under the new standard. Related to this, Alfonso et al. (2011) compared goodwill impairment accounting in four European countries during the periods 2001-2004 and 2005-2010, representing periods before and after adoption of IFRS in Europe, respectively, and found evidence indicating decreased value relevance under post-IFRS adoption era, also linking this result to the fair-value nature of the new standards governing goodwill accounting.
Aside from the US and UK-based studies mentioned, Lapointe-Antunes et al. (2009) and Escaffre & Sefsaf (2010) have investigated datasets for Canada (year 2002) and France (2007-2008), respectively, and also found evidence of the negative value impact of goodwill
12
impairment. Broader samples encompassing the US and several European countries were examined by Knauer and Wöhrmann (2012) for years 2005-2009 and Laghi et al. (2013) for 2008-2011, although evidence supporting value relevance of goodwill impairment was found for the first study but not consistently among the geographical and time-based subsamples examined by the latter.
AbuGhalazeh et al. (2012) looked at the UK market using an accounting-based market value model, but limited the dataset to the top 500 firms in terms of market capitalization and to only two years 2005-2006. They found evidence in support of the value relevance of goodwill impairment during the early years of adoption of IFRS 3, but concluded that more years of study would be useful in testing the long-term effects of this current standard.
Lastly, to note that out of the twelve related studies as described above, half utilized either a classical or modified event study framework while the other half employed market value and other types of models.
In summary, existing literature is currently dominated by evidence supporting the negative association between goodwill impairment charges and market value of firms. However, there is some degree of variation in the findings when considering certain firm and announcement-specific attributes, pre-and post-announcement effects, as well as different geographies, time periods and prevailing accounting rules.
4. Methodology and hypothesis
This study employed a classical event study framework in line with the work of Hirschey and Richardson (2003) and Li et al. (2010), and following the key principles described in “The Econometrics of Financial Markets” by Campbell, Lo and MacKinlay (1997). The main steps under this approach are: event definition, selection criteria, normal and abnormal returns, estimation procedure, testing procedure, and finally empirical results, interpretation and conclusion. The event is defined as the goodwill impairment announcement made by a firm to the public. The event windows, representing the periods in time within which any abnormal returns are measured, were set to be the standard 3-day, 11-day and 20-day windows around announcement date. However, to provide greater granularity, these are further split into windows
13
representing -10, -5, and -1 days before and +1, +5, and +10 days after the event date similar to the approach of Li et al. (2010). Furthermore, to test for any underreaction in the market, longer windows spanning 55 and 250 days after the event (t = +10 to t= +65 and +260) were also examined. As elaborated in the Data section of this paper, the sample for this study was derived from the list of the top 500 firms listed in the London Stock Exchange (excluding Financials) that have made a goodwill impairment announcement within the period 2010 to 2013.
“Normal” or expected returns were calculated using daily share price data within a one-year estimation period prior to the event date in line with Hirschey and Richardson (2003), Li et al. (2010) and Campbell et al. (1997). As noted by Li et al., there exists a tradeoff between the number of observations used for estimation and possible structural change in the firm as the length of the estimation period becomes greater; however, the one-year estimation period was selected as it is common practice in the literature. Following Campbell et al. (1997), a gap between the estimation and event periods was set so as to lessen the possible impact of the event on the estimation of normal returns. This led to the choice of a 255-day estimation period which starts 300 trading days (t = -300) and ends 45 trading days (t = -45) prior to the event date. For calculating normal returns, the main model used is the single-index market model (or “market model”) which is most popular in the literature. Compared to other possible models (See Robustness section), the advantage of this employing this model is that it leads to a reduction in the abnormal return variance and thus can result in more accurate inferences (Campbell et al. (1997)). The model is based on the assumption that the relationship between the return of the security ( and the market ( is linear:
Market model :
For this study, the market index is taken as the FTSE All-Share Index, which is the broadest index representing the UK equity market. Alternatives are the FTSE 100, FTSE 350 and FTSE SmallCap indices, which comprise the top 100, top 350 and the smallest UK firms in terms of market capitalization, respectively; however, given that sample for this study consists of the top 500 UK firms, the FTSE All-Share index is deemed most appropriate. To note that the returns of the FTSE All-Share and FTSE 350 have been largely in line within the sample period, in contrast with the other two indices (See Figure 1).
14
Abnormal returns ( ) were then calculated as actual ( less normal return ( ), and cumulated within the desired event windows as follows:
Abnormal return :
Aggregation for each firm :
Aggregation across time :
The null hypothesis is that or the cumulative abnormal return across all firms over each specified event window is zero (See Kothari and Warner (2006) and Campbell et al. (1997)), which is equivalent to stating that the event (goodwill impairment announcement) does not exert a significant impact on the value of a firm.
and
Finally, the test statistic was computed using the equation below and compared to the appropriate critical values in order to test the null hypothesis under various significance levels. Following Hirschey & Richardson (2003) and Henderson (1990), among others, a t-test is applied to assess the stated hypothesis.
Test statistic:
where:
Among the possible firm- and announcement specific attributes which might influence the relation between goodwill impairment, three attributes were selected based on deemed importance and implementability, namely the influence of simultaneous announcements (particularly operating earnings and other disclosures), materiality of goodwill impairment and information asymmetry (analyst coverage). The full sample was then segregated into subsamples
15
based on these attributes for which calculation and testing was done for each, employing a quartile-based segregation approach for the latter two attributes in line with Li et al. (2010).
5. Data
The base sample was drawn from the top 500 UK-listed firms in terms of market capitalization (excluding Financials) that have recorded goodwill impairment charges during the period 2010 to 2013. In contrast to other types of events which are available through regulatory filings or databases, goodwill impairment announcements are less readily available, thus the dataset of announcement dates had to be constructed. Examining the full list of all UK-listed firms would have been preferred as it represents a more balanced sample in terms of firm size and market capitalization; however, working with the top 500 (as with AbuGhazaleh et al. (2012)) is more feasible given the limited timeframe and the effort required to determine announcement dates. Table 2 summarizes the sample selection process. Starting with the 385 non-financial firms belonging to the top 500 UK-listed firms as at 31 March 2014 as provided by the Financial
Times, the subset of firms that had goodwill impairment during the sample period was identified.
This was done by downloading actual yearly goodwill impairment charges reported in the firms’ income statements from Datastream (DS Code WC 18225), which resulted in the selection of 95 firm observations representing 25% of the initial list of firms. Thereafter, the actual goodwill impairment announcement dates for such subset of firms were determined. As highlighted by Campbell et al. (1997), pinpointing the precise event or announcement date is critical for the success of an event study, thus a considerable amount of time was devoted to this phase. Announcement dates were identified by utilizing the news search tool LexisNexis which provides access to newspapers, newswires and the Regulatory News Service (RNS) provided by the LSE. As with existing studies, the search words “goodwill”, “impairment”, “write off” and “charge” were used. Once announcement dates have been extracted, the exact dates for the estimation periods and event windows around each announcement date were determined. Subsequently, share price data as well as firm-specific attributes (goodwill impairment over total assets and analyst coverage) for those periods were extracted from Datastream to enable abnormal return calculations and testing for the entire sample and each subsample or quartile. Due to missing announcement dates in LexisNexis and/or price data from Datastream, an
16
additional 20 firms were excluded resulting in a final count of 75 firm observations. Of these 75 firms, some have reported goodwill impairment related to multiple business segments for more than one year during the sample period 2010 to 2013, resulting in 129 goodwill impairment announcements or observations which represent the final sample for this study.
The mean market capitalization of the 75 firm observations included in the final sample as at 31 March 2014 is GBP3.8 billion, ranging from GBP177 million to GBP58 billion, while total assets average at GBP5.4 billion with a wider range of GBP10 million to GBP 140 billion. These values are similar to the original top 500 (excluding Financials) list, albeit somewhat smaller in terms of mean market value but larger in terms of total assets or balance sheet size (See Table 3). In terms of sector composition, the sample firms are also in line with the top 500 (ex. Fin) firms although over-represented in Aerospace & Defense, Media and Support Services but underweighted in Mining (See Table 4), indicating sizeable goodwill impairment charges for the first three sectors during the sample period. Compared to the overall UK market, the sample firms necessarily have an upside bias in terms of market capitalization due to the choice of utilizing the top 500 UK-listed firms as base sample.
The 129 goodwill impairment observations are spread across the sample period 2010 to 2013 in terms of number and value, albeit a sizeable increase is seen in 2013 (See Table 5). Excluding outliers (mostly Vodafone impairment charges ranging from GBP2-8bn each year, and two other large impairments in 2013), the upward bias is decreased although still exists for that year. The average annual goodwill impairment loss for the sample firms within the sample period is GBP105 million, but note that this mean figure takes into account years wherein the firms had no goodwill impairment (about 55% or half of the time) and that yearly goodwill impairment charge had a wide range starting at around GBP0.12 million to as high as GBP7.7 billion. In relation to total assets (before impairment), these annual impairment charges average at 3% and varies between 0.01% to 24% of total asset levels.
Certain firm- and announcement-specific attributes were extracted for the purpose of exploring whether these attributes influence the relation between goodwill impairment announcements and firm market value. For announcement-specific attributes, the types of announcement in terms of “messiness” were determined in line with Hirschey and Richardson (2003). Announcement
17
observations were classified into either “simple” announcements, wherein only the goodwill impairment was announced without any other material disclosures, or “messy” announcements which were accompanied by confounding information, either in the form of positive operating earnings, negative operating earnings or other miscellaneous information. Note that operating earnings were taken as figures before any goodwill impairment in order to measure the operating performance of the rest of the company and assess its influence, if any, on the direction and magnitude of the share price impact of the goodwill impairment announcement.
In the studies performed by Hirschey & Richardson (2003), Francis et al. (1996) and Bens et al. (2007), it was noted that majority of goodwill impairment disclosures are “messy” announcements in that they are disclosed alongside other firm announcements most typically earnings disclosures. For the data examined in this study, all of the announcement observations in fact turned out to be “messy” (See Table 6). Out of the 129 announcement observations, N=108 (84%) were accompanied by positive earnings disclosure, N=10 (8%) with negative operating earnings disclosure and N=11 (9%) with other miscellaneous information such as impairment of other intangibles and assets, cash and non-cash restructuring charges, loan refinancing, management changes etc. Moreover, of the 118 observations with simultaneous operating earnings disclosures, 80% came in the form of annual or interim results announcements while the remaining 20% were preliminary announcements or management discussion ahead of final results. As noted by Kothari (2001), the presence of such “confounding events” affects the inferences that can be drawn from an event study, hence this is taken into account in interpreting the findings as explained in the Results section.
As for firm-specific attributes, the materiality of goodwill impairment as well as level of information of asymmetry for each observation were examined in line with the work of Li et al. (2010) and Bens et al. (2007), respectively. For the first attribute, the ratio of goodwill impairment charge (DS Code WC18225) to total assets (DS Code WC02999) was computed. Full-year goodwill impairment amounts were considered in relation to year-end total assets in the year in which the impairment announcement was made, as implemented by Li et al. (2010). Although goodwill impairment charges may be disclosed prior to year-end either as separate announcements or with interim results release (and thus may differ from full-year reported figures), the above procedure is deemed to provide a good measure of goodwill impairment
18
materiality and greatly reduced the complexity of data extraction. However, to note that an additional step was made by adding back the goodwill impairment charge to year-end total assets in the denominator in order to achieve a better approximation of the asset or balance sheet size upon announcement and before the impairment charge. For information asymmetry, the number of analysts following a firm as provided by IBESv was considered as with Bens et al. (2007). This in turn is approximated by the variable “Earnings Per Share Total Number of Estimates in the Mean FY1” which is available in Datastream (DS Code EPS1NE). For simplicity, only the latest value for this variable was considered on the assumption that the degree of analyst following and thus the level of information asymmetry does not vary materially over the sample period of four years. On the basis of these two firm characteristics, the full sample was divided into quartiles (of N=33 observations each) and testing was conducted on the top and bottom quartiles in line with Li et al. (2010) in order to estimate the impact of these attributes on the valuation effects of goodwill impairment.
6. Results
Before proceeding with the formal results of the event study, it is important to note the auxiliary finding that simple or pure goodwill announcements appear to be rare since the sample turned out to be entirely composed of “messy” announcements. That is, goodwill impairments have a tendency to be disclosed alongside other important firm announcements, on the basis of the observed characteristic of the particular sample dataset, time period and prevailing accounting regime examined in this study. The precise reason for this observation is unknown. It may be rooted to the current accounting standard in effect during the sample period which mandates periodic goodwill impairment testing. This may have rendered impairment charges less novel or newsworthy versus goodwill impairment occurring under previous regimes of goodwill amortization or optional impairment which were the subject of earlier studies. Hirschey and Richardson (2003) commented that the “adoption of FASB Statement No. 142 promises to make goodwill write-offs routine corporate events” and such routine nature may be the reason for the reduced attention on goodwill write-offs. Another possible explanation may be the relatively stable economic conditions during the sample period 2010 to 2013, leading to less material goodwill impairment losses incurred by firms. This may have foregone the need for firms to
19
make separate announcements regarding goodwill impairment compared to earlier periods of more difficult economic conditions and accompanying elevated levels of goodwill impairment charges. Interestingly though, it can be noted that Hirschey and Richardson (2003), Francis et al. (1996) and Bens et al. (2007) also worked with samples which were predominantly composed of goodwill impairment announcements with confounding events, even though their sample period occurred prior to the current mandatory goodwill impairment era (except Bens et al. who also looked at pre-, post- and transitionary periods). Whether this is an inherent limitation in examining goodwill impairment announcement data, or arises due to the current accounting regime and/or macroeconomic setting, it is important to note that such confounding events may impact the conclusions that can be derived from an event study (Kothari (2001)) and hence should be kept in mind in interpreting the results.
Table 7 – Panel A summarizes the results of the event study for the full sample and each subsample for various event windows using the single-factor market model. Figures 2 to 9 present graphical representations of the cumulative abnormal returns for short and long windows for the total sample and subsamples.
Looking at the event period CARs for the full sample (See Table 7 and Figure 2), the results appear to be inconclusive – statistically significant positive CARs were observed for certain segments of the short event window around announcement date, but negative CARs were also found although not significant. Taking the immediate 3-day window surrounding the event (days -1 to +1), a positive and significant CAR was observed, albeit of a low magnitude at +0.29%. For the other segments within the short window, the magnitude of the CARs, both in the positive or negative direction, were also on the low end ranging from -0.06% to 0.64%. Overall, these results seem to suggest that investors react slightly positively or almost neutrally to announcements of goodwill impairment due to the small magnitude of CARs (albeit not statistically significant throughout the short window). This is contrary to the predominant finding in existing literature that goodwill impairments are linked to significant negative share price reactions upon announcement. Moreover, it supports the notion that goodwill impairments have a “confirming rather than predicting value” and that its information value is limited due to the complexity and significant management discretion involved in the implementation of the current impairment-only standard. Related to the findings of Bens et al. (2007) and Z. Li et al. (2011) on
20
the impact of the implementation of SFAS 142, the small magnitude of event window CARs (positive or negative) observed may as well be due to the more regular recording and smaller magnitude of losses and/or the greater amount of discretion involved in implementing the current goodwill accounting rule. However, as noted, results should be interpreted with caution due to the existence of confounding events, in particular given the fact that majority of the observations were disclosed along with positive operating earnings information (See Table 6).
For the subsamples decomposed on the basis of the type or “messiness” of announcements (see Table 7 and Figure 3), it can be seen that firms that announced positive operating earnings at the same time as goodwill impairment experienced positive and significant CARs around the announcement period although of relatively low magnitude in the region of about 1% (+0.90% in days -1 to +1). This may be an indication that the positive earnings disclosure has a dilutive effect on any possible adverse signal conveyed by goodwill impairment announcements. To note however that the magnitude of such positive CARs are lower than those found by Hirschey and Richardson (2003) and Li et al. (2010) within the same event windows, though the reason for this unknown. Looking at firms that announced negative operating earnings, statistically significant negative CARs were found, in particular at -1.76% in the 3-day window surrounding the announcement. Similarly, firms that disclosed other miscellaneous announcements (such as impairment of other intangibles and assets, cash and non-cash restructuring charges, loan refinancing, management changes etc.) also displayed negative and significant CARs but of higher magnitude (-3.80% for days -1 to +1). For these two subsamples, the disclosure of negative operating earnings and other announcements which are typically negative in nature may have an aggravating effect on any adverse impact of a goodwill impairment announcement. The results found for this subsampling based on messiness of announcements are largely in line with the findings of Hirschey and Richardson (2003). Li et al. (2010) also obtained similar results, in particular observing positive or attenuated negative CARs around announcement date for firms reporting positive EBITDA, in contrast with negative and aggravated negative CARs for firms with negative EBITDA. All in all, the results are consistent the idea that the market interprets disclosures of goodwill impairment in the context of other important firm announcements as explained by Hirschey and Richardson (2003). However, the interpretation of above results is again not completely straightforward due to the difficulty in decomposing the share price impact
21
between any portion caused by the goodwill impairment and any that arises from the operating earnings and other simultaneous disclosures.
Turning to the subsamples divided according to materiality of goodwill impairment, measured in terms of the goodwill impairment amount over total assets before impairment (See Table 7 and Figure 4), it can be seen that firms belonging to the top quartile exhibited statistically significant negative CARs within the short event window (-2.92% within days -1 to +1), in contrast to the bottom quartile which displayed positive significant CARs albeit of low magnitude (+0.67% for days -1 to +1). Graphically, it can be seen in Figure 4 that firms that incurred more material goodwill impairments suffered a sharper drop in returns around announcement date. These results are in line with the findings of Li et al. (2010) who found that the adverse share price impact of goodwill impairment announcements tend to be larger for firms with a higher proportion of assets carried as goodwill. It is also related to the findings of Z. Li et al. (2011) and Escaffre & Sefsaf (2010) who found that share price reactions are linked to the size of the goodwill impairment. In contrast, it does not support the notion that it is the fact of a goodwill impairment and not its size that determines any market value impact as found in the studies by Hirschey and Richardson (2003) and Francis et al. (1996).
Lastly, short-window CARs for firms belonging to the highest information asymmetry group (or least analyst following) are found to be more negative, albeit slightly, than the low information asymmetry group at -1.41% and -1.26% in days -1 to +1, respectively (See Table 7). In addition, the CAR for the former group appear to be more volatile or fluctuating in the short-term as depicted in Figure 5. This points to a possible greater “surprise” factor among market participants for the high information asymmetry (low analyst coverage) group particularly within the immediate days surrounding the announcement (days -1 to +5). Bens et al. (2007) also had similar findings, explaining that such results are “consistent with the goodwill impairment accrual being more informative when there are fewer analysts generating publicly available information about the firm.” Also as noted in the KPMG study (2014), “Some markets may be better in anticipating impairments causing the actual impairment to be less important than others” which may be the case for the high analyst coverage (low information asymmetry) group. It is also related to the results of the study of Li et al. (2010), who found that firms that did not disclose any prior information about the goodwill impairment suffered more adverse share price
22
impact upon actual announcement of the write-off, also providing evidence that the information environment that a firm is in affects any potential impact of goodwill impairment on share prices. The last two columns of Table 7 summarizes the long window CARs, intended to measure possible post-announcement effects, of goodwill impairment announcements derived using the market model. On average, positive and statistically significant CARs were observed for both the 55-day and the longer 250-day windows. These results point to the recovery of share prices within the one-year period following the announcement date, as illustrated graphically in Figure 6. This finding is consistent with that of Li et al. (2010) who also looked examined CARs within 65 days post-announcement. Moreover, it does not support the notion of “underreaction” among market participants towards goodwill impairment announcements found in the study of Hirschey and Richardson (2003), who noted that the adverse valuation effects of goodwill impairment announcements are not completely realized by the end of the announcement window as further market value deterioration occurs one year after the announcement. However, note that there was no elaborate explanation provided in the study by Li et al. (2010) regarding the possible reasons for the divergence with Hirschey & Richardson’s findings in their earlier study. Hirschey and Richardson (2003) also noted that determining the cause of such observed negative post-announcement effects was beyond the scope of their study and only highlighted that investors need to be aware of such observed phenomenon. Indeed, it is difficult to determine the possible reasons and interpret the findings for the post-announcement period as with the two papers cited. Due to the nature of the research design of this study and the previous studies which employed event study methodology, there exists a tradeoff in that it allows measurement and testing of the long-term trend in CARs but not determination of the cause of such trend.
Additionally, tests for post-announcement effects were also applied to the subcategorization of the sample according to announcement and firm specific attributes (See Table 7 and Figures 7, 8 and 9). Interestingly, there appears to be greater recovery of CARs for firms that announced negative operating earnings alongside goodwill impairment (versus those who disclosed positive earnings and other information), those that incurred more material goodwill impairment (versus low materiality group) and those within the higher information asymmetry or lower analyst following group (versus low information asymmetry group), both for the intermediate window of -10 to +65 days and the longer window of up to 250 days after event date. These findings are in
23
line with those of Li et al. (2010) who also noted more pronounced increases in CAR in the post-announcement period for the first two subsamples (operating earnings backdrop and goodwill impairment materiality). A possible reason may be the link to “big bath” behavior, in that firms experiencing negative operating earnings may choose to report large goodwill impairment charges at present in order to display a more positive earnings trend in the future (See Z. Li et al. (2011) and Zhang (2011)). In the case of the high information asymmetry firms, which tend to be smaller firms with low analyst coverage, the more positive performance reflected by the greater recovery of CAR within the long window may as well be linked to the well-documented “small firm anomaly” that small firms tend to outperform larger firms. However, same as Li et al. (2010) who did not attempt to investigate the causes of such observations in their study, it is again difficult to interpret the results and perform root cause analysis in the context of the event study framework employed.
In summary, the examination of the full sample yielded evidence of both positive and negative CARs, albeit of low magnitude and differing statistical significance, within the short window surrounding the goodwill impairment announcement date. It was also found that simultaneous announcements of positive operating earnings in contrast to negative earnings or other miscellaneous information appear to have a diluting and aggravating effect, respectively, on any adverse share price impact of goodwill impairment. Moreover, more negative share price impact within the short window surrounding the event were found for firms announcing more material impairment in relation to total assets (higher impairment materiality) and those that have lower analyst coverage (higher information asymmetry). Within the post-announcement period, positive and significant CARs were found on average suggesting recovery of share price. However, such trend of recovery appear to be more pronounced for the subset of firms that made simultaneous announcement of negative operating earnings, and those that belong to the high goodwill impairment materiality and high information asymmetry group.
7. Robustness
For robustness, two other models of calculating normal returns were utilized in order to verify the results obtained using the market model. These are defined as follows:
24
Mean-adjusted return model :
Market-adjusted return model :
The first model takes the mean return of the security ( ) as constant through time (i.e. the security is expected to generate the same return as the average during the estimation period), whereas the second assumes that α=0 and β=1 in the market model (i.e. the security is expected to generate a return equal to the market during the event window in the absence of news). Other more sophisticated models are also possible, for instance multi-factor models which incorporate other regressors in addition to the market return and economic models such as the Capital Asset Pricing Model (CAPM) and Arbitrage Pricing Theory (APT). However, these were not employed due the limited marginal explanatory power they offer as well as the unnecessary complication that they may add to the study (Campbell et al. (1997)).
The results derived using the mean-adjusted and market-adjusted models are summarized in Table 7, Panels B and C. As shown, they lead to the same conclusions as the market model both for the short and long event windowsvi, thus confirming the key findings discussed under the Results section.
The testing procedures utilized in this study are considered “parametric” in nature, meaning that they hinge upon specific assumptions regarding the distribution of returns. Non-parametric tests such as the sign test and rank test, which do not depend on such assumptions, may also be used together with parametric testing to check for robustness (Campbell et al. (1997). The sign test, for instance, may be useful in verifying that results do not depend on outliers (Cowan (1992)). However, the test may not work well if the distribution of abnormal returns is skewed as with daily data (Campbell et al. (1997) which is the case for the daily returns data used in this study. Furthermore, many other studies have analyzed the additional usefulness of using non-parametric tests in conjunction with parametric testing but with conflicting conclusions (See for instance Campbell et al. (1997) and Henderson (1990)). Lastly, to note that out of the six existing studies which have investigated the association of goodwill impairment and market value using event study methodology as reviewed in this paper, only one has employed a single non-parametric test (Wilcoxon signed-ranks test) which corroborated the findings of the parametric testing used (See
25
Escaffre & Sefsaf (2010)). Thus, in view of the above and the limited timeframe, additional non-parametric tests were not included in this study.
8. Conclusion
This paper extends prior research by exploring the relation between goodwill impairment announcements and any potential impact on market value. This was done within the context of the prevailing international accounting standard on goodwill accounting mandating regular impairment testing as opposed to previous regimes of amortization and/or optional impairment. While existing literature is dominated by evidence supporting the negative impact on market value of goodwill impairment, which is line with the view that such disclosures provide new and valuable information to the market regarding a downward revision in management’s estimates of future profit-generating potential, alternative views remain that goodwill impairment should have limited valuation impact due to its “confirming rather than predicting value” and the level of complexity and management discretion involved in implementing the current impairment-only standard. Moreover, some authors conjecture that more regular recording of impairments and resulting smaller magnitude of losses booked and more managerial discretion involved in financial reporting may lead to lower value impact of impairment announcements, as could have possibly resulted from the shift to the current standard.
The research question was investigated within an event study framework applied on a sample drawn from the UK market within the period 2010 to 2013, representing the latest available dataset that was unaffected by major market crises and occurred within the post-IFRS adoption era in Europe. To provide richer insight into the goodwill impairment-market value relation, the impact of additional firm and announcement attributes were explored as well any post-announcement effects within longer time periods following post-announcement. To note that this study was not intended to be a comparative study of pre- and post-adoption of the current standards governing goodwill, but instead sought to investigate the research question within the context of the prevailing accounting regime.
The examination of the full sample yielded slightly inconclusive results as both positive and negative CARs, albeit of low magnitude and differing statistical significance, were observed
26
within the short window surrounding the goodwill impairment announcement date. In terms of “messiness” of announcements, it was found that simultaneous announcements of positive operating earnings result in positive event window CARs, in contrast to negative earnings or other miscellaneous information which are linked to negative CARs. These point to possible diluting and aggravating effects of simultaneous positive and negative earnings announcements, respectively, albeit such results should be interpreted with caution due to the difficulty in decomposing the share price impact between any portion arising from the goodwill impairment and that resulting from the other disclosures. Stronger inferences can be drawn from the examination of the other subsamples, in particular that firms announcing more material impairment in relation to total assets (higher materiality) and those that have lower analyst coverage (higher information asymmetry) tend to exhibit more negative share price impact within the short window surrounding the event. As for the longer window spanning one-year after the announcement date, positive and significant CARs were observed on average, pointing to recovery of share prices in the post-announcement period and in contrast with evidence from a previous study indicating underreaction of market participants to goodwill impairments upon announcement. Finally, an interesting but unexplained finding is that such recovery of share prices were found to be more pronounced for the subset of firms that made simultaneous announcement of negative operating earnings, and those that belong to the high goodwill impairment materiality and high information asymmetry (low analyst coverage) group. These results are robust to the choice of the normal returns model used within a parametric event study framework.
The finding of both small positive and negative CARs for the full sample within segments of the short event window seem to suggest that investors react slightly positively or almost neutrally to announcements of goodwill impairment. However, such interpretation is limited by the fact that the magnitude of the CARs observed were very low and not consistently significant across window segments, as well as the presence of confounding events in particular of the predominance of simultaneous positive earnings disclosures. Although the study did not result in strong evidence supporting the debate on whether or not goodwill impairment exerts an adverse impact of stock price on the basis of the full sample, it does provide corroborating evidence to previous research that find links between the share price impact of goodwill impairment and simultaneous announcements (in particular of the operating earnings background), as well as the
