Investors anticipate goodwill impairment
J. J. Levert
2008
University of Groningen
Faculty of Economics and Business
MSc BA Finance Master Thesis
Student number: 1342673
Abstract
This paper investigates whether investors anticipate goodwill impairments using event study methodology. Impairment of goodwill refers to the write down of the book value of goodwill to its fair value. The period from acquisition to impairment of goodwill is used to analyze investors’ anticipation of this goodwill impairment. The results indicate that the majority of goodwill impairment amount does not contain additional value information to the investors. An average impairment amount equal to 51% of a company’s market capitalization on average results in only a 4.4% fall in the company’s stock price. Investors recognize value-destroying acquisitions both at acquisition and in the period between acquisition and impairment, leading to a respectively 1.3% and 68% decreases in stock price compared to the market return. The total impairment amount is significantly and negatively related with both the abnormal returns at acquisition and before impairment. Finally, it is found that impairment of recently acquired goodwill results in significantly more negative abnormal returns (-6.5%) compared to goodwill acquired longer ago (-0.9%).
Keywords: announcement returns, event study, goodwill, impairment
J. J. Levert Europaplein 33-II 1078 GT, Amsterdam
1) INTRODUCTION 4
2) THEORETICAL FRAMEWORK AND LITERATURE REVIEW 7
2.1) ACQUISITION AND TREATMENT OF GOODWILL 7
2.2) MERGERS AND ACQUISITIONS 8
2.2.1) HUBRIS 8
2.2.2) AGENCY THEORY 10
2.2.3) INVESTORS’ REACTION TO ACQUISITIONS AND IMPAIRMENTS 11
3) DATA & METHODOLOGY 17
3.1) SAMPLE CONSTRUCTION 17
3.2) SENSITIVITY ANALYSES 21
3.2.1) SIMPLE ANNOUNCEMENTS 21
3.2.2) EARLY WARNING 21
3.2.3) ‘RECENT’ VERSUS ‘OLD’ GOODWILL 21
3.3) EVENT STUDY 22
3.3.1) EVENT DATE AND EVENT WINDOW 22
3.3.2) ESTIMATION WINDOW 24
3.3.3) NORMAL RETURN 25
3.3.4) MARKET AND RISK ADJUSTED MODEL 25
3.3.5) MARKET ADJUSTED MODEL 26
3.3.6) DESCRIPTIVE STATISTICS 26
3.4) STATISTICAL TESTS 27
3.4.1) PARAMETRIC TEST 27
3.4.2) NON-PARAMETRIC TEST 28
3.4.3) REGRESSION 30
3.4.4) MANN-WITHNEY-WILCOXON TEST 31
4) RESULTS 33
4.1) ACQUISITION ANNOUNCEMENT, RUN-UP PERIOD AND IMPAIRMENT ANNOUNCEMENT33
1) Introduction
Over the last decade many companies impaired large amounts of goodwill: KPN (2001, USD 14b), Vivendi (2002, USD 18b), Cable & Wireless (2002, USD 2b) and Vodafone (2006, USD 34b). Impairment of goodwill is the write down of the book value of goodwill, in this paper referred to as “accounting impairment”. Accounting impairment is the consequence of value-destroying acquisitions. Both overpayment and unrealized synergies could cause an acquisition to be value-destroying. Overpayment causes value destruction because the premium paid exceeds the synergies expected. In case synergies do not materialize the acquisition proves to be value-destroying ex post. In general, companies engage in mergers and acquisitions (M&A) because the value of the combination is expected to be higher than the sum of the parts. In other word, the acquiring firm recognizes synergies that can create incremental cash flows. Agency theory (Coase, 1937; Jensen and Meckling, 1976) and hubris theory (Roll, 1986) explain why M&A do not always create value.
Previous research found that the announcement of accounting impairment causes a significant fall in stock prices. The abnormal returns caused by the announcement of accounting impairment were relatively small compared to the amount impaired (Bartov et al. 1998; Hirschey and Richardson, 2002). The authors proposed that investors might anticipate accounting impairments and already devalued the share price before the impairment announcement. The decrease in share price because investors recognize that an acquisition destroys value is referred to in this paper as “economic impairment”. The anticipation of accounting impairment would lead to the relatively small abnormal returns at the announcement of accounting impairment. Decreasing share prices prior to accounting impairment corresponded with investors’ anticipation of accounting impairment (Bartov et al. 1998; Hirschey and Richardson, 2002). Both articles did not show whether decreasing share prices are related with the amount impaired. The fall in stock prices at accounting impairment announcement suggest that the announcement contains information value. According to Bartov et al. 1998, investors fail to incorporate all value relevant information into the share price before accounting impairment is announced.
announcements will result in negative abnormal returns. When investors indeed recognize value-destroying acquisitions, it is expected that economic impairment occurs prior to accounting impairment. In this paper it is analyzed whether investors on average recognize value-destroying acquisitions and economically impair an acquisition before accounting impairment is announced. Furthermore, it is investigated whether the accounting impairment announcement holds relevant information value for investors. Data used by previous research does not include the complete period where economic impairment might occur. The data merely includes share prices of the accounting impairment announcement and of a one or two year period preceding that announcement. Economic impairment is expected to occur in the period from the announcement of acquisition of goodwill up to the accounting impairment announcement. To perform the empirical tests for this study, a unique data-set was composed that links the announcement of accounting impairment directly with the acquisition that created the impaired goodwill. With this data-set it is possible to analyze share prices from acquisition announcement up to and including accounting impairment announcement. To my knowledge no other study has linked the acquisition directly with impairment and neither has the complete period between acquisition and accounting impairment been analyzed before.
The composed data-set includes 62 European companies that announced accounting impairment between 2001 and 2007. The acquisitions included in the data-set proved value-destroying due to the fact that accounting impairment followed the acquisition. Using event study methodology the abnormal returns from acquisition up to and including accounting impairment are examined to test whether economic impairment preceded accounting impairment. An OLS regression tests whether abnormal returns of economic impairment explain the amount of accounting impairment.
relatively small announcement returns on accounting impairment found in previous research. A novelty of this study is the fact that the focus will be on the European market and to date this market has not empirically been analyzed for announcement effects of goodwill impairment.
The infrequent nature of accounting impairments and the vague communication of information concerning the accounting impairment, confronts investors with difficulty when assessing a company’s performance and value (Bartov et al. 1998). In order to standardize the treatment of goodwill and goodwill impairments, policymakers introduced IFRS 3, IAS 36 and IAS 38 in 2004. This paper is interesting for organization that set accounting principles, such as the International Accounting Standards Board, as it shows whether investors were able to economically impair acquisitions even without the new standards, this can be important when designing improved standards. Furthermore this paper shows whether the introduction of impairment testing is supported by the results.
This paper is interesting for investors as well, since it will show whether information value is entrenched in accounting impairment announcements, or that accounting impairments are anticipated and incorporated in share prices.
2) Theoretical Framework and Literature Review
2.1) Acquisition and treatment of goodwill
Goodwill is recognized in the books of an acquirer upon acquisition of a target company. The carrying book value of goodwill from a certain acquisition is calculated as the fair value of all the identifiable net assets after an acquisition less the equity purchase price. See appendix I for a detailed explanation and a schematic overview of the creation of goodwill.
Before the IFRS standards, goodwill was amortized over a fixed number of years. Under IFRS standards, goodwill has an indefinite lifetime and companies have to perform an annual impairment test to estimate the value goodwill. An impairment test assesses the recoverable amount of goodwill. When the carrying amount (i.e. book value) of goodwill exceeds the recoverable amount, the goodwill has to be impaired. The amount of goodwill impairment is equal to the carrying amount less the recoverable amount. The recoverable amount is the higher of the fair value of a company’s net assets less cost of selling these assets or the value in use. Both are based on the present value of future cash flows. The fair value of a company’s net assets less cost of sales takes a market perspective, i.e. what is the value of the assets for other companies. The value in use determines what the present value of future cash flows are for the company as a stand alone enterprise. Typically companies engage with an independent third party to perform impairment tests, this will provide comfort to the company’s external auditor with respect to the accuracy of the test outcomes. In addition to annual impairment testing, companies have to perform an impairment test after triggering events as well. A triggering event is an event that significantly affects the current profitability of the asset or significantly decreases the future profitability of the asset and thereby the recoverable amount.
2.2) Mergers and Acquisitions
In general, companies engage in mergers and acquisitions (M&A) because the value of the combination is expected to be higher than the sum of the parts. The improvement that the two companies plan to achieve compared to the stand alone situation is referred to in this paper as synergies. In some theories the managers of a firm act in the interest of the owners of a company. Shareholders are the formal owners of a company and they pursue value maximization. On average shareholders see value maximization as earning a high return on their invested capital. Managers can attain the shareholders’ goal of value maximization by increasing the share price in a sustainable manner. Following the value maximization goal, managers should not undertake value-destroying acquisitions and thus create value with every acquisition they do. However, there are several theories, e.g. hubris- and agency theory, that explain why managers might not always act in the best interest of the shareholder. These theories explain how shareholder value might be destroyed in M&A.
First, the hubris theory, developed by Roll (1986), argues that managers overestimate their ability to manage the target firm and therefore, on average, overpay in an acquisition. Second, agency theory, assumes that managers act in their own interest and therefore might destroy shareholder value by overpaying in M&A deals. Both theories explain why overpayment could occur.
2.2.1) Hubris
Several authors tested the Hubris theory of Roll (1986) on different samples and in various settings. Hubris theory in its strictest form assumes that no synergy gains exist in acquisitions. In case no synergies exist, acquisitions are driven by valuation mistakes. When valuation mistakes drive acquisitions, acquisitions merely occur when the valuation mistake causes overvaluation (Roll, 1986).
measured as the abnormal share price returns of the target and total gains as the combined abnormal share price returns of target and acquiring shareholders. Since the authors find a significant positive relation between target and total gains they conclude that on average synergies are present in acquisitions. In acquisitions purely driven by synergies, it is expected that both firms can reap the benefits (Berkovitch and Narayanan, 1993). In case both firms can enjoy the benefits of synergies, a positive relation between acquiring and target gains is expected. In acquisitions driven by hubris, a negative relation between acquiring and target shareholder gains is expected (Berkovitch and Narayanan, 1993). Note that the expected relation between acquiring and target gains is opposite for synergies and hubris. The authors find no significant evidence for any relation between acquiring and target gains, indicating that hubris off-sets the synergistic gains found in the sample. Berkovitch and Narayanan (1993) prove that hubris is present in M&A. The finding that hubris off-sets the expected acquiring shareholders gains caused by synergies, shows that on acquisition date investors realize that besides synergies, hubris is present as well and hubris negatively influences investors’ value expectations of the acquisition.
2.2.2) Agency Theory
Agency theory is a component of contract theory which was introduced in 1937 by Coase and further developed by Jensen and Meckling (1976). According to contract theory a problem exists between the providers of funds - investors - and the managers who control the funds. The problem is caused because investors cannot force managers to use the funds exactly the way investors would like and therefore cannot ensure a proper return on their investment. Furthermore, the division of ownership by the investors and control by managers of the company causes information asymmetry between investors and managers. The information asymmetry leads to agency risk for the investor. Agency risk is the possibility that managers abuse their position to appropriate managerial rents. An example of agency risk is moral hazard and occurs when managers employ investors’ funds for their own benefit. Examples of moral hazard are empire building and disproportionate use of company’s perks, e.g. usage of the company’s private jet for private purposes (Tirol, 2006). Moral hazard problems might lead to value-destroying acquisitions and thereby causing shareholder losses.
motive. The finding that acquisitions into unrelated areas leads to negative abnormal returns for the acquiring shareholders, shows that at acquisition the investors recognize the existence of the agency problem and the value destructive nature of the acquisition and economically impair the acquisition.
Furthermore, Shleifer and Vishny (1989) describe that the agency theory might very well be present in companies through managers who not only choose investments on a basis of value maximization but also to entrench themselves deeper into the company. Entrenchment makes it more costly to replace managers and therefore they are able to extract higher management rents from the shareholders. This means that acquisitions driven by managers that wish to entrench them into the company are potentially value-destroying for investors. This is merely an example of an agency problem that could arise with M&A.
The agency theory shows that managers may act in their own interest and thereby destroy shareholder value in M&A. Especially Morck et al. (1990) prove that investors, at least in some cases, are able to recognize moral hazard motives and economically impair those acquisitions. In the next section it is discussed whether and when investors recognize value-destroying acquisitions.
2.2.3) Investors’ reaction to acquisitions and impairments
The Hubris and Agency theories provide possible explanations for the fact that value-destroying acquisitions occur. This paper analyzes whether investors are able to recognize and economically impair value-destroying deals before accounting impairment takes place. The expectation is that there are two moments for investors to recognize acquisitions that destroy value. These two moments are at acquisition announcement and the period between acquisition announcement and accounting impairment announcement. Analyzing these two moments completely independently will be difficult since economic impairment at acquisition will affect market expectations in the second period.
expected synergies. In some cases investors might believe that the synergies are overstated and/or the premium paid is too high. When investors on average believe that synergies are lower than the premium paid, the acquisition announcement will lead to negative abnormal returns. This would indicate that investors economically impair overpayment and thereby goodwill upon the acquisition announcement based on their expectations at that moment.
Henning et al. (2000) use 1576 US acquisitions between 1990 and 1994 to show that investors do indeed impair overpayment upon acquisition. In the article the goodwill component is decomposed into three parts: going concern, synergies and residual. The going concern is measured as the fair market value of the target’s assets less the market value of the target before the acquisition announcement. The fair market value of assets is the opening balance of the target’s assets on the balance sheet of the acquirer. Synergies are calculated as the sum of the cumulative absolute announcement returns. Absolute announcement returns are the dollar amount change in market capitalization – share price times outstanding shares – of both the target and acquirer. The residual is computed by subtracting going concern and synergies from the acquired goodwill. Calculating the residual in this manner Henning et al. (2000) capture the overpayment part of the acquisition. The authors use the acquirer’s market value of equity after the acquisition as dependent variable and regress it against the three previously mentioned components of goodwill. They find positive relations between the market value of equity and both the going concern and synergies. The market value of equity is significantly negative related with the residual component. This shows that investors recognize overpayment in transactions and economically impair acquired goodwill when they believe the company overpaid for the acquisition. The article of Henning et al. (2000) proves that the acquisition announcement is an important moment to analyze when considering economic impairment of an acquisition by investors.
profitable ex post. Over time more information about the performance of the acquired firm will become public and investors are increasingly able to assess the incremental value resulting from the acquisition. The companies included in the sample of this paper1 are all companies that impaired their goodwill. In this case the news that becomes public over time is expected to be negative and the likelihood that the acquisition will be profitable ex post decreases over time. Overtime investors realize the increased likelihood that the acquisition will indeed be value-destroying, therefore investors will devalue the stock price, i.e. economically impair the acquisition. Even for investors that economically impaired at acquisition, negative news results in a decreased likelihood of their most positive scenario or a downward revision of potential scenarios and thus causes a further economical impairment.
Hirschey and Richardson (2002) analyzed 80 US companies that were listed on the NYSE, AMEX or on the National Association of Securities Dealers Automated Quotation System (NASDAQ) and announced goodwill impairment during 1992 – 1996. They found that in the period preceding accounting impairment, companies significantly underperformed the market by 42%. This shows that investors do also use information before the accounting impairment to economically impair an acquisition. Economic impairment might not just occur in the year before impairment. In this paper the whole period between acquisition announcement and the announcement of accounting impairment is analyzed. This whole period is analyzed to discover whether investors react on the negative information related to the acquisition that becomes public over time. Overall two hypotheses can be posted which are not entirely independent of each other. Henning et al. (2000) showed that investors are able to economically impair value-destroying deals at acquisition. The first hypothesis is therefore:
H1: The acquisition announcement returns of the acquirer are negative
Since this paper uses a sample of companies that impaired their goodwill, it is expected that over time negative information about the acquisition became public. Investors would realize that the likelihood that the acquisition would be profitable ex post was decreasing. Therefore negative abnormal returns are expected over time as well. The period between acquisition announcement and impairment announcement is called the run-up period.
H2: The abnormal returns over the run-up period are negative
Hypotheses 1 and 2 merely hypothesize a certain movement in stock prices surrounding an acquisition and in the period between acquisition and accounting impairment for value-destroying deals. The hypotheses do not state a certain relation between the scale of economic impairment and the amount of the accounting impairment. When investors indeed recognize value destruction, it is expected that there should be a relation between the economic impairment and the accounting impairment. The more value destructive a deal is the more economical impairment by the investors is expected. When economic impairment indeed precedes accounting impairment, the amount of economic impairment is expected to be related to the amount of accounting impairment. Economic impairment includes both the abnormal returns2 at acquisition (Acq-R) and the abnormal returns over the run-up period (Run-Up). Both are expected to have a negative relation with the accounting impairment amount. More specifically, when the acquisition abnormal returns decrease, the accounting impairment amount is expected to increase and also in case the run-up abnormal returns decrease, the accounting impairment amount is expected to increase.
H3: The impairment amount is negatively related with abnormal returns at acquisition
announcement
H4: The impairment amount is negatively related with abnormal returns over the
run-up period
Hypotheses 1 to 4 discuss that investors are able to recognize the value destruction of an acquisition and therefore economic impairment precedes accounting impairment. This suggests that at accounting impairment announcement investors already impaired the acquisition and are therefore not surprised by the accounting impairment. Previous research showed that this is not true since accounting impairment announcements result in negative abnormal returns. Hirschey and Richardson (2002) found significant negative abnormal returns of 2-3% at the announcement of accounting impairment. The most
recent goodwill are more negative compared to old goodwill, since there is more information value in recent goodwill. Table 1 provides an overview of all the hypotheses. H5: Recent goodwill leads to more negative abnormal returns at accounting
impairment announcement than old goodwill
H6: Abnormal returns at accounting impairment announcement are negative Table 1: Overview hypotheses
Hypotheses
H1: The acquisition announcement returns of the acquirer are negative H2: The abnormal returns over the run-up period are negative
H3: The impairment amount is negatively related with abnormal returns at acquisition announcement H4: The impairment amount is negatively related with abnormal returns over the run-up period
H5: Recent goodwill leads to more negative abnormal returns at accounting impairment announcement than old goodwill
3) Data & Methodology
3.1) Sample Construction
In this section the method of data selection and sample structuring is discussed. The sample includes listed companies across Europe. Companies in this sample are listed at either the European Euronext markets of Amsterdam, Brussels, Lisbon or Paris, furthermore listings are from the London Stock Exchange, Alternative Investment Market and Deutscher Aktienindex. A total of 3667 listed companies were included in this initial database.
Different criteria are applied to create this database. The Capital IQ3 database provided the information on whether a firm impaired goodwill between 2001 and 2007. In the selected period 822 companies impaired goodwill.
In this paper the period from acquisition to impairment is analyzed. Therefore it is necessary to know to which acquisition the impaired goodwill is related to. Furthermore, the acquisition has to have sufficient size to significantly influence the share price of the company. In case the acquisition is too small, the performance in the period between acquisition and impairment of goodwill might not be significantly influenced by the acquisition. For the sample of 822 companies it is not known which acquisition led to impairment. As proxy for the size of the original acquisition, goodwill impairment as percentage of market capitalization at impairment is used. Where, market capitalization is the Capital IQ market capitalization of the company in the corresponding year of the impairment. Only companies that impaired goodwill ≥ 10% of market capitalization were included in the sample. This arbitrary proxy implicates that goodwill impairment of ≥ 10% of market capitalization is the result of a large acquisition in the past that could significantly affect the performance of the company over the period from acquisition to impairment. After applying this restriction 151 companies remain.
The Bloomberg database was used to find both the announcement of accounting impairment and to find which acquisition led to the impairment of goodwill. The Bloomberg database also shows the amount of goodwill impairment. Within the sections of (important) company news, searches on “goodwill”, “impairment”, “write off” and
“write down” have been performed. All the articles on the companies included in the sample which contained any of these search-criteria have been analyzed. The announcement dates of goodwill impairments were found for the whole sample. Difficulties with finding the acquisition that led to impairment decreased the sample with approximately 45% to 77 companies. This is caused by the fact that companies are only required to announce which cash generating unit (CGU) is subject to goodwill impairment, the company is not required to publish to which acquisition the impairment is related to. Knowing merely the CGU is not enough to find the acquisition.
The acquisition dates of the 77 acquisition were retrieved from Capital IQ, Mergermarket and Zephyr. Some of these acquisitions, mostly the acquisition that occurred relatively long before the impairment, did not include sufficient share price returns in the estimation window to estimate a correct normal return4 and were excluded. Furthermore, companies that did acquisitions in multiple stages or announced impairment of multiple acquisitions were excluded, since no isolated acquisition or impairment return could be formed. The final sample includes 62 acquisitions that can be linked directly to the impairment.
At announcement date, information becomes public and investors might adapt their expectations about the company’s future cash flows. By calculating abnormal return around the announcement date it is possible to analyze whether the announcement had a positive/negative effect on investors’ expectations. In order to calculate abnormal returns one needs the actual daily stock returns of the companies and of a benchmark market. According to Koller et al. (2005), a true benchmark portfolio includes all assets, i.e. both traded and non-traded assets. Since this market portfolio is not available, a proxy for this portfolio is used. Following Koller et al. (2005) this study will use the MSCI World. The correlation between the MSCI Europe and World is high, approximately 0.94, and therefore merely the MSCI World is used. Furthermore, the MSCI World is assumed to be a well diversified portfolio and a proxy for mere market risk and no unique risk. Rational investors are expected to be well diversified as well and therefore the MSCI World is most suitable as a benchmark index. Country indices are not fully diversified and include unique (country) risk, therefore country indices are not the most suitable benchmarks for well diversified investors. The actual daily stock returns of both the company and MSCI world are calculated by dividing the closing price of that day by the closing price one day before. The daily stock prices are collected from the Bloomberg
database and are corrected for stock splits and dividend payments. Daily stock prices were used as they provide more explanatory power than weekly or monthly intervals, while shorter intervals do not provide any proven advantage (Brown and Warner, 1985; MacKinlay, 1997; Morse, 1984). Non-synchronous trading might be an issue when using daily closing prices to estimate market model coefficients. According to MacKinlay (1997) the differences between abnormal returns adjusted for non-synchronous trading and OLS derived abnormal returns are minimal. Brown and Warner (1985) state that there is no evidence that, in case of non-synchronous trading, other procedures, such as the Scholes- Williams or Dimson procedures, improve the specification or the explanatory power of the tests compared to the Market and Risk Adjusted Model. No adjustments are made for non-synchronous trading in this paper.
As was stated in part two the abnormal return at acquisition (Acq-R) and run-up abnormal returns (Run-Up) are hypothesized to be negatively related to the impairment amount. The impairment amount (Imp) is calculated as the amount of goodwill impairment divided by the market capitalization. The market capitalization is calculated by multiplying the closing price of the company posted by Bloomberg ten days before the impairment announcement by shares outstanding at that date. The closing price ten days before the impairment announcement is used to ensure that the market capitalization is unaffected by the announcement.
The acquisition of goodwill sometimes occurred before 1999, in this period the Euro was not yet used. To prevent biases due to currency differences, all the financial information was converted to dollars.
Table 2: Summary of the construction of the data sample
Criteria Number
Initial companies within selected markets/indices 3667 Goodwill impairment announcement 2001-2007 822 ≥ 10% of market capitalization 151
Acquisition traced 77
Too little data in acquisition announcement estimation window 15 Final sample goodwill impairment linked to acquisition 62
sample compared with the total companies of that specific stock exchange or index. The “% Sample” provides the percentage that specific stock exchanges or indices represent in the final sample. Of the Deutscher Aktienindex, Euronext Lisbon and Euronext Brussels no impairments are included in the sample. Eight companies have two accounting impairments between 2001 and 2007 included in this sample, all these impairments were caused by different acquisitions and were announced in different years. Although no extreme differences are shown in the “% Total” column, the English stocks are well represented in the final sample. In table 4 the distribution over different industries is given. The industries are based on GICS codes and collected in the CapitalIQ database. The majority of the sample comes from Information Technology, Consumer Discretionary and Industrials. As Hirschey and Richardson (2002) proved that no significant difference exists between industries, no sensitivity analyses across industries are done.
Table 3: Summary of the geographical division of the sample companies
Stock Exchange / Index Total companies Sample
companies % Total % Sample
Euronext Amsterdam 189 4 2.1% 7.4%
Euronext Paris 768 4 0.5% 7.4%
London Stock Exchange 1109 30 2.7% 55.6% Alternative Investment Market 1293 16 1.2% 29.6%
Deutscher Aktienindex 30 0 0 0%
Euronext Lisbon 56 0 0 0%
Euronext Brussels 222 0 0 0%
Total 3667 54 1.5% 100%
The average and median amount of goodwill impairment over the 62 sample companies are USD 1,6b and USD 23m respectively. The average and median impairment amount as percentage of market capitalization are 51% and 34% respectively.
Table 4: Overview of sample division by industry
3.2) Sensitivity analyses
Some issues might cause biases in testing results. In order to prevent these issues to cause biased test-results, samples are formed based on potential bias issues. Statistical tests are applied to conclude whether significant difference exist between samples.
3.2.1) Simple announcements
When performing an event study, it would be ideal to test data points which are free of any other occurrences which could potentially cause abnormal returns that are not related to the event. When collecting the data, part of the companies which announced goodwill impairment announced their quarterly, half year or annual report simultaneously. These earning announcements potentially influence share price, in those cases abnormal returns do not isolate the effect of the accounting impairment announcement. The sample will be split in one sub-sample where the accounting impairment announcement comes in isolation (simple) and another where simultaneous earnings announcement are done (multi). Earnings announcements might also affect the company’s share price before or after the release date, therefore the multi-sample exists of companies that announce earnings in t = [-3, 3] around the accounting impairment announcement. Splitting the sample in this way, this paper follows Hirschey and Richardson (2002). Of the total sample of 62 accounting impairments, 13 were accompanied by earning announcements.
3.2.2) Early warning
The abnormal returns of accounting impairment announcements could be biased when a company already warned the investor about possible impairment before the official announcement. Some companies for example stated that an impairment of an unknown amount was going to be announced. The essence is to measure the abnormal return caused by the accounting impairment announcement. Warning investors before that date could have created announcement returns on the ‘warning’ date and therefore a significant reaction on the actual announcement date may less pronounced. For the total sample of 62 goodwill impairment announcements, 6 were preceded by a warning.
3.2.3) ‘Recent’ versus ‘old’ goodwill
longer ago does not. The sample is split into two samples to analyze hypothesis 5. The first sample is recent goodwill and includes goodwill acquired less than three years ago upon impairment announcement. The second sample includes old goodwill, i.e. goodwill acquired more than three years ago. To determine how long ago goodwill was bought, the difference in days between acquisition and impairment announcement was calculated. Then the days were divided by 365, all numbers < 3 were included in Recent and all numbers ≥ 3 in Old. The sub-samples Recent and Old include 39 and 23 respectively. 3.3) Event study
Event study methodology is used frequently in financial literature to examine the influence of a certain event on stock prices. This methodology tests whether return anomalies exist around a certain event date. According to Binder (1998), Fama et al. (1969) caused a methodological revolution in the world of accounting, economics and finance with their paper on event studies.
In order to calculate whether an event has caused abnormal returns three steps need to be followed. First it is necessary to search the exact event date and stipulate the estimation period and event window. Second, the abnormal return in the event window is calculated per share and aggregated for the sample. Finally, statistical tests are applied to test whether the actual values differ significantly from the predicted values (MacKinlay, 1997).
3.3.1) Event date and event window
A crucial point in event studies is to correctly identify the event date t = 0. This should be the exact date when the announcement about the event reaches the financial market, i.e. becomes public information. An event might not only influence the share price on t = 0, but also the days surrounding the event. Rumors could cause a run-up in share prices preceding the official announcement date or the announcement could cause abnormal returns over a longer period than merely at t = 0. In order to include all days that are affected by an event, event studies use an event window that includes all days affected by the event. In figure 1 an overview of the abnormal returns5 surrounding acquisition and impairment announcements is given over a period of t = [-9, 9]. T = 0 is always included
in the event window and days around t = 0 that are significantly6 influenced by the event are included in the event window as well.
The impairment announcement merely affects t = 0 significantly, therefore t = 0 is used as event window for the impairment announcement. For the acquisition announcement both t = -1 and t = 1 are significantly affected and therefore included in the event window. T = 0 is in itself not significantly influenced by the acquisition announcement, but since this is the actual announcement date it is included in the event window. The event window for acquisition announcement is t = [-1, 1].
Figure 1: The market and risk adjusted abnormal returns of both impairment and acquisition announcements over an event window of t = [-9, 9].
For the run-up abnormal returns the abnormal returns between acquisition and impairment are used as an event window. To make sure that neither the acquisition nor the impairment influences the run-up abnormal returns, the run-up event window begins 10 days after the acquisition and ends 10 days before the impairment. According to Brown and Warner (1985) using a long event window could cause autocorrelation, but the benefits of adjusting for autocorrelation are very limited. Furthermore, longer event windows could cause a significant power decrease of the statistical test (Brown and Warner, 1985). A long event window is still used in this paper because investors can economically impair the acquisition at any point between acquisition and impairment and the event date for economical impairment is not known with certainty. The consequence of limiting the event window to a shorter period could result in incomplete measurement of economic impairment over the run-up period.
6 A 95% confidence interval is used to decide whether a date was significantly influenced by the announcement. The test statistic was calculated using equation 8 which can be found in part 3.4.1
3.3.2) Estimation window
An estimation window is the period over which the expected normal return of the shares of a company is estimated using the Market Adjusted Model and the Market and Risk Adjusted Model7. The Constant Mean Return Model is not used in this paper since its power significantly decreases when there is clustering (Brown and Warner, 1980 and 1985). In this paper clustering might occur, especially in the run-up period where overlapping event windows could occur. According to Brown and Warner (1980 and 1985), adjusting for clustering does not improve the results and might in some cases even be harmful compared to procedures that assume independence. In this paper no adjustments are made for possible clustering and thus independence of event dates is assumed.
The estimation window should normally not include the event window, since the event might influence the estimations of the normal return model. On the other hand, the estimation window should be close to the event date to make the best estimation for normal return. Following Bartov et al. (1998) 250 trading days are used in the estimation window and stocks where more than 40 trading days are missing from the estimation period are eliminated. In this study the estimation window is t = [-260, -10], where t = 0 is the event date of the acquisition. This estimation window is used for all three moments in time; the acquisition, the run-up period and the accounting impairment. By using this estimation window for all three moments the implicit assumption is made that the risk profile of the company versus the market remains equal over the whole period. This is a strong assumption and could potentially affect our outcomes, therefore the Market Adjusted Model is also used to check the sensitivity of the outcomes. In the figure 2, an overview of the event windows and estimation window is given for goodwill impairment announcement returns, run-up period and acquisition announcement returns. The time between acquisition and impairment differs from company to company and therefore IMP represents the accounting impairment announcement and IMP-10 the end of the run-up event window. The event window is therefore t = [IMP-1, IMP+1], which includes the day prior to impairment announcement, the impairment announcement day and the day after.
Figure 2: The estimation window is the same for all event windows. Event window 1 is the acquisition announcement, event window 2 is the run-up period and event window 3 is the accounting impairment announcement. Since the run-up period is not the same for all companies, the impairment announcement day is defined as IMP and for example IMP-1 is the day before the impairment announcement.
3.3.3) Normal Return
The goal of an event study is to measure whether an event creates abnormal return. To calculate abnormal return, it is first necessary to calculate normal return. The definition of normal return is the expected return of the share prices of the company without conditioning on the analyzed event. The definition of normal return varies across models, this will be discussed in the next two sub-sections. Actual return less the expected normal return is defined as abnormal return. Equation 1 shows the calculation for abnormal returns:
Equation 1: Ai,t Ri,t E(Ri,t Xt)
8
Where Ai,t, Ri,t and E (Ri,t|Xt) are the abnormal return, actual return and the expected
return respectively. The E (Rit|Xt) is dependent on which model (Xt) is used to specify the
normal return of the security.
3.3.4) Market and Risk Adjusted Model
The Market and Risk Adjusted Model removes the part of the return which is the result of the variation in the market’s return. As a result the variance of the abnormal return decreases. Lower variation offers the benefit of predicting more accurate normal returns (MacKinlay, 1997).
Equation 2: Ai,t Ri,t i iRm,t
Rm,tis the return at time t for the MSCI world, i.e. the market return. Furthermore, αi, βi
are the OLS parameters which are calculated by regressing the company’s return against the market return over the estimation window.
8 The equations from part 3.3.3 up to and including part 3.4.1 are based on Brown and Warner (1980 and 1985)
-260 -10 0 1
Estimation Window Event Window 1 Event Window Event Window 3
10 IMP-10 IMP-1 IMP IMP+1
3.3.5) Market Adjusted Model
The market adjusted model calculates the abnormal return as the difference between the market return and the company’s stock return. By applying this model as well, it is analyzed whether the outcomes of the Market and Risk Adjusted Model are very sensitive to model changes. The expected abnormal return is calculated in Equation 3: Equation 3: Ai,t Ri,t Rm,t
3.3.6) Descriptive statistics
This part will show and discuss the descriptive statistics of the data sample. First, the mean, skewness, kurtosis and JarqueBera values over the estimation window t = [260, -10] are calculated for all companies separately. Second, over those four statistics the mean, median, minimum and maximum over all companies are calculated. In table 5 the results are showed.
Table 5: The descriptive statistics of the abnormal returns, calculated with the Market and Risk model, over the estimation period.
Mean Median Min Max
Mean 0,01% 0,00% 0,00% 0,48%
Skewness 0,71 0,76 1,78- 2,47
Kurtosis 5,59 5,12 0,80 16,93
Jarque-Bera 246,84 99,25 19,71 1.465,33
On average the companies have positive abnormal returns over the estimation window. The combination of positive average abnormal returns and a median average abnormal return close to zero, leads to skewed distribution with long tails on the positive abnormal returns side. Furthermore, the distribution is leptokurtic meaning that the distribution of the abnormal returns of the average company has fatter tails and is more peaked around the mean compared to normal distribution. The combination of both a skewed and leptokurtic distribution leads to higher Jarque-Bera scores than would be the case with normal distribution9.
Although Brown and Warner (1980, 1985) state that non-normality is not an issue in event studies, non-parametric test, which do not assume normal distribution, will test the
robustness of the parametric test applied in this paper. Both the parametric and the non-parametric will be discussed in the next sub-section.
3.4) Statistical Tests
3.4.1) Parametric Test
Abnormal returns alone do not prove anything. Statistical tests check whether abnormal returns are significantly different from zero. In order to test if the abnormal returns are significant, the abnormal returns are standardized. Standardization is done by dividing the abnormal returns by their standard deviation.
Equation 4: A'i,t Ai,t /S(Ai,t) Equation 5: ( ) 10( ) 250 260 2 * , ,
t i t i t i A A A S Equation 6:
10 260 , * 250 1 t t i i A AWhere A'i,t is the standardized abnormal return, S(Ai,t)is the standard deviation of
company i at time t and *
i
A is the average abnormal return over the estimation window t = [-260, -10]. For event window longer than one day A'i,t is calculated using equation 7.
Equation 7: 5 , 0 , 2 , , ( ) '
y x t t i y x t t i t i A S A AX and y, respectively, represent the beginning day and end day of the multiple day event window. The test statistic is calculated using equation 8.
Equation 8: T-test = 2 1 1 , ( ) '
t N i t i N A0,05) and 99% (probability < 0,01) confidence interval. Equation 8 is used to test hypotheses 1, 2 and 6.
3.4.2) Non-Parametric Test
Non-parametric tests do not make specific assumptions concerning the distribution of abnormal returns. Although Brown and Warner (1985) find that non-normality does not matter for the power of parametric tests in event studies, non-parametric test are applied as a robustness check in this paper. Two types of non-parametric tests will be used: the generalized sign test (Cowan et al. 1990) and the rank test (Corrado, 1989).
The generalized sign test of Cowan et al. (1990) compares the proportion of positive abnormal returns in the event window with the proportion of positive abnormal returns in the estimation period. The abnormal returns are calculated with the Market and Risk Adjusted Model and thus as in equation 1. In case the event window exists of more than one day, the abnormal returns are aggregated using equation 9. The cumulative average abnormal returns are calculated using equation 10.
Equation 9: CARi,(x,y)
, y x t t i A Equation 10: CAAR 1 1 ,( , ) y) (x,
N i y x i CAR NWhere CARi,(x,y) and CAAR(x,y) are the event window’s cumulative abnormal returns and
Equation 11:
N i t t i D N p 1 10 260 , 250 1 1 Equation 12: otherwise A if A if D t i t i t i , , , 0 0 1P is the proportion of sample returns on a random day that are expected to be negative. To calculate the general sign test z-value, equation 13 is used.
Equation 1310: 5 , 0 )]^ 1 ( [Np p Np w ZGeneralsign
Where w is the number of companies that have negative cumulative abnormal returns over the event window.
The Corrado rank test merely shows how the rank test can be used in situation where the event window is a single day. Cowan (1992) extends the Corrado rank test to event windows entailing more than one day by assuming independence of abnormal returns. The author finds that the model is powerful for event windows of one or two days but its power sharply declines when the event window increases. This is the reason why the rank test designed by Cowan (1992) is only used to check the robustness of hypotheses 1 and 2.
The abnormal returns of the estimation window (same as used for the parametric test) and event window are taken together and ranked. For the acquisition and impairment announcement this means that the complete period is ranked from 1 to 253 and 251, respectively, where rank 1 represents the largest abnormal return.
Equation 14: 2 1 260 ' 2 5 , 0 / ) (
x t t q Rank TRP R K R K q ZWhere Kq is the mean rank across q days in the event window and n companies. R is
the mean rank over the total ranking period, the mean rank is 127 for the acquisition
announcement and 126 for the impairment announcement. K is the average rank on day t
t across n companies in the ranking period. TRP are the total days included in the ranking period, i.e. the days of the estimation period plus the days of the event window. X’ is the beginning of the event window of either the acquisition announcement or the impairment announcement. The total ranking period for the acquisition announcement entails the three days of its event window plus the estimation window. The total ranking period for the impairment announcement entails the one day of its event window plus the estimation window.
3.4.3) Regression
For hypotheses 3 and 4, OLS regression analysis is used to conclude whether the expected relationship between variables exists. OLS assumes that residuals are uncorrelated and homoskedastic. Since in cross-sectional regressions there is no reason to assume that the residuals are homoskedastic, both unadjusted results and White’s heteroskedastic-consistent coefficient covariance are showed (MacKinlay, 1997). Autocorrelation is checked using the Durbin-Watson test. In figure 3 the overview is given from when autocorrelation exists. The critical values dL and du depend on both the
number of observations and the number of explanatory variables included in the regression, excluding the constant. The table presented in Brooks (2006, p. 674) is used to find the corresponding dL and du.
Figure 3: The schematic overview of the critical values of the Durbin-Watson test
Equation 15 shows the regression that is used to test hypotheses 3 and 4. Equation 15: Imp = α + β1 * Acq-R + β2 * Run-Up
0
Positive autocorrelation Inconclusive No evidence of autocorrelation Inconclusive Negative autocorrelation
Where Imp is the amount of goodwill impaired as percentage of market capitalization11. Acq-R and Run-Up are the abnormal returns of the acquisition and impairment announcement respectively. The regression will both be performed for the abnormal returns calculated using the Market Adjusted Model and the Market and Risk Adjusted Model.
3.4.4) Mann-Withney-Wilcoxon Test
Hypothesis 5 hypothesizes that recent goodwill will create significantly more negative abnormal returns on accounting impairment announcement compared to old goodwill. Since the abnormal returns of the impairment announcement sample are non-normal, a nonparametric method that does not rely on the assumption of normality is used.
H0: Abnormal returns at accounting impairment announcement are equal for Recent
and Old goodwill (Recent = Old)
Ha: Abnormal returns at accounting impairment announcement are more negative for
Recent compared to Old goodwill (Recent < Old)
To test this hypothesis certain steps must be followed to calculate the critical value. First, the two samples are placed together and are ranked. Second, the ranks are added for the two samples separately.
Equation 1612:
nx i x i x Rank R 1 ,X is either 1 or 2 representing the Recent or Old sample. Rx represents the sum of the
ranks. The value nx represents sample size. The z-value for the sample Recent (x=1) is
calculated using equation 17.
Equation 17: u u MWW s m R z ( 1 ) Equation 18: 2 ) 1 ( 1 2 1 n n n mu
11 Calculation of Imp is explained in sub-section 3.1
Equation 19: 12 ) 1 ( 1 2 2 1 nn n n su
4) Results
4.1) Acquisition announcement, run-up period and impairment announcement
In table 6 the results of testing hypothesis 1, 2 and 6 are presented. Both parametric and non-parametric results show that the acquisition announcement returns are significantly negative, thus the fact that the data was not normally distributed has not significantly affected the results. Although different mean returns for the Market model and the Market and Risk model are found, both are significantly negative. This implicates that investors on average recognize a value-destroying deal at acquisitions and, at least partly, economically impair the acquisition at announcement. To my knowledge this is a new finding based on a sample including acquisitions where accounting impairment follows. The finding demonstrates that on average investors are able to recognize overpayment upon acquisition. Although on average investors recognize overpayment, still 44-46% of the acquisition announcement returns are positive. This either means that investors did not recognize the value destructing nature of all acquisition or that the value destruction was only evident for some acquisitions. Furthermore, in some cases changes after acquisition could have led to unrealized synergies and value destruction. In that case positive acquisition returns are not surprising.
impairment announcement. This result leads to the acceptance of hypothesis 2. Note that especially the average 128% negative return compared to the market is an extreme number and it is not obvious that this is a mere result of one (large) acquisition. For future research it might be necessary to find a better proxy for the performance of the target after acquisition.
Table 6: Both the parametric and non-parametric test results of acquisition announcement returns, accounting impairment announcement returns and the run-up return
Acquisition Returns Run-up Goodwill Impairment Returns
Market & Risk
Market Market & Risk
Market Market & Risk
Market
N 62 62 62 62 62 62
Mean -0.58% -1.31% -128% -68% -4.65% -4.41%
T-test probability 0.04** 0.02** 0.00*** 0.00*** 0.00*** 0.00***
General sign probability 0.08* 0.04** 0.00*** 0.00*** 0.04** 0.05**
Rank test probability 0.08* 0.02** 0.00*** 0.01***
Positive 44% 46% 27% 19% 28% 28%
***, **, * indicates a confidence level of respectively 99%, 95% and 90%
From table 6 it is also shown that investors react significantly negative on impairment announcements. On average, stocks have a significant fall in stock price on the announcement day. The fall in stock price is much less than the average goodwill impairment amount as percentage of market capitalization of 51%. This implicates that the grand majority of the impairment does not surprise the investors. This result is in line with the results of Hirschey and Richardson (2002), they find an average fall of 3.5% on impairment announcement with an average impairment of 42% goodwill to market capitalization.
Hypotheses 1 and 2 showed that investors recognize value destruction before accounting impairment. The results of hypothesis 5 will provide more information about the cause of the significant fall in share prices on impairment announcement showed by the results of hypothesis 6.
4.2) Old versus Recent
that stock prices fall on impairment announcement simply because of the negative news effect, does not seem to hold. When the fall is purely caused by the negative news effect, on average no significant difference is expected between recent and old goodwill. The fact that recent goodwill responds significantly more negative to impairment announcements compared to old goodwill, implicates that the theory of Bartov et al. (1998) is favored over the negative news effect. Bartov et al. (1998) proposes, but do not test, that investors anticipate impairments but fail to fully incorporate the value information in a timely manner into the share prices. With these results it can at least be concluded that impairment announcement of recent goodwill contains significantly more unexpected negative information value than impairment announcements of old goodwill. With this result hypothesis 5 is accepted.
Table 7: Result of the Mann-Withney-Wilcoxon test between sub-samples Recent and Old on goodwill impairment announcement returns
Market & Risk Market
Recent Old Recent Old
N 39 23 39 23
Mean -6.8% -0.9% -6.5% -0.9%
MWW Probability 0.00*** 0.00***
***, **, * indicates a confidence level of respectively 99%, 95% and 90%.
4.3) Sensitivity analysis
Two sensitivity analyses are performed. The first is whether impairment announcement returns are influenced when accompanied earning announcements. In table 8a and 8b it is shown that although on average simple impairment announcements do result in more negative returns, they do not differ significantly from multi announcements.
The second sensitivity is to test whether impairment announcement returns are significantly affected by previous warnings. It is shown that although warnings on average even seem to lead to more negative impairment returns it is not significantly different from announcement returns without warning.
Both a warning before the actual announcement and simultaneous earnings announcement do not significantly influence the impairment announcement returns.
Table 8a: Results of the Mann-Withney-Wilcoxon test between sub-sample Single and Multi on goodwill impairment announcement returns
Market & Risk Market
Simple Multi Simple Multi
N 49 13 49 13
Mean -4.9% -3.6% -4.7% -3.1%
Table 8b: Results of the Mann-Withney-Wilcoxon test between sub-samples Warning And No Warning on goodwill impairment announcement returns
Market & Risk Market
Warning No Warning Warning No Warning N 6 56 6 56 Mean -6.3% -4.5% -5.6% -4.3% MWW Probability 0.29 0.33
***, **, * indicates a confidence level of respectively 99%, 95% and 90%.
4.4) Regression
In table 9 an overview of the OLS regression is given. Both the acquisition returns and the run-up returns are significantly and negatively related with the impairment amount. Consequently, when acquisition returns are more negative that the impairment amount is higher. Similar, in case run-up returns or more negative the impairment amount is higher. Only the acquisition returns using white’s heteroskedastic-consistent coefficient covariance is somewhat insignificant when considering the Market and Risk model. The fact that acquisition returns and the impairment amount are negatively related, indicates that investors recognize the scale of value destruction upon acquisition. It also indicates that for positive acquisition returns, i.e. for those acquisitions that seemed fairly good at acquisition, the ultimate impairment amount is lower compared to acquisition where value destruction was clear at acquisition.
Table 9: Results of regression: Imp = α + β1 * Acq-R + β2 * Run-Up
Market & Risk Market
β Probability (white’s adjustment probability) β Probability (white’s adjustment probability) Constant 0.45 0.00*** (0.00***) 0.32 0.00*** (0.00***) Acq-R -1.82 0.01*** (0.11) -1.69 0.01** (0.08*) Run-Up -0.04 0.02** (0.09*) -0.25 0.00*** (0.02**) Durbin-Watson 2.29 2.27 Adjusted R2 12.7% 20.3%
***, **, * indicates a confidence level of respectively 99%, 95% and 90%.
include much more information than merely investors’ economical impairment of an acquisition. In section 5.1 it was already noted that the negative returns over the run-up period seem too extreme to be caused merely by one value-destroying deal. The noise included in this proxy could have caused a low R2. Although the R2 is low, the acquisition returns and the run-up period are important variables in explaining impairment amount as they are significantly related to the impairment amount.
4.5) Future Research
In table 10 an overview of the outcomes for all hypotheses stated in part two is given. The results of this paper give several incentives for future research. First of all the fact that, although on average investors recognize value destructing deals at acquisition, approximately 45% of the acquisition in the sample leads to positive announcement returns. Previous research has showed several factors that cause negative acquisition announcement returns. As a result of this study it is maybe interesting to analyze which factors lead to an increased likelihood of impairment. Furthermore, this research proves that on average acquisition returns for companies that impair goodwill after acquisition are significantly negative. Comparing a control group of acquisitions without impairments with a sample that has announced impairment could show whether a significant difference exist between acquisition announcement returns of the two groups.
Table 10: An overview of the outcomes from all hypotheses
Hypotheses Result
H1: The acquisition announcement returns of the acquirer are negative Accepted H2: The abnormal returns over the run-up period are negative Accepted H3: The impairment amount is negatively related with abnormal returns at acquisition
announcement Accepted
H4: The impairment amount is negatively related with abnormal returns over the run-up period Accepted H5: Recent goodwill leads to more negative abnormal returns at accounting impairment
announcement than old goodwill Accepted H6: Abnormal returns at accounting impairment announcement are negative Accepted
Future research could also explore other possible explanatory variables of the amount of goodwill impaired.
5) Conclusion
This paper analyzes whether investors are able to recognize the value destruction related to an acquisition. Value destruction either occurs upon acquisition, by means of overpayment, or after the acquisition, because expected synergies do not materialize. In this paper it is hypothesized that investors recognize value destruction prior to accounting impairment.
A sample of 62 European goodwill impairments between 2001 and 2007 are used to analyze the period from acquisition to impairment of goodwill. The focus is on three periods, the period surrounding the acquisition announcement, the run-up period and the announcement of goodwill impairment.
The results show that economic impairment mainly occurs prior to accounting impairment. First, both the acquisition returns and the run-up returns of acquisitions that lead to goodwill impairments are significantly negative. Second, an impairment amount averaging 51% of the market capitalization merely leads to an average abnormal return of -4.65%. Both results show that economically impair covers the majority of the accounting impairment prior to the announcement of accounting impairment.
The abnormal returns at acquisition and in the run-up period are significantly related with the impairment amount. In case acquisition returns become more negative, i.e. investors believe the acquisition destroys more value, the total amount that is impaired increases. This indicates that investors are, on average, able to distinguish to what extent an acquisition destroys value prior to accounting impairment.
One of the results is that on average acquisition returns are significantly negative. However, approximately 45% of the acquisition returns is positive. Apparently for some acquisitions it is difficult to recognize value destruction at acquisition. On the other hand it is interesting that acquisitions with positive acquisition returns turn out to be less value-destroying, i.e. result in a lower amount of accounting impairment, compared to acquisitions with negative acquisition returns. This still indicates that investors are, on average able to recognize the magnitude of value destruction.
at finding more explanatory variables or improve the proxies used in the regression of this paper. On the basis of this research it is shown that acquisitions and run-up return form a significant base to explain the impairment amount.
The observation that stock prices significantly fall at accounting impairment announcement implicates that the announcement contains unexpected negative value information. This paper showed that negative announcement returns are not merely the consequence of the negative news effect. Impairment of recently acquired goodwill leads to significantly more negative returns compared to old goodwill. This implicates that relatively more negative information value is included in recently acquired goodwill. This finding supports, but does not prove, the proposed theory of Bartov et al. (1998) that investors fail to fully incorporate value information in a timely manner. Additional research is required to conclude what causes the more negative impairment announcement of recent goodwill.
For organizations such as the International Accounting Standards Board, the results of this paper could be the results support the change in the treatment of goodwill from annual amortization to annual impairment testing. This paper shows that old goodwill has lower value information for investors compared to recent goodwill. Annual impairment testing of goodwill ensures that goodwill is written down at the moment it cannot be recovered by the company.
For investors the results are interesting as they indicate that, although the majority of the accounting impairment amount is already reflected in the stock price, the impairment announcement on average still leads to negative stock returns. These negative returns show that the write down of goodwill contains value information about future cash flows, which holds to a stronger extent for recently bought goodwill.
impairment of more recently acquired goodwill results in more negative impairment announcement returns.
