Testing if winners underperform losers by using the novel
identification strategy of Malmendier, Moretti and Peters (2014),
with an increase in the long duration contested sample;
accomplished by increasing the number of mergers classified as
contested. Empirical study United States 1985-2011.
Master thesis
Msc Business Economics Specialization: Finance
Supervisor: Dhr. Vladimir Vladimirov
Robin de Blieck – 10282939
July 7, 2015
2 Statement of originality
This document is written by Robin de Blieck who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no other sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
In this paper the novel identification strategy of Malmendier, Moretti and Peters (2014) is used to estimate the long-run abnormal returns to mergers. They exploit bidding contests to compare the post-merger performance of winners and losers. In long duration contests the loser’s post-merger performance is used to construct the counterfactual performance of the winner had he not won the contest. This research uses a larger long duration contested sample than in the paper of Malmendier, Moretti and Peters (2014). First, this is done by increasing the number of mergers classified as contested. This is accomplished by using the new measure of takeover competition of Boone and Mulherin (2007). Furthermore, this is done by splitting the total sample of mergers into terciles and not in quartiles as in the paper Malmendier, Moretti and Peters (2014). This paper also tests whether there are measurable pre-existing differences between winners and losers that could bias the estimate of the causal effect of the merger. The pre-existing characteristics studied are book-to-market ratio, acquirer size and method of payment. Besides, this study tests whether merger-induced changes in leverage are linked to the acquirers’ post-merger underperformance. In this paper winners and losers returns are not closely aligned in the years before the contest. By controlling for acquirer size and leverage, differences between winners and losers are minimized in the pre-contest period. With the use of these control variables winners underperform losers by 17.22 percent in the post-merger period.
4
Table of contents
1. Introduction ...5
2. Literature review...8
2.1 Novel identification strategy Malmendier, Moretti and Peters (2014) ...8
2.2 New measure of takeover competition ...9
2.3 Comparison of findings of Malmendier, Moretti and Peters (2014) with empirical studies ... 10
3.Hypotheses ... 13
3.1 Main hypothesis ... 13
3.2 Other hypotheses ... 13
4.Data and descriptive statistics ... 15
3.1 Data collection ... 15 3.2 Summary Statistics ... 19 5.Methodology ... 20 6.Results ... 23 Hypothesis 1 ... 24 Hypothesis 2 ... 24 Hypothesis 3 ... 25 Hypothesis 4 ... 27 7. Robustness check ... 28 8.Conclusion ... 29 9.References ... 31 10.Appendix ... 32
5
1. Introduction
Does an acquisition improve the post-merger performance of the acquiring firm or do acquirers overbid and destroy shareholder value? The long-run underperformance of acquiring firms (Agrawal, Jaffe and Mandelker, 1992) and the negative announcement effects documented for a large number of U.S. mergers and acquisitions (Moeller, Schlingemann, and Stulz, 2005) seems to indicate overbidding.
A major problem in estimating the value created or destroyed is the difficulty of obtaining unbiased estimates. Completely unbiased estimates of the long-run
underperformance could be obtained in an ideal world, in which there is information about what has happened to companies that merger in the absence of the merger. Comparing this information to the post-merger performance results in estimates that exactly measure the causal effect of the merger. In the real world this information is not available and therefore other methodologies have to be applied. Such as the comparison of companies that have merged to companies that have not merged. This can result in unbiased estimates, because the merged companies are likely to have performed differently even in the absence of the merger. In their paper Malmendier, Moretti and Peters (2014) therefore say that acquiring firms are a selected group and engage in mergers at selected points in time. This makes it difficult to find a valid control group.
According to Malmendier, Moretti and Peters (2014) already established
methodologies do not succeed in obtaining unbiased estimates. Estimates based on changes in stock prices on the day of the announcement may be biased due to price pressure around mergers, information revealed in the merger bid, or market inefficiencies (Malmendier, Moretti and Peters, 2014). Also, estimates based on long-run abnormal returns could be biased due to unobserved differences between the firms that merge and those that do not (Malmendier, Moretti and Peters, 2014).
Malmendier, Moretti and Peters (2014) have tried to solve this by generating a novel approach to estimate the long-run abnormal returns after a merger. In their paper they use bidding contests in which at least two bidders have a similar chance of winning the contest. This is the case in long duration bidding contests. They use the post-merger performance of the loser to calculate the counterfactual performance of the winner had he not undertaken the merger. This new identification strategy allows to control for acquiring firms (winners) at
6
specific points in time that are hard to control for with the usual market-, industry-, and firm-level observable characteristics (Malmendier, Moretti and Peters, 2014).
In the paper of Malmendier, Moretti and Peters (2014) they only find data for 90 mergers with both winning bidders and losing bidders. This sample is split into quartiles in order to identify bidding contests with a long duration. This results in only about 20 mergers with a long contest duration. This is a small sample compared to other studies that
investigate the post-merger performance of the acquiring firms. In the paper of Agrawal, Jaffe and Mandelker (1992) they use a sample of 937 mergers to be able to investigate the post-merger performance of the acquiring firms. The main focus of this paper is therefore in increasing the long contest duration sample.
First, this is done by increasing the total sample of mergers classified as contested. This is accomplished by using the new measure of takeover competition of Boone and Mulherin (2007). In their paper they show that mergers into 1990s appear to be non-competitive as measured by number of bidders that publicly attempt to acquire the target. By incorporating pre-public, private bids they show that there actually was a highly
competitive takeover market.
In the study of Malmendier, Moretti and Peters (2014) they have only incorporated public and binding bids in their merger database. They do this, because in this way bidders are identified that are seriously interested in the acquisition and they are therefore more likely to be similar ex-ante. This is important for the novel identification strategy in order to estimate the causal effect of the merger. However, bidders in the private takeover process are also seriously interested in the target firm, but the competition takes only place before the public announcement. That they are seriously interested is apparent from the kind of target information they receive. The bidders contacted by the target firm that sign
confidential agreements receive non-public information about the firm (Hansen, 2001). If these bidders are not seriously interested they do not get this kind of information.
Otherwise the target firm risks misuse of the non-public information. Therefore bidders in the private takeover process should also be taken into account in classifying a merger as contested. This does not case a failure of the novel identification strategy of Malmendier, Moretti and Peters (2014). This makes it possible to include mergers in the merger database that only have a competition in the private takeover process.
7
splitting the total sample of mergers into terciles and not in quartiles as is done in the paper of Malmendier, Moretti and Peters (2014). According to Malmendier, Moretti and Peters (2014) alternative sample splits yield very similar results. It should therefore be possible to increase the long duration contested sample in this way.
However, it is unlikely that long duration contests minimize all the pre-existing differences between winners and losers in all the bidding contests. In addition, bidding contests could have a long duration for reasons unrelated to intensive competition. In their novel identification strategy Malmdendier, Moretti and Peters (2014) make the assumption that bidders know their own valuation of the target. In reality this should not be the case. If they do not know their own valuation it could take a lot of time for them to file an offer. The offers by different bidders could be totally different. In these contests the differences
between winning bidders and losing bidders are not minimized in the pre-contest period, even though there is a long contest duration. It is true, that the new identification strategy minimizes the differences between winners and losers generally. But, in specific contests this is not the case. These merger contests can bias the results of Malmdendier, Moretti and Peters (2014). This paper therefore controls for characteristics that according to prior literature are significantly associated with long-term post-merger performance. The
characteristics investigated are book-to-market ratio, acquirer size and method of payment. In addition, long duration contests cannot control for factors in the post-merger period that bias the causal effect of the merger. In their paper Malmdendier, Moretti and Peters (2014) they find evidence that merger-induced changes in leverage are linked to the acquirers’ post-merger underperformance. In their paper they find that winners have significantly higher leverage ratios than losers in the post-merger period. The market sees this excessive leverage as potentially harmful for the long-term health of the company. This can result in a post-merger underperformance of winners relative to losers not solely caused by the merger.
This paper proceeds as follows. Chapter 2 gives a guide throughout the most
important literature. Chapter 3 provides an description of the hypotheses investigated in this paper. In chapter 4 the data used is described and in chapter 5 the methodology is
presented. In chapter 6 the results of this paper are shown and in chapter 7 robustness checks are done to estimate the causal effect of the merger. In chapter 8 the conclusion of
8
this paper is formulated and in chapter 9 the references are presented. Chapter 10 contains the appendix with figures and tables.
2. Literature review
This chapter provides background information about the findings and the new identification strategy of Malmendier, Moretti and Peters (2014). In addition, the new measure of
takeover competition of Boone and Mulherin (2007) is discussed. Also, the findings of Malmendier, Moretti and Peters (2014) for the post-merger performance of the acquiring firm are compared to the findings in other empirical studies. First, the novel identification strategy of Malmendier, Moretti and Peters (2014) is discussed in more detail. Thereafter, the new measure of takeover competition of Boone and Mulherin (2007) is explained. After that the findings of Malmendier, Moretti and Peters (2014) are compared to the findings in other empirical studies.
2.1 Novel identification strategy of Malmendier, Moretti and Peters (2014)
In their new identification strategy they exploit bidding contests to compare the post-merger performance of winners and losers. They use contests where, ex-ante at least two bidders have a similar chance of winning. This is the case in long duration contests when both bidders have the same valuation for the target. In their paper they assume that bidders know their own valuation of the target, and that they have the same belief about the value of the target. If they have different valuations, the bidder with the higher valuation is the winner (Malmendier, Moretti and Peters, 2014). The contests duration is short, because it ends when the winner’s bid just exceed the valuation of the losing bidder. However, if both bidders have similar valuations for the target, the bidding contest is likely to take longer (Malmendier, Moretti and Peters, 2014). This is so, because more bidding rounds are needed until the bidder with the slightly lower valuation drops out of the contest. Malmendier, Moretti and Peters (2014) expect in these contests maximum similarity between winners and losers in the pre-contest period. Hence, losers are a valid counterfactual for winners in these contests.
Malmendier, Moretti and Peters (2014) tests this by comparing the valuation paths of winners and losers in the three years leading up to the bidding contests for contests with a different duration. According to them similar performance trends suggest that the market
9
expectations regarding the future performance of winners and losers are comparable. The under- or outperformance of the acquiring firm will be caused by the merger in this case. However, different trends in abnormal performance before the merger contest could lead to different performance after the merger not caused solely by the merger. Losers do not provide to be a good counterfactual in this case and the estimate of the causal effect of the merger is biased.
For the long duration contests Malmendier, Moretti and Peters (2014) indeed find that winners’ and losers’ abnormal returns closely track each other during the 20 months before the merger announcement. They therefore use in these contests the post-merger performance of the losers to calculate the counterfactual performance of the winners had they not undertaken the merger.
2.2 New measure of takeover competition
Malmendier, Moretti and Peters (2014) classify a merger as contested only on the number of bidders in the public takeover process. This in conjunction with the standard classification of a takeover auction (Boone and Mulherin, 2007). According to this classification a takeover auction is a takeover with more than one publicly announced bidder (Boone and Mulherin, 2007). Such auction terminology is used in empirical studies by Schwert (1996, 2000) and Moeller, Schlingemann and Stulz (2004).
In contrast Boone and Mulherin (2007) increases the number of mergers classified as contested by also taking the pre-public, private takeover process into account. In figure 1 the difference is illustrated between the private takeover process and the public takeover
process.
Figure 1: Timeline of the takeover process.
This figure shows a timeline of the takeover process. The private takeover process is the period from the private initiation of the takeover to the first public announcement of the takeover. The public takeover process is the period from the first public announcement of the takeover to the resolution of the takeover.
10
In the paper of Hansen (2001) a model of the private takeover auction is described. The private takeover process starts when the target firm or selling firm hires an investment banker (Hansen, 2001). The target firm and the investment banker together determine the number of potential bidders that should be contacted. The contacted bidders are asked to sign confidential agreements whereby the bidders receive non-public information (Hansen, 2001). The bidders that sign confidential agreements are asked to submit preliminary indications of interest. A part of the bidders signing these indications are asked to submit binding sealed offers from which the winning bidder is determined (Hansen, 2001). An example of this process is shown at the end of the appendix. It shows the sales process of Applied Signal Technology Inc.
To classify each takeover as either an auction or a negotiation, Boone and Mulherin (2007) focus on the potential buyers contacted and the potential buyers signing confidential agreements. According to Boone and Mulherin (2007) in an auction multiple bidders are contacted and sign confidential agreements. On the other hand, in an negotiation the selling firm deals with only a single bidder.
In order to obtain data for the number of bidders in the private takeover process for each takeover attempt in the sample, they review filings from the EDGAR system of the U.S. Securities and Exchange Commission (for takeovers in 1993 and later). Besides, they consult the LexisNexis and Laser disclosure database (for takeover in the 1993 and earlier). They obtain information on the details of the process for each for each takeover attempt from the background section of 14A and S-4 filings.
2.3 Comparison of findings of Malmendier, Moretti and Peters (2014) with other empirical
studies
The main finding of the paper of Malmendier, Moretti and Peters (2014) is that after the merger, losers significantly outperform winners. Depending on the measure of abnormal performance and definition of the contested merger, losers outperform winners by 25% to 50% over the three years following a closely contested merger. This evidence suggest that, for the sample of contested mergers, “winning means losing:” The shareholders of the acquiring company would be better off had their company lost the merger contests.
They also find that the underperformance of the winners does not depend on pre-merger differences between winners and losers. Differences between hostile and friendly acquisitions, variation in acquiring Q, the number of bidders, diversifying and concentrating
11
mergers, variation in target size or acquirer size, or differences in method of payment. This is in contrast with the papers of Rau and Vermaelen (1998), Moeller, Schlingeman, and Stulz (2004), Loughran and Vijh (1997).
In the study of Rau and Vermaelen (1998) they investigate the effect of the book-to-market before the merger on the post-merger performance. According to their paper there are two types of firms with different book-to-market ratios. There are ‘glamour’ firms and ‘value’ firms. Glamour firms are firms with high past stock returns and high past growth in cash flow and earnings (Rau and Vermaelen, 1998). These firms have low book-to-market ratios before the merger. According to the paper of Rau and Vermaelen (1998) the managers of these firms are more likely to overestimate their own abilities to manage an acquisition. This is due to the good past performance. Due to this, the board of directors and large shareholders are more likely to give the management the benefit of doubt and approve its acquisitions plans (Rau and Vermaelen, 1998).
On the other hand there are value firms. Value firms are firms with a bad past performance. These firms have high book-to-market ratios before the merger (Rau and Vermaelen, 1998). In these firms managers, directors and large shareholders are more critical before approving a merger transaction. If the merger is not a success it can decline the chance of survival of the company. The merger has therefore to add shareholder value which increases the survival of the company (Rau and Vermaelen, 1998).
These findings are in line with the performance extrapolation hypothesis according to the paper of Rau and Vermaelen (1998). This hypothesis means that the market only
gradually reassesses the quality of the bidder as the results of the merger become clear. Hence, glamour bidders have in the short-run higher returns than value bidders (Rau and Vermaelen, 1998). The short-run is the period around the announcement of the acquisition. In the long-run this reverses and value bidders have higher returns than glamour bidders.
The paper of Moeller, Schlingeman, and Stulz (2004) provides evidence for a poor post-merger performance for large acquirers. Their results indicate that small firms fare significantly better than large firms when they make an acquisition announcement. The abnormal returns associated with acquisition announcements for small firms exceeds the abnormal returns associated with acquisition announcements for large firms by 2.24
percentage points. Large firms experience losses when they announce acquisitions of public firms irrespective of the method of payment. Small firms gain significantly when they
12
announce an acquisition, except for acquisitions of public firms paid for with equity.
According to their findings acquisitions result in losses in the aggregate, because the losses incurred by the large firms are much larger than the gains realized by small firms.
Loughran and Vijh (1997) investigate the effect of the method of payment on the post-merger performance. They find a relationship between the post-merger returns and the mode of the merger and the form of the payment. During a five year period after the merger, acquirers stock returns are 61.7 percent higher than matching stock returns in cases where a cash tender offer is made (Loughran and Vijh, 1997 ). The acquirer stock returns are 25 percent smaller than matching stock returns for cases in which a stock merger offer is made (Loughran and Vijh, 1997 ).
In the paper of Asquith, Bruner and Mullins (1990) they also find the same
relationship between method of payment and underperformance. Their paper demonstrates that the form of merger financing affects the market‘s reaction to a merger announcement. The announcement returns are positive for cash bids and significantly smaller for stock financed bids. In their study the found that after correcting for stock financed mergers, the total returns are positive.
Malmendier, Moretti and Peters (2014) find that the observed underperformance of the acquirer in the post-merger period is not due to a higher offer premium and differences in operating performance. However, they find that merger-induced changes in leverage are linked to the acquirers’ post-merger underperformance. This is according to them in
particular the case in cash financed deals, where acquirers finance their mergers with debt. The market may see this excessive leverage as potentially harmful for the long term health of the company (Malmendier, Moretti and Peters, 2014).
That leverage has a negative effect on future performance is in conjunction with the empirical evidence of Penman, Richardson and Tuna (2007). In their paper they decompose the book-to-price ratio (B/P ratio). The book-to-price ratio can be decomposed into an enterprise book-to-price and a leverage component (Penman, Richardson and Tuna, 2007). The enterprise book-to-price component reflects operating risk. The leverage component reflects financing risk. In their analysis they find that the enterprise book-to-price ratio is positively related to subsequent stock returns. The leverage component of the B/P ratio is negatively associated with future stock returns (Penman, Richardson and Tuna, 2007).
13
3.Hypotheses
This chapter focuses on the formulation of the hypotheses. First the main hypothesis is discussed and thereafter the other hypotheses.
3.1 Main hypothesis
As stated in the introduction the paper of Malmendier, Moretti and Peters (2014) uses a small long duration contested sample. The long duration contested sample only contains about 20 mergers. This paper increases the long duration contested sample.
First, this is done by using the new takeover measure of Boone and Mulherin (2007). According to Malmendier, Moretti and Peters (2014) only public bidders are seriously interested in the merger. Therefore, according to them only in these bidding contests the differences between winning bidders and losing bidders are minimized in the pre-contest period. However, bidders in the private takeover process that sign confidential agreements receive non-public information about the target firm. This implies that these bidders are also seriously interested in the target firm. It is therefore possible to classify a merger as
contested based on the number of bidders in the private takeover process without damaging the new identification strategy of Malmendier, Morretti and Peters (2014).
Secondly, the long duration contested sample is increased by dividing the total sample of mergers into terciles. By doing this, the contest duration of the long duration sample decreases. This can possibly result in a failure of the new identification strategy. But, according to Malmendier, Moretti and Peters (2014) alternative sample sizes yields similar results. Therefore the main hypothesis is:
Main hypothesis: The causal effect of the merger is correctly estimated by the new identification strategy of Malmendier, Moretti and Peters (2014) when the long duration sample is increased.
3.2 Other hypotheses
The increase in the long duration contested sample does not cause a failure of the new identification strategy. Therefore losers should be a good counterfactual for winners. The first hypothesis therefore is:
14
In this paper about the same time period is investigated as in the study of Malmendier, Moretti and Peters (2014). In the paper of Malmendier, Moretti and Peters (2014) the time period studied is from January 1, 1985 to December 31, 2009. In this paper the time period investigated is from January 1, 1985 to December 31, 2011. According to the main
hypothesis the increase in the long duration contested sample does not causes a failure of the new identification strategy. Therefore the second hypothesis specified below should hold.
H2: Losers outperform winners in the post-merger period.
There are pre-existing differences between winners and losers that can bias the estimate of the causal effect of the merger. The pre-existing differences investigated in this paper are the book-to-market ratio, acquirer size and method of payment. However, according to Malmendier, Moretti and Peters (2014) there is no empirical evidence that pre-existing differences bias the estimates. According to the main hypothesis the increase in the long duration does not damage the new identification strategy. Therefore, the hypothesis specified below should hold.
H3: There are no measurable pre-existing differences between winners and losers that bias the estimates.
The long duration contest only minimizes the differences between winners and losers in the pre-contest period. After the merger there could be differences between winners and losers characteristics that influence post-merger performance. Difference in characteristics such as leverage. Therefore, the fourth hypothesis specified below should hold.
H4: Merger induced changes in market leverage are linked to the acquirers’ post-merger performance.
15
4.Data and descriptive statistics
This chapter describes how the data are collected in order to provide an answer to the hypotheses. The first section deals with the data collection and describes the databases. The second section shows the most important descriptive statistics.
3.1 Data collection
The merger data are collected from the Thomson One Mergers and Acquisitions database. The start date is set at January 1, 1985 and the end date is set at December 31, 2011. This is about the same period as in the paper of Malmendier, Moretti and Peters (2014) such that results are comparable. The acquirer nation (code) is set at the United States of America. The Form of the deal (Code) is set at Merger (Stock or assets) to eliminate all acquisitions that are not mergers. The deal status (code) is set at completed. This study eliminates mergers in which the acquirer or the target firm are financial or utility firms. To accomplish this mergers are eliminated where the acquirer or the target have a primary SIC code
starting with a 6 or 49. This is done, because these firms behave very differently than normal firms. The data items that are selected for the mergers in the Thomson One Mergers and Acquisitions database are the six-digit CUSIP code for the acquirer. This is the ACU code in the Thomson One database. Furthermore, data are collected for the competing bidders in the bidding contest. In Thomson One this is called COMPETEACU and it represents all the CUSIP codes of the bidders competing for the target. Also, data are collected for the number of bidders in a contests. This is variable called the BIDCOUNT in the Thomson One Mergers and Acquisitions database. In addition, data are collected for the public announcement date and the effective date. These variables are named DE and DA in Thomson One Mergers and Acquisitions database.
Two databases are generated: one without the competitive flag and one with the competitive flag. The one with the competitive flag is about the same sample as
Malmendier, Moretti and Peters (2014) have used in their study. These mergers are only classified as contested based on the number of bidders in the public takeover process. When this option is selected in the Thomson One Mergers and Acquisitions database this result in a merger database of 330 mergers. For all bids in this database is checked whether they are for the same target. This result in a loss of 8 bids that are not for the same target. Duplicate CUSIP codes are removed for each contested merger. This is done, because the interest is in
16
the firms that have bid for the target and not in the number of bids by the same firm for the same target.
The database without the competitive flag is used to search for contested mergers that only have an active competition in the private takeover process. When this option is not selected in the Thomson One Mergers and Acquisitions database a merger database is
generated of 21239 mergers. For the uncontested mergers1 in this database, data are
collected for the number of bidders in the private takeover process. These data are obtained from the S-4 and 14A EDGAR SEC files. The background section of these filings contain the data needed. These EDGAR SEC filings are available at the EDGAR SEC file archive database. The EDGAR SEC files are only available for the time period from 1994 to 2011. In this way 30 SEC files are collected of mergers that only have a competition in the private takeover process.
The SEC files are only collected for target firms, because only these files contain information about the number of bidders contacted. Also, data for the private date is collected. This is the date when the sales process begins. Unfortunately, no data for the losing firms is available in the EDGAR SEC filings. Only data are available for the winning firms. See for an example the sales process of Applied Signal Technology Inc. at the end of the appendix section. To obtain an indication of the post-merger performance of the losing firms after these contests data for similar firms in the same industry are collected. This is accomplished by matching the SIC codes of the losing firms in these contests to the SIC codes of a similar firm in the same industry. To classify a merger as contested the same definition is used as Boone and Mulherin (2007). According to them in an auction multiple bidders are contacted and sign confidential agreements. On the other hand, in an
negotiation the selling firm deals with a single bidder.
To collect the PERMCO codes of the bidders the CRSP translate to PERMCO/PERMNO tool is used. The CUSIP codes are used to obtain the PERMCO codes. This is done by
matching the first six digits of the CUSIP and the first six digits of the NCUSIP code in the PERMCO/PERMNO convertor. Some PERMCO codes where not found automatically by combining the databases. These PERMCO codes are manually collected by searching for the company names in the PERMCO/PERMNO database. In this way the PERMCO codes are
1
Classified as uncontested based on the number of bidders in the Thomson One Mergers and Acquisitions database. Thus, when the number of bidders in the public takeover process is equal to one.
17
collected for 44 bidders. The PERMCO codes are needed, because the CUSIP codes can change after the merger. The PERMCO codes remain unchanged.
To calculate the market-adjusted buy-and-hold return, stock data and benchmark data are collected. The database that is consulted for the stock data is the CRSP database. In the CRSP database the quarterly updated stock/security files are selected. After that, the monthly stock file is selected. The range date is set from January 1, 1981 and December 31, 2011. This timeframe is needed to be able to calculate the buy-and-hold return in the three years before- and in the three years after the contest. The data items that are selected are holding period return (RET) and price alternate (ALTPRC). The holding period return contains the change in value of common stock over a monthly time period. Price alternate is a
monthly price derived from daily prices. The price alternate is used to calculate the return over the period from t = 0 to t=1.
The benchmark data are also collected from the CRSP quarterly updated database. From stock/security files the stock market indexes option is selected. Data are collected for the value-weighted return including all distributions (VWRETD) and total market value (TOTVAL). The value-weighted return including all distributions (VWRETD) contains the monthly returns on a value-weighted market portfolio. The total market value (TOTVAL) exist of the total market value for a given market in $1000. The total market value is used to calculate the return over the period from t = 0 to t = 1.
To calculate the forecasted earnings-to-price ratio, earnings per share data are collected from the I/B/E/S database. In this database the summary history file option is selected. The forecasted period end date is set for the period January 1, 1981 and December 31, 2014. The two years ahead mean earnings per share forecast data are obtained by selecting fiscal year (2) and mean estimate. The monthly forecasted earnings-to-price ratio (FE/P) is calculated by dividing mean estimate of earnings per share forecast by monthly stock prices. For the monthly stock prices the price alternate (ALTPRC) is used.
The calculate the book-to-market ratio the following formula is used: Book-to-market ratio = (Book value of the firm)/(Market value of the firm)
The book value of the firm is calculated by multiplying the book value per share with the shares outstanding. The market value of the firm or the market capitalization is calculated by multiplying the share price with the number of shares outstanding.
18
The data to calculate the book-to-market ratio is collected from the COMPUSTAT database. In the COMPUSTAT database the COMPUSTAT monthly updates are selected and the North America data. The fundamental annually option is selected. Again, the data date is set from January 1, 1985 to December 31, 2011. The book-value per share (BKVLPS) is collected from this database. Data for the number of shares outstanding are collected from the CRSP monthly database. This is the same database as the monthly stock data. The number of shares outstanding is the variable SHROUT. SHROUT indicates the number of shares
outstanding in thousands. SHROUT is multiplied by book-value per share (BKVLPS) to get the book value of the firm. This value is divided by 1000 to obtain the book value of the firm in millions. The market value of the firm is calculated by multiplying the number of shares outstanding (SHROUT) by the price alternate (ALTPRC). Again, this value is divided by 1000 to obtain the market value of the firm in millions. The book value of the firm is divided by the market value of the firm to obtain an indication of the book-to-market ratio on a monthly
basis. The market
leverage ratio is calculated as follows:
Market leverage = (Total debt)/(Market value of the firm)
Total debt is calculated by summing short-term debt and long-term debt. The market value of the firm is calculated by summing total assets and market equity and deducting book equity. Market equity is calculated by multiplying the stock price with shares outstanding. The book value of equity is calculated by summing book value of shareholder equity and balance sheet deferred taxes and investment tax credit. After that, the book value of preferred stock is subtracted to get the book value of equity.
Data to be able to calculate the market leverage ratio is collected from the quarterly COMPUSTAT database. The data date is set from January 1, 1985 to December 31, 2011.To calculate the total debt, data for the short-term debt and long-term debt is collected. For short-term debt the variable current liabilities (LCTQ) is used. For long-term debt the variable long-term debt (DLTTQ) is collected. These two variables are summed to calculate the total debt. Both variables are in millions, so in this way the total debt in millions is obtained.
To calculate the market value of the firm the following data is collected. First, data are obtained for total assets (ATQ). This value is in millions. Thereafter, data for the
19
calculation of book value of equity are collected. The data needed is the book value of shareholder equity (SEQQ), balance sheet of deferred taxes (TXDITCQ) and the book value of preferred stock (PSTQ). Again, these variables are in millions. Market equity is calculated on a monthly basis. This is done by calculating the market capitalization. The market
capitalization is computed by multiplying the number of shares outstanding (SHROUT) by the price alternate (ALTPRC). This value has to be divided by 1000 to get the value in millions, because SHROUT is the number of shares outstanding in thousands. The calculation of the market equity value in this way gives a good indication of the market value of leverage on a monthly basis.
To see if large acquirers underperform small acquirers data for size are needed. The market capitalization is used as a measure of the size. The same as in the paper of
Malmendier, Moretti and Peters (2014). Again, market capitalization is computed by multiplying the number of share outstanding (SHROUT) with the price alternate (ALTPRC). Again, this value is divided by 1000 to get market capitalization in millions.
Data that are needed for the calculation of the method of payment is collected from the Thomson One Mergers and Acquisitions database. The following variables are collected percentage cash (PCT_Cash). This is the percentage of the transaction value offered in cash. Data for (PCT_Stock). This is the percentage of the transaction value offered in stock. Data for (PCT_OTHER). This is the percentage of the transaction value not offered in cash and stock. These variables are used to distinguish between all-cash financed merger, all-stock financed mergers and mixed financed mergers.
3.2 Summary Statistics
To identify long duration contests the total merger sample is split into three subsamples. This is a different strategy than in the paper of Malmendier, Moretti and Peters (2014). In their paper they split the sample into quartiles. But, this results in a small sample of only 20 merger contests for the long duration contest subsample. This database is too small to obtain meaningful results.
In this paper the long duration subsample is increased. This is partly accomplished by splitting the total merger contest sample into three parts or terciles. However, by splitting the sample into three parts the contest duration in the long duration subsample is decreased compared to a split into quartiles. As is done in the paper of Malmendier, Moretti and Peters
20
(2014). In table 1.1 the percentile values of contest duration for the total merger sample are presented. Contest duration is measured on a monthly basis in this table. The long duration tercile has a duration of 6.2 months or higher. The second tercile has a contest duration between 3 months and 6.2 months. The first tercile has a contest duration smaller than 3 months. In the paper of Malmendier, Moretti and Peters (2014) their merger contests last two to four months in the first quartile; those in the second quartile last five to seven months; those in the third quartile last eight to eleven months; and those in the fourth quartile last one year or more. The contest duration in the long duration subsample in this research is smaller than in the paper of Malmendier, Moretti and Peters (2014). This has probably a negative effect on the validity of their new identification strategy.
Table 1.2 shows the number of mergers in the total sample and in the long duration contests subsample. The total sample consist of 80 mergers. Of this total sample 28 mergers have a long duration contest.
The the amount of mergers which have only a competition in the private takeover process is 10 in the total merger database. The other mergers classified as contested in this way (19) are lost due to difficulties in obtaining the stock data for the +/- three-year event time window.
5.Methodology
In this chapter the methodology is described to investigate the hypotheses.
To investigate if the findings of the new identification strategy are valid with an increase in the merger database the following methodology is applied.
To investigate the differences in abnormal performance between winners and losers in the three years before (hypothesis 1) and the three years after the merger contest
(hypothesis 2) a controlled regression framework is used. In this framework the buy-and-hold returns (BHARs) are calculated for each bidder for each month in the three years before- and after the merger contest.
To calculate the buy-and-hold return an event time t variable is constructed that counts the months relative to each contest. Time t = 0 indicates the end of the month preceding start of the contest. The end of the month before that is set to t = -1 etc. Time t = 1 indicates the end of the month in which the merger is completed. The contest start date is
21
the announcement date for the mergers classified as contested in the Thomson One Mergers and Acquisitions database. The contest ends at the effective date. See figure 1a. The contest start date is the private date for mergers that only have an competition in the private takeover process. Again, the contest end date is set at the effective date. See figure 1b.
The buy-and-hold return is calculated as the difference between the cumulated bidder stock return and the cumulated benchmark return (Malmendier, Moretti and Peters, 2014). The return is normalized to zero in the month preceding the start of the contest.
The formula for the buy-and-hold return after the merger is as follows: (1)
Where i denotes the bidder, j the bidding contest, t and s index event time, rijs is the stock
return of the bidding firm earned in event period s and rijs (bm) is the benchmark return in
event period s (Malmendier, Moretti and Peters, 2014). S = 1 indicates that the contest starts at t = 1. The buy-and-hold return is calculated until month t = 36.
The equation of the buy-and-hold return before the merger is:
(2)
S = 0 indicates that the contest starts at time t = 0. The buy-and-hold return is calculated until month t = -35.
To investigate if losers are a valid counterfactual for winners (hypothesis 1) the pre-merger contest similarity between winners and losers is compared. This is done by using data for stock prices and analyst expectations. First, this is investigated by comparing the average abnormal stock returns of winners and losers in the three years leading up to the merger contest. The average pre-merger performance trend of the winners should be similar to the pre-merger performance trend of losers. According to Malmendier, Moretti and Peters (2014) similar trends prior to the merger suggest that market expectations regarding the future performance of winners and losers are comparable. However, different trends
22
possibly indicate differences in characteristics that might affect performance even without
the merger (Malmendier, Moretti and Peters, 2014).
Besides this is tested by comparing forecasted earnings-to-price ratios winners and losers. Similar forecasts in the three year period before the merger contest implies the same analyst expectations regarding the future performance of winners and losers (Malmendier, Moretti and Peters, 2014).
To statistically test if winners underperform relative to losers (hypothesis 2) the following equation is used:
(3)
This is a more advanced difference-in-difference regression that account for linear time trends in the pre-contest and post-merger period. The variable W is a dummy variable that indicates whether bidder i is a winner in merger contest j. If it is a winner it has the value 1. Period (t) is a variable counting event time. Thus, the time period from t = -35 to t= 36. The variable Post is a dummy variable that indicates whether period t is in the post-merger period. It has the value 1 if it is in the post-merger period. This is the period from t=1 to t=36. The variable n indicates contests fixed effects. This controls for differences between
contests.
In this specification the coefficient a0 measures the average performance level of
losers before the contest. The coefficient a4 shows the difference in average performance for
losers in the post-merger period relative to the average performance for losers in the
pre-merger contest period. The coefficient a1 shows the average performance difference
between winners and losers in the pre-contest period. The coefficient a5 measures the
average performance difference of winners from the pre-contest period to the post-merger
period relative to average performance shift in outcome for the losers. The coefficient a2
measures the linear time trend of losers in the pre-contest period and the coefficient a6 is
the difference in trend for losers in the post-merger and pre-contest period. The coefficient
a3 shows the difference in the linear time trend for winners and losers in the pre-contest
period and the coefficient a7 measures the difference in trend between winners and losers in
23
The value effect of the mergers is estimated as the long-run performance difference
between winners and losers at t = 36. To tests this the following specification is used a1 + a5 +
35*(a3 + a7).
Thereafter, this paper continues into whether there are measurable pre-existing differences between winners and losers that bias the estimates of the causal effect of the merger. If there are big differences between losers and winners in the pre-contest period, the comparison of winners and losers in the post-merger period is difficult. The pre-existing differences that are investigated are book-to-market ratio, acquirer size and method of payment. This is investigated by checking if any of such characteristics are correlated with post-merger performance and if they constitutes a large fraction of the long duration contested subsample. They have an effect if they are correlated with post-merger performance and if such characteristics are concentrated in long-duration contests (Malmendier, Moretti and Peters, 2014).
To investigate this the total sample of mergers is split in two samples. For book-to-market ratio the total sample is split into acquirers with a low book-to-book-to-market ratio and a high book-to-market ratio. For acquirer size the total sample is split into acquirers with a large size and acquirers with a small size. For method of payment the total sample is split into cash financed mergers and stock financed mergers. Graphs are generated that show the market adjusted buy-and-hold returns for the winners and the losers for the two
subsamples. Again, these graphs show the market adjusted buy-and-hold returns (BHARS) for the +/- three-year event window around the merger contests.
6.Results
This part of the paper shows the results for the hypotheses specified in the hypothesis section. The main sample of interest is the long duration subsample. First, is tested whether losers are a valid counterfactual for winners. Secondly, is tested whether losers outperform winners. Thereafter, is tested whether there are measurable pre-existing difference
between winners and losers that bias the estimates. After that, test are done to investigate whether merger induced changes in market leverage are linked to the acquirers’ post-merger performance. The figures and tables of the results are available at the appendix at the end of this paper.
24
Hypothesis 1: Losers are a valid counterfactual for winners
In figure 2a the graphical results are presented for the abnormal stock returns (BHARs) for the months in the three-year period before the merger contest. The sample studied is the long duration subsample. Until month 20 winners and losers have the same performance trend. After month 20 the performance trend between winners and losers is totally different. Losers have a steep declining abnormal performance trend and winners have a
small declining abnormal performance trend. The coefficient a3 which measures the
difference in the linear trend between winners and losers in the pre-contest period is positive and significant at the 10 percent level. See table 2. This implies that winners significantly outperform losers in the pre-contest period. This is in conjunction with the visual evidence in figure 2a. This causes a failure of the new identification strategy of Malmendier, Moretti and Peters (2014), because winners and losers should have the same performance trend in the pre-contest period to unbiasedly estimated the causal effect of the merger.
Figure 3 shows the graphical results for the average forecasted earnings-to-price ratios (FE/P ratios) for the subsample of long contests duration and total sample of mergers. In the long duration contest subsample the winners and the losers do not closely track each other in the 36 months leading up to the merger fight, this suggest that analyst were
expecting a different performance for winners and losers. However, in the total sample winners and losers closely track each other in the 36 months leading to the bidding contest. This suggest that analyst were expecting a similar performance for winners and losers. This is in contrast with the findings of Malmendier, Moretti and Peters (2014). In the long duration contest subsample winners and losers have the same forecasted earnings-to-price ratio (FE/P) trend in the pre-contest period. In the total sample winners and losers have a
different trend. The results of this section indicates that losers are not a valid counterfactual for winners in the long duration subsample. This is not in conjunction with hypothesis 1. Hypothesis 2: Losers outperform winners in the post-merger period
Figure 2a presents graphically the results for the post-merger performance for winners and losers in the long contest duration subsample. The figure shows that after month 10 winners underperform losers. Winners display negative abnormal performance and a downward trend throughout the post-merger period. In contrast losers have a positive abnormal performance and a flat trend.
25
In the unadjusted buy-and-hold return graph winners underperform losers after month 10. However, winners display a positive abnormal performance and a flat trend in the post-merger period. Again, losers have a positive abnormal performance, but they have an
increasing trend in the post-merger period. The underperformance of the winners relative to losers does not seem to be related to the choice of the benchmark. This seems to suggest that for the long duration subsample ‘’winning means losing’’. This implies that the shareholders of the acquiring company would be better off had their company lost the merger contests.
Table 2 presents the results for the coefficient estimates for regression (3). The subsample of interest is the long duration tercile (T3), but the results for the two other terciles are also presented. The effect on the acquiring firm’s long-run abnormal
performance is reported in the lower part of the table and labelled ‘Merger effect’. Winners underperform losers by 17.80 percent in the post-merger period. The effect is statistically significant at the 1 percent level. Thus, winners fare significantly worse than losers in the long duration sample. This is in conjunction with hypothesis 2. However, as already noted
the coefficient a3 is statistically significant at the 10 percent level. This implies that it is
difficult to infer from the losers the counterfactual performance for the winners.
Therefore, control variables have to be added to the regression in order to control for the differences between winners and losers in the pre-contest period. The other terciles T2 and T1 also have significant differences in abnormal performance in the post-merger period. In T2 the effect is positive and significant at the 1 percent level. In T1 the effect is negative and significant at the 5 percent level.
Hypothesis 3: There are no measurable pre-existing differences between winners and losers that bias the estimates.
This section shows the results of whether the pre-existing differences book-to-market ratio, acquirer size and method of payment explain the difference in post-merger performance between winners and losers. If they do, winners and losers differences could be attributed to these characteristics . This can results in biased estimates of the causal effect of the merger. As already told in the methodology section, these pre-existing differences explain the post-merger performance difference if these characteristics were correlated with long-term performance and if they are a large fraction of the long duration subsample. In order to investigate this the total merger sample is split in two subsamples.
26
Figure 4 shows graphically the results of the market adjusted buy-and-hold returns (BHARs) for the sample split based on the acquirers’ value of the book-to-market ratio before the merger contests. In the high book-to-market ratio graph winners underperform losers after month 10. But, they differences in post-merger performance are small. In the low book-to-market ratio graph winners also underperform losers after month 10. The
underperformance is greater than in the high book-to-market ratio graph. In table 3 the results are presented for the coefficients of regression (3). The estimate of the causal effect of the merger for the high book-to-market ratio subsample is -0.952. This implies that winners underperform losers by 0.952 percent in the post-merger period. In the low book-to-market ratio subsample winners underperform losers by 12.63 percent. These findings are in conjunction with the graphical evidence in figure 4. This is in conjunction with the findings of Rau and Vermaelen (1998). According to their findings acquirers’ with a low book-to-market ratio underperform acquirers’ with a high book-book-to-market ratio in the
post-merger period. However, the number of post-mergers in which the acquirer has a low book-to-market ratio in the long duration subsample is only small. See table 4. It is therefore unlikely that they explain the underperformance of the winners relative to losers in this subsample.
In figure 5 the graphical results are presented of the market adjusted buy-and-hold returns (BHARs) for the sample split based on acquirer size. In the large acquirer size graph the differences between winners and losers in the post-merger period are great. Winners underperform losers over the entire three year period after the merger. Also, in the small acquirer size graph the differences between winners and losers are huge. However, losers underperform winners over the entire three year period after the merger. Again, in table 3 the results are presented for regression coefficient of regression (3). The causal effect estimate of the merger is -18.70. This implies that winners underperform losers by 18.70 percent in the postmerger period. The coefficient for the small acquirer size sample is -3.334.
This provides evidence that large acquirers underperform small acquirers in the post-merger performance. This is in line with the empirical evidence of Moeller, Schlingeman, and Stulz (2004). Also, the number of mergers in which the acquirer has a large size is about one-third of the long duration subsample. See table 4. They could therefore have a significant effect on post-merger performance differences between winners and losers.
27
In figure 6 the graphical results are presented for all-cash financed mergers and stock financed mergers. In the all-cash financed mergers subsample acquirers have a small negative abnormal stock return while in all-stock financed mergers subsample acquirers have a large negative abnormal stock return. However, there are no big differences between winners and losers in both graphs. It is therefore unlikely that method of payment
differences influence the post-merger underperformance of winners relative to losers. In table 3 all-cash financed mergers have a coefficient of 5.326 and all-stock financed mergers have a coefficient of 10.65. This means that winners outperform losers by 5.326 percent and 10.65 percent respectively. This is not in conjunction with the graphical evidence in figure 6.
The conclusion from this section is that differences in acquirer size can bias the estimate of the causal effect of the merger. This in contrast with hypothesis 3.
Hypothesis 4: Merger induced changes in leverage are linked to the acquirers’ post-merger performance.
This section studies if merger induced changes in leverage are linked to the acquirers underperformance in the post-merger period. In figure 7 the graphical results are presented for the total sample split based on acquirers with a great change in leverage and acquirers with a small change. There are great differences between the two graphs. In the graph for acquirers with a large leverage change winners underperform losers in the post-merger period. In contrast winners outperform losers in the post-merger period in the graph for acquirers with a small change in leverage. The coefficient estimates of regression (3) show that acquirers with a large change in market leverage underperform losers by 22.89 percent in the post-merger period. On the other hand acquirers with a small change in market leverage outperform losers by 15.04 percent in the post-merger period. However, mergers in which the acquirer has a great change in market leverage represents only a small fraction of the long duration subsample. See table 4.
Thus, there seems to be a link between merger induced changes in leverage and the acquirers’ post-merger performance. This is conjunction with hypothesis 4.
28
7. Robustness check
In order to evaluate the robustness of the estimate of the causal effect of the merger, regressions are done with the addition of some control variables. In these regressions there is controlled for pre-existing difference between winners and losers. The pre-existing differences that are controlled for are book-to-market ratio, acquirer size and method of payment. Besides, there is controlled for leverage to test whether merger induced changes in leverage have a significant effect on post-merger underperformance of the acquiring firm relative to losers. The results are presented in table 5. In these regressions all the estimates of the causal effect of the merger are significant at the 1 percent level.
In regression 1 the control variable book-to-market ratio is added to the difference-in-difference regression. The coefficient that measure the effect of the merger on the acquiring firm’s long-run abnormal performance is -23.50. This implies that winners underperform losers by 23.50 percent in the post-merger period. This variable is labelled
‘Merger effect’ at the bottom of table 5. The coefficient a3 is positive and statistically
significant at the 5 percent level. There are therefore significant differences between
winners and losers in the pre-contest period. The estimate of the causal effect of the merger could therefore be biased, because losers do not provide to be a good counterfactual for winners.
Regression 2 controls for differences in acquirer size. The coefficient that measures the percentage underperformance of the winners relative to the losers is 17.21. The
coefficient a3 is positive and not statistically significant. This implies that there are no
significant differences between winners and losers in the pre-contest period and therefore the estimate of the causal effect of the merger is not biased. Thus, losers are a good counterfactual for winners when is controlled for acquirer size.
In regression 3 is controlled for differences in method of payment. The coefficient of the ‘Effect Merger’ is -17.80. This means that winners underperform losers by 17.80 percent
in the post-merger period. Again, the coefficient a3 is positive and statistically significant.
This implies that losers are not a good counterfactual for winners.
Regression 4 controls for leverage. The coefficient that measures the percentage
29
positive and not significant. Thus, when is controlled for leverage losers are a good counterfactual for winners.
In regression 5 there is controlled for the pre-existing differences book-to-market ratio, acquirer size and method of payment. The coefficient that measures the effect of the merger on the acquiring firm’s long-run abnormal performance is -22.59. This implies that
winners underperform losers by 22.59 percent in the post-merger period. The coefficient a3
is positive and significant. This means that losers are not a good counterfactual for winners when is controlled for pre-existing differences.
Regression 6 controls for the pre-existing differences book-to-market ratio, acquirer size and leverage. It also controls for market leverage. The coefficient that measures the causal effect of the merger is -22.59. This is the same value as in regression 5. Again, the
coefficient a3 is positive and statistically significant. Therefore, the losers are not a good
counterfactual for the winners in this regression.
In regression 7 there is controlled for acquirer size and leverage. The coefficient that measures the percentage underperformance of winners relative to losers is 17.22. The
coefficient a3 is positive and not significant. Thus, losers are a good counterfactual for
winners when is controlled for acquirer size and leverage.
This means that the control variables acquirer size and leverage have to be added to the regression in order to get unbiased estimates of the causal effect of the merger.
8.Conclusion
The main focus of this paper is in estimating the causal effect of the merger by increasing the long duration contested sample. This is accomplished by using the new identification
strategy of Malmendier, Moretti and Peters (2014). In this subsample winners underperform losers by 17.80 percent in the post-merger period. However, losers are not a good
counterfactual for winners in this subsample. First, the coefficient a3 is positive and
significant at the 10 percent level. Also, winners and losers have a different performance trend in the forecasted earnings-to-price ratio (FE/P) graph. Therefore, the estimate of the causal effect of the merger is biased.
In order to obtain unbiased estimates, this paper also investigated whether there are pre-existing differences between winners and losers that bias the causal effect of the
30
merger. The differences between them in the post-merger period are only small for acquirers with a low- or high value of the book-to-market ratio. Therefore, differences in book-to-market ratio for acquirers do not bias the estimate of the causal effect.
However, there are large differences between winners and losers for acquirers with a different size. Winning bidders with a large size underperform the losers in the post-merger period. In contrast small winning bidders outperform the losers in the post-merger period. Furthermore, when there is controlled for acquirer size there are no significant differences in linear trend between winners and losers in the pre-contest period.
By contrast, there are no large differences between winners and losers for different ways of financing a merger. Therefore, differences in method of payment cannot bias the estimate of the causal effect of the merger.
There are large differences between winners and losers in the post-merger period for large- and small merger induced changes in leverage. Besides, when is controlled for
leverage there are no significant differences in the linear trend between winners and losers in the pre-contest period.
Therefore, the control variables acquirer size and leverage have to be added to the regression in order to get unbiased estimates of the causal effect of the merger. When there is controlled for those variables winners underperform losers by 17.22 percent. This
evidence suggest that, for the sample of contested mergers , “Winning means losing”. The shareholders of the acquiring company would be better off had their company lost the merger contests.
A constraint of this study is that the EDGAR SEC files are only available for the time period from 1994 to 2015. Reduction of the study to this time period results in a small of mergers with only a public takeover competition. At a later point in the future it should be possible to increase the database with merger with only a competition in the private takeover process for the entire time period under study.
Another problem is the difficulty to obtain the losing bidder names in the EDGAR SEC files. It is therefore hard to increase the merger database without impairing the new
identification strategy of Malmendier, Moretti and Peters (20140. In further research this issue can possibly be solved.
31
9.References
Agrawal, A., Jaffe, J. F. & Mandelker G. N. (1992) ‘The Post-Merger Performance of Acquiring Firms: A Re-examination of an Anomally ’, The Journal of Finance, vol. 47, no. 4, pp. 1605 - 1621
Asquith, P., Bruner R. F. & Mullins D. W. (1987) ‘Merger returns and the form of financing’, Journal of Financial Economics, vol. 11, pp. 121-139.
Boone, A.L. & Mulherin, J.H. (2007) ‘How Are Firms Sold?’, Journal of Finance, vol. 62, no. 2, pp. 847-875.
Hansen R.G. (2001) ‘Auctions of companies’, Economic Inquiry, vol. 39, no. 1, pp. 30-43
Loughran, T. & Vijh, A.M. (1997) ‘Do long-term shareholders benefit from corporate acquisitions?’, Journal of Finance, vol. 52, no. 5, pp. 1765-1790.
Malmendier, U., Moretti, E. &Peters, F. (2012) ‘Winning by Losing: Evidence on the Long-run Effects of Mergers’, National Bureau Of Economic Research, no. 18024.
Moeller, Sara B., Schlingemann F. P. & Stulz R. M. (2004) ‘Firm size and the gains from acquisitions’, Journal of Financial Economics, vol. 73, 201-228.
Moeller, Sara B., Schlingemann F. P. & Stulz R. M. (2005) ‘Wealth Destruction on a Massive Scale? A Study of Acquiring-Firm Returns in the Recent Merger Wave ’, The Journal of Finance, vol. 60, no. 2, pp. 757-782.
Penman, Stephen H., Richardson S. & Tuna I. (2007) ‘The book-to-price effect in stock returns: Accounting for leverage’, Journal of Accounting Research, vol. 45, 427-467.
Rau, P.R. & Vermaelen T. (1998) ‘Glamour, value, and the post-acquisition performance of acquiring firms’, Journal of Financial Economics, vol. 49, no. 2, pp. 223-253.
Schwert, G. W. (1996) ‘Markup pricing in mergers and acquisitions’, Journal of Financial Economics, vol. 41, no. 2, pp. 153–192.
Schwert, G. W. (2000) ‘Hostility in takeovers: In the eyes of the beholder?’ Journal of Finance, vol. 55, pp. 2599–2640.
32
10. Appendix
Figure 1: Construction of event time
These figures show the construction of event time for the +/- three-year event window around merger contests. In figure a the event time construction for the sample of firms with a public takeover competition is presented. In figure b the event time construction for firms with only a private takeover
competition is shown.
a: Event time construction for sample of firms with public takeover competition
33 -5 0 -4 0 -3 0 -2 0 -1 0 0 10 20 30 40 50 BH AR -40 -20 0 20 40 period Winners Losers -5 0 -4 0 -3 0 -2 0 -1 0 0 10 20 30 40 50 BH AR -40 -20 0 20 40 period Winners Losers Figures 2
These figures show the results of the buy-and-hold abnormal returns (BHARs) for each month in the +/- three-year event window around the merger contests. Figure a shows the market adjusted buy-and-hold return. The market adjusted buy-and-hold return is calculated as the cumulative difference between the holding period return (RET) and the return on a value weighted index (VWRETD). Figure b shows the unadjusted buy-and-hold return. The unadjusted buy-and-hold return is determined by computing the cumulative holding period return (RET), but the benchmark is set to zero.
34 0 .0 2 .0 4 .0 6 .0 8 .1 .1 2 .1 4 F o re ca st e d Ea rn in g s-t o -Pri ce R a ti o -40 -20 0 20 40 period Winners Losers 0 .0 2 .0 4 .0 6 .0 8 .1 .1 2 .1 4 F o re ca st e d Ea rn in g s-t o -p ri ce R a ti o -40 -20 0 20 40 period Winners Losers Figure 3
These figures show the average forecasted earnings-to-price ratio (FE/P) of winners and losers for each month in the +/- three-year event window around merger contests. Figure a shows the results for the long contest duration tercile. Figure b shows the results for the total sample of 80 mergers.
a)Long contest duration tercile b) Total sample
35 -5 0 -4 0 -3 0 -2 0 -1 0 0 10 20 30 40 50 BH AR -40 -20 0 20 40 period Winners Losers -5 0 -4 0 -3 0 -2 0 -1 0 0 10 20 30 40 50 BH AR -40 -20 0 20 40 period Winners Losers Figure 4
These figures show the results of the market adjusted buy-and-hold abnormal returns (BHARs) for each month in the +/- three-year event window around the merger contests. The market adjusted buy-and-hold return is calculated as the cumulative difference between the holding period return (RET) and the return on a value weighted index (VWRETD). In figure a the market adjusted buy-and-hold returns are shown for acquiring firms with a high book-to-market ratio before the merger contest. Figure b shows the market adjusted buy-and-hold returns for acquiring firms with a low book-to-market ratio before the merger contest.
36 -5 0 -4 0 -3 0 -2 0 -1 0 0 10 20 30 40 50 BH AR -40 -20 0 20 40 period Winners Losers -5 0 -4 0 -3 0 -2 0 -1 0 0 10 20 30 40 50 BH AR -40 -20 0 20 40 period Winners Losers Figure 5
These figures show the results of the market adjusted buy-and-hold abnormal returns (BHARs) for each month in the +/- three-year event window around the merger contests. The market adjusted buy-and-hold return is calculated as the cumulative difference between the holding period return (RET) and the return on a value weighted index (VWRETD). In figure a the market adjusted buy-and-hold returns are shown for acquiring firms with a large size before the merger contest. Figure b shows the market adjusted buy-and-hold returns for acquiring firms with a small size before the merger contest.
