Universal Investors and M&A Peers Ben Branch, Ning Pu 11th December 2012
Abstract
Institutional investors need to decide whether to favor or oppose a proposed acquisition. Large institutional investors such as pension funds may own the stocks and bonds of both target and bidder companies as well as investments in their peer companies, such as other potential bidders, customers, suppliers and rivals. In this paper we explore how the portfolio of a widely diversi ed investor may be aected by a proposed acquisition. Based on short-horizon and long-horizon analysis, we found takeovers tend to create value in the wealth of stockholders and bondholders of the peer companies. Our results are largely supportive of the e ciency argument that suggests mergers and acquisitions are driven by perceived e ciency increase. This study contributes to the existing literature on M&A since it is the rst empirical study exploring takeovers' impact on peer- rm bond abnormal returns. Consider the following example of a proposed takeove's wider impact: On August 15, 2011, Google Inc. (NASDAQ:GOOG) and Motorola Mobility Holding, Inc. (NYSE:MMI) announced that they had entered into a de nitive agreement under which Google will acquire Motorola for $12.5 billion, or $40 per share by cash, with a premium equaling 63% markup to the closing price of Motorola shares on the previous trading day. As a result, Motorola's stock price jumped more than 53% on a single day responding to the announcement.
A series of dramatic price reactions also
occurred to several peer companies of the two merging rms. The share price of Nokia (NYSE: NOK), the Finnish cellphone manufacturer, jumped more than 15% on the day of the announcement, while the share price of Research In Motion (RIM) (NASDAQ: RIMM), the Canadian company that develops BlackBerry phones, hiked up almost 10%. The root of Google's attempt to acquire Motorola concerns patents. In combining with Motorola, Google would automatically claim ownership over 17,000 patents currently held by Motorola an act that would largely augment Google's current reservoir of patents a liated with its Android operating system (a little over 1,000 patents), thereby enabling Google to compete and defend itself more eectively in the smart-phone business.
Google's biggest competitors in the
smart-phone business, Apple and Microsoft, had previously led a consortium of six companies (Apple, EMC, Ericsson, Microsoft, Research In Motion, and
1
Sony) in June to procure more than 6,000 patents from Nortel Networks for $4.5 billion. Clearly the warfare over patents in the mobile phone industry is ongoing, and the most eective defensive tactic to circumvent lawsuits on intellectual property infringement is through acquiring companies that already own a large amount of relevant patents.
Consequently, whoever holds patents with high
likelihood of selling becomes an attractive takeover target. RIM and Nokia, both companies with well-developed smart-phone operating systems, large customer bases, and relatively cheap stock prices, became likely targets for the next round of acquisitions. Their stock prices jumped as a result. While it is a well-documented fact, established from the early 1980s, that acquirers of takeovers gain moderately (if at all) within the merger proposition announcement window, whereby targets of takeovers gain large and positive returns, it's less clear what price impacts takeovers are likely to incur on the peer companies of the merging rms. In the case of the Google and Motorola merger, any peer companies replete with relevant patents for smart-phone operating systems would become likely takeover targets in the future, such is the case of RIM and Nokia, and hence their share prices go up. For mergers that are driven by synergies, as is often the case for M&A, what price impacts are the merger announcements likely to have on the peer companies? From the standpoint of universal investors, such as big pension funds and endowment funds, who may be exposed to many components of a particular product market, including but not limited to competitors, suppliers and retailers, an understanding of the short- and the long-term wealth- eects of proposed mergers is critical. This is the issue which we explore herein.
1 INTRODUCTION Institutional investors often need to decide how to vote on a proposed acquisition, or whether to accept or reject or tender in a takeover attempt. In isolation, such decisions are usually straightforward. If a signi cant premium is oered, a yes vote is probably indicated. Large institutional investors, however, often not only own the shares of both the bidder and target companies and their bonds, but also may have stakes in other related parties, such as bidders' and targets'
2
peers. Clearly, the voting decision needs to consider the impacts on all parties surrounding a particular takeover in which the investor has a stake. Previous literature provided ample evidence that bidders' shareholders tend at most to gain moderately from merger announcements. Target shareholders are the main bene ciaries of any value created by the transactions. We found several product-market analysis papers that, in order to decipher the motives of takeovers for regulatory purposes, studied takeovers' impact on shareholders of rivals, customers, and suppliers of the merging rms. We are not aware, however, of any literature which has evaluated takeovers' impacts on peer company bondholders.
Nor are we aware of any study of takeovers' impacts on share
and bond holders of peer rms from a universal investor's standpoint. In this study, we explore how takeovers impact share and bond holders of peer rms of bidders and targets.
Our research should help universal investors, such as
CalPERS and TIAA-CREF, make better informed decisions. On balance we nd that the impacts of takeover announcements on peer stock and bond holders tend to be small but either positive or neutral. In the short run the shareholders of both acquirer and target peers tend to bene t from the announcements. In the longer run the impact seems to largely disappear for the acquirer peers but remain for the target peers. The peer bond holders of both the target and acquirer tend to bene t modestly from the announcement.
2 LITERATURE REVIEW Eckbo (1983) is one of the rst papers that studied the impact of takeovers on horizontal rivals of target rms. He proposes a method to identify rival rms for each target rm by matching the four-digit SIC code and the rms' primary products. He found generally positive and signi cant abnormal returns for the shareholders of rival rms over the merger proposal announcement window. Further, the magnitude of the rival's abnormal returns tend to persist over the antitrust complaint announcement window, which supports the line of thought that most takeover gains are driven by increases in productive e ciency, as opposed to decreased competitive pressures. Song and Walkling (2000) examine the wealth eects of takeover announcements on the merging rms and their rivals, suppliers, and corporate customers. They de ne rivals as rms in the same Value Line industry (an alternative to SIC codes) as the initial industry target at the time of the target's announcement date. Similar to Eckbo (1983), abnormal returns in Song and Walkling (2000) are computed using the market model (estimated from daily returns 300 days to 61 days before the announcement date of the initial industry target). Also consistent with Eckbo (1983), Song and Walkling (2000) found positive and signi cant CAAR to the shareholders of rival rms over the short-term announcement window. Their ndings largely support the e ciency considerations, rather than oligopoly or buyer power motives, as the prime motive that drives the average horizontal takeover in their sample. Fee and Thomas (2004) examine the eect of horizontal takeovers on ri-
3
vals, suppliers, and corporate customers. The results are consistent with Eckbo (1983) in that rivals exhibit positive abnormal returns on merger announcements, but they do not experience negative abnormal returns when antitrust authorities challenge the mergers. The positive returns would re ect expected gains due to either anticompetitive impacts or e ciency, but the lack of negative returns is inconsistent with the anticompetitive hypothesis. Shahrur (2005) also independently examine the eect of horizontal takeovers on the same industry players, i.e. suppliers, rivals, and customers. The only dierence between the two papers lies in the methodology used to identify suppliers and customers; but nevertheless, the two papers' results are very similar. Speci cally, Shahrur (2005) tests the collusion motive by examining the takeover wealth eects on supplier and customer rms. Their approach complements the methodology in Eckbo (1983), who test the collusion motive by examining the wealth eects of merger and antitrust announcements on rival rms. Consistent with the ndings of Eckbo (1983), Song and Walkling (2000), and Fee and Thomas (2004), Shahrur (2005) nds that rivals and corporate customers earn positive abnormal returns during the merger announcement period. Also similar to the previous literature, Shahrur (2005) nds extensive evidence that supports the e ciency hypothesis versus the collusive hypothesis. Bhattacharyya and Nain (2011) nd strong evidence that horizontal mergers create buying/market power within the industry, which should also have an eect on rival rms. Though their study is not primarily concerned with rivals, they quote Eckbo (1983), Fee and Thomas (2004), Shahrur (2005) in their discussions on rival rms. Clougherty and Duso (2009) use a novel approach for determining rivals by basing their merger database on transactions analyzed by the European Commission for antitrust implications (meaning that EC experts have carefully analyzed the market to identify rivals). Unlike Song and Walkling (2000) and Shahrur (2005), they opt for a shorter 3-day window centered on the merger announcement date, but Clougherty and Duso also verify their results against those papers' 11-day window. Again, their results include signi cant evidence in favor of positive abnormal returns for rivals; they show that rivals perform better than acquirers (but worse than targets) on average. Additionally, they nd that positive rival abnormal returns are not only driven by the merger's signal of "future acquisition probability;" instead, these returns are mostly driven by competitive eects. Clougherty and Duso (2010) use the eects of the merger on merging and rival rms as a reliable basis for classifying the merger itself. In particular, the merger is deemed as synergistic if the merging rms have positive abnormal returns, but it's considered either collusion-based or e ciency-based depending on the rival abnormal returns (where positive rival returns indicate collusion, and negative indicates e ciency). Similarly, if the merging rms have negative abnormal returns, the merger is classi ed as non-synergistic if the rival earns positive returns, and pre-emptive if the rival returns are negative. Akhigbe, Borde, and Whyte (2000) build on Eckbo (1983), Chatterjee (1986), and especially the acquisition probability hypothesis discussed in Song and
4
Walkling (2000). Like Song and Walkling (2000), Akhigbe, Borde, and Whyte use Value Line industries to identify rivals. They nd that merger announcements produce gains for both targets and target rivals (as predicted by the acquisition probability hypothesis), yet merger terminations negatively impact targets while still producing signi cantly positive abnormal returns for rivals. This implies that that merger-induced abnormal returns are caused by signaling, rather than market power or competitive advantages, since in either case, rival rms appear more likely to become targets themselves. Easterbrook and Fischel (1982) argue that unequal allocation of gains from corporate control transactions helps motivate certain value-increasing transactions which might otherwise not take place.
This has a profound eect on
investors, who often diversify their holdings in several dierent companies, and may in fact nd themselves with a stake in both sides of a merger or takeover. In this situation, such investors would prefer any transaction which produces a net gain across their holdings in both corporations to no transaction at all, even if the gains from one side vastly outweigh the gains (or even losses) from the other side. To that end, these investors would be in favor of unequal allocations insofar as they drive unequally pro table transactions which might not have occurred otherwise. Similar to Easterbrook and Fischel (1982), Hansen and Lott (1996)Hansen and Lott (1996) argue that U.S. companies are held by well-diversi ed shareholders who tend to be indierent to how any gain from an acquisition is to be divided. In particular, they show that investors with stock in both the acquiring and target rms in a takeover will be in favor of an agreement which maximizes value across all holdings, even when there's wealth transfer in the process. Speci cally, the article provides empirical evidence that higher levels of cross-ownership in an acquisition is directly correlated with higher bids by acquirers, which is justi ed by this additional value being transferred to the target, and therefore back to the cross-investor.
Furthermore, they nd evi-
dence that excess returns for the acquiring rm's shareholders are dependent on whether the target is publicly or privately held. This eect of diversi cation has important implications for industries such as computers and automobiles, according to the article, since cross-ownership is prevalent in such industries. Matvos and Ostrovsky (2008) arrive at ndings similar to the previous two articles, and nd that, once adjusted for cross-ownership, the mean abnormal return to institutional shareholders around merger announcements is statistically indistinguishable from zero. They demonstrate empirically that a subgroup of institutional shareholders, the cross-owners, in fact realize positive returns, even in equity mergers. This suggests that, with su ciently large stakes in the target, cross-shareholders are likely to cast a yes vote in a merger even when shareholders who hold only shares in the acquirer are losing money. They demonstrate further that the con ict of interest arises only in mergers with negative returns in the acquirer. Matvos and Ostrovsky (2008) is the rst paper that evaluates the relationship between abnormal returns and the voting decision for a merger from the angle of cross-ownership.
5
3 DATA & METHODOLOGY Target rms usually specialize in a single line of business which accounts for the majority of their sales; therefore, peer companies to targets in our de nition are likely to be target rms' close competitors. In contrast, the acquiring rms are, on average, much more diversi ed.
Therefore, the peer rm assigned to each
acquirer following our methodology should include but not be limited to rival rms, and thus we call them peer rms, despite the great similarities these rms share with their matching pairs. The stock return data we used are the split-and-dividend-adjusted daily stock returns from CRSP from 2002 to 2010.
Bond daily pricing data were
obtained from TRACE, which records bond pricing data from 2002 to present. M&A related data were downloaded from Thomson's SDC Platinum database. All of the accounting measures, including annual revenue, total asset, total liability, and market capitalization data, for all the companies involved were obtained from COMPUSTAT. The 90 transactions underlying our analysis are the only Mergers and Acquisitions transactions that have complete data within the 11 day time-window [5 days before the proposal announcement day (i.e. day 0) up until 5 days after the proposal announcement day] in all our required databases (SDC, CRSP, COMPUSTAT, and TRACE). The sample was derived from merging the universes of SDC, CRSP, COMPUSTAT, and TRACE, and the 90 M&A transactions have the complete matches we require in all four databases. After obtaining these 90 transactions, we identify matching peers for each transaction's acquirer and target. We call acquirers and targets involved in these 90 transactions base acquirers and base targets to distinguish from acquirer peers and target peers (which we refer to later). We de ne peer company as the company, under the same four-digit Standard Industry Classi cation (SIC) code as its matching base, that has the closest annual revenue stream as compared to its base (which could either be an acquirer or a target) in the scal year prior to the year when takeover proposal announcement takes place. With this de nition, we found one matching peer company for each acquirer and target involved in these 90 transactions.
To ensure that these peer companies do
indeed match their base companies well, we downloaded three more accounting measures from COMPUSTAT [total assets, total equity (proxy for Market Cap), and total liability] for the 180 base companies and their 180 matching peers. A comparison for the average values of these four accounting measures (annual revenue, total assets, total equity, and total liability) for each of the base versus matching peer groups is presented in Table 1. To verify that the identi ed peer rms are good matches we performed the "Two-sample Assuming Unequal Variance t-test." The null hypothesis for this test assumes that population means of the two samples are the same.
The
test results show very large p-value and thus we don't have enough evidence to reject the null. Therefore, we conclude that the 180 peers are, on average, good matches to their base companies. Next, we calculate the abnormal returns for the stocks of the 360 companies
6
(180 bases plus 180 peers). For this, we used the Return on S&P Composite Index from CRSP. The abnormal return,
ARi,t ,
is simply the excess return a
given company's stock return has over on the daily return of the S&P Composite Index, as follows:
ARi,t = Ri,t − Rm,t ARi,t is the abnormal return of stock i on day t, Ri,t is the gross stock Rm,t is the gross return of S&P Composite Index on day t. The cumulative abnormal return of stock i on each day t, or CARi,t , is the sum of abnormal returns over T consecutive trading days, as follows: where
return,
T X
CARi,t =
ARi,t
1 The cumulative average abnormal return for
N
stocks on day
t
is given by:
N X CAARt = ( CARi,t )/N i=1 To evaluate the long-term eects takeovers have on the stock performance of the merging rms and their matching peer companies, we computed the buy-and-hold
CAARt
of each company that has complete stock return informa-
tion available in the one-year post-announcement window. The buy-and-hold
CAARt ,
or
ematically,
CBHARt , was computed CBHARt is given by:
based on daily abnormal returns. Math-
N X CBHARt = ( BHARi,t )/N i=1 where
BHARi,t =
t Y
(1 + ARi,s ) − 1
s=1 Note that the formula for
BHARi,t
speci es that cumulative buy-and-hold
abnormal return at month t is obtained from the product of one plus daily abnormal return until the end of month t, minus 1. The above summarizes our steps for calculating equity-related abnormal returns. Below we present the methodology by which we compute abnormal bond returns. The absence of bond analysis in contrast to the ample supply of equity analysis in the M&A literature suggests that the problem may be due to the scarcity of bond data.
Clearly bonds, as a whole, tend to be much less
liquid than stocks. To confound this problem, TRACE is the only bond pricing database we are aware of that provides comprehensive current and historical
7
daily bond pricing data (TRACE records daily bond pricing data for approximately 52,438 issuance of bond a liated with 4463 companies from 2002 to 2010). Therefore, the results for our bond analysis largely rely on the quality and the completeness of TRACE. In our eort to match the 180 peer companies previously identi ed with TRACE, we were able to match 105 peers (58 acquirer peers and 47 target peers) by ticker symbols. Removing the problematic data and saving only TRACE data that meet our requirements (detailed below) resulted in 83 peers (51 acquirer peers and 32 target peers).
Our nal plots
for bond abnormal returns involve comparing the cumulative weekly buy-andhold CAAR for `base companies' with their matching `peer companies' within the span of 10 weeks (i.e.
from 5 weeks leading to the `announcement date'
to ve weeks after the `announcement date').
Among the 83 peers, 81 peers
found matching base companies that have complete data covering the required 10 weeks window. Our plots for bond abnormal returns are based on these 81 peer-base pairs. We decided to use the cumulative weekly buy-and-hold CAAR versus the more conventional daily CAAR as the basis for comparison because the average bond in TRACE has a recorded price available every 7 calendar days. We believe this is partially due to the illiquid nature of bond trading and partially due to the endogenous issues with the way TRACE records bond prices. Regardless of the causes, the lack of bond prices renders the calculation of bond CAAR on a daily basis impossible. Accordingly, we thought of aggregating the weekly buy-and-hold CAAR for each bond issuance in our sample, so that each data point (now weekly) contains at least one valid bond price from TRACE that would enable us to compute returns. A more detailed explanation of the steps is as below: 1. First, we matched the 180 peer companies with the TRACE universe by ticker. Note that some companies among our 180 peers occurred more than once. That is, some companies become matched as peer companies to more than one base company.
To avoid confusion, we created a unique identi er
for each of the 180 peer companies. This way we have 180 unique peer identi ers in our beginning data set. Out of the 180, 105 peer identi ers became matched with TRACE by ticker. 2. Out of these 105 matches, 22 identi ers dropped out because their pricing data in TRACE did not exist in our event window, leaving a total of 83 identi ers (or peers) in the nal output le (51 acquirer peers and 32 target peers). 3. For those 83 peer identi ers, we tried to mend the missing data problem by setting the bond prices that are missing to the previous available bond price. We repeated this procedure to ll in blank for each trading day that the bond prices are not available in TRACE, until all the missing prices for these 83 peers were interpolated.
8
4. We then proceed to join LBBD investment-grade corporate-bond index with our lled 83-peers data set on each trading day.
This allows us
to compute the weekly buy-and-hold CAAR for each week, whereas the weekly buy-and-hold CAAR is given by the excess weekly buy-and-hold return of each bond has over the weekly buy-and-hold return of LBBD index in the same week. The formula for computing the weekly buy-andhold return of each bond is as follows:
BPt − BPt−7 BPt−7
BHBRt = where
BPt
before day
is the bond price at day
t, BPt−7
is the bond price a week
t.
The weekly buy-and-hold return of LBBD index is calculated in the same way, where
BHBRtLBBD
is given by:
BHBRtLBBD =
LBBD BPtLBBD − BPt−7 LBBD BPt−7
The buy-and-hold abnormal return for each peer,
BHARt ,
is calculated
as follows:
BHARt = BHBRt − BHBRtLBBD Consequently we obtained the result for the average buy-and-hold abnormal return for each peer
i
at time
t, BHARi,t ,
by averaging the number
of bond under each peer i.
PN CBHARi,t =
j=1
BHARi,j,t N
Note that the time point t is a string that is arbitrarily set equal to {-4,-
t = −4, CBHARi,−4 stands for the CAR for the fourth week leading the announcement date. When t = 0, CBHAR0 stands for the buy-and-hold CAR for the one week leading to the announcement date. When t = 1, CBHAR1 stands for the 3,-2,-1,0,1,2,3,4,5}, such that when buy-and-hold
buy-and-hold CAAR for the one week after the announcement date. 5. We followed the same procedures to computing for the
CBHARi,t
for the
base companies. 6. After
CBHARi,t
for base companies are computed, we subset the number
of peers that found matching peers in the peer data set in step 5. We then
CBHARi,t for both series, i.e. peer and CBHARi,t , or CU M.CBHARi,t is given
computed the cumulative weekly base.
The cumulative weekly
by:
CU M.CBHARi,t =
t=5 X
BHARi,s
s=−4 7. Finally, we plotted the peer vs. base are presented in Figures 6 and 7.
9
CU M.CBHARi,t
series. The results
4 EMPIRICAL RESULT The past literature on M&A has extensively documented the patterns for acquirer and target stock abnormal returns around the merger proposal announcement date. Eckbo (1983) and Ellert (1976), for example, presented plots that show takeovers create large economic rents for target shareholders, while leaving the acquirer shareholders relatively unaected. Our base group results are consistent with this pattern.
Figure 1:
Stock CAR Plots for Base Group
This plot shows that before the proposal announcement, the
CAR
of both
bidders and targets exhibited relatively little change, with slight upward trending patterns that signal information leakage.
On the day of the proposal an-
nouncement, however, we see a huge jump in the (about 12%), and a dip (~1-2%) in the however, the
CAR
CAR
CAR
of the target group
of the acquirer group. Overall,
of the acquirer group remains stable around 0% after the
announcement. Most of the historical literature on rivals focus on rivals to target companies, for the reason that most target companies specialize in one line of business; and thus, their rivals are easily identi ed under the same industry classi cation, most commonly by the four-digit primary SIC code (Eckbo 1983, Fee and Thomas 2004, Shahrur 2005) or Value Line industry classi cations (Song and Walkling 2000, Clougherty and Duso 2009). We only found one study to that examined the impact of takeovers on acquirer rivals Walker & Sue 2007. But all of the rival literature agrees on the positive economic rents takeover announcements impart on the rival companies.
Again, our ndings are consistent with the
previous rival literature.
10
Figure 2:
Peer Stock CAR Acquirer Peers v. Target Peers
CARs before CARs around the announcement is less protrusive, but the acquirer peer group's CARs tend to jump to a higher level. In addition, the target peer group's CARs also rose to a higher level about 15 days after the proposal announcement. These positiveCARs may We see from this plot that both sets of peers exhibit positive
and after the proposal announcement.
The change in
re ect the market's suspicion that more industry consolidation and restructuring will ensue, and thus a higher level of industry-wide e ciency will be achieved. This should send positive signals to peer groups' stock prices, for both acquirer peers and target peers, as we see in our plot. In sum, we would expect universal investors to generate positive signi cant
CARs
from their stock holdings in both base (acquirer and target) and peer
(acquirer and target peer) groups across the 63-day announcement event period. This gain can be enhanced if the universal investors happen to be holding stakes in both target and acquirer peers of M&A transactions. The previous results tell us that universal investors would do well by holding stocks in peer and base groups within the 63-day proposal announcement window.
To explore what
happens to universal investors' stock positions in the long-run, if they continue to hold on to these stocks, we examined the 12-month stock
BHAR for acquirers
in the base group, and acquirer peers and target peers in the peer group. We estimated that, on average, most Mergers and Acquisitions have a dealcompletion period of six to seven months.
Therefore, we would expect stock
return information of target rms to cease to exist six or seven months after the announcement date, given successful deal completion. In the following, we
BHAR for acquirers for whom the deals are completed, and also BHAR of both peer groups corresponding to all acquirers and targets
plot the stock the stock
included in our dataset. Walker & Sue 2007 is the only paper we know of that has looked at the
11
long-term impact takeovers has on rival companies. They examined buy-andhold abnormal returns for acquirer rivals to a maximum of 5 years post the announcement date. But they only documented post-announcement period in order to shows the frequency distribution of the post-acquisitions periods, rather than focusing on the actual on the long-horizon
BHAR
BHAR
for the rivals. Herein, we not only focused
or base group, but also compared the long-horizon
BHAR of the acquirer rival group with that of the target rival group, and nally BHAR of the acquirer rival group with their matching acquirer
the long-horizon base group.
Figure 3:
Acquirer BHAR 12 Months After M&A Announcements
Data obtained from CRSP. Market returns are proxied by S&P500 index returns from CRSP. Total number of acquirers included in this plot is 75 (out of 90 acquirers). Among these 90 acquirers, 15 acquirers dropped out as the deals fell through.
This plot is based on the acquirer stock
CARs
for the 75 completed deals.
We see that the acquirers suered about ve months after the announcement
CARs become positive. This patCARs might be a result of the market's
date, but after nine months, the acquirers' tern is understandable, as the drop in
speculation on whether the deals will go through.
After the deals are closed
successfully, however, the market starts to restore its faith in the acquirer rms. Therefore, the
CARs
of the acquirers start to increase, and eventually become
positive again.
12
Figure 4:
Peer StockBHAR 12 Months After M&A Announcements
Data obtained from CRSP. Market returns are proxied by S&P500 index returns obtained from CRSP. Total number of peers included in the plot is 153 (out of 170 peers matched with CRSP by CUSIP6).
Among the 153 matched peers,
72 are acquirer peers and 81 are target peers.
17 peers dropped out due to
inadequate or erroneous data.
To examine the long-term stock returns of holding peer rms (both acquirer peers and target peers), we plot the 12-month We see that initially the
BHAR
BHAR
for the two peer groups.
of acquirer peers outperforms the
BHAR
of
target peers, but the patterns reverses at around 6 months after the announcement dates.
But the dierence between acquirer peers and target peers are
small. The maximum
BHAR
dierence within one-year window is about 4%.
13
Stock BHAR (Acquirers v. Matching Acquirer Peers) 12 Months After M&A Announcements
Figure 5:
Data obtained from CRSP. Market returns are proxied by S&P500 index returns obtained from CRSP. The 59 acquirers included in the plots are the 59 acquirers to whom the mergers are complete and also their matching peers (i.e., 59 peers) have complete data within the one-year post announcement data time frame.
To see the long-term eect the takeovers have on acquirers and their matching peers, we plot the
BHAR
of these two groups within the one-year-after-
announcement-date window. Note that the
BHAR
of the acquirers' 59 match-
ing peers are persistently positive, while the 59 acquirers suer. This could be the result of the market continuing to view acquirer peers favorably in light of their future acquisition potential, until that sanguine outlook starts to wear o as nothing materializes within one year of the announcement date.
4.1
Bond Abnormal Returns
Our survey of the past literature turned up no papers that have attempted to evaluate the impact that takeovers have on bonds of peers. As explained before, we suspect this phenomenon has a lot to do with the limitation in quality bond data. It is true, based on our analysis, that the calculation of abnormal bond returns on the daily basis is almost impossible, due to the sparse data. But in our case, we are still able to make comparison for bond abnormal returns of the peer group to that of their matching base group through cumulative weekly BH CAARs. Our results aligned with our prediction that bond holders of acquirer peer and target base bene ted prominently from the takeover announcements. Below we present our results. We predict that peer acquirer will outperform the acquirer base, as acquirer peer might see the announcement as a signal that it's also time for them to
14
resort to takeover as a means to shore up their level of e ciency (Eckbo 1983). Therefore, bond holders of the acquirer peer would bene t, as the probability for the rating of the bond to go up is enhanced. The bond abnormal return for the acquirer peer group should remain relatively unaected, since the prospect on whether this merger will be successful and value-adding is still unclear at the point of takeover announcement.
Figure 6:
ers)
Bond Cumulative Weekly BHAR (Acquirer Peers v. Acquir-
These results accord with our expectations. The data-point at time `0' signi es the cumulative
BHAR
from the beginning up until one week prior to the
announcement date. We see from the acquirer peer group a large jump (~3%) from t = -1 to t = 0. This jump can only happen when the bond price on the takeover announcement date is much higher than its price one week prior. In the target group, we predict that the target base is the winner relative to target peer, but the
BHAR
for target peer should also be positive (Song
and Walkling 2003, Clougherty and Duso 2009), as the takeover announcement might serve as a harbinger boding that peers could be themselves become actual targets in the next rounds of takeovers. expectations.
15
Again, our results conform to our
Figure 7:
Bond Cumulative Weekly BHAR (Targets v Target Peers)
Like the acquirer peer in the previous chart, we see a large jump (~8%) for target base from t =-1 to t = 0. This result is also found in the rst part of the analysis on takeover base companies.
We see that the BH CAAR for
target peer rises a little (~1%) in the one week after the announcement, which is in line with our prediction. This price jump could re ect an increase in the probability that the target peers will become actual targets. What entails, if target peers really emerged to become actual targets, is a likely increase of their bond ratings. Therefore, this 1% increase after announcement for the peer target is not di cult to explain.
4.2
PEERS CORRELATION
What is the impact on peers when the transactions are or are not value producing? To explore this question, we compute Pearson correlation coe cients for the paired base vs. peer subgroups. Our results, in addition to addressing the question above, should also help explain any potential wealth transfers that may exist amid takeovers between base company group (i.e. acquirers and targets) and their peer company group (peers of acquirers and peers of targets). The steps for this test are as follows: 1. We divide the sample into transactions that create value and those that don't. 2. We compute the average buy-and-hold CAAR (BH.CAAR) for the various subgroups of peers. 3. We run correlations between various subgroups of peers and their matching base companies (either acquirer or target).
16
If the correlation coe cient is positive for a particular base vs. peer subgroup, then we tentatively conclude that takeovers have similar impacts on the stock prices of peer and base companies within this subgroup.
If the correlation
coe cient is negative for a particular base vs. peer subgroup, then we tentatively conclude that takeovers tend to have opposite impacts over the stock prices of peer companies and base companies within this subgroup. We have 81 takeover transactions in our sample with complete information to compute enterprise
BHAR
for base companies within the 3-day announce-
ment date window [AD: AD+2]. Based on these 81 takeover transactions, we identi ed 73 matching peer companies (49 acquirer peers vs.
24 base peers)
BHAR. For both the base enterprise BHAR for each company
with complete information to compute enterprise and the peer groups, we then compute the
in these groups over a certain holding period (3-day holding period for base acquirers and base targets; 30-day holding period for acquirer peers and base peers). The enterprise
BHARs
for both groups are then divided, among their
component groups (acquirers vs. targets; acquirer peers vs. target peers), into two subgroups, i.e. `value-creating' and `value-destroying', based on the signs of enterprise
BHAR.
If the sign of, say, an acquirer enterprise
BHAR
is positive,
then we assume that takeover transaction has created value in this acquirer company, and vice versa. The formula used to compute enterprise return is as follows:
Enterprise.BHAR = where
EM V
BM V EM V ×BHARequity + ×BHARbond (EM V + BM V ) EM V + BM V
stands for equity market value (given by number of share out-
BM V stands for bond market value (proxLastP rice ×(LT.Debt+Current.Debt) ; BHARequity is given by the dier100 ence in equity prices between the end-of-period equity price and the beginning-
standing multiplied by share price); ied by
of-period equity price over the beginning-of-period equity price. Similarly, we obtain tional
BHARbond . Note that we choose to use BHAR in lieu of the convenCAR because of data problem inherited in our raw data. Many of our
peer companies, especially target peers, are small companies whose stocks are sparsely traded. Confounding the liquidity problem in stock data is the even more severe liquidity problem in bond data, as most companies in our sample, both acquirer peers and target peers, have bonds traded on irregular time intervals. We believe that using
BHAR
BHAR
help remedy the sparse data problem, as
is supposed to factor in everything that has happened within a certain
holding period. Subsequently, we are able to match 38 peer companies to the 81 M&A transactions originally available in our sample. The table below presents a summary outline of the 38 peer companies remained in our nal analysis:
Number of Acquirer Peers
Number of Target Peers
Total
Create Value
8
4
12
Destroy Value
20
6
26
Total
28
10
38
17
The categorization of `create value' versus `destroy value' is based on whether
BHAR of a given company is above or below zero. Speci cally, if BHAR' is greater than zero (base enterprise BHAR > 0), then
the enterprise
`base enterprise
we regard the transaction as creating value to the base company. If base enterprise
BHAR
is less than zero (base enterprise
CAR
< 0), then we say that the
transaction as destroying value to the base company. The same token applies for the acquirer companies. In this manner, we obtained four subgroups, labeled respectively `value-creating acquirers', `value-destroying acquirers', `value-creating targets', `value-destroying targets'. For each subgroup, we calculate the average of enterprise
BHAR
for its matching peer company group, the results of are
presented below:
Average BHARpeers
Acquirer
Acquirer
Target
Target
(Value-creating)
(Value-destroying)
(Value-creating)
(Value-destroying)
(0.0380)
0.0208
(0.0033)
0.0148
This table shows that when takeovers create value for acquirers, they tend to destroy values in peers of acquirers. This could be the case when acquisitions help the acquiring rms emerge as more e cient and thus stronger competitor relative to its peers, through realizing synergy with the target rms, for example. In sharp contrast, the peer group to acquirers that destroy value result in positive
BHAR.
This evidence is also consistent with the e ciency hypothe-
sis, since acquisitions that do not help acquirers emerge as more e cient and thus stronger competitor, the acquisitions actually make the acquirers weaker as compared to its peers, through loss in takeover premium, for instance. For the target group, when the transaction creates value to the target companies, the average
BHAR
to its matching peer group is negative. This is understandable,
as the acquisition of a target company that creates value make the target company a more e cient competitor relative to its peers, thus the negative eect on target peers from tougher competition override the positive eect of the sig-
BHAR. BHAR of the peer companies to targets that do not
naling eect of the mergers on target peers, hence the negative average The result that the average
create value is positive also makes sense, as a target company that destroys value is weakened by the takeover, and thus the competition become more advantageous towards the target peer companies, hence the positive
BHAR
within the
announcement window. Consequently, this result also support the competitive advantage hypothesis versus either the market power or signaling hypothesis. To explore the degree of wealth transfer between base companies and peer companies, we conducted correlation tests. The results of the correlation tests are as follows:
18
Acquirer Peer
Acquirer Peer
Target Peer
Combined Peer
Acquirer
Target
Combined Base
Enterprise CAR
Enterprise CAR
Enterprise CAR
Enterprise CAR
Enterprise CAR
Enterprise CAR
1.000
NA
NA
(0.3022)
NA
NA
NA
1.000
NA
NA
0.0050
NA
NA
NA
1.000
NA
NA
(0.2197)
(0.3022)
NA
NA
1.000
NA
NA
NA
0.0050
NA
NA
1.000
NA
NA
NA
(0.2197)
NA
NA
1.000
Enterprise CAR Target Peer Enterprise CAR Combined Peer Enterprise CAR Acquirer Enterprise CAR Target Enterprise CAR Combined Base Enterprise CAR
The results show that the enterprise
CAR
of acquirer peers and base ac-
quirers are negatively correlated at -0.3022, indicating that the stock prices of acquirers and their peer companies move inversely.
The negative correlation
for the acquirers and their peers is consistent with the proposition that the more value created by the transaction, the more e cient and thus stronger the competitor thereby created, and thus its peers are likely to face more vigorous competition. The enterprise CAR of target peers and base targets are positively correlated at 0.0050, suggesting that any positive price impacts on target companies would also bring about marginally positive price impacts on the peers of the target companies. The positive correlation between targets and their matching peers is consistent with the signaling eect proposed in Song and Meckling (2004) hypothesizing transactions that create value bode well for the peer companies of the targets, in the way that they may also emerge as potential targets in the future who will be bought out at a premium. The combined enterprise CAR of peers and base companies are negatively correlated at -0.2197, showing that, on average, takeovers' impacts on the stock prices of base companies move inversely with those of their peer companies. Speci cally, the negative correlation between the combined enterprise CAR of peers and base companies mean that at the junctures when the stock prices of base companies outperform, the stock prices of the matching peer companies tend to underperform. Therefore, universal investors should allocate their assets so as to allow for dierent exposures (long vs. short) in base and peer companies, buying equal amounts of investment in all parties of takeovers (i.e. base companies and peer companies) would generally result in straight losses. This very last result should provide useful guidance to universal investors while making decisions on asset allocation and risk hedging. Overall, our results from the correlation test are largely consistent with the preponderance of M&A literature that bolsters both the e ciency hypothesis and positive signaling eects for takeovers that create values.
19
5 CONCLUSION Previous M&A literature has largely focused on the pattern of changes takeovers brought to stocks of acquirers and targets during the proposal announcement window.
From a universal investor's point of view, however, the pro ts and
losses resulting from the equity exposures to the acquirers and the targets are only part of the equation. What ultimately matters most to a universal investor is the total impact of a takeover, given their widely diversi ed holdings of the parties aected by a takeover, not just from the merging rms but also the peer groups, in all securities, not just the stocks but also the bonds.
This
study provides a comprehensive analysis to help universal investors make better informed voting decision in light of a takeover proposal. Our work contributes to the M&A literature in two main dimensions. First, the previous literature has focused on the short-term stock CAARs within a takeover-related event window(s).
In addition to analyzing the short-term
(wherein our results are consistent with the literature), we explore the long-run impact. We found evidence that long-term stock bondholder CAR typically sees a turn in abnormal returns around the 5-6 month juncture, for both the acquirers and the acquirer peers in our sample. We suspect this is related to whether the deals are successfully completed, but what exactly are the factors that may have caused this pattern is a question to be explored.
Second, we may well
be the rst study to provide analysis of takeovers' impact on bond abnormal returns for peers. In this, we studied takeovers' impacts on cumulative weekly buy-and-hold bond CAAR, for both merging rms and their peers in our sample, as opposed to calculate the daily bond CAAR, in order to avoid the sparse bond data problem engrained in TRACE. Our ndings show that takeovers tend to impact the bond abnormal returns positively for acquirer peers and target bases, while leaving the bond abnormal returns of acquirer base and target peers relatively unaected. A further look at the impact on peers as it relates to whether the takeover creates or destroys value is intriguing as well. Speci cally, we nd that value creating mergers are bad for the acquirer's peers but good for the target's. Overall, our results shed light on how universal investors should cast their votes in light a takeover proposal. Speci cally, if the stakes are held exclusively in the winner groups, namely, the acquirer peers and the target base groups, then they should vote in favor of the takeover. On the other hand, if the stakes are held exclusively in the non-winner groups, namely, the acquirer base and the target peer groups, the vote, in either way, should not result in much of a gain or loss, as these two groups are expected to be generating modest gains in the short-run and random returns in the long-run, given equal exposures in both parties and equal exposures in stocks and bonds within these two parties. As this paper is among the rst attempts to tap into the peer and bond analysis in the M&A literature, much can be done in the future to re ne our knowledge in the takeover-announcement-driven peer and bond performances. We have not analyzed the factors that drive the stocks' long-term CARs. Ideally, we would also want to obtain understanding as to bond's long-term CARs
20
following the merger proposal announcement.
Eventually, we would want to
analyze the actual returns to universal investors from takeovers given the data on the actual weights allocated in dierent asset groups are obtainable.
21
Table 1
Matching Statistics Acquirer Base Acquirer Peer Target Base Target Peer
Annual Revenue
Mean Total Assets
Total Equity
Total Liabilities
29,887.84
129,894.58
19,300.01
110,238.96
26,635.81
127,167.62
16,100.85
109,832.35
5,684.51
37,600.69
4,152.88
33,268.50
4,962.68
40,850.40
3,626.02
37,073.42
22
References R.G. Hansen and J.R. Lott. Externalities and corporate objectives in a world with diversi ed shareholder/consumers. Journal of Financial and Quantita-
tive Analysis, 31(1), 1996. 13f, CDA/Spectrum. G. Matvos and M. Ostrovsky. Cross-ownership, returns, and voting in mergers.
Journal of Financial Economics, 89(3):391403, 2008. M.H. Song and R.A. Walkling. Abnormal returns to rivals of acquisition targets: A test of the 'acquisition probability hypothesis'. Journal of Financial
Economics, 55(2):143171, 2000.
23
![[Uwe_Schöning]_Logic_for_Computer_Scientists_(Mod(BookFi.org ...](https://p.pdfkul.com/img/300x300/uweschaninglogicforcomputerscientistsmodbookfiorg-_59d308691723dde389357142.jpg)
