M&A League Tables - Valuable Proxy or False Advertising?
Paul Nap
∗20 June, 2014
Abstract
This paper investigates the empirical validity of M&A league tables, i.e. the ranking of banks deal volume advised. It offers evidence on how top advisers select the M&A deals they advice on, i.e. how they set their deal acceptance stringencies. Is the M&A adviser merely an enabler of deals, as the passive-execution hypothesis suggests? This thesis investigates whether the adviser passively executes any deal, whether it selects them and whether it provides valu-able advice. To address mandate acceptance stringency, several measures of ’good’ corporate governance are related to the likelihood of firms choosing a top adviser by market share and subsequently we test whether top advisers create abnormal returns for their clients. This re-search finds that in the 2008-2013 period, transactions led by top tier advisers are associated with statistically and economically significant postive returns, compared to transactions done hiring tier 2 advisers or using inhouse capabilities. The aggregate outperformance of tier 1 ad-visers diminishes with increasing transaction value. This research offers evidence for a skilled advice hypothesis and points at a potential presence of moral hazard among advisers. In the aggregate, there is weak positive association between corporate governance and the choice of M&A adviser.
JEL classification: G34
Keywords: Adviser selection, Agency, Mergers and Acquisitions
1
Introduction
Aswath Damodoran noted on the proposed management buyout of Dell, that the acquisitions done
by Dell in the last five years were largely wasteful and caused losses of over $7bn. Damodoran
also notes that the same management allegedly responsible for engaging in wasteful M&A behavior
(empire building) may be able to buy out shareholders at the price they, themselves, depressed.
This anecdote illustrates an agency problem, set in the context of M&A: If both principal and agent
maximize their respective utilities, even under efficiency, it is likely that there is some divergence of
interest (Jensen and Meckling, 1976). In many cases, firm size remains an important determinant
for executive compensation. From this, a manager may get an incentive to grow a firm, where
this growth is not warranted or indeed, valuable. Secondly and importantly, how and why has the
adviser decided to execute these - allegedly - such wasteful corporate undertakings? If the client
turns to its adviser for skilled advice, absent moral hazard, the adviser should have turned down
these deals. Whether there are differences in the stringency with which advisers accept deals, and
whether they truly can provide skilled advice, is the subject of this research.
Jensen and Ruback (1983) argue that the source of takeover gains is still (in 1983)
undocu-mented. Indeed, studies on M&A paint a picture that is at best inconclusive about the resulting
value creation. While the value to target firms shareholders is widely documented (e.g. Huang
and Walkling, 1987), returns to the acquiring firms shareholders are not generally documented to
be positive in the aggregate (i.e. Jensen and Ruback, 1983; Andrade et al., 2001 and Moeller et
al., 2005). From the empirical literature, which is more thoroughly explored in the literature
re-view, it seems that M&A is one of the areas where shareholders stand to lose value. This seems
to be particularly so for the recent merger wave. Between 1998 and 2001, the total loss resulting
from acquisitions to acquiring firm shareholders is reported to amount to $240bn, in the US alone
(Moeller, Schlingemann and Stulz, 2005). While research is inconclusive about the aggregate of
value destroyed or created, it is certainly the case that bad deals have been done and will be done.
Moeller et al. (2005) report a number of deals that destroyed such value, that the acquiring firm
Moral hazard is a term widely used to describe phenomenons where negative effects of an action
may not accrue to the agent. One look at league tables - the M&A industry standard to assess
’the most fit adviser’ for transactions, learns that it is denominated in total deal value dollars,
rather than percentage returns. From this very high level view, it may be suspected that incentives
could become skewed. Indeed, one might argue that the sheer scale of value destruction reported
by Moeller et al. (2005) casts doubt on either an adviser’s incentive to close a deal - any deal,
its ability to assess quality or just how skilled its advice is. If the bank is viewed as a player in
a repeated game where it balances short-term profits with long-term reputation, it should attach
significant value to its reputation in undertaking deals that create value. In turn, it is also not
well understood how management sets their adviser in an undertaking that very few managers
have experience with, and that may result in significant value destruction to their shareholders;
not to mention that a particularly bad deal induces a probability of being fired. The ’quality’ of
takeovers i.e. the value creation to shareholders - is empirically reported to increase with certain
corporate governance measures (e.g. Masulis, Wang and Xie, 2006). Several others scholars have
related corporate governance measures to firm values (e.g. Gompers, Ishii and Metrick, 2003). The
literature so far has by and large seen the adviser as an ’enabler’ of a deal, and rarely attribute
explanatory power of announcement returns to this key player. This thesis aims to contribute to
research by investigating the skill of and the stringency with which advisers accept mandates. To
this end, several corporate governance characteristics of firms are studied and taken as proxies for
the adviser’s ex ante deal screening. It is studied whether the top tier M&A advisers in the league
tables are likely to be advising different firms and whether CARs of undertaken transactions are
stronger (weaker) for these top advisers than for tier 2 advisers and firms not hiring any adviser.
Lastly it is argued whether it is likely that this difference is attributable to a different screening
profile, or actual skill differences among advisers throughout the M&A process itself.
This research entails investigation of the following, generally associated with ’good corporate
governance’ and a higher values for Tobins Q in empirical literature: antitakeover provisions (e.g.
Gompers, Ishii and Metrick, 2003), cash on hand measures such as free cash flow (e.g. Jensen, 1986),
Datta, Iskander-Datta and Raman, 2001), product market competition (Herfindahl Hirschman
in-dices) and CEO stock ownership (e.g. Morck, Shleifer and Vishny, 1987). Additional control
variables are introduced in the data section. These variables are used to investigate the profile of
acquirers. Whether corporate governance is associated with different (types of) advisers, deals and
returns constitute the building blocks of the main question:
Are the most reputable advisers associated with a strong corporate governance
profile among their M&A clients, and does this profile result in higher returns? The
answer to this question is important to academics and professionals alike, to further assess the
con-sequences of firms’ corporate governance and adviser selection in takeover situations and to more
accurately assess the independence of the advice itself. The hypotheses are stated below:
1) The most reputable advisers are associated with a strong corporate governance profile among
their clients
2) The most reputable advisers make more strongly significantly positive returns for their clients
in M&A transactions, relative to tier 2 advisers and firms hiring no adviser
3) The most reputable advisers make these more strongly significantly positive returns for their
clients due to skilled advice - it is not (exclusively) a result of a different corporate governance
profile.
4) Larger deals make less positive returns.
The availability of governance data (Riskmetrics) makes the US the designated research area
to study this question. The sample period is set from 2008-2013, as it is for the years 2007 - 2012
that Riskmetrics reports data on directors and antitakeover provisions, and the control variables
are lagged one year. In addition, data is obtained from ThomsonOne, Compustat, and Execucomp.
Stock movements are taken from Eventus.
This thesis proceeds as follows: in chapter 2, previous literature on M&A and its resulting value
loss is briefly addressed. Consequently, previous research on how adviser selection happens and how
it should happen are discussed. From this basis a theoretical model for M&A adviser performance
is established. The data and methodology are discussed in chapter 3, resulting in construction of
Chapter 5 concludes and discusses.
2
Literature Review
In this literature review, the theoretical framework and empirical findings relevant to this research
are presented. Firstly, M&A volumes and the performance of M&A in general are addressed.
Secondly, the variables that have been connected to M&A performance give a good starting point
what variables to model for adviser selection - some findings are briefly discussed. Finally, previous
research that investigates the role of the adviser with M&A transactions is discussed.
2.1
Mergers and acquisitions
Acquisitions are one of the largest and most apparent ways corporate investment takes place
(Ma-sulis, Wang and Xie, 2007). They are conducted for a number of reasons and can be loosely
divided into disciplinary and synergistic (Morck, Shleifer and Vishny, 1988). Unlocking anticipated
synergies are crucial for the success of the synergistic acquisition. The ’synergies’ of a synergistic
acquisition may, however, accrue to the manager instead of to the shareholder, presenting an agency
problem. M&A is a corporate undertaking that is an especially good source of information on the
agency problem, as these endeavors are relatively visible to an outsider.
The agency problem entails that managerial discretion grows with entrenchment and as such, so
will management’s power to maximize their own utility at expense of the shareholder, and empire
build. Mechanisms to prevent entrenchment exist internally as well as externally and have been
widely documented. E.g. monitoring, aligning of financial interests or making it less costly or
difficult to fire managers.
This paper proceeds by discussing previous empirical work to highlight the economic magnitude
of the agency problem in an M&A context and discuss research on some of the previously indicated
2.2
M&A performance
Among the research articles published on M&A performance are articles by (but not limited to)
Morck, Shleifer and Vishny (1990), Lang, Stultz and Walkling (1991), Datta, Iskandar-Datta and
Raman (2001), Bliss and Rosen (2001), Moeller, Schlingemann and Stultz (2005)) and Masulis,
Wang and Xie (2007). This paper proceeds by highlighting findings from the latter two.
Moeller, Schlingemann and Stultz (MSS) (2005) conduct comprehensive research covering the
takeover wave of the late 1990s and early 2000s. Compared to the takeover wave of the 1980s,
acquiring firm shareholders suffered substantial and significantly negative wealth effects in the
1990s. The losses to acquiring firm shareholders are especially notable in the period 1998-2001,
vis--vis acquiring firm shareholders in the 1980s. MSS research shows an average loss in acquirer
equity value of 12% of the deal value in the 1998-2001 period, compared to a 1.6% loss in the 1980s.
From 1991 to 2001, shareholders of acquiring firms lost a total of $216 billion, up from $4 billion
from 1980 to 1990. MSS arrive at this estimate through measuring ARs on the 3 days surrounding
the announcements of M&A transactions. The 50-fold increase is attributable to an ’only’ 6-fold
increase in total capital spent on acquisitions. Notable is that despite this substantial destruction
of shareholder wealth, the mean abnormal returns found are still positive. This points to a skewed
distribution. Indeed, the skewness of the distribution of abnormal returns has greatly increased
(MSS). Moreover, the dollars spent on the worst acquisitions have increased substantially more
than dollars spent on the other draws from the sample. E.g., the proportion of dollars spent on
acquisitions in the first percentile of returns increases from 13.68% for 1980 to 1997, to 32.74% for
1998 to 2001 (pp. 759). The wealth destruction is mainly attributable to the worst deals. Deals
with losses in excess of $1bn the authors name ’large loss deals’. these deals account for 43.4% of
capital spent on M&A and $397 billion of aggregate value loss. The abnormal return of these large
loss deals are -10.6%. Shareholders lose an average of $2.31 per dollar spent on large deals from
1998 to 2001 (pp. 765). This finding is notable as it implies that simply burning the cash paid in
the acquisition would have yielded a better NPV. MSS argue that the acquisition forces the market
acquisition as an alarming signal.
Masulis, Wang and Xie (MWX) (2007) conduct comprehensive research on the effect of the
agency problem on M&A quality for 3333 acquisitions by 1268 firms in the US between January
1st, 1990 and December 31st, 2003. As proxies for managerial security, they use 3 indices, 1) the
Gompers, Ishii and Metrick index: an index compiled of a large number of antitakeover provisions
used by US firms, 2) the subset of the GIM index used by Bebchuk, Cohen and Ferrell (BCF)
(2004), consisting of six provisions which BCF argue are the most important, and 3) The one key
provision, used by Bebchuk and Cohen (BC) (2005): staggered boards. The results show that a
part of the depressed Q’s found by GIM for firms with a larger number of ATP provisions are
explained by relatively poorer acquisitions. MWX note that the role of acquisitions is unlikely to
be the only explanation for the 8.5% buy and hold return difference GIM find year on year for
firms in the highest decile of shareholder rights strength versus the lowest decile of shareholder
rights strength. MWX control for a myriad of factors, amongst which deal control variables and
managerial incentives. Their results hold robust for the relationship between antitakeover provisions
and abnormal returns following deal announcement, when implementing these controls.
After these basics on M&A activity and factors that are associated with its performance, this
paper proceeds by looking at the role of the adviser. The literature on M&A discussed up until
this point has willingly or unwillingly accepted the ’passive-execution’ hypothesis, in that it has
not modeled any role for the adviser in the returns of the target. Whether this is an unjustifiably
crude assumption is the focus of this research. The next section studies the role of the adviser in
corporate takeovers.
2.3
The role of an adviser in M&A
The adviser in an M&A transaction may fulfill different roles. He may actively pitch takeover
candidates to clients - in this case the bank is entirely responsible for the CARs resulting from
the deal announcement, as it selects the target and advises throughout the process. A firm may
approach the bank offering a non-specific mandate, that it is looking to acquire according to some
deal announcement abnormal returns, as the same logic applies as in the first case. The third case
is where a client instructs the adviser to specifically acquire a set target. In this case it is argued
that the adviser is only responsible for the orthogonal component of the CARs (Bao and Edmans,
2011). However, the adviser does have the option to ’vote with its feet’ and decide to not do the
deal. These three ways in which a mandate may materialize may result in different CARs, due to
varying weights on the role of adviser and firm.
As stated, previous literature on M&A transactions has, in many cases, not considered the role
or skill of the adviser, e.g. MWX. Research specifically into the role of an adviser has mostly
attributed the deal’s CAR entirely to an adviser (e.g. Bowers and Miller (1990) Rau (2000) and
Hunter and Jagtiani (2003). This research arguably over-attributes statistical power to advisers,
in particular for fixated clients transactions - as the target is set by the client. Other research
has controlled for a number of deal characteristics and thus stripped explanatory power from the
adviser, which may play a substantial role in the selection of deal criteria. This likely leads to
under-attribution as the adviser may recommend deal characteristics, or choose to not accept
man-dates with undesirable characteristics (e.g. Servaes and Zenner (1996) and Kale, Kini and Ryan
(2003)) (Bao and Edmans, 2011).
The role of the M&A adviser has been described using different hypotheses (Bao and Edmans,
2011). The skilled-advice hypothesis states that banks have qualities in recommending and
execut-ing deals that lead to synergistic benefits. The adviser offers capabilities that firms do not have
in-house.
Secondly, the passive execution hypothesis states that advisers are enablers, they provide no
or little advice and do deals on client initiative. Thirdly, the deal completion hypothesis, which
dictates that banks’ end-all goal is to do a deal. Lastly, the limited capacity hypothesis states that
advisers vary not in ability but in capacity. If an adviser constrained by capacity can do only the
best deals, the least frequent advisers should be shown to make the best returns. Whichever the
valid hypothesis (hypotheses), in a selection process there is the importance of reputation and
rela-tionships. This paper proceeds by addressing findings from different investment banking activities:
2.4
Reputation and relationships in bond offerings
A relationship component within adviser selection has been widely documented in various settings.
Yasuda (2005) shows positive and significant effects of bank relationships on the choice of
under-writer, in the bond market. The relation he finds is strongest for junk-bonds (high yield) and
first time bond issuers. The effect of a pre-existing relationship onto fees is ambiguous - Yasuda
shows that with commercial banks, the pre-existing relationship leads to lower fees, although the
association is not as strong as between relationship and underwriter choice. However, strikingly,
investment banks are shown to increase their fees with relationship. Not all relationships matter
- only those where banks played an informational (signalling) role of crucial importance (Yasuda,
2005).
Fang (2005) studies investment bank reputation and its relation to the price and quality of
bond offerings. He finds that banks with a better reputation (he uses a binary variable for high
market share) command lower yields for their clients. The fees are shown to increase with
rep-utation - explained as an economic rent on reprep-utation or a ’costly signal’. The net proceeds for
clients are higher - the lower yields more than offset the higher fees commanded. These results hold
true for his tests for endogeneity, i.e. the firms have not self selected themselves onto reputable
banks. This paper proceeds by addressing documented results from reputation and equity offerings.
2.5
Reputation and relationships in equity offerings
The relation between reputation and equity offerings has been documented at least since the work
of Beatty and Ritter in 1986. The hypothesis and findings state that a bank has incentives to
underprice just enough, as underpricing too much results in a loss of firms and underpricing too
little in loss of investors. Additional research has been done by e.g. Carter (1990, 1998), Booth
(1996), Schadler (1994), Chemmanur and Fulghieri (1994) (more on Chemmanur and Fulghieri’s
Ljunqvist et al. (2006) study the effect of relationship and reputation in the awarding of equity
and debt underwriting mandates by US firms. Their main topic of research is the behavior of equity
research analysts that issue recommendations that are aggressive relative to the consensus. It has
been widely suggested that investment bankers ask research analysts to give a more positive view
of a firm, for which they hope to receive an M&A mandate. Ljunqvist et al. find no evidence for
these allegations. This result is somewhat remarkable as regulatory investigations have led to a
restriction in the extent to which research analysts can pitch deals. This critique is parried by the
researchers, as they point to the credibility of an underwriting signal, which is arguably negatively
impacted by overly optimistic recommendations. The research does show an important component
for previous equity offerings undertaken for the firm by an investment bank in the awarding of
future mandates, relationships matter.
In M&A, no actual underwriting is done, and the role of the adviser is different. Below, this
thesis addresses reputational and relationship effects discerned by prior literature within the context
of M&A.
2.6
Reputation in M&A
Compared to equity and debt offerings, considerably less research has been done on the
reputa-tion and relareputa-tionship coefficients between firms and their advisers. Research has largely implicitly
followed the passive execution hypothesis. A relatively young stream of literature devotes more
at-tention to portraying the role of an adviser as an active player within a repeated game, as scholars
have done with market offerings since the 80’s. Three notable contributions to this literature are
addressed below.
Rau (2000) researches 372 mergers and 388 tender offers. He documents an insignificant
re-lationship between adviser market share and prior acquirer returns. I.e., an uncertain effect on
resulting market share following a more value creative (destructive) transaction. Rau formulates
two hypotheses about the performance of advisers and their resulting market share. The superior
deal hypothesis, which states that the performance of an acquirer in an M&A transaction
yearly deal value). It also predicts that acquirers advised by banks with a high market share may
do better deals. The deal completion hypothesis, conversely, states that a bank will just do a deal
-any deal. This implies that there is no link between transaction excess returns and market share of
an investment bank. According to Rau (2000), bulge brackets charge a relatively larger percentage
of their total fees on a contingent basis than lower tiered investment banks.
Rau (2000) finds no relation between acquirer CARs and adviser market share. As such, he
ar-gues he is forced to accept the deal completion hypothesis. This conclusion holds for longer horizons
and use of annual instead of semiannual CARs. He also finds that bulge brackets, charging higher
relative contingent fees, have a higher deal completion rate (86% vs 75%). Moreover, acquirers that
use first-tier investment banks earn lower excess returns than acquiring firms advised by second- or
third-tier banks. Thirdly, Rau (2000) finds that acquisition premiums are higher for deals
under-taken with first- and second-tier banks (58% and 56%, respectively) versus third-tier banks (38%).
This offers one possible solution for the observed first two phenomenons: with a higher acquisition
premium the deal is more likely to be closed, and the high acquisition price invites low abnormal
returns. Finally, Rau finds no positive relation between post acquisition abnormal returns and
proportion of contingent fees. In fact, he documents a strongly negative relation between these two
for tenders: the higher the proportion of contingent fees (based on completion), the worse is the
post-acquisition performance. These four findings point to merit of the deal-completion hypothesis.
Rau concludes questioning why the market fails to recognize that incentives to close a transaction
do not necessarily yield value maximizing transactions.
Bao and Edmans (2011) document a number of findings. They find fixed effects for hiring an
adviser in the announcement returns of M&A deals. The inter-quartile range of the fixed effects
(i.e. the difference between the 25th and 75th percentile) is 1.26% (0.74% if the sample is limited to
banks averaging at least 3 deals per year)- this compares to the full sample average return of 0.72%.
Stated differently, there are differences in the announcement returns for deals where an adviser was
hired. These differences are not likely the result of measurement bias due to infrequent advisers.
The result is economically significant, considering the mean bidder size is about $10bn. The fixed
case for the skilled advice hypothesis. The correlation between market share and bank fixed effects
is small and insignificant, a similar finding to Rau (2000). The advisers with the worst performance
in terms of CAR are not the most frequent advisers. This finding, they argue, disproves the
lim-ited capacity hypothesis: market share and performance should be significantly inversely related,
according to this hypothesis. Advisers’ market shares are independent of past CARs, but depend
on past market share. Moreover, they find that market share negatively predicts future CARs.
The selection of advisers based on historical market share (league tables) is not necessarily
inef-ficient. Clients may build a relationship with the adviser, which may in turn lead to better expected
performance in M&A transactions or in other types of transactions. However, Bao and Edmans
document a negative effect of retaining a past adviser on M&A CARs. Of the 15 largest banks,
UBS has the worst fixed effects (-0.12%), Bank of America the best (1.47%). Bao and Edmans also
research the performance from an ex ante point of view (i.e. if a firm had to invest in a given year,
with information up to that point) find significant persistence in CARs in 8 out of 9 horizons. Yet,
they find that clients do not chase past returns: returns are an insignificant predictor of market
share.
Golubov, Petmezas and Travlos (GPT) (2012) research whether adviser reputation (the top-8
banks based on value of deals) is more important in M&A transactions in which the target is a
publicly traded firm. They find that top-tier advisers are associated with an 1.01% higher abnormal
returns than lower-tier advisers, upon their client announcing acquisitions of a public firm. This
translates into shareholder value of $65.83 million. There is no such difference when announcing
the takeover of a private firm. Top-tier advisers are also associated with higher fees. GPT find
evidence that top-tier banks can better identify and structure synergies. Also, top-tier banks can
command a larger share of the synergies, unless the adviser to the target is also a top-tier bank.
Lastly, deals advised by the top-tier banks take less time to complete. Endogeneity concerns are
raised on account of GPT finding that top-tier banks are regularly hired by larger firms with higher
book-to-market ratios, firm-specific volatility, lower upwards stock momentum and firms with
ap-petite for a larger target. From aforementioned findings, the OLS-based finding of top-tier adviser
top-tier and non top-tier advisers are switched, the results indicate that had a firm hired a
non-top-tier firm adviser, rather than a non-top-tier adviser, the CAR would be 1.24% lower. Conversely,
had a firm hired a top-tier adviser rather than a non-top-tier adviser, its CAR would on average be
1.01% higher. These findings hold true at the 1% significance level. The top-tier advisers do charge
premium fees for their services. Interestingly, GPT find no strong relation between top-tier advisers
and deal completion - which Rau (2000) did find. Now that some determinants that have been
theoretically associated with the abnormal returns of deal announcements have been covered, this
thesis proceeds by addressing theory about if and how the adviser incorporates this information
into their deal acceptance. A firm may approach the adviser, wanting to acquire or even having
specific target set, or the adviser may recommend a deal to a firm. From the data, one cannot
distinguish which mechanism is at play. Consequently, the origination of an M&A deal must be
treated homogeneously, and the term ’stringency’ incorporates willingness to pitch or execute deals.
The origination of the mandates may play an important, but not readily observed role. The ways in
which a mandate may materialize are treated homogenously. In the next subsection, a theoretical
framework is written out, to which the paper will regularly reference.
2.7
The stringency of deal acceptance
In a world with good and bad acquisitions, shareholders have a belief system about the quality
of a deal. We may assess the quality of a deal in the following way, taken from Chemmanur and
Fulghieri’s (1994) work on equity offerings and modified to be applicable to M&A: let ϕ denote the
fraction of deals that is ’good’, relative to all deals undertaken. A good deal, for simplicity, yields a
shareholder 1, a bad deal yields 0. Investment banks allocate resources to do the research on firms.
They may be of a high cost type I=H, or a no cost type I=N. α0 ∈ (0,1) depicts the probability
investors attach to the advising bank to be of the no cost type at t=0. In the presence of costs
for the high cost type, other things being equal, this type has a larger incentive to close a deal to
generate cash flow. As such, the no cost type has a larger incentive to set strict evaluation standards
transaction becomes, the better the CARs the deal yields. I.e.: as investors update their belief
about the cost type of the bank, the announcement returns continuously adjust to reflect investor’s
perception of the quality of a new deal. The probability of being the no cost type may be seen as
a proxy for reputation. The banks, like investors, cannot a priori identify a bad deals, and banks
can only set standards for doing a deal. It can identify truly good projects and will always market
these. The stricter the deal acceptance standards, the lower the probability that the adviser does
a transaction that is wasteful to shareholders, denoted by r. If it is in fact the case that some
advisers select better targets, are more stringent in agreeing to do a deal or negotiate better terms,
then gradually, the managers acting in shareholder interest may favor this adviser. Then we may
hypothesize that managers chase this performance and we should see market share build up with
these performers. The belief system depicted below (equation 1-3) describes the updating belief
of investors that an adviser may do a good deal. It is a proxy for a reputation coefficient within
past performance of banks: the perceived chance that a bank is less likely to execute a deal that
destroys value to acquiring firm shareholders. Assuming no principal-agent interest divergence,
and no lag in adviser selection, management should judge the transaction history of an adviser in
the exact same way and value past performance. It should approach this adviser. As managers
increasingly do so, market share should build up with the best performing advisers. This ex ante
hypothesis empirically does not seem to be the case, this is addressed in the discussion. Using a
modified version of Chemmanur and Fulghieri’s (1994) equation to model the market’s assessment
of an equity offering undertaken by a certain underwriter, we may write the expected value of a
deal to an investor as:
V0= ϕ α 0 ϕ + rN 0 (1 − ϕ) + (1 − α0) ϕ + rH 0(1 − ϕ) (1)
The reputation proxy, or the probability of a bank being of the no-cost type, updates according to
Bayes’ rule:
αG1 = [ϕ + (1 − ϕ)r
H 0]α0
ϕ + (1 − ϕ)[α0rH0 + (1 − α0)r0N]
(2)
For a ’bad’ deal:
αB1 = rN 0 α0[ϕ + r0H(1 − ϕ)] ϕ[rN 0 α0+ rH0 (1 − α0)] + r0NrH0 (1 − ϕ) (3) With, P rob(e = G|f = G) = 1; P rob(e = G|f = B) = r ∈ [p, 1], p > 0 (4)
The belief system summarized in equations 1 through 3 models a reputational component in
the expected returns for M&A transactions. It illustrates the strategic positions of advisers within
the context of trying to obtain a mandate. From the destroyed value that has been reported, it is
obvious that the value of a bad deal is not just zero. The results of this paper are related back to
these equations. An updating market belief system regarding the role of the adviser is in
accor-dance with the stickiness observed in abnormal returns by advisers, by Bao and Edmans (2011),
even though it predates their research significantly. Bao and Edmans (2011) show that abnormal
announcement returns in the past have significantly positive predictive power over future returns.
This is also what above equations imply, as belief systems ’update’ and do not haphazardly change.
It predicts stickiness in the value the market percieves, i.e. adviser deal performance and/or
accep-tance. If the equation stated above applies to management as well as investors, and if taking the
positive association between past and future returns into consideration, one may hypothesize that
past returns should draw in interest from companies in using this bank as an adviser. α increases as
the bank does a deal with positive NPV. If diminishing returns to scale are assumed and managers
ability (Bao and Edmans, 2011). Managers would flock to the best performing adviser, and the
diminishing returns to scale would dictate a drop in the returns. Seeing how the returns are shown
to be sticky, perhaps management teams use other measures to assign a mandate, or, in terms of
the framework, assign a higher probability to a certain adviser being of the no cost type. This
would imply manager - shareholder interest divergence. The results come back to this.
In the industry, league tables play a role in obtaining a mandate. Indeed, Bao and Edmans
(2011) find that past market share is a predictor for future market share, despite past market share
being a significantly negative predictor for future returns. This observed discrepancy between the
theoretical prediction that past returns should induce higher deal flow and empirical findings of
how mandates are awarded merits further investigation. At first glance, it seems that management
awards mandates in an inefficient manner, the aggregate showing no sign of market share building
up with the best performer, despite past performance being a predictor for future returns. However,
if equation 1 is accurate, we may see management abiding more by it the closer they become to
(being) shareholders. In other words, with convergence of interests between management and
share-holders, reality may move towards this theoretical prediction. I.e. corporate governance measures
may have an effect on the selection of an adviser, as it does with M&A profile in general (MWX).
In return, the adviser may respond by accepting or rejecting a deal, based on these corporate
gover-nance measures. As each adviser balances its long-term reputation with its short-term earnings, an
equilibrium (albeit a highly dynamic one) may originate, where (in the aggregate) the most strongly
governed firms select themselves onto the most stringent advisers. It must be noted, however, that
any finding of difference in corporate governance profiles of acquiring firms per adviser may still be
attributed to either of the four hypotheses. I.e.: any difference may result from pure self-selection
or from turning down bad deals. Endogeneity concerns are addressed later on. Other concerns are
related to data-mining, i.e. confounding ex-post information systems with ex-ante decision making.
Perhaps management does award mandates efficiently ex ante. Corporate governance is measured
through studying several measures of ’good corporate governance’ that are associated with higher
values of Tobin’s Q in existing literature. This paper comes back to this model later on. Firstly,
3
Data
The research entails a wide array of data from different sources. The sources and the selection
of the data are discussed in this section. The research is on US firms, as the US knows the most
thorough disclosure of corporate governance provisions and executive compensation. The databases
used are ThomsonOne for data on the M&A transactions, Eventus for stock returns; Compustat
for company-specific information; Execucomp for information on managerial stock holdings and
information and Riskmetrics for data on corporate governance provisions and director status. Data
is accumulated for the years 2007-2012. The data from Riskmetrics and its accumulation changed in
2007 and this is the constraining factor. Following the acquisition of the provider of the Riskmetrics
data IRRC (Investor Responsibility Research Center) by ISS (Investor Shareholder Services) the
method of data collection changed to conform to follow ISS specifications. In this paper it is argued
that it is unwise to mix the 2007 onwards data with data reported in prior years. As such, the
timespan is set at 6 years, and independent data is extracted for 2007-2012 for US firms, in order to
assess deal returns for 2008-2013, as several of the acquirer characteristics are lagged. Continuous
data on the independent variables are winsorized at cuts of 1 and 99. The databases and the
extracted information of interest are discussed below.
3.1
ThomsonOne
Data on the dependent variables is extracted from ThomsonOne. It shows the adviser for the
deal, deal size and SIC codes. Any deal under $1 million is dropped, as are minority stakes and
acquisitions of targets in which the acquirer already had a controlling stake, pre-transaction. Any
announced deal is compliant for which the acquirer is public. Ownership of the target includes
the whole spectrum. It is not limited to a public target as in prior work. Financial advisers must
have conducted an average of more than one deal per year, in order to not bias research based
the requirement that the target is public. For the deals, the following criteria have been used: 1)
The deals is done between January first, 2008 and December 31st, 2013. 2) The acquirer is a US
based, public firm. 3) Deal value exceeds $1m. 4) A change in effective ownership results from the
deal. 5) All data on the company and the transaction are available from Riskmetrics, Thomsonone,
CRSP, Compustat and Execucomp for the prior year. To research the association of corporate
governance measures and the M&A adviser, all the deals in the 2008 - 2013 time period are merged
with the dataset containing the independent variables. The full set for which all the data is present
contains 2302 observations. I.e., for 2302 deals, all data is present for the acquiring firm and its
stock price returns. This data contains the full supporting set of antitakeover provisions, executive
compensation and ownership, firm capital structure characteristics, director data, industry codes,
et cetera.The sample construction is shown below:
Table 1: Construction of the sample
Constraint (cum.) Number of observations Acquirer nation: US 287533
Acquirer status: public 148835 Percent owned post-transaction >50.0001 89398
Deal value >$1m 45598 Date 01 Jan 2007 - 31 Dec 2012 7028
No pre-deal majority and
all data availabe 2302
The way that is the most traditional to evaluate the performance of an acquisition is to
deter-mine the cumulative abnormal returns (CARs) introduced by Brown and Warner (1985). Buy and
hold returns introduce more statistical problems than CARs (Datta, Iskandar-Datta and Raman,
2001). New listing bias (sampled firms have a long post-event history, the benchmark incorporates
new listings), rebalancing bias (the benchmark is rebalanced, the sampled firm is not) and skewness
bias (long-run returns are positively skewed) are three of the culprits that introduce problems in
long run returns. Long run abnormal returns yield positively biased statistics, buy and hold returns
All things considered, CARs as a dependent with a short event horizon serve this papers
pur-poses best, as it is minimally confounded by statistical issues. The event window is set from 2
days prior to announcement, to 2 days after announcement, as in MWX. The window is set in
this way for the following reasons: 1) According to Fuller, Netter and Stegemoller (2002), SDC
announcement dates are right in 92.6% of cases, and off by no more than 2 days for the remaining
observations 2) The longer the event window, the greater the chance of noise and bias. This leads
to the definition for CARs employed in this paper:
\ CARi= 2 X t=−2 ˆ eit (5) , with, ˆ
eit= Rit− E[Rit] = Rit − [ ˆαit+ ˆβ ∗ Rmarket+ t] (6)
The market model parameters are defined over a 200 day period, 203 to 3 days prior to
an-nouncement.
3.2
Compustat
Compustat is used for company-specific information. From this database the control for industry
code is extracted, by SIC code, for fixed effects. Moreover, calculations of leverage, Tobin’s q,
free cash flow and Herfindahl indices are done using Compustat data items. These variables are
included to control for debt ratio, proxy for firm prospects, proxy for cash on hand for management
and the product market competitions, respectively. Below, an overview of the construction of these
variables is provided for replication purposes. The regressors are generated using the following
calculations, where the ’items’ indicate the Compustat line item numbers:
- Tobin’s q: (Book value of assets (item 6) book value of common equity (item 60) + common
shares outstanding (item 25) x market value of equity (item 199)) / book value of assets (item 6)
- Leverage: (Long term debt (item 9) + short term debt (item 34) ) / (Book value of assets
value of equity (item 199))
- Free cash flow: operating profits before depreciation (item 13) interest expense (item 15)
capital expenditures (item 128) change in net working capital (item 180) - income taxes (item 16)
- Herfindahl Index: HHI = " n X i=1 F irmisales(item12) T otalindustrysales #2 (7)
The extent of agency problems outside of the free cash flow hypothesis is further researched by
looking into proxies for managerial entrenchment by means of antitakeover provisions, which render
the market for corporate control a less effective disciplinary mechanism. In addition to antitakeover
provisions, director independence is studied. These data are extracted from riskmetrics.
3.3
Riskmetrics
Riskmetrics reports data on directors and governance. The data on governance is reported from
2007 to 2012 and is the constraining factor on sample size for this research. The 2006 and prior
governance data are dubbed legacy and constitute the original G index used for earlier years,
com-piled by Gompers, Ishii and Metrick (2003). This data is accumulated differently from the 2007
onwards data. As the differences are a black box to the author of this paper, data prior to 2007 are
not included. The six antitakeover provisions used by BCF are taken from the governance database
and the data on directors is taken from the directors database. What the different antitakeover
provisions entail is explained in the appendix.
3.4
Execucomp
The antitakeover provisions are hypothesized to increase the magnitude of the principal-agent
con-flict, by making it costlier to replace management. By making it costlier, the manager is free
him. However, there are also tools at shareholder disposal to try and reduce the agency conflict,
for instance by tying managerial financial interest to their own. Data on executive equity-based
compensation and managerial stock holdings is extracted from Execucomp. Execucomp reports the
amount of shares owned by the top 5 executives of a firm. The percentage of these holdings is taken
to total shares outstanding, from Compustat. Equity based compensation is also processed into a
percentage, by comparing it to total compensation (tdc1). Listed below is the table of variables.
The variables from the datasets of the independents are lagged one year, as a corporate takeover
can induce quite a shock to cash flows and capital structure. Additional controlling is done through
setting the dependent variable as abnormal returns, thus controlling for a myriad of share price and
company related factors, assuming an efficient market hypothesis to some (semi-)strong extent. The
corporate governance characteristics shown above are strongly correlated. Pairwise correlations are
shown in the appendix.
4
Methodology
The literature demonstrates that several acquirer (and deal) characteristics have been linked to
M&A performance. It is not likely that advisers are oblivious to these findings. If the bank’s
rep-utation is something it considers and values, these acquirer characteristics may be associated with
decisions made in the ex ante deal screening. This thesis investigates whether investment banks truly
have skill, whether they deny certain deals and whether the selection of an adviser is associated with
this skill. To model this, firm characteristics of the acquiring firm must be investigated, to establish
whether there exists a difference in the profiles of the clients of different advisers. This helps in
discerning quality of execution from the stringency of the mandate acceptance. Consequently, it is
investigated whether fixed effects are discerned for firms undertaking a transaction and hiring a (tier
1) adviser and whether the historical market share (deal returns) are affiliated with lower (higher)
ca-Table 2: Overview of the variables
Variables & Description
M&A performance - ThomsonOne
Cumulative abnormal returns Abnormal returns using market model, with parameters defined over 200 days and the event window set from two days prior to two days after the announce-ment.
Adviser The advising bank to the acquirer.
Antitakeover provisions - Riskmetrics
Limited ability to amend Shareholders ability to amend a firms governing bylaws and charter documents is limited.
Classified (staggered) board A provision where directors are allocated to different classes with terms that overlap.
Golden parachute Agreements that compensate managers in the event of a demotion, termination or resignation following a change in corporate control.
Poison pill Poison pills allocate rights to the target firms to acquire more of its own shares, or shares in the raider at a sizable discount.
Supermajority provisions Requirements for mergers to be approved by more than the majority required by state law, in order to pass.
Board characteristics - Riskmetrics Independence ratio Ratio of independent directors to total directors
Compensation - Execucomp Equity-based compensation Ratio of EBC to total compensation
Managerial stock holdings Ratio of shares held by manager to total shares outstanding (Compustat)
Boardsize Number of board members
Firm items - Compustat
Tobin’s Q Replacement value of assets
Free cash flow Cash flow available to managers
Leverage Ratio of debt to equity
Firm size Balance sheet total
pabilities of firms or even their tier 2 peers. Deal CARs are equally weighted. This method would be
flawed if one assumes that larger deals require more skill, which this research does not. This paper
attributes equal weight to all the deals as the variable of interest is the individual deal performance.
In order to investigate the selection of an adviser, this thesis takes several steps: 1) Advisers
are divided into three categories, tier 1 adviser - the 8 largest by total deal value over the
sam-ple period, tier 2 adviser - the other advisers and, third, inhouse - where no adviser was hired,
2) The corporate governance profiles of firms advised by a tier 1 adviser are investigated, to see
whether these clients are different from the full sample. If the firms are different, this may be the
result of varying mandate acceptance stringencies 3) Tier 1 and tier 2 adviser are related to CARs,
with and without acquirer characteristics. With these three steps, it is attempted to work towards
the validity of the adviser role hypotheses laid out below. I.e.: does the adviser passively execute
a deal, or can it provide skilled advice - and if it can, which adviser provides the most skilled advice.
Due to bankruptcies and takeovers, the league tables are slightly different from the traditional
top 8. The 8 banks that are traditionally considered to be the ’bulge brackets’ are Goldman Sachs,
Merrill Lynch, Morgan Stanley, JP Morgan, Citi, Credit Suisse, Lehman Brothers and Lazard. Bear
Stearns, Lehman Brothers and Merrill Lynch are now part of JP Morgan, Barclays and Bank of
America, respectively. Throughout this paper, a comparable definition is employed, however, the
top 8 of the 2008 to 2013 sample period are titled tier 1 adviser. The other advisers are titled tier
2 adviser. The third group of transactions have no adviser. This paper proceeds by addressing the
results.
5
Results
This section shows the results of the findings for associations between tier 1 advisers and acquiring
the firms that are in the aggregate associated with selecting a tier 1 adviser, tier 2 adviser or
no adviser. Secondly, this section addresses the CARs, whether is is associated with corporate
governance characteristics of the adviser’s clients and/or whether it is plausible that some of the
aggregate value creation (destruction) is attributable to the relatively skilled advice of the adviser,
compared to the whole sample.
5.1
Cumulative deal value
As the industry standard for demonstrating quality are the league tables, this is the first area this
research investigates acquiring firm profiles. The summary statistics for transaction values in the
research period are reported below, by adviser.
Table 3: ’League table’ summary statistics for 2008 - 2013
Reported below are the summary statistics for deal activity in the 2008 - 2013 time period. For deals where more than 1 adviser was involved, the transaction value is fully added to both
advisers.
Rank Adviser Average deal value Smallest Largest # Deals Total deal value
($m) ($m) ($m) ($m) 1 JP Morgan 3,228.3 55.9 67,285.7 92 297,003.6 2 Bank of America 2,787.2 82.82 67,285.7 100 278,720 3 Goldman Sachs 2,800 11.3 67,285.7 85 238,000 4 Barclays 3,963.7 65 67,285.7 60 237,822 5 Citi 3,132.9 87.8 67,285.7 58 181,708.2 6 Morgan Stanley 1,871.8 60.2 23,500 80 149,744 7 Credit Suisse 1,712.8 77 16,053.8 50 85,640 8 Deutsche Bank 1,447.2 44 16,053.8 44 63,676.8 Tier 1 advisers 1,890.5 11.3 67.285.7 514 971,717
The summary statistics show some of the aspects that a typical league table shows, i.e. the
total value of the deals undertaken by the adviser, a perceived signal of quality. The largest deal
done within the sample period is the $67bn deal in which Pfizer acquired Wyeth, announced on
January 26th, 2009. 5 of the banks in this league table got a piece of that deal. Alarmingly, this
deal shows a (-2, +2) CAR of -14.4% for Pfizer. This deal would likely qualify for the title ’large
loss deal’ by MSS. Abnormal returns are investigated further in their own subsection. This thesis
company or deal that these advisers are more likely to accept. To estimate whether there are firm
characteristics that are associated with being advised by a tier 1 adviser, we may estimate the
following model, with tier 1 advisers being the 8 advisers depicted above:
T ier1 adviser = ˆα + ( ˆβ1, · · · , ˆβn) ∗
Acquirer characteristics
+ (8)
The model is estimated with and without controls for deal value, in different specifications, and
Table 4: Corporate governance profile for firms advised by tier 1 advisers (the top 8 by aggregate deal value) Characteristic (1) (2) (3) (4) (5) (6) ATP index 0.00910 0.0268 0.0468 0.0938* 0.0495** 0.0884** (0.0389) (0.0695) (0.0303) (0.0539) (0.0249) (0.0429) Q -0.00447 0.0143 -0.00211 0.0216 0.00183 7.73e-05 (0.0532) (0.0946) (0.0528) (0.0934) (0.0433) (0.0783) Leverage 0.796*** 1.537*** 0.789*** 1.537*** 0.147 0.316 (0.245) (0.438) (0.244) (0.436) (0.197) (0.341) FCF to sales 4.92e-05** 7.61e-05* 4.94e-05** 7.59e-05* 1.28e-05 1.84e-05
(2.11e-05) (3.92e-05) (2.09e-05) (3.89e-05) (1.41e-05) (2.34e-05) HHI 0.426** 0.789** 0.415** 0.767** 0.372** 0.635** (0.185) (0.322) (0.184) (0.320) (0.152) (0.260) Ratio EBC -0.0718 -0.148 -0.0577 -0.123 0.0142 0.0378 (0.126) (0.228) (0.123) (0.225) (0.101) (0.168) Board size 0.0833** 0.152** 0.0859** 0.157** 0.0335 0.0554 (0.0389) (0.0699) (0.0384) (0.0692) (0.0305) (0.0536) Board independence 0.679* 1.114 0.619 1.023 0.616** 1.120** (0.390) (0.689) (0.386) (0.682) (0.313) (0.551) Chairman CEO -0.192 -0.408 -0.194 -0.408 -0.186 -0.331 (0.150) (0.271) (0.148) (0.268) (0.118) (0.207) Managerial ownership -0.000825 -0.00132 -0.000589 -0.000975 -0.000232 -0.000396 (0.000892) (0.00160) (0.000865) (0.00157) (0.000675) (0.00124) Log total assets -0.128*** -0.237*** -0.119*** -0.220*** 0.208*** 0.353***
(0.0312) (0.0556) (0.0304) (0.0540) (0.0219) (0.0377) Log deal value 0.754*** 1.351*** 0.751*** 1.346***
(0.0356) (0.0689) (0.0354) (0.0684)
Constant -5.119*** -9.096*** -5.306*** -9.480*** -3.492*** -5.997*** (0.471) (0.868) (0.438) (0.807) (0.343) (0.609)
Year dummies YES YES NO NO NO NO
Model Probit Logit Probit Logit Probit Logit Observations 2,264 2,264 2,264 2,264 2,264 2,264
These results show what firm characteristics influence the decision of a tier 1 adviser to accept
or suggest a deal. Most expectedly, large advisers select large deals and large clients. Essentially,
these large advisers are in many cases banks that have the balance sheet and the network to
under-take these large transactions. It also likely comes with a greater fee. Throughout the specifications,
significantly associated with being adviser by a tier 1 adviser are board size and the Herfindahl
index (5%). When dropping year dummies, the antitakeover provision index becomes positive at
the 5% significance level and increases the likelihood of an adviser with a large market share
ad-vising. This coefficient loses its significance when dropping deal value controls. When dropping
the deal value component, other corporate governance characteristics turn significant.
Character-istics of ’strong’ or ’weak’ corporate governance, respectively, do not seem to drive the regression
results in one direction, i.e. away or towards tier 1 advisers. The same output but for acquiring
firms that decide not to hire and adviser is shown in the appendix. Replacing the bucket of tier
1 adviser with individual advisers from the bucket does not change the picture. The difference
in statistical significance of corporate governance characteristics when dropping certain controls
is striking and stresses the importance of estimating the proper functional form when looking to
isolate the weights of the individual characteristics. As this paper treats these characteristics as
more of control variables, more detailed research about the proper functional form is left for future
research. This thesis proceeds by addressing the abnormal returns resulting from M&A transactions.
5.2
Abnormal returns
A mandate being accepted or not, is - even for the best adviser - no guarantee regarding the quality
of the execution. The quality of execution is defined here as the wealth created by the deal in terms
of equity capital. Whether the top advisers produce higher CARs is the subject of this section.
The example of the largest deal being so wealth destructive is a reminder that equally weighting
announcement effects may not capture the true aggregate of value creation (destruction). Indeed,
equally weighting the abnormal returns for acquirers in M&A transactions, the average CAR is a
Table 5: CARs for transactions with (a) tier 1 adviser(s) and having no adviser Average CAR Median Standard deviation Min Max Tier 1 adviser 1.68% 0.50% 0.181 -24.92% 369.46%
No adviser 0.20% 0.02% 0.048 31.58% 32.25% Total 0.64% 0.37% 0.095 -31.58% 369.46%
The summary statistics show that the average CARs are positive, but the median indicates
positive skewness. This finding underlines findings concerning large loss deals by MSS, where large
deals on average do worse. The average returns for firms advised by a tier 1 adviser seem better
than firms using in house capabilities, which is investigated further in this section. Below, the
regression is stated that estimates the value of tier 1 advisers in deal announcement returns.
\ CARi= ˆα + ˆβ1∗ T ier1adviser + ( ˆβ2, · · · , ˆβn) ∗ Acquirer characteristics + year (9)
The model is estimated four times with different specifications, with and without industry fixed
Table 6: Regression output for cumulative abnormal returns for deals undertaken 2008 - 2013 Characteristic (1) (2) (3) (4) (5) (6) (7) (8) Tier 1 adviser 0.0145** 0.0140** 0.0184*** 0.0161*** 0.00968* 0.00938 0.0164*** 0.0124** (0.00620) (0.00618) (0.00529) (0.00521) (0.00582) (0.00581) (0.00499) (0.00484) ATP index 0.000177 -0.00115 -0.00111 -0.00124 -0.00166 -0.00161 (0.00229) (0.00178) (0.00178) (0.00199) (0.00160) (0.00160) Q 0.00160 0.00165 0.00168 0.00102 0.00122 0.00122 (0.00304) (0.00301) (0.00302) (0.00267) (0.00265) (0.00265) Leverage 0.0692*** 0.0706*** 0.0680*** 0.0638*** 0.0657*** 0.0625*** (0.0195) (0.0194) (0.0193) (0.0130) (0.0130) (0.0129) FCF to sales -6.36e-07 -6.68e-07 -7.29e-07 -3.30e-07 -3.64e-07 -4.82e-07 (1.02e-06) (1.02e-06) (1.02e-06) (9.17e-07) (9.14e-07) (9.14e-07) HHI 0.0450** 0.0471** 0.0471** 0.0448*** 0.0456*** 0.0454*** (0.0197) (0.0197) (0.0197) (0.0101) (0.0101) (0.0101) Ratio EBC 0.00114 0.000831 0.00105 0.00222 0.00165 0.00200 (0.00615) (0.00614) (0.00614) (0.00603) (0.00602) (0.00602) Board size -0.00181 -0.00242 -0.00256 -0.00178 -0.00240 -0.00266 (0.00217) (0.00214) (0.00214) (0.00197) (0.00195) (0.00194) Board independence 0.00436 0.00874 0.00869 0.0314 0.0321* 0.0320 (0.0233) (0.0230) (0.0230) (0.0197) (0.0195) (0.0195) Chairman CEO 0.00204 0.00101 0.000835 0.000500 -0.000546 -0.000839 (0.00755) (0.00751) (0.00751) (0.00747) (0.00742) (0.00742) Managerial ownership 9.40e-05* 8.86e-05* 8.79e-05* 5.23e-05 4.95e-05 4.87e-05
(4.81e-05) (4.79e-05) (4.79e-05) (3.97e-05) (3.95e-05) (3.96e-05) log total assets -0.00564*** -0.00609*** -0.00506*** -0.00901*** -0.00930*** -0.00768***
(0.00191) (0.00188) (0.00173) (0.00165) (0.00163) (0.00147) log deal value 0.00186 0.00226 0.00331** 0.00365**
(0.00166) (0.00165) (0.00157) (0.00155)
Constant 0.00298 0.0104 0.0134 0.00292 0.0155 0.0179 0.0230 0.00371* (0.0284) (0.0268) (0.0267) (0.00218) (0.0229) (0.0211) (0.0210) (0.00225)
Year dummies YES NO NO NO YES NO NO NO
SIC fixed effects YES YES YES YES NO NO NO NO
Observations 2,302 2,302 2,302 2,302 2,302 2,302 2,302 2,302 R-squared 0.025 0.022 0.022 0.005 0.037 0.034 0.032 0.003 Number of AcquirerSIC 296 296 296 296
The model is estimated 8 times, to control for different aspects. The models incorporating
industry fixed effects, as Bao and Edmans (2011) estimated the models, seem the best
approxi-mation of reality. These models find the statistically strongest association between tier 1 adviser
and abnormal returns. First and foremost, this output shows the coefficient on Tier 1 adviser is
significant at the 5% level throughout the fixed effects specifications and some of the regular OLS
estimates. This implies that for the deals contained in the sample, an acquiring firm could expect
to achieve around 1.5% better return around the announcement period than a firm who uses a tier
2 adviser or uses inhouse capabilities to execute the M&A transaction. This result holds robust
throughout the fixed effect specifications, and are weakened some when pooling industries. In
ad-dition, leverage, the Herfindahl index and total assets are significant and positive, as for the first
two cases, theory predicts they should. Year dummies do not seem to enhance the model by much.
Interestingly, in the most basic models, the coefficient on tier 1 adviser is higher than in models
that more widely control. This could imply negative self selection onto tier 1 advisers, i.e. overall
worse acquirer profiles. This deduction hinges on that the expected abnormal return, controlling
for the firm profiles, is lower in widely controlled estimations than estimations not controlling for
firm profiles. If so, it could point to execution skill. Leverage and product market competition are
significant throughout estimations. The other corporate governance characteristics do not seem to
play a large role in deal announcement returns. Interestingly, in the fixed effect estimates, deal value
is not a component that is significantly associated with abnormal returns. This leaves us still in
search of an explanation behind the large loss deals and an explanation for the value weighted and
equally weighted averages of CARs displaying distinctly different characteristics. The hypothesis
that banks may set looser mandate acceptance stringencies for deals that contribute more to their
top line is worth exploring as a possible explanation. This is addressed in the next section.
5.3
Performance and transaction size
The theoretical model of section 2.7 sets a single set parameter for mandate acceptance stringency
have been offered, resulting from moral hazard. The standard assessment of adviser performance is
measured in total transaction size and the compensation for larger deals typically is larger -
induc-ing incentives to prefer doinduc-ing larger transactions. Alternatively, deals may increase in complexity
with their size. From this, this section goes into the possibility that advisers are more willing to
do large deals. This may entail that these deals are with less certainty a good deal, with increasing
transaction size. Alternatively, these large transactions may be more complex, and advisers may
be less capable to succesfully close these larger transactions. From either hypothesis, it follows
that average CARs may fall with increasing deal value. However, as transaction size increases, so
generally does publicity. This may drive results in the opposite direction, i.e. shrinking the desire
to do lacklustre return, high value deals. The transaction value is interacted with the dummy for
large adviser, to estimate continuous movements throughout increasing transaction value. Some
considerations must be made before looking at this subset. As the first regression shows, the chance
of selecting a tier 1 adviser is correlated with transaction size, and tier 1 adviser also is correlated
with higher returns. This could bias the findings upwards - i.e. dampen the interaction term when
estimating continuously across all observations. From the nature of the data and the findings thus
far, one may hypothesize that the relationship between transaction size and CARs is complex to
estimate and readily subject to noise through known and unknown correlations when using an unfit
model. To estimate whether there is an association between higher transaction values and CARs,
different models are estimated. Simply adding the interaction term of ’tier 1 adviser * transaction
size’ to the CARs model (fixed effects specification) previously estimated returns a significantly
negative coefficient for the interaction term, despite the expected counterdirection of the bias in
Table 7: Regression output for cumulative abnormal returns for deals undertaken 2008 - 2013, including inter-action between tier 1 adviser and transinter-action value
Characteristic (1) (2) (3) (4) Tier 1 adviser 0.0733*** 0.0749*** 0.0670*** 0.0700***
(0.0246) (0.0245) (0.0233) (0.0232) ATP index 2.86e-05 -0.00113 -0.00145 -0.00173 (0.00229) (0.00178) (0.00199) (0.00160) Q 0.00190 0.00193 0.00109 0.00129
(0.00304) (0.00301) (0.00267) (0.00265) Leverage 0.0703*** 0.0718*** 0.0644*** 0.0662*** (0.0195) (0.0194) (0.0130) (0.0129) FCF to sales -6.12e-07 -6.37e-07 -3.06e-07 -3.27e-07 (1.02e-06) (1.02e-06) (9.16e-07) (9.13e-07) HHI 0.0443** 0.0464** 0.0441*** 0.0448*** (0.0197) (0.0197) (0.0101) (0.0101) Ratio EBC 0.000543 0.000229 0.00151 0.000949 (0.00615) (0.00613) (0.00603) (0.00602) Board size -0.00177 -0.00235 -0.00184 -0.00240 (0.00216) (0.00214) (0.00197) (0.00194) Board independence 0.00590 0.0100 0.0322 0.0327* (0.0233) (0.0229) (0.0197) (0.0195) Chairman CEO 0.00229 0.00132 0.000893 -5.60e-05 (0.00755) (0.00750) (0.00746) (0.00741) Managerial ownership 9.89e-05** 9.44e-05** 5.45e-05 5.24e-05
(4.80e-05) (4.79e-05) (3.96e-05) (3.95e-05) log total assets -0.00557*** -0.00598*** -0.00893*** -0.00918***
(0.00191) (0.00188) (0.00165) (0.00162) log deal value 0.00340* 0.00383** 0.00494*** 0.00532***
(0.00178) (0.00176) (0.00169) (0.00167) log deal value x tier 1 adviser -0.00940** -0.00971** -0.00922** -0.00973***
(0.00380) (0.00379) (0.00363) (0.00361) Constant -0.00605 0.000636 0.00844 0.00951
(0.0286) (0.0270) (0.0230) (0.0213) Observations 2,302 2,302 2,302 2,302
R-squared 0.028 0.026 0.040 0.037
Fixed effects YES YES NO NO
Year dummies YES NO YES NO
The table above is a little difficult to interpret. The CARs improve with a tier 1 adviser advising
on the transaction. Seperately, assuming monotonical relationships, CARs continuously improve
with a higher transaction size. However, the signs significantly invert when tier 1 adviser and deal
value interact, indicating that for the larger transactions it is not as clear whether with increasing
transaction sizes a tier 1 adviser still does better. This table suggests that a tier 1 adviser is a
good choice, but a worse one with increasing transaction value. The assumption of monotonicity
in these relationships is a strong one, as we know that tier 1 advisers generally do not do deals
below a certain transaction size, i.e. are relatively clustered around larger deal values. The tier 1
advisers are correlated with deal value itself, but also returns (are not homogenously represented
among deal values and achieve better CARs). To relax the assumption of the interaction being
homogenous across the observations, below, the top decile of deals by transaction size is
investi-gated. The general association between tier 1 advisers and returns for this right tail of the sample
may offer further, more robust evidence on mandate stringency with increasing deal value. The top
decile of deals is in 80% of cases advised by at least a tier 1 adviser. As such, top decile deals are
strongly correlated to the tier 1 adviser dummy. The top decile dummy is therefore included in
every estimate. The output is shown below.
Interestingly, for the top decile, the estimate shown below indicates that the returns a tier 1
adviser produces for these largest of deals nearly eliminates the percentual abnormal return increase
from hiring a tier 1 adviser in the first place, controlling for top decile. This strengthens the
in-terpretation of the first estimate. While top decile itself knows an insignificant positive coefficient
and tier 1 adviser knows a significantly positive coefficient, in this model too, the interaction is
significantly negative. For the largest of deals the positive effect associated with a tier 1 adviser
all but disappears. Striking is the sign of the top decile dummy, despite it not being significant:
the largest of deals are not significantly associated with lower returns in the aggregate, except for
when a tier 1 adviser is involved. It hints at returns not suffering from additional complexity with
increasing deal value, as CARs would be lower across the board. This finding may point to moral
Table 8: Regression output for cumulative abnormal returns for deals undertaken 2008 - 2013, for the top decile of transaction sizes
Characteristic (1) (2) (3) (4) Tier 1 adviser 0.0264*** 0.0267*** 0.0242*** 0.0247***
(0.00624) (0.00623) (0.00587) (0.00587) Top decile x tier 1 adviser -0.0342** -0.0342** -0.0353** -0.0359** (0.0171) (0.0171) (0.0167) (0.0167) Top decile 0.0111 0.0115 0.0126 0.0131 (0.0147) (0.0147) (0.0144) (0.0144) ATP index 0.000318 -0.00109 -0.00104 -0.00163 (0.00228) (0.00178) (0.00199) (0.00160) Q 0.00185 0.00183 0.00116 0.00128 (0.00304) (0.00301) (0.00267) (0.00265) Leverage 0.0677*** 0.0688*** 0.0613*** 0.0631*** (0.0194) (0.0193) (0.0130) (0.0129) FCF to sales -6.40e-07 -6.81e-07 -3.50e-07 -4.00e-07
(1.02e-06) (1.02e-06) (9.16e-07) (9.13e-07) HHI 0.0441** 0.0464** 0.0442*** 0.0448*** (0.0197) (0.0197) (0.0101) (0.0101) Ratio EBC 0.000738 0.000431 0.00184 0.00126 (0.00615) (0.00614) (0.00604) (0.00603) Board size -0.00189 -0.00259 -0.00202 -0.00276 (0.00216) (0.00214) (0.00197) (0.00194) Board independence 0.00425 0.00885 0.0305 0.0316 (0.0233) (0.0230) (0.0197) (0.0195) Chairman CEO 0.00161 0.000545 4.73e-05 -0.00109 (0.00755) (0.00750) (0.00746) (0.00742) Managerial ownership 9.70e-05** 9.11e-05* 5.30e-05 4.95e-05
(4.81e-05) (4.79e-05) (3.97e-05) (3.96e-05) log total assets -0.00439** -0.00469*** -0.00708*** -0.00726***
(0.00178) (0.00176) (0.00153) (0.00151) Constant 7.68e-06 0.00983 0.0150 0.0204
(0.0284) (0.0268) (0.0229) (0.0211)
Fixed effects YES YES NO NO
Year dummies YES NO YES NO
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1