CEO Overconfidence, CEO Dominance and Corporate Acquisitions

CEO Overconfidence, CEO Dominance and Corporate Acquisitions
Rayna Brown*
Neal Sarma
Department of Finance
The University of Melbourne
Victoria 3010
AUSTRALIA
Keywords: overconfidence, dominance, corporate acquisitions, independent board
JEL codes: G34, G38
Acknowledgments: We thank Rob Brown, Bonnie Buchanan, Edward Lee, David Reeb, Kim
Sawyer, Ian Sharpe and an anonymous referee for helpful comments on earlier drafts. We are
also indebted to seminar participants at the Annual AIBF Banking and Finance Conference
(2005, Melbourne), Bangor Business School at the University of Wales and the Annual
Meeting of the FMA (2006, Salt Lake City). For technical assistance we are grateful to Philip
G. Brown and Kim Sawyer. All remaining errors are ours.
Rayna Brown also wishes to thank the School of Accounting and Finance, the University of
Manchester, for support during a sabbatical leave.
November 20, 2006
*Corresponding author
Ph:
61 3 8344 7661
Fax: 61 3 8344 6914
e-mail: [email protected]
CEO Overconfidence, CEO Dominance and Corporate Acquisitions
Abstract
This study investigates the role of CEO overconfidence (hubris) and CEO dominance in the
firm’s decision to undertake an acquisition. We argue that it is important to capture not only the
extent of overconfidence but also the ability of the CEO to impose his or her views on the
firm’s decisions. We test this approach using logistic regression and Australian data. The
results suggest that both CEO overconfidence and CEO dominance are important in explaining
the decision to acquire another firm. When compared to existing US studies, the evidence on
CEO overconfidence is robust across two different financial and corporate governance systems.
Our results also indicate that CEO dominance is at least as significant as CEO overconfidence
in the decision to undertake an acquisition.
2
1.
Introduction
The mergers and acquisitions literature suggests that there are three main motivations for
takeovers. The first motivation is the creation of synergies so that the value of a new combined
entity exceeds the sum of its previously separate components. The second motivation arises due
to agency conflicts between managers and shareholders. Jensen (1986) suggests that managers
may rationally pursue their own objectives at the expense of shareholders’ interests. Finally, the
third motivation for takeovers is managerial hubris (Roll 1986). Roll’s hubris hypothesis
suggests that managers of acquiring firms make valuation errors because they are too optimistic
about the potential synergies in a proposed takeover. As a result, they overbid for target firms
to the detriment of their stockholders.
Thus, there are two main theories – rational responses to agency costs and non-rational
managerial hubris – that have been suggested to explain why managers make value-destroying
acquisitions. Although the hubris hypothesis has considerable intuitive appeal, and has been
discussed in the literature for two decades, it has only infrequently been subjected to direct
empirical testing. Behavioral assumptions such as investor overconfidence have become
common in the asset pricing literature but the corporate finance literature has largely neglected
behavioral assumptions in models of managerial decision making (Barberis and Thaler, 2003).1
There has been only a limited amount of theoretical research and very few empirical studies.
Moreover, such evidence as does exist concentrates on the United States (Hayward and
Hambrick, 1997; Malmendier and Tate, 2004). The US evidence suggests that overconfident
managers are more likely than other managers to destroy value. We argue that it is important to
capture not only the extent of CEO overconfidence but also the effect of CEO dominance,
1
This pattern has emerged even though objections to behavioral finance, such as arbitrage and learning,
tend to be more persuasive in asset pricing than in corporate finance (Heaton, 2002).
3
which is the ability of the CEO to impose his or her overconfident views on the decisions of the
firm. We develop a measure of CEO dominance that is based on executive compensation. An
empirical test of our approach is conducted using Australian data for the period 1994-2003.
Although the US and Australian financial systems have much in common, there are
significant differences that may affect the influence of managerial overconfidence. The first
concerns corporate governance regulations. Anand (2005) classifies countries into three groups
depending upon their corporate governance compliance requirements. Using this classification,
Australia, along with 13 other countries including Germany, Italy and the United Kingdom,
falls into the group of countries that has voluntary governance guidelines but mandatory
disclosure of governance practices. The US is in a group of only two countries that have
mandatory governance practices and mandatory disclosure of governance practices. A second
important difference relates to the relative reliance on intermediary-based and capital marketbased financing. Amongst large industrialized countries the US is at one extreme with less than
30% of financing allocated through intermediaries, while Germany is at the other extreme with
75% allocated through intermediaries. Like most developed countries, Australia falls between
these extremes (Reserve Bank of Australia, 2000). Third, among Australian listed companies,
controlling blockholders are quite common.2 Shleifer and Vishny (1986) argue that external
blockholders reduce the scope of CEO opportunism and lower agency costs. Brailsford et al.
(2002) suggest that managerial stock ownership and external block ownership interact, such
that at high levels of managerial stock ownership, managerial entrenchment competes with
monitoring activities to reduce substantially the significance of external block ownership.
2
Many of these blockholders are pension funds known as “superannuation funds”. Under Australian law,
employee, employer and government contributions to superannuation must be equal to at least 9% of
every individual’s salary. Most of the superannuation funds which receive the contributions are privately
run and managed. Assets of superannuation funds are equivalent to approximately 80% of Australia’s
GDP and approximately 40% of all Australian superannuation fund assets are invested in Australian
equities. Edey and Simon (1996) provide a detailed description of Australia’s compulsory
superannuation scheme.
4
Lamba and Stapledon (2001) report that in Australian companies lacking a blockholder, only
37% of the share capital was voted on director-election resolutions. External blockholders are
more likely to vote in director elections, thus increasing the chance of CEO opportunism being
mitigated by an independent board.
Our results suggest that CEO overconfidence and CEO dominance affects corporate
behavior as revealed in acquisition decisions. Overconfident CEOs are more likely to make
acquisitions – especially diversifying acquisitions – than other CEOs. However, we also find
that CEO dominance is as least as significant as CEO overconfidence. Evidence that CEOs may
destroy firm value also poses the question of how to rein in an overconfident CEO. Our results
indicate that having an independent board of directors assists in achieving this goal. The
evidence provided by this study should assist in attempts to mitigate the destructive effects of
CEO behavior through stronger corporate governance regulation.
The structure of the paper is as follows. Section 2 provides a brief overview of the
literature on the wealth effects of mergers and acquisitions. In Section 3 the theoretical
background and measurement of our proxies for CEO overconfidence and CEO dominance are
discussed. The methodology is outlined in Section 4, while Section 5 documents our data
sources and provides descriptive statistics. Our results are presented in Section 6. Some
concluding comments are presented in Section 7.
2.
The wealth effects of mergers and acquisitions
There are two broad streams in the literature on mergers and acquisitions.3 The first
stream investigates the motives for undertaking acquisitions, which are traditionally considered
to be either the maximization of shareholders’ wealth or managerial hubris. The second stream
3
For simplicity, in the remainder of the paper we will use the term ‘acquisition’ to include merger.
5
investigates the wealth effects of acquisitions. If managers act to maximize shareholders’
wealth, then an acquisition can be seen as adding value to both target and acquirer through the
creation of synergies that are expected to produce economic gains and hence increase wealth.
However, there is a consensus amongst empirical studies that acquisitions are valueenhancing for stockholders in target firms but on average are at best value-neutral for
stockholders in acquiring firms. In their survey of US evidence, Andrade et al. (2001) find a
positive abnormal return of 16% to targets that is remarkably consistent over time, and a
negative, but insignificant abnormal return to acquirers. Walter and da Silva Rosa (2004)
survey the Australian evidence and report similar conclusions. In the case of targets, “the
evidence is unequivocal … target firm shareholders benefit considerably” (p. iv), whereas “the
share price performance of acquirers around the bid period is difficult to reconcile with the
value-increasing hypothesis” (p. vi).
Several explanations have been offered for this disappointing outcome for acquirers. If
the market for potential targets is sufficiently competitive, then the benefit of a proposed
acquisition should be competed away, leading to a mean return of zero for acquirers. A
negative return to stockholders in acquiring forms could be explained by agency costs: that is,
the manager(s) of acquiring firms favor takeovers because their power, wealth and status are
increased. Such managerial behavior is rational but not in the interests of the stockholders.
Alternatively, a negative return to acquiring stockholders may be explained by hubris or
overconfidence on the part of the CEO of the acquiring firm. This explanation suggests that the
CEO may sincerely believe that a merger is in the best interests of the stockholders but that this
belief is not rationally based.
Agency costs and/or managerial hubris may be more likely in the case of diversifying
acquisitions. Morck et al. (1990) find that a significant negative abnormal return accrues to
bidding firms upon the announcement of a diversifying acquisition. Maquiera et al. (1998) and
6
Bhagat et al. (2004) provide further empirical evidence that acquiring firm stockholders gain
less from diversifying acquisitions than from non-diversifying acquisitions. In addition, there is
evidence that diversified firms trade at a discount to stand-alone entities in the same line of
business (Lang and Stulz, 1994; Berger and Ofek, 1995; Servaes, 1996). The existence of a
diversification discount has often been interpreted as evidence that diversification destroys
value. Scharfstein and Stein (2000) suggest that there may be increased agency costs in
diversified firms. Findings on the diversification discount have recently been the subject of a
debate that has been well summarized in Martin and Sayrak (2003). Diversifying acquisitions
have, therefore, been linked to the existence of agency costs as diversification may benefit
managers (Morck et al. 1990), and to the existence of managerial overconfidence (Malmendier
and Tate, 2004).
3.
CEO overconfidence and CEO dominance: theory and measurement
3.1
CEO overconfidence
‘Overconfidence’ is defined as an overestimation of one’s own abilities and of outcomes
relating to one’s own personal situation (the ‘better-than-average’ effect) (Langer, 1975).4 The
hypothesis of overconfidence in finance is based upon an extensive literature in psychology
which finds that people are generally overconfident (Frank, 1935; Weinstein, 1980). For
example, people tend to overestimate their abilities relative to the average when assessing their
relative skill (Larwood and Whittaker, 1977). Roll (1986) was the first study in the corporate
finance literature to investigate the effects of managerial overconfidence. Gervais et al. (2003)
4
The term ‘optimism’ is sometimes used to describe the ‘better-than-average’ effect. Following
Malmendier and Tate (2004), we use the term ‘overconfidence’ to refer to both the ‘better-than-average’
and ‘narrow-confidence-intervals’ effects. ‘Optimism’ is defined as a general overestimation of
exogenous outcomes, such as may occur at the outbreak of a war.
7
argue that managers may be more overconfident than the general population because of
selection bias. That is, people who seek managerial positions are more likely to be those who
are overconfident about their ability as a future manager.
There are two main objections to the proposition that managers are overconfident. The
first objection is that irrational managers will be “arbitraged” away through takeovers or other
mechanisms. However, corporate takeovers involve extremely high transaction costs and
arbitrageurs will need to bear large idiosyncratic risks, thus severely limiting the power of
arbitrage (Heaton, 2002). Moreover, if managerial irrationality is a widespread phenomenon,
then there is no guarantee that the replacement manager will be rational (Paredes, 2004).
Further, a firm’s internal incentive mechanisms may not eliminate managerial irrationality
(Goel and Thakor, 2000; Heaton, 2002). The second objection is that irrational managers will
learn from experience to become rational. However, the feedback from corporate financial
decisions is typically infrequent, slow and noisy. Under these circumstances, it is less likely
that agents will learn from experience (Brehmer, 1980; Heaton, 2002). Importantly, both
objections are weaker in a corporate finance setting than in the setting of financial markets
(Heaton, 2002; Gervais et al., 2003).
The empirical evidence on the role of overconfidence in acquisition decisions is limited.
Lys and Vincent (1995) adopt a case study approach to analyze AT&T’s acquisition of NCR.
They suggest that the massive value destruction that resulted from that acquisition could be
attributed to managerial hubris. Hayward and Hambrick (1997) test Roll’s hubris hypothesis.
They argue that the psychological effects of strong recent firm performance, media praise for
the CEO, and high relative CEO compensation will result in hubris. They find strong evidence
that the hubris of CEOs leads them to overbid for targets.
8
Malmendier and Tate (2004) study the relationship between managerial overconfidence
and acquisitions.5 They assume that there exist only two types of CEOs: rational (nonoverconfident) CEOs and overconfident CEOs.6 They argue that the behavior of overconfident
CEOs differs from the behavior of rational CEOs in two ways. First, overconfident CEOs
overestimate the potential synergies of a proposed acquisition because they believe that their
leadership skills are “better than average”. They may also overestimate potential synergies
because they fail to perceive some of the risks involved in an acquisition due to the “illusion of
control” over its outcome. Second, overconfident CEOs mistakenly believe that their
company’s equity is undervalued by the market. This belief arises because overconfident CEOs
overestimate the future returns that could be generated under their leadership.
A rational CEO will decide to acquire another firm if the value of the synergies that will
accrue to the acquiring firm’s stockholders is greater than zero. The rational CEO is also
indifferent between financing the merger with cash, equity or a combination of cash and equity.
In contrast, the acquisition decision of an overconfident CEO depends on the means of
financing, due to the perceived cost of external finance. An overconfident CEO will decide to
acquire whenever perceived merger synergies exceed the perceived loss from issuing
undervalued equity. Therefore, Malmendier and Tate do not predict an unambiguous
relationship between CEO overconfidence and corporate acquisitiveness. However, in their
empirical work, Malmendier and Tate find strong evidence of higher average acquisitiveness
among overconfident CEOs. This finding is consistent with Roll’s (1986) hubris hypothesis,
5
Heaton (2002) had previously analyzed the effect of managerial overconfidence on corporate
investment and managerial resistance in takeovers.
6
It is also assumed that managers invest in all projects that they believe have a positive net present value
and never invest in projects that they believe have a negative net present value.
9
which unambiguously predicts that overconfident CEOs will make more acquisitions than
rational CEOs.7
Our proxy for CEO overconfidence relies on trait theory, which is regularly used by
psychologists to measure and explain personality. Traits constitute underlying personality
dimensions on which individuals vary. Allport and Odbert (1936) compiled a list of 18,000
words from Webster’s dictionary that could be described as traits. Over the years, researchers
reduced the number of traits in the list using factor analysis. Most trait theorists agree that the
original list can be reduced to just five traits, known as the Big Five Factors or the Five Factor
Model (FFM) (Goldberg, 1981, 1993; John, 1990; McCrae and Costa, 1990, 1997).
The FFM has been arrived at by many independent studies using different data sets and
has been found to be universal across cultures. This conclusion has prompted some
psychologists to claim that they have uncovered general laws of personality structure. The five
factors are openness, conscientiousness, extroversion, agreeableness and neuroticism. Each of
the factors represents several highly correlated sub-factors or traits. Each factor is measured on
a continuous, normally distributed scale. The factors and traits are listed in Table 1.
INSERT TABLE 1
Quantifying overconfidence is problematic as there is no instrument to directly measure a
personality trait. Hayward and Hambrick (1997) use three proxies: recent stockholder returns to
measure recent organizational success; content analyses of major newspapers and magazines
about CEOs to measure media praise for the CEO; and CEO compensation relative to the
second-highest paid officer to measure CEO self-importance. Malmendier and Tate (2004) use
7
In Roll (1986), the hubris of managers does not result in managers believing that their firm’s equity is
undervalued by the market.
10
two measures; the first is based on how long a CEO holds company options and the second is a
press coverage proxy.8
The proxy we use for CEO overconfidence is based on media coverage.9 To classify
CEOs as overconfident or rational, data were collected on how the leading business press in
Australia portrayed each individual CEO during the sample period. For each individual CEO,
five separate searches were conducted of the Factiva database.10 The searches were for specific
personality traits based on the Five Factor Model of Personality outlined in Table 1. For each
CEO in the sample, a record was made of the number of articles during the sample period that
portray the CEO as (a) “confident” (b) “optimistic” and (c) “reliable”, “cautious”,
“conservative”, “practical”, “frugal”, “disciplined”, “conscientious”, “not confident” or “not
optimistic”. Also recorded was the total number of articles which mention the CEO during the
sample period.
Like Hayward and Hambrick (1997), we have chosen to construct a continuous variable
and, because some CEOs are mentioned in the press more often than others, our measure of
overconfidence (oc) is expressed in relative terms:
oc =
( a ) + (b )
(c )
(1)
A potential limitation of this measure of overconfidence is that managers may attempt to
project an aura of false confidence to the press in order to mislead investors and keep their
stock price high (Malmendier and Tate, 2004). However, Malmendier and Tate suggest that it
would seem unlikely that managers would pursue this strategy in the long term because
eventually the manager’s credibility would be questioned. A second potential limitation is that
8
See Malmendier and Tate (2005) for a review of overconfidence measures.
Data limitations prevent the use of a proxy based on the length of time a CEO holds company options.
10
The Factiva search was conducted via the Westlaw database which has a subscription to Factiva. The
publications searched were The Age (Melbourne), Australian Financial Review, Sydney Morning Herald
and Business Review Weekly. A more detailed description of the press search is given in Appendix B.
9
11
managers may try to “hype” major corporate events such as acquisitions to improve their
chances of success. Malmendier and Tate (2004) contend that for managerial hyping to be
successful, the CEO would need to be mentioned as confident or optimistic in the press a
relatively large number of times. To control for managerial hyping, we include in the regression
a control variable (total), which is the number of articles that mention the CEO during the
sample period.
3.2
CEO dominance
Following Haleblian and Finkelstein (1993), we define ‘dominance’ (or ‘power’) as the
capacity of an individual to exert their will. CEO dominance may be an important factor in
acquisition behavior since the CEO is typically the most powerful member of the corporate
elite (Jensen and Zajac, 2004).
Dominance differs from overconfidence. Overconfidence is a personality trait and
therefore is intrinsic to the individual. Dominance is in principle an objective fact of behavior.
It is the demonstrated ability of one person to impose their will on others. Hence, dominance
has meaning only in a social or organizational context.11 Dominance may follow from
overconfidence, but not all overconfident CEOs will be able to be dominant. In a corporate
context, a decision in which a dominant individual is very likely to wish to exert their
dominance is in the determination of their personal compensation.
In their analysis of the relationship between governance structures and acquisition
behavior, Jensen and Zajac (2004) include a control variable for CEO power. They argue that
this is necessary to prevent differences in effects across governance positions being confounded
11
To illustrate, Robinson Crusoe could have been overconfident before the arrival of Man Friday but he
could not have been dominant.
12
by differences in CEO power. In similar vein, we argue that a variable for CEO dominance is
needed to prevent the effects of different levels of CEO overconfidence being confounded by
different levels of CEO power. Thus, we argue that both CEO overconfidence and CEO
dominance must be included when testing for the significance of CEO hubris in corporate
acquisition behavior.
The annual compensation of the CEO may be considered an estimate agreed to by the
board of the value of that person’s contribution to the firm for the year. Paredes (2004) argues
that large executive compensation packages are paid against the backdrop of a corporate
governance system which is characterized by deference to the CEO. As noted in his summary of
the normative executive compensation debate, Paredes (2004, p. 32) observes that according to
one stream of the literature, “huge” CEO pay reflect a board that is shirking its responsibility by
not exercising due care in overseeing and negotiating executive pay.
Our main proxy for CEO dominance (dom1) is the natural logarithm of the ratio of CEO
total annual remuneration to the firm’s total assets:
⎛ ceo remuneration ⎞
⎟
⎝ total assets
⎠
dom1 = log ⎜
(2)
CEO remuneration is the most significant validation and form of recognition a chief executive
receives, and high compensation is more salient than other possible measures of a CEO’s
success and value to the firm (Paredes, 2004). CEO remuneration is calculated as base salary +
directors fees + performance bonuses + allowances and non-cash benefits. Total assets is a
measure of the size of the firm. A high ratio of CEO compensation to total assets indicates that
the firm expects a very large contribution from that person compared to the size of the firm
and/or that the CEO has considerable influence over the decisions of the board.
A possible limitation of this measure is that it is based on the assumption that CEOs who
exert their power in one area (determination of their compensation) will exert their power in
13
another area (acquisition decisions). While CEOs are usually concerned about their personal
compensation, it is of course possible that a CEO might care little about personal compensation
but be enthusiastic about acquisitions: megalomaniacs do not necessarily want to be rich.
Whether this possibility arises frequently enough to be a problem is an empirical question.
Although dom1 is our preferred proxy for CEO dominance, as a robustness test we also
proxy CEO dominance with a non-continuous (ordinal) variable:
dom2 = observations of dom1 ranked in ascending order
(3)
Results using dom2 will test whether our main results are driven by outliers or other
discontinuities in the data.
4.
Methodology
We have two main hypotheses: (i) an overconfident CEO has a positive effect on the
probability of the firm conducting an acquisition and (ii) a dominant CEO has a positive effect
on the probability of the firm conducting an acquisition. We test these hypotheses using logistic
regression and pooled cross-sectional time series data. The main dependent variable is acq,
which equals 1 if the CEO made at least one successful acquisition in a particular firm-year. In
subsequent tests we also employ two other dependent variables: (i) dacq, which equals 1 if the
firm made at least one successful diversifying (unrelated) acquisition during a particular firmyear and (ii) racq, which equals 1 if the firm made at least one successful related (nondiversifying) acquisition during a particular firm-year.12
In order to isolate the effects of CEO overconfidence and CEO dominance, it is
necessary to control for the confounding effects of firm characteristics and other potentially
12
An acquisition is defined as diversifying if the acquiring and target firms do not share a primary 2-digit
Standard Industry Classification (SIC) code. The results for racq are not reported as the results are not
significant.
14
important factors in the decision to acquire. We therefore include Tobin’s Q as a control for
growth and investment opportunities, the proportion of independent directors as a control for
effective corporate governance, and cash flow as a control for different levels of internal
resources. Managers’ stock and option ownership is included as a control for its incentive
effects on managers. The variable size (natural logarithm of the book value of assets) is
included to control for firm size.13 To control for the possibility of merger waves in particular
years we include year dummies.14 Similarly, we used dummy variables to control for possible
industry effects. We classified all firms into one of four major industry groups: financial
services (industry group 1), construction and manufacturing (industry group 2), transport and
retail services (industry group 3) and mining (industry group 4).15
The logistic identity (random effects) to be estimated is:
I = β0 + β1oc + β2dom + β3gov + β4q + β5cf + β6owner + β7size + β8total +
β9 D1994 + β10D1995 + β11D1996 + β12D1997 + β13D1998 + β14D1999 +
β15D2000 + β16D2001+ β17D2002 + ε
(4)
where
I
= acq, dacq or racq
oc
= proxy for CEO overconfidence (equation 1)
dom
= dom1 or dom2, which are proxies for CEO dominance (equations 2 and 3)
gov
= proxy for effective corporate governance, defined as the proportion of nonexecutive directors on the board
q
= proxy for growth and investment opportunities, defined as the market value of
assets divided by the book value of assets
13
Moeller et al. (2004) argue that agency problems and hubris may be more prevalent in larger firms.
It is widely accepted that mergers tend to occur in waves. Gorton et al (2005) provide a review of this
literature.
15
Originally, the industry classifications were based on the 7 SIC categories. These results were not
significant. We then combined categories to reduce the number of industry groups to 4 and reran all
regressions. The results remained insignificant but are reported in Tables 6 and 8. The reference industry
for the dummy variables is financial services.
14
15
cf
= cash flow, which is a proxy for corporate resources, and is defined as net profit
after tax before abnormal items plus depreciation, all divided by the book value
of assets
owner
=
size
= natural logarithm of the book value of assets
total
= total number of articles that mention the CEO during the sample period
control for ownership incentive effects, defined as the number of ordinary shares
of the company in which the CEO has a beneficial interest, whether through
partly paid shares, fully paid shares, or stock options, divided by the total number
of shares outstanding
D(year) = dummy variables which equal 1 if an acquisition occurred in the year specified,
where year equals 1994, 1995, …, 2002.
Robustness is tested in two ways. First, industry dummies were added to equation (4) to test for
industry effects. Second, an alternate measure of CEO dominance was included (dom2).
The sample period is January 1, 1994 to December 31, 2003. The regression is estimated
on unbalanced panel data for 65 firms. The initial sample consisted of all firms that were
included in the S&P/ASX 50 Index at the start of the sample period. Firms which did not have
annual report data in the Connect 4 database for at least two years were removed. Firms that
were subsequently included in S&P/ASX 50 Index during the sample period were added to the
sample in the year that they were included in the index. Firms that were excluded at any time
from the S&P/ASX 50 Index were not removed from the sample unless they were delisted. If a
firm was delisted it was removed from the sample in the year in which it was delisted. For each
firm in the sample acquisitions data were collected on an annual basis using Thomson Financial
Securities Data Corporation (SDC) database.
16
5.
Data sources and descriptive statistics
5.1
Data sources
The following data were collected from annual reports accessed through the Connect 4
database.
The name of the CEO: The CEO for each firm during each year of the sample period was
identified. If a firm temporarily did not have a CEO at the time of publishing its annual report,
the previous year’s CEO was assumed to still be the CEO.16
Remuneration data: Data were collected on the total compensation of each CEO for each
year of the sample period.17 Contractual termination payouts were excluded from the measure
of total compensation. If it was unclear whether there had been a termination payout, or if there
was uncertainty about the amount of a payout, no data were recorded for that observation.
Before 1998, CEO and executive officer compensation were reported in bands of $10,000,
beginning at $100,000 for most firms. Following Defina et al. (1994), the mid-point of the
relevant compensation band was recorded and it was assumed that the highest paid company
director was the CEO.
Share ownership: Annual data relating to each CEO’s beneficial interest in the firm’s
ordinary share capital were collected from the annual reports.
Board structure: The proportion of independent directors on each firm’s board of
directors was recorded. An independent director is defined as a non-executive director (i.e. not
a current employee of the firm).
Other data were obtained from a variety of sources, as follows.
16
If remuneration and ownership data regarding that CEO were unavailable in the annual report for that
year, then that observation was dropped.
17
‘Firm’ is defined here as the consolidated entity.
17
Market data: Market data were obtained from IRESS.
Press coverage variables: A CEO is classified as overconfident depending upon his/her
press coverage throughout the sample period and hence his/her classification does not change.
However, if a firm changes its CEO, then its classification may change from one managed by
an overconfident CEO to one not managed by an overconfident CEO or vice-versa.
All financial variables are constructed from annual observations for each firm and each
CEO during the sample period. Detailed definitions of all variables are provided in Table 2.
INSERT TABLE 2
5.2
Descriptive statistics
Table 3 presents a correlation matrix of all variables.
INSERT TABLE 3
The measure of CEO overconfidence (oc) has a correlation with the measure for CEO
dominance of only 0.2430 This finding indicates that the proxy for CEO dominance captures
different attributes to the proxy for CEO overconfidence and therefore is not merely an
alternative proxy for CEO overconfidence. The data in Table 3 also suggest that larger firms
have a higher proportion of independent directors, while CEOs with higher levels of stock
ownership are associated with a less independent board of directors.
Descriptive statistics of the data used to construct all variables are presented in Table 4.
INSERT TABLE 4
Panel A of Table 4 provides descriptive statistics for the main dependent and independent
variables. In Panel B an overview of the press data is presented. The mean number of articles
which mention the CEO is 509 and the median 304, which indicates that we have an adequate
number of articles from which to make a classification regarding overconfidence for the vast
majority of CEOs. Summary statistics of the acquisitions data are presented in Panel C.
18
6.
Results
For each regression, parameter estimates, p-values and the exponential of β are reported.
In section 6.1 results for all acquisitions are discussed and in section 6.2 results for diversifying
acquisitions are discussed.
6.1
All acquisitions
The results presented in Table 5 are based on the total sample of acquisitions. They
provide direct empirical evidence on the extent and importance of CEO overconfidence and
CEO dominance in firm acquisition behavior. The results of robustness tests using industry
dummies and an alternative proxy for CEO dominance (dom2) are presented in Table 6.
INSERT TABLES 5 AND 6
Our central results are shown in Table 5 under specification 1, which includes both the
overconfidence proxy (oc) and the dominance proxy (dom1). The likelihood ratio statistic is
highly significant (1% level), while the overconfidence proxy (oc) is significant at the 5% level
and the dominance proxy (dom1) is significant at the 1% level. In the robustness test reported
in Table 6 (Panel A) industry dummies are included. The rationale for this test is that some
industries might attract overconfident and/or dominant CEOs more often than others, in which
case it could be argued that our proxies for overconfidence and dominance are merely picking
up an industry effect. In specification 1A the significance of the overconfidence proxy (oc)
decreases to 10% and the significance of the dominance proxy (dom1) remains at 1%. None of
the industry dummies is significant. In the other robustness test reported in Table 6 the
19
alternative proxy for CEO dominance (dom2) is used. The significance of the overconfidence
proxy (oc) is 5% and the significance of the dominance proxy (dom2) is also 5%.
The importance of including proxies for both CEO overconfidence and CEO dominance
is demonstrated by examining the results for specifications 2A and 3A in Table 5. In
specification 2A, CEO dominance (dom1) is excluded and CEO overconfidence (oc) is found to
be significant at the 1% level. In specification 3A, CEO overconfidence (oc) is excluded and
CEO dominance (dom1) is found to be significant at the 1% level. The addition of industry
dummies (specifications 2A and 3A in Table 6) reduces the significance of CEO
overconfidence to 5% but the significance of dominance remains at 1%. Again, none of the
industry dummies is significant.
The relative importance of CEO overconfidence and CEO dominance in acquisition
behavior is best demonstrated through the effect that a change in each variable has on the odds
and probability of an acquisition. Considering specification 1, for overconfidence (oc), the
effect is to increase the odds by a factor of 1.09. A 1-unit change in dominance (dom1)
increases the odds by a factor of 4.5. The probability of a firm undertaking an acquisition is
calculated at 23.23%.18 If a firm acquires an overconfident CEO, the effect is to increase the
probability of it making an acquisition by 1.6 percentage points (from to 23.23%. to 24.83%).
For a 10% increase in the variable to proxy CEO dominance (dom1), the probability of a firm
making an acquisition increases by 2.5 percentage points from 23.23% to 25.89%.19
When the influence of CEO overconfidence is considered without a proxy for CEO
dominance (specification 2) the acquisition of an overconfident manager increases the odds by
a factor of 1.12. However, when CEO dominance is considered without a proxy for CEO
overconfidence (specification 3) a 1-unit change increases the odds by a factor of 5.8. Thus, for
18
All probabilities are calculated at the means of the variables as reported in Table 4.
The 10% increase was calculated as a 10% increase in the ratio of CEO compensation to total assets
and then converted to the natural logarithm.
19
20
a 10% increase in the variable (dom1) the probability of an acquisition increases by 3.2
percentage points from 23.5% to 26.7%.
Several of the control variables in specification 1A are also significant. Larger firms and
firms with higher values of Tobin’s Q are more likely to make acquisitions. This result is
expected as larger firms should be less financially constrained and should have a greater
capacity than smaller firms to make an acquisition. Cash flow is found to have a significantly
negative effect on acquisitiveness.20 This result is unexpected as cash flow should be an
indicator of internal resources. However, further robustness tests show that this result is
sensitive to the definition of cash flow.21 If cash flow is normalized by capital instead of assets,
the effect becomes marginally positive but is not significant. Using cash balances instead of
cash flow as an indicator of internal resources results in a highly significant positive effect on
acquisitiveness. Overconfidence and dominance remain highly significant under both
alternative specifications.
The CEO’s stock and option ownership levels (owner) are not found to have a significant
effect on acquisitiveness. This finding may indicate the ineffectiveness of stock and option
holdings as an incentive mechanism, but this interpretation is subject to two caveats. First, we
are unable to differentiate between value-creating and value-destroying acquisitions. Second, as
discussed by Sanders (2001), the risk-return characteristics of stock ownership and stock option
ownership are fundamentally different. Sanders acknowledges that a common theme in the
literature is to view CEO stock option ownership as a substitute for CEO stock ownership.
However, Sanders argues that the different risk-return characteristics may have different effects
20
21
As the coefficient of cash flow (cf) is very large, its significance may be biased upwards.
Detailed results of these robustness tests are presented in Appendix A.
21
on acquisition activity and his results indicate a negative (positive) association between CEO
stock (stock option) ownership and firm acquisition activity respectively.
An important finding is that effective corporate governance, as measured by a higher
proportion of independent directors on the board (gov), significantly mitigates acquisitiveness.
Previous research (Heaton, 2002), has suggested that an independent board of directors may be
an effective way to mitigate CEO overconfidence. The findings provide empirical support for
that proposition.
6.2
Diversifying acquisitions
As discussed in section 2, the theoretical expectation is that agency costs and/or
managerial hubris are likely to be more important in the case of diversifying takeovers. Our
empirical analysis confirms this expectation. The main results for diversifying acquisitions are
presented in Table 7 and the robustness tests in Table 8. Following the format used in the
previous subsection, results are presented for three specifications in Table 7 and for
specifications 1 and 3 in Table 8. Equation (4) is estimated with the dependent variable (dacq)
taking the value 1 if the CEO completed a diversifying acquisition in a particular firm-year.
INSERT TABLES 7 AND 8
The likelihood ratio statistics reported in Table 7 are highly significant for all three
specifications. In specification l, the proxy for overconfidence (oc) is no longer significant, (pvalue = 0.15), whereas the proxy for dominance (dom1) is significant at the 1% level. In
specification 2, the proxy for CEO overconfidence (oc) is significant at the 5% level. In
specification 3, the proxy for CEO dominance (dom1) is significant at the 1% level.
In Table 7, specification 1, a 1-unit change in CEO dominance (dom1) increases the odds
by a factor of 13.4, a much larger factor than that estimated for CEO overconfidence (1.08).
22
From the coefficients reported in specification 1, the probability of a firm undertaking a
diversifying acquisition is 11.5%. If a firm acquires an overconfident CEO the effect is to
increase the probability of a diversifying acquisition by 0.8 of a percentage point (from to
11.5%. to 12.3%). For a 10% increase in the variable to proxy CEO dominance (dom1), the
probability increases by 2.8 percentage points from 11.5% to 14.3%. However, the influence of
CEO dominance in a diversifying acquisition is most clearly demonstrated in specification 3.
For a 10% increase in the variable to proxy CEO dominance (dom1), the probability of a firm
completing a diversifying acquisition increases by 7.8 percentage points (from 12.9% to
20.7%).22
Our results support the proposition that the dominance variable captures the ability of the
CEO to impose his/her views on firm decisions. The less justifiable the acquisition (as is
arguably the case for diversifying relative to related acquisitions), the more important is
dominance relative to overconfidence. It indicates that studies attempting to measure the effects
of CEO overconfidence should control for CEO dominance, especially in the case of
diversifying acquisitions.
7.
Implications and conclusion
This study investigates the roles of CEO overconfidence and CEO dominance in the
decision to undertake an acquisition in the Australian market place. We argue that it is
important to capture not only the extent of overconfidence but also the likelihood that the CEO
will be able to impose his or her overconfident views on the firm’s decisions. The results
22
As a robustness test an estimation of Equation (4) was also conducted for related acquisitions. The
results were not significant and are not reported. Acquisitions are classified as related if the acquirer and
target share a primary 2-digit SIC code.
23
suggest that both CEO overconfidence and CEO dominance are important in explaining the
decision to acquire another firm. CEO dominance is particularly important in the case of
diversifying acquisitions, with the probability of a diversifying acquisition almost doubling
with a 10% increase in CEO dominance. When compared to existing US studies, the evidence
on CEO overconfidence is robust across two different countries, two different time periods and
two different financial and corporate governance systems. Our results also indicate that CEO
dominance is at least as significant as CEO overconfidence in the explanation of the acquisition
decision. Therefore, future research into the role of CEO overconfidence in the acquisition
decision should control for CEO dominance.
A higher proportion of independent directors on the board mitigates the effect of CEO
overconfidence and CEO dominance and reduces the probability of the firm deciding to make
an acquisition. If the effect of CEO overconfidence in making potentially value-destroying
acquisitions is a concern to stockholders and corporate regulators, then the findings suggest that
a possible solution may be to ensure that there is an independent board of directors.
24
References
Allport, G.W. and Odbert, H.S. (1936.) Trait names: a psycho-lexical study. Psychological Monographs
47 (211).
Anand, A., (2005). Voluntary vs mandatory corporate governance: Towards an optimal regulatory
framework, American Law and Economics Association Annual Meetings, 15th Annual Meeting,
May, New York.
Andrade, G., Mitchell, M. and Stafford, E. (2001.) New evidence and perspectives on mergers. Journal
of Economic Perspectives 15: 103-120.
Barberis, N. and Thaler, R. (2003.) A survey of behavioral finance. In: Handbook of the Economics of
Finance. North Holland, Amsterdam, pp. 1052-1121.
Berger, P. and Ofek, E. (1995.) Diversification's effect on firm value. Journal of Financial Economics
37: 39-66.
Bhagat, S., Dong, M., Hirshleifer, D. and Noah, R. (2004.) Do tender offers create value? New methods
and evidence. Dice Center Working Paper 2004-4, Ohio State University, Fisher College of
Business.
Brailsford, T., Oliver, B. and Pua, S. (2002.) On the relationship between ownership structure and capital
structure. Accounting and Finance 42: 1-26.
Brehmer, B. (1980.) In one word: not from experience. Acta Psychologica 45: 223-241.
Defina, A., Harris, T. and Ramsay I. (1994.) What is reasonable remuneration for corporate officers? An
empirical investigation into the relationship between pay and performance in the largest
Australian companies. Company and Securities Law Journal 12: 341-356.
Edey, M. and Simon, J. (1996.) Australia’s retirement income system: Implications for saving and capital
markets. Reserve Bank of Australia, Research Discussion Paper 96-03.
Frank, J.D. (1935.) Some psychological determinants of the level of aspiration. American Journal of
Psychology 47: 285-293.
Gervais, S., Heaton, J. and Odean, T. (2003.) Overconfidence, investment policy, and executive stock
options. Unpublished working paper. Duke University, The Fuqua School of Business.
Goel, A. and Thakor, A. (2000.) Rationality, overconfidence, and leadership. Unpublished working
paper, Michigan Business School.
Goldberg, L. (1981.) Language and individual differences: the search for universals in personality
lexicons. In: Wheeler, L. (ed.) Review of Personality and Social Psychology. Sage Publications,
Beverly Hills, CA.
Goldberg, L. (1993.) The structure of phenotypic personality traits. American Psychologist 48: 26-34.
Gorton, G., Kahl M. and Rosen R. (2005.) Eat or be eaten: A theory of mergers and merger waves. The
Rodney L. White Center for Financial Research, Wharton School, Working Paper 18-05.
Haleblian, J. and Finkelstein, S. (1993.) Top management team size, CEO dominance and firm
performance: the moderating roles of environmental turbulence and discretion. Academy of
Management Journal 36 (4): 844-863.
Hayward, M. and Hambrick, D. (1997.) Explaining the premiums paid for large acquisitions: Evidence of
CEO hubris. Administrative Science Quarterly 42: 103-127.
Heaton, J. (2002.) Managerial optimism and corporate finance. Financial Management 31: 33-45.
Jensen, M. and Zajac, A.J. (2004.) Corporate elites and corporate strategy: How demographic
preferences and structural position shape the scope of the firm. Strategic Management Journal
25: 507-524.
John, O. (1990.) The big five factor taxonomy: dimensions of personality in the natural language and in
questionnaires. In: Pervin, L. (ed.), Handbook of personality: Theory and research. Guilford
Press, New York, pp. 66-100.
Lamba, A. and Stapledon, G. (2001.) The determinants of corporate ownership structure: Australian
evidence. The University of Melbourne Public Law Research Paper No. 20.
Lang, L. and Stulz, R. (1994.) Tobin's Q, corporate diversification, and firm performance. Journal of
Political Economy 102: 1248-1280.
Langer, E. (1975.) The illusion of control. Journal of Personality and Social Psychology 32: 311-328.
Larwood, L. and Whittaker, W. (1977.) Managerial myopia: Self-serving biases in organizational
planning. Journal of Applied Psychology 62:, 194-198.
25
Lys, T. and Vincent, L. (1995.) An analysis of value destruction in AT&T's acquisition of NCR. Journal
of Financial Economics 39: 353-378.
Malmendier, U. and Tate, G. (2004.) Who makes acquisitions? CEO overconfidence and the market's
reaction. NBER Working Paper 10813, National Bureau of Economic Research, Cambridge,
MA.
Malmendier, U. and Tate, G. (2005.) Does Overconfidence Affect Corporate Investment? CEO
Overconfidence Measures Revisited, European Financial Management, 11: 649-659.
Maquiera, C., Megginson, W. and Nail, L. (1998.) Wealth creation versus wealth redistributions in pure
stock-for-stock mergers. Journal of Financial Economics 48: 3-33.
Martin, J. and Sayrak, A. (2003.) Corporate diversification and shareholder value. Journal of Corporate
Finance 9, 37–57.
McCrae, R. and Costa, P. (1990.) Personality in adulthood. Guilford Press, New York.
McCrae, R. and Costa, P. (1997.) Personality trait structure as a human universal. American Psychologist
52: 509-516.
MIT Laboratory for Financial Engineering (2004.)
forbin.mit.edu/RiskAndPreferences/personalitymodels.jsp
Moeller, S., Schlingemann, F. and Stulz, R. (2004.) Firm size and the gains from acquisitions. Journal of
Financial Economics 73: 201-228.
Morck, R., Shleifer, A. and Vishny, R. (1990.) Do managerial objectives drive bad acquisitions?. Journal
of Finance 45: 31-48.
Paredes, T. (2004.) Too much pay, too much deference: Is CEO overconfidence the product of corporate
governance? Unpublished working paper, Washington University School of Law.
Reserve Bank of Australia (2000.) Australian financial markets: Looking back and looking ahead,
Bulletin, March.
Roll, R. (1986.) The hubris hypothesis of corporate takeovers. Journal of Business 59: 197-216.
Sanders, W. G. (2001.) Behavioral responses of CEOs to stock ownership and stock option pay.
Academy of Management Journal 44: 477-492.
Scharfstein, D. and Stein, J. (2000.) The dark side of internal capital markets: divisional rent seeking and
inefficient investment. Journal of Finance 55: 2537-2564.
Servaes, H. (1996.) The value of diversification during the conglomerate merger wave. Journal of
Finance 51, 1201-1225.
Shleifer, A. and Vishny, R. (1986.) Large shareholders and corporate control. Journal of Political
Economy 94 (3): 461-488.
Walter T. and da Silva Rosa, R. (2004.) Australian mergers and acquisitions since the 1980s: What do we
know and what remains to be done? Australian Journal of Management 29: Editorial, R1.
Weinstein, N. (1980.) Unrealistic optimism about future life events. Journal of Personality and Social
Psychology 39: 806-820.
26
Table 1
The Five Factor Model
The first row identifies the five factors. The next five rows contain the five lower-order traits
(“sub-factors”) for each factor. Within each factor, traits are highly correlated; across factors,
they are not.
Source: MIT Laboratory for Financial Engineering (2004).
Openness
Imagination
Artistic Interests
Emotionality
Conscientiousness
Competence
Orderliness
Dutifulness
AchievementAdventurousness
striving
Intellect
Self-discipline
Liberalism
Cautiousness
Extroversion
Friendliness
Gregariousness
Assertiveness
Agreeableness
Trust
Straightforwardness
Altruism
Activity level
Compliance
Excitement
seeking
Cheerfulness
27
Modesty
Neuroticism
Anxiety
Hostility
Depression
Selfconsciousness
Impulsiveness
Tender-mindedness Vulnerability
Table 2
Variable Definitions
Panel A contains the variable definitions for the 3 dependent variables used when estimating
Equation (4). Panel B contains the definitions of the proxies for CEO overconfidence and CEO
dominance. In Panel C all control variables are defined. Panel D contains the definitions of the
search terms used to construct the proxy for overconfidence. Panel E contains the definitions of
subsidiary variables used in the analysis.
Panel A: Dependent variable
Variable Name
Definition
acq
Binary variable equal to 1 if the firm made at least one eventually successful
acquisition during a particular year.
Binary variable equal to 1 if the firm made at least one eventually successful
diversifying acquisition during a particular year. An acquisition is defined as
diversifying if the acquiring and target firms do not share a primary 2-digit SIC
code.
Binary variable equal to 1 if the firm made at least one eventually successful
related acquisition during a particular year. An acquisition is defined as related
if the acquiring and target firms share a primary 2-digit SIC code.
dacq
racq
Panel B: Measures of overconfidence and dominance
Variable Name
oc
dom1
dom2
Definition
The ratio of the number of “confident” plus “optimistic” mentions divided by
the number of “reliable,” “conservative,” “practical,” “frugal,” “disciplined,”
“conscientious,” “not confident,” and “not optimistic” mentions
natural logarithm of CEO compensation (ceo_pay) divided by book value of
assets
dom2 = observations of dom1 ranked in ascending order
Panel C: Control Variables
Variable Name
Definition
cashnorm
Cash divided by book value of assets
Net profit after tax before abnormal items plus depreciation, normalized by
book value of assets
Net profit after tax before abnormal items plus depreciation, normalized by
capital
Market value of assets divided by book value of assets
Natural logarithm of book value of assets
The number of non-executive directors on the board divided by the total
number of directors on the board
The number of ordinary shares of the company in which the CEO has a
beneficial interest, whether through partly paid shares, fully paid shares, or
stock options, divided by the total number of shares outstanding.
Total number of articles that mention the CEO during the sample period.
cf
cf2
q
size
gov
owner
total
28
Panel D: Press Variables
Variable Name
Definition
confident
optimistic
Number of articles that portray the CEO as “confident”.
Number of articles that portray the CEO as “optimistic”.
Number of articles that portray the CEO as “reliable,” “cautious,”
“conservative,” “practical,” “frugal,” “disciplined,” “conscientious,” “not
confident,” “not optimistic”.
cautious
Panel E: Other variables used to construct independent variables
Variable Name
Definition
market value of assets
market value of equity
Market value of equity plus book value of assets minus book value of equity.
Fiscal year closing price multiplied by total number of shares outstanding.
Total compensation of the CEO in a particular year calculated as base salary +
directors fees + performance bonuses + allowances and non-cash benefits.
Book value of property, plant and equipment
ceo compensaton
capital
29
Table 3: Correlation coefficients
All variables are defined in Table 2.
dacq
racq
oc
dom1
gov
owner
q
size
total
cf
cf2
cashnorm
acq
0.7594
0.6097
0.1017
-0.0422
-0.0669
0.1335
0.0012
0.1737
0.1793
-0.1732
0.1425
0.1223
dacq
racq
1
0.0463
0.0491
-0.0899
-0.0635
0.1443
-0.0679
0.2425
0.2416
-0.2193
0.2459
0.0932
1
0.0639
-0.0010
-0.0278
0.0810
0.0475
0.0410
0.1217
-0.0468
-0.0927
0.0805
oc
1
0.2430
0.0792
0.0541
0.0604
-0.1607
-0.1411
0.0589
0.0306
0.0047
dom1
1
-0.2337
0.1555
0.5512
-0.8647
-0.3366
0.5797
-0.1836
0.2116
gov
owner
1
-0.3862
-0.2096
0.3011
0.0222
-0.0460
-0.0025
-0.0799
30
1
0.2176
-0.0370
0.3670
-0.0394
0.0320
0.0561
q
1
-0.5212
-0.0923
0.5742
-0.0849
0.1562
size
1
0.5447
-0.5474
0.2035
-0.1822
total
1
-0.1776
0.0364
-0.0332
cf
1
-0.2310
0.1473
cf2
1
-0.0230
Table 4
Descriptive statistics
All variables are defined in Table 2.
Financial variables are reported in AUD in Panels A and B and USD in Panel D.
Panel A: Dependent variables and main independent variables
Dependent variables
acq
dacq
0.2745
0.1791
0
0
1
1
0
0
0.4467
0.3839
Mean
Median
Maximum
Minimum
Standard deviation
Main independent variables
oc
dom1
2.7927
-3.6530
2
-3.5353
18
-1.9009
0
-5.2363
2.7822
0.5815
Panel B: Firm variables
Mean
Median
Maximum
Minimum
Standard deviation
gov
0.7604
0.8000
0.9286
0.1250
0.1413
owner
0.0170
0.0014
0.4487
0.0000
0.0619
q
0.9555
0.7960
8.4401
0.0308
0.8805
cf
0.0771
0.0758
0.4465
-0.0187
0.0520
cf2
cashnorm
0.5302
0.0411
0.1680
0.0273
11.0930
0.3977
-3.1131
0.0001
1.0919
0.0502
size
22.6587
22.4463
26.6565
19.1028
1.4812
Panel C: Summary statistics of press data
Mean
Median
Maximum
Minimum
Standard deviation
cautious
7.1558
5
51
1
9.1364
optimistic
2.8884
2
20
0
3.5415
Panel D: Summary statistics of acquisitions data
Number of acquisitions
Number of diversifying acquisitions
Mean deal value US$m
Median deal value US$m
Standard deviation
Stock offers (%)
312
78
380
132
105
13
31
confident
9.8930
8
38
0
8.4865
total
602.1651
378
3981
13
680.1659
Table 5
CEO overconfidence and acquisitiveness
This table presents results for the estimation of Equation (4) using logistic regression (random
effects). The dependent variable is binary where 1 indicates that the firm completed an
acquisition in a particular firm-year. All variables are defined in Table 2. Sample size is 430.
acq = β0 + β1oc + β2dom + β3gov + β4q + β5cf + β6owner + β7size + β8total +
β9 D1994 + β10D1995 + β11D1996 + β12 D1997 + β13D1998 + β14 D1999 +
β15D2000 + β16D2001 + β17 D2002 + ε
Variable
constant
oc
dom1
gov
owner
q
cf
size
total
D1994
D1995
D1996
D1997
D1998
D1999
D2000
D2001
D2002
LR statistic
observations
with acq=1
observations
with acq=0
Specification 1
-11.0528
(0.0091)***
0.0878
(0.0402)**
1.5067
(0.0057)***
-2.1033
(0.0446)**
-1.0397
(0.6404)
0.3602
(0.0489)**
-11.0652
(0.0031)***
0.7567
(0.0026)***
0.0001
(0.5791)
-0.6703
(0.3897)
-0.5060
(0.4723)
-0.8701
(0.2046)
0.4108
(0.4311)
0.1985
(0.6997)
0.5132
(0.3050)
-0.2012
(0.6954)
0.7324
(0.1238)
0.0080
(0.9868)
Exp(β)
0.0000
73.9857
(0.0000)***
312
Specification 2
-3.5609
(0.2571)
0.1150
(0.0059)***
….
….
-1.9418
(0.0522)*
-0.2172
(0.9188)
0.3714
(0.0357)**
-10.1510
(0.0060)***
0.1797
(0.1895)
0.0004
(0.1093)
-1.3352
(0.0666)*
-1.1253
(0.0862)*
-1.4325
(0.0269)**
-0.0194
(0.9683)
-0.1499
(0.7599)
0.0875
(0.8521)
-0.4540
(0.3628)
0.5257
(0.2567)
-0.1098
(0.8166)
65.8078
(0.0000)***
312
118
118
1.0918
4.5116
0.1221
0.3536
1.4336
0.0000
2.1312
1.0001
0.5115
0.6029
0.4189
1.5080
1.2196
1.6706
0.8178
2.0801
1.0080
Exp(β)
0.0284
1.1218
….
0.1435
0.8048
1.4498
0.0000
1.1968
1.0004
0.2631
0.3245
0.2387
0.9808
0.8608
1.0914
0.6351
1.6916
0.8960
Specification 3
-11.7795
(0.0004)***
….
….
1.7623
(0.0005)***
-1.8219
(0.0745)*
-0.2896
(0.8870)
0.3359
(0.0608)*
-10.9748
(0.0024)***
0.8315
(0.0000)***
….
….
-0.6042
(0.4263)
-0.4182
(0.5432)
-0.7217
(0.2789)
0.5437
(0.2854)
0.3204
(0.5246)
0.6526
(0.1795)
-0.0748
(0.8817)
0.8432
(0.0711)*
0.0598
(0.9000)
69.8392
(0.0000)***
312
Exp(β)
0.0000
….
5.8258
0.1617
0.7485
1.3991
0.0000
2.2968
….
0.5465
0.6582
0.4859
1.7224
1.3777
1.9205
0.9279
2.3238
1.0617
118
* significant at 10%; ** significant at 5%; *** significant at 1%; p-values in parentheses.
32
Table 6
Robustness tests: Industry effects and CEO dominance in acquisitions
Panel A provides the results when industry dummies are included in Equation (4). Panel B provides the results when the alternative proxy for CEO
dominance (dom2) is included Equation (4) in place of dom1. In both panels, the dependent variable is binary where 1 indicates that the firm
completed an acquisition in a particular firm-year. All independent variables are defined in Table 2. All estimations are made using logistic regression
(random effects). Sample size is 430.
acq = β0 + β1oc + β2 dom + β3gov + β4q + β5cf + β6owner + β7size + β8total + β9 D1994 + β10D1995 + β11D1996 + β12 D1997 + β13D1998 + β14 D1999
+ β15D2000 + β16 D2001 + β17 D2002 + β17 Dind2+ β18Dind3+ β19 Dind4+ε
Variable
constant
oc
dom1
dom2
gov
owner
q
cf
size
total
D1994
Specification
1A
-11.4997
(0.0113)**
0.0796
(0.0728)*
1.4780
(0.0070)***
….
….
-2.2130
(0.0373)**
-0.9003
(0.6953)
0.3664
(0.0471)**
-10.7944
(0.0110)**
0.7732
(0.0030)***
0.0001
(0.6452)
-0.6708
Panel A
Robustness test using industry dummies
Exp(β)
Specification
Exp(β)
Specification
2A
3A
0.0000
-4.4557
0.0116
-12.4716
(0.2116)
(0.0002)***
1.0828
0.1043
1.1100
….
(0.0165)**
….
4.3842
….
….
1.6923
….
(0.0011)***
….
….
….
….
….
….
0.1094
-2.1186
0.1202
-2.0317
(0.0379)**
(0.0516)*
0.4065
0.0151
1.0152
-0.1825
(0.9945)
(0.9293)
1.4425
0.3837
1.4677
0.3475
(0.0316)**
(0.0542)*
0.0000
-10.2520
0.0000
-10.4135
(0.0143)**
(0.0122)**
2.1667
0.2214
1.2478
0.8530
(0.1549)
(0.0000)***
1.0001
0.0004
1.0004
….
(0.1946)
….
0.5113
-1.3200
0.2671
-0.6110
33
Exp(β)
0.0000
….
5.4322
….
0.1311
0.8332
1.4155
0.0000
2.3466
….
0.5428
Panel B
Robustness test using alternative proxy (dom2)
Specification
Exp(β)
Specification
Exp(β)
1B
3B
-11.5207
0.0000
-13.7225
0.0000
(0.0116)**
(0.0004)***
0.0940
1.0986
….
….
(0.0268)**
….
….
….
….
….
….
….
0.0045
1.0045
0.0055
1.0055
(0.0146)**
(0.0021)***
-2.2988
0.1004
-2.0484
0.1289
(0.0251)**
(0.0392)**
-0.9922
0.3708
0.3520
1.4219
(0.6508)
(0.8577)
0.3819
1.4651
0.3630
1.4376
(0.0308)**
(0.0376)**
-11.0271
0.0000
-10.4494
0.0000
(0.0033)***
(0.0038)***
0.4994
1.6477
0.5911
1.8060
(0.0088)***
(0.0002)***
0.0003
1.0003
….
….
(0.2686)
….
-0.9621
0.3821
-0.8821
0.4139
D1995
D1996
D1997
D1998
D1999
D2000
D2001
D2002
Dind2
Dind3
Dind4
LR statistic
observations
with acq=1
observations
with acq=0
(0.3901)
-0.5065
(0.4724)
-0.8657
(0.2083)
0.4311
(0.4108)
0.2152
(0.6769)
0.5275
(0.2930)
-0.1884
(0.7141)
0.7394
(0.1207)
0.0091
(0.9848)
0.2020
(0.6257)
0.0137
(0.9753)
-0.1020
(0.8353)
74.4995
(0.0000)***
312
118
0.6026
0.4208
1.5389
1.2401
1.6946
0.8283
2.0947
1.0092
1.2239
1.0138
0.9030
(0.0707)*
-1.1148
(0.0900)*
-1.4237
(0.0285)**
0.0108
(0.9825)
-0.1238
(0.8017)
0.1163
(0.8051)
-0.4290
(0.3909)
0.5431
(0.2421)
-0.1026
(0.8283)
0.3331
(0.4094)
0.0681
(0.8750)
-0.0080
(0.9868)
66.7549
(0.0000)***
312
0.3280
0.2408
1.0108
0.8836
1.1234
0.6511
1.7214
0.9025
1.3953
1.0705
0.9921
(0.4236)
-0.4240
(0.5398)
-0.7208
(0.2829)
0.5636
(0.2697)
0.3337
(0.5081)
0.6583
(0.1764)
-0.0699
(0.8896)
0.8428
(0.0719)*
0.0555
(0.9072)
0.2947
(0.4560)
0.0055
(0.9886)
-0.2041
(0.6670)
75.2115
(0.0000)***
312
118
118
* significant at 10%; ** significant at 5%; *** significant at 1%; p-values in parentheses.
34
0.6544
0.4864
1.7570
1.3961
1.9316
0.9325
2.3229
1.0571
(0.2017)
-0.7546
(0.2665)
-1.0867
(0.1028)
0.2245
(0.6553)
0.0157
(0.9750)
0.3566
(0.4621)
-0.3128
(0.5355)
0.6624
(0.1584)
-0.0418
(0.9301)
0.4702
0.3373
1.2517
1.0158
1.4285
0.7314
1.9393
0.9591
(0.2320)
-0.6590
(0.3241)
-0.9134
(0.1614)
0.3674
(0.4581)
0.1481
(0.7637)
0.5153
(0.2793)
-0.1582
(0.7497)
0.7959
(0.0855)*
0.0207
(0.9650)
1.3427
1.0055
0.8153
71.9797
(0.0000)***
312
65.4568
(0.0000)***
312
118
118
0.5174
0.4012
1.4439
1.1596
1.6741
0.8537
2.2164
1.0210
Table 7 Diversifying acquisitions
This table presents results for the estimation of Equation (4) using logistic regression (random
effects). The dependent variable is binary where 1 indicates that the firm completed a
diversifying acquisition in a particular firm-year. Acquisitions are classified as diversifying if
the acquirer and target did not share a primary 2-digit SIC code. All variables are defined in
Table 2. Sample size is 430.
dacq = β0 + β1oc + β2dom2 + β3gov + β4q + β5cf + β6owner + β7size + β8total +
β9 D1994 + β10D1995 + β11D1996 + β12 D1997 + β13D1998 + β14 D1999 +
β15D2000 + β16D2001 + β17 D2002 + ε
Variable
constant
oc
dom1
gov
owner
q
cf
size
total
D1994
D1995
D1996
D1997
D1998
D1999
D2000
D2001
D2002
LR statistic
Specification 1
-19.5014
(0.0006)***
0.0748
(0.1423)
2.5935
(0.0003)***
-2.2681
(0.0941)*
-1.0922
(0.6661)
0.2969
(0.1526)
-13.8624
(0.0032)***
1.2895
(0.0001)***
0.0000
(0.8784)
-0.3284
(0.7380)
-0.5104
(0.5921)
-0.3814
(0.6384)
0.1653
(0.7971)
0.5101
(0.3879)
0.4541
(0.4450)
-0.3288
(0.5936)
0.7753
(0.1476)
-0.2660
(0.6369)
84.0839
(0.0000)
Exp(β)
0.0000
1.0777
13.3759
0.1035
0.3355
1.3457
0.0000
3.6311
1.0000
0.7201
0.6003
0.6829
1.1798
1.6654
1.5748
0.7198
2.1712
0.7665
Specification 2
-5.3914
(0.1453)
0.1113
(0.0199)**
….
….
-2.4879
(0.0407)**
-0.1209
(0.9598)
0.3382
(0.1141)
-12.8999
(0.0062)***
0.2650
(0.1032)
0.0005
(0.0676)*
-1.5107
(0.0871)*
-1.6095
(0.0626)*
-1.3730
(0.0639)*
-0.5774
(0.3234)
-0.1050
(0.8472)
-0.2910
(0.5913)
-0.7339
(0.2145)
0.3963
(0.4375)
-0.4819
(0.3804)
68.8144
(0.0000)
Exp(β)
0.0046
1.1177
0.0831
0.8861
1.4024
0.0000
1.3034
1.0005
0.2207
0.2000
0.2533
0.5614
0.9004
0.7475
0.4800
1.4864
0.6176
Specification 3
-18.9121
(0.0000)***
…..
…..
2.7020
(0.0000)***
-2.0643
(0.1183)
-0.6506
(0.7787)
0.2780
(0.1812)
-13.9400
(0.0023)
1.2845
(0.0000)***
….
….
-0.3471
(0.7155)
-0.5089
(0.5829)
-0.3327
(0.6715)
0.2235
(0.7198)
0.5651
(0.3249)
0.5248
(0.3595)
-0.2552
(0.6711)
0.8410
(0.1076)
-0.2304
(0.7039)
82.0954
(0.0000)
observations with
353
353
353
dacq=0
observations with
77
77
77
dacq=1
significant at 10%; ** significant at 5%; *** significant at 1%; p-values in parentheses.
35
Exp(β)
0.0000
….
14.9096
0.1269
0.5217
1.3205
0.0000
3.6130
….
0.7067
0.6011
0.7170
1.2504
1.7597
1.6901
0.7747
2.3188
0.7942
Table 8
Robustness tests: Industry effects and CEO dominance in diversifying acquisitions
Panel A provides the results when industry dummies are included in Equation (4). Panel B provides the results when the alternative proxy for CEO
dominance (dom2) is included Equation (4) in place of dom1. In both panels, the dependent variable is binary where 1 indicates that the firm
completed a diversifying acquisition in a particular firm-year. All independent variables are defined in Table 2. All estimations are made using logistic
regression (random effects). Sample size is 430.
dacq = β0 + β1oc + β2 dom + β3gov + β4q + β5cf + β6owner + β7size + β8total + β9 D1994 + β10 D1995 + β11D1996 + β12 D1997 + β13D1998 + β14 D1999
+ β15D2000 + β16 D2001 + β17 D2002 + β17 Dind2+ β18Dind3+ β19 Dind4+ε
Variable
constant
oc
dom1
dom2
gov
owner
q
cf
size
total
D1994
Specification
1A
-22.1754
(0.0006)***
0.0435
(0.4189)
2.6948
(0.0003)***
….
….
-2.6969
(0.0622)*
-1.3830
(0.6047)
0.2340
(0.2713)
-7.5341
(0.1633)
1.4340
(0.0001)***
0.0000
(0.9445)
-0.2494
(0.7991)
Panel A
Robustness test using industry dummies
Exp(β)
Specification
Exp(β)
Specification
2A
3A
0.0000
-7.2639
0.0007
-21.3623
(0.1054)
(0.0000)***
1.0445
0.0789
1.0821
….
(0.1164)
….
14.8028
….
….
2.7346
….
(0.0001)***
….
….
….
….
….
….
0.0674
-2.8543
0.0576
-2.5675
(0.0277)**
(0.0685)*
0.2508
-0.1077
0.8979
-1.2877
(0.9651)
(0.5873)
1.2636
0.2867)
1.3320
0.2233
(0.1960)
(0.2921)
0.0005
-8.8505
0.0001
-7.3336
(0.1092)
(0.1653)
4.1956
0.3559
1.4275
1.4066
(0.0775)*
(0.0000)***
1.0000
0.0004
1.0004
….
(0.2141)
….
0.7793
-1.4118
0.2437
-0.2909
(0.1116)
(0.7624)
36
Exp(β)
0.0000
….
15.4041
….
0.0767
0.2759
1.2502
0.0007
4.0821
….
0.7476
Panel B
Robustness test using alternative proxy (dom2)
Specification
Exp(β)
Specification
Exp(β)
1B
3B
-17.8486
0.0000
-20.4544
0.0000
(0.0016)***
(0.0000)***
0.0814
1.0848
(0.0960)*
….
….
….
….
….
….
0.0070
1.0070
0.0079
1.0080
(0.0024)***
(0.0004)***
-2.8706
0.0567
-2.7880
0.0615
(0.0238)**
(0.0222)**
-1.1704
0.3102
0.3762
1.4567
(0.6366)
(0.8616)
0.3388
1.4033
0.3025
1.3532
(0.1256)
(0.1819)
-13.9246
0.0000
-12.8321
0.0000
(0.0041)***
(0.0056)***
0.7583
2.1346
0.8720
2.3917
(0.0012)***
(0.0000)***
0.0003
1.0003
….
….
(0.2400)
….
-0.9804
0.3752
-0.8710
0.4185
(0.2909)
(0.3344)
D1995
D1996
D1997
D1998
D1999
D2000
D2001
D2002
Dind2
Dind3
Dind4
LR statistic
observations with
acq=1
observations with
acq=0
-0.4451
(0.6395)
-0.2643
(0.7459)
0.3258
(0.6184)
0.5936
(0.3215)
0.5560
(0.3525)
-0.2968
(0.6318)
0.7883
(0.1462)
-0.2856
(0.6153)
0.0134
(0.9787)
-0.1752
(0.7530)
-19.5832
(0.9969)
99.8923
(0.0000)***
353
77
0.6408
0.7677
1.3851
1.8106
1.7436
0.7432
2.1997
0.7516
1.0134
0.8393
0.0000
-1.5370
(0.0760)*
-1.2513
(0.0935)*
-0.3929
(0.5096)
0.0568
(0.9188)
-0.1288
(0.8151)
-0.6631
(0.2646)
0.4439
(0.3900)
-0.4514
(0.4136)
0.2455
(0.6063)
-0.0252
(0.9615)
-19.2940
(0.9970)
83.7040
(0.0000)***
353
0.2150
-0.4666
(0.6184)
-0.2517
(0.7537)
0.3423
(0.5898)
0.6073
(0.2979)
0.5726
(0.3241)
-0.2709
(0.6562)
0.8135
(0.1273)
-0.2753
(0.6257)
0.0159
(0.9728)
-0.2641
(0.5674)
-19.6710
(0.9969)
99.2481
(0.0000)***
353
0.2861
0.6751
1.0584
0.8792
0.5152
1.5587
0.6368
1.2783
0.9751
0.0000
77
77
* significant at 10%; ** significant at 5%; *** significant at 1%; p-values in parentheses.
37
0.6271
0.7775
1.4082
1.8354
1.7729
0.7627
2.2557
0.7593
1.0161
0.7679
0.0000
-1.0764
(0.2305)
-0.8702
(0.2582)
-0.2468
(0.6841)
0.1252
(0.8233)
0.0990
(0.8603)
-0.5417
(0.3644)
0.5951
(0.2524)
-0.3889
(0.4820)
….
….
….
….
….
….
78.7356
(0.0000)***
353
77
0.3408
0.4189
0.7813
1.1333
1.1041
0.5818
1.8133
0.6778
….
….
….
-0.9585
(0.2754)
-0.7009
(0.3516)
-0.1094
(0.8546)
0.2531
(0.6475)
0.2589
(0.6399)
-0.3780
(0.5185)
0.7284
(0.1556)
-0.3158
(0.5642)
….
….
….
….
….
….
75.2442
(0.0000)***
353
77
0.3835
0.4961
0.8963
1.2880
1.2955
0.6852
2.0718
0.7292
….
….
….
Appendix A – Table A1: Robustness tests for the definition of cash flow
This table presents the results of two robustness tests for the measurement of cash flow. The
dependent variable is binary where 1 indicates that the firm made at least one successful
acquisition in a particular firm-year. The variables are defined in Table 2. In specification 1, the
measure of internal resources is cf2. In specification 2, the measure of internal resources is
cashnorm. Sample size is 430.
acq = β0 + β1oc + β2dom + β3gov + β4q + β5cf + β6owner + β7 size + β8total +
β9 D1994 + β10 D1995 + β11D1996 + β12 D1997 + β13D1998 + β14 D1999 +
β15D2000 + β16 D2001+ β17 D2002 + ε
Variable
constant
Specification 1
Specification 2
-13.0869
-15.8036
(0.0016)***
(0.0002)***
oc
0.0864
0.1024
(0.0429)**
(0.0170)**
dom1
1.2779
1.1982
(0.0162)**
(0.0259)**
gov
-2.4266
-2.6432
(0.0181)**
(0.0107)**
owner
0.3270
0.5944
(0.8772)
(0.7827)
q
0.0945
0.1035
(0.5557)
(0.5113)
cf2
0.2142
….
(0.0305)**
….
cashnorm
….
7.3849
….
(0.0015)***
size
0.7897
0.8957
(0.0016)***
(0.0004***)
total
0.0001
0.0000
(0.7584)
(0.9311)
D1995
-0.6737
-0.6473
(0.3436)
(0.3592)
D1996
-0.9506
-0.9184
(0.1739)
(0.1846)
D1997
0.4574
0.3732
(0.3801)
(0.4812)
D1998
0.2453
0.2257
(0.6312)
(0.6612)
D1999
0.5366
0.5788
(0.2801)
(0.2485)
D2000
-0.1016
-0.1501
(0.8403)
(0.7683)
D2001
0.7982
0.8036
(0.0896)*
(0.0897)*
D2002
0.0908
0.0170
(0.8489)
(0.9719)
LR statistic
68.5313
74.0311
(0.0000)***
(0.0000)***
Observations with acq = 0
312
312
Observations with acq =1
118
118
* significant at 10%; ** significant at 5%; *** significant at 1%; p-values in parentheses.
38
Appendix B – Description of Press Search
To classify CEOs as overconfident, data was collected on how the leading business
press in Australia portrays each individual CEO during the sample period. The publications
searched were The Age, Australian Financial Review, Sydney Morning Herald, and Business
Review Weekly.
For each individual CEO five separate searches were conducted in the Factiva
database. The search was conducted via the Westlaw database which has a subscription to
Factiva. Factiva on Westlaw is fully searchable using Westlaw search commands. The total
number of articles that referred to the CEO during the sample period was found using the
following search command:
“CEO’s Full Name” & “Company’s Name” & DA(AFT 01/01/1994 & BEF 31/12/2003)
This gave the total number of articles that had both the CEO’s name and the name of the CEO’s
company in the same article, while restricting the search to articles published during the sample
period. For each CEO, the search initially used the full name of the CEO that was reported in
the annual report. However, the results of each search were then checked to ensure that the
CEO in question was not commonly being referred to by another name. For example, Foster’s
CEO Edward Kunkel is invariably mentioned as Ted Kunkel. Similarly, Gerald Harvey of
Harvey Norman, is often referred to as Gerry Harvey. For these CEOs, the number of articles
that mentioned the CEO’s nick-name far exceeded the number of articles that mentioned the
CEO’s actual name. Therefore, for these CEOs, the number of total articles mentioning the
CEO was found by searching for their commonly used full name.
The number of articles that referred to the CEO as being “confident” was found using
the following search command:
“CEO surname” /s “confident” & “CEO’s Full Name” & “company name” & DA(AFT
01/01/1994 & BEF 31/12/2003) % “not confident”
This gave the number of articles which had the CEO’s surname in the same sentence as the
word “confident” in any article which contained both the full name of the CEO and the name of
the CEO’s company. Articles in which the CEO was described as “not confident” were filtered
out of the results of this search. Similar searches were conducted for each of the personality
traits used to construct the variable (oc); “optimistic”, “reliable”, “cautious”, “conservative”,
“practical”, “frugal”, “disciplined” and “conscientious”. A review of over one hundred articles
suggested that the results obtained were highly accurate.
39