Articol HUSON MALATESTA SI PARRINO 2004 PDF

ARTICLE IN PRESS
Journal of Financial Economics 74 (2004) 237275
Managerial succession and rm performance$

Mark R. Husona, Paul H. Malatestab, Robert Parrinoc,*
a
Department of Finance and Management Science, School of Business, University of Alberta,

Edmonton, Alta., Canada T6G 2R6
b
Department of Finance and Business Economics, University of Washington Business School,
University of Washington, Seattle, WA 98195, USA
c
Department of Finance, McCombs School of Business, University of Texas at Austin,
Austin, TX 78712, USA
Received 25 April 2003; accepted 19 August 2003
Abstract
We examine CEO turnover and rm nancial performance. Accounting measures of
performance relative to other rms deteriorate prior to CEO turnover and improve thereafter.
The degree of improvement is positively related to the level of institutional shareholdings, the
presence of an outsider-dominated board, and the appointment of an outsider (rather than an
insider) CEO. Turnover announcements are associated with signicantly positive average
abnormal stock returns, which are in turn signicantly positively related to subsequent
changes in accounting measures of performance. This suggests that investors view turnover
announcements as good news presaging performance improvements.
r 2004 Elsevier B.V. All rights reserved.
JEL classification: G30; G32; G34
Keywords: Corporate governance; CEO turnover; CEO succession; Firm performance
We thank Jonathon Karpoff, Charles Kahn, seminar participants at University of Alberta, Arizona
State University, Clemson University, University of Illinois, Indiana University, University of Maryland,
Rutgers University, University of Texas at Austin, and University of Western Ontario, and an anonymous
referee for helpful comments. Huson acknowledges nancial support from the Canadian Utilities
Fellowship at the University of Alberta and Parrino acknowledges nancial support from the Bureau of
Business Research at the University of Texas at Austin.
*Corresponding author. Tel.: +1-512-471-5788; fax: +1-512-471-5073.
E-mail address: parrino@mail.utexas.edu (R. Parrino).
0304-405X/$ - see front matter r 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jneco.2003.08.002
ARTICLE IN PRESS
238
M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275
1. Introduction
In the year 2000, the largest 500 publicly traded companies in the United States
had combined revenues of $8.4 trillion. Their assets exceeded $21 trillion, and they
employed more than 30 million people.1 Moreover, these gures only hint at the
overall economic signicance of the public corporate sector and the importance of
decisions made by the top managers of public corporations. Given the sheer size of
these rms, there can be little doubt that decisions made by their principal ofcers
can create or destroy wealth on a vast scale. These decisions also have considerable
ramications for the millions of employees and billions of customers.
Because of the key economic role played by top corporate managers, there is a
great deal of interest in how the managerial labor market functions in general, and in
the causes and consequences of managerial succession in particular. An extensive
scientic literature has evolved on the determinants of top management turnover and
stock price reactions to turnover events. Much of this literature debates the
effectiveness of corporate boards in monitoring managerial behavior and in
executing the critical functions of hiring and ring top managers. In this paper,
we contribute to the debate by examining rm performance changes ensuing from
top management turnover.
There is a general consensus that the likelihood of management turnover is
negatively related to rm performance. Warner et al. (1988), for example, nd that
rms with low stock returns are more likely to change their CEO, president, or board
chairman than are other rms. Fee and Hadlock (2003) report that industry-adjusted
rm stock returns are negatively related to CEO turnover and that the likelihood of
turnover among the top ve executives below the CEO level is signicantly higher
when the CEO is dismissed. Coughlan and Schmidt (1985) and Weisbach (1988)
report similar evidence. Kim (1996) derives a simple yet elegant model relating CEO
turnover to performance and demonstrates empirically that rm stock returns have a
persistent negative effect on turnover probability. Weisbach (1988) shows that a
measure of industry-adjusted rm earnings is negatively related to top management
turnover as well. Overall, previous results are consistent with the proposition that
boards of directors monitor corporate performance and act to replace managers of
poorly performing rms.
The evidence on the consequences of these managerial replacement decisions is not
so clear. Nearly all of the evidence comes from event studies of the stock market
reaction to news about top management turnover. Unfortunately, the underlying
theory pertaining to succession events lacks sufcient structure to provide
unambiguous predictions about stock price reaction to turnover news. One might
argue, for example, that turnover presages improved management and higher cash
ow. Firm value, therefore, should increase on the news of management turnover.
It is also plausible, however, that turnover signals that recent management
decisions have proven unsound. To the extent that investors did not know about
1
The gures on revenues, assets, and employment are scal year 2000 gures obtained from the
Standard & Poors Compustat research database.
ARTICLE IN PRESS
239
managements poor performance, rm value might fall on the news of management

turnover. This can occur even if investors believe that the incoming manager will be
superior to the outgoing manager. Hence, it is not surprising that the results of event
studies on management turnover are mixed. Though Bonnier and Bruner (1989) and
Weisbach (1988) observe signicantly positive stock price reactions to turnover
news, Khanna and Poulsen (1995) nd just the opposite result. Reinganum (1985)
and Warner et al. (1988) report small, statistically insignicant price changes
associated with turnover events. On balance, the event study evidence suggests that
investors believe that corporate boards promote benecial changes through their
management succession decisions. This evidence is not, however, especially
compelling.
Despite the potential importance of managerial succession decisions made by
corporate boards, there is very little direct evidence regarding the subsequent
outcomes of these decisions. Stock price reactions around the time of management
turnover reect investors expectations regarding these outcomes but do not
reveal the outcomes themselves. Weisbach (1995) shows that CEO turnover
prompts rms to divest poorly performing business units. However, only Hotchkiss
(1995) and Denis and Denis (1995) examine the relation between turnover and
subsequent changes in operating performance measured using accounting numbers.
Hotchkiss analyzes 197 rms that emerge from Chapter 11 bankruptcy. She
nds that rms whose pre-bankruptcy management retains control are more likely to
le for a second bankruptcy and to have negative operating income after
reorganization. Her evidence indicates that management turnover improves future
performance. We cannot draw general conclusions about turnover from these
results, however, because her sample includes only extremely poorly performing
rms.
Denis and Denis study 908 management succession events occurring between 1985
and 1988. For each event, they analyze the operating rate of return on total assets
(OROA) over the seven-year period centered on the year of the management change.
They nd that average and median industry-adjusted OROA increase over periods
starting one year before and ending two or three years after the management change.
Performance improvements appear to be somewhat larger when directors force
managers out than when managers retire under normal circumstances. Performance
nevertheless improves in both cases.
The results reported by Denis and Denis suggest that management turnover tends
to enhance corporate performance. Moreover, the results are consistent with the
notion that the positive abnormal stock returns observed by some researchers
around succession events reect rational anticipation by investors of subsequent rm
performance improvements. Several issues, however, remain unresolved. First, the
time period spanned by the Denis and Denis sample is very short and corresponds
with a period of a very active takeover market (Comment and Schwert, 1995; Huson
et al., 2001). It is unclear whether the relations they nd apply to other time periods.
Second, Denis and Denis rely on unadjusted OROA and industry-adjusted OROA
to measure performance changes. Hence, the performance improvements they
observe could be due to mean reversion of the accounting performance time series
ARTICLE IN PRESS
240
rather than to management turnover.2 Finally, Denis and Denis contrast forced
resignations and normal retirement but do not further analyze the determinants of
turnover-related rm performance changes, nor do they provide evidence on the
relation between performance changes and turnover announcement stock returns.
In this paper, we extend the research on post-turnover performance in several
ways. Denis and Denis discuss the importance of the external takeover market and
monitoring by inuential shareholders in precipitating CEO turnover. We examine
the actual empirical relation between institutional shareholdings, rm-related
takeover activity, successor CEO origin, board composition, and post-turnover
performance. We also explore whether other publicly available information helps to
predict performance changes. We use the turnover announcement abnormal stock
return as a proxy for publicly available information about the turnover.
We develop predictions about the cross-sectional relation between performance
improvements and rm or succession characteristics such as board composition,
ownership structure, and external takeover activity. We test these predictions using
several approaches, including cross-sectional regression. In our empirical work, we
examine more CEO successions and a longer time period than Denis and Denis and
report results using the performance-based control group matching method
advocated by Barber and Lyon (1996). This method controls for potential mean
reversion of the accounting performance time series, which may affect measures of
performance change around management turnover events. Finally, we employ
econometric methods that control for the survival bias encountered when examining
determinants of post-CEO turnover performance changes.
Our results support the view that deteriorating rm performance triggers
management turnover. On average, unadjusted, industry-adjusted, and control
group-adjusted OROA exhibit statistically signicant declines from three years
before through one year before the turnover year. We nd also that average control
group-adjusted OROA increases signicantly from one year before to three years
after the turnover year. This latter result tends to refute the notion that observed rm
performance improvements are attributable to mean reversion of accounting
performance time series. Instead, the improvements seem to stem from management
turnover and improved managerial quality. The evidence also indicates that postturnover performance improvements were greater in the 19831994 period than in
the 19711982 period, suggesting that improvements in managerial quality were
greater in the latter period.
Our cross-sectional analyses indicate that post-turnover changes in rm OROA
are positively related to institutional ownership and are greater when the board is
dominated by outside directors and when the successor CEOs are rm outsiders. The
relation between board composition and post-turnover changes in OROA strengthens over time, suggesting that outside directors made better CEO replacement
decisions near the end of the period we examine. There is also some evidence that
performance changes are positively inuenced by external takeover pressure. We nd
Denis and Denis comment on this possibility, but address it only indirectly; see p. 1,045 of their paper.
ARTICLE IN PRESS
241
no reliable difference between post-turnover performance changes for forced and

voluntary successions.
We also examine the connection between performance changes and turnover
announcement abnormal stock returns. The evidence shows that announcement-date
abnormal returns are signicantly positively related to subsequent changes in rm
operating performance. Indeed, the abnormal return is signicant in multiple
regressions that include the structural variables we consider. This result suggests that
investors use information beyond that contained in the structural variables to
forecast turnover-induced performance changes.
The rest of the paper proceeds as follows. Section 2 discusses the theories and
hypotheses that we examine. The data are described in Section 3 and the empirical
evidence is presented in Section 4. Finally, conclusions and implications are
discussed in Section 5.
2. Theory and hypotheses

As in Kim (1996), we assume that rm performance over any time interval is the
sum of manager quality and a random component arising from chance. This random
component, which can be thought of as arising from industry and rm-specic (or
manager-specic) shocks, is assumed to have a mean of zero and to be serially
independent. This implies that shocks are transitory and that changes in
performance are negatively related to earlier shocks. The chance-driven component
of performance is mean-reverting.
2.1. Forced turnovers
2.1.1. Improved management hypothesis
The improved management hypothesis holds that forced management turnover
tends to increase managerial quality and therefore expected rm performance. Under
this hypothesis, quality, which is not directly observable, varies across managers.
Firm directors attempt to infer quality from realized performance. If performance is
sufciently poor, the board infers that the incumbent manager is of low quality and
that the expected benet of replacing him exceeds the expected cost. Another
manager is installed whose expected quality exceeds that of his predecessor.
Moreover, poor performance tends to coincide with bad luck as well as low manager
quality. Thus, future performance is expected to increase following the change in
management for two reasons: the expected increment in manager quality is positive
and manager luck is expected to revert to normal.
2.1.2. Scapegoat hypothesis
. (1979),
The scapegoat hypothesis is based on the agency models of Holmstrom
Shavell (1979), and Mirrlees (1976). The hypothesis holds, in contrast to the
improved management hypothesis, that quality does not vary across managers. Poor
performance under the scapegoat hypothesis arises from chance alone rather than
ARTICLE IN PRESS
242
low managerial quality. In other words, poor performance results from bad luck, not
bad management. Under the scapegoat hypothesis, managers are all alike but supply
quality as a function of effort, which is not directly observable. Managers dislike
effort so they must be threatened with dismissal if performance is low. In
equilibrium, all managers supply the same effort (quality) and only those who are
unlucky are red. Boards of directors understand that all managers are alike, but
must re managers of poorly performing rms to induce other managers to provide
the desired level of effort.
Since replacement candidates are of the same quality as the outgoing manager,
turnover itself does not increase managerial quality or expected rm performance.
Consequently, a manager who is red for poor performance can be viewed as a
scapegoat. Even though turnover does not increase managerial quality, the expected
change in rm performance following turnover is positive. This is because turnover is
triggered by improbably poor performance outcomes arising from chance.
Subsequent performance should revert to mean levels.
2.2. Voluntary turnover
Voluntary turnover can arise in the course of normal CEO retirement due to age
or, for example, when a CEO leaves to manage another rm or a government
agency. Such turnover need not be associated with poor prior performance. We do
not suppose that CEOs who leave voluntarily are necessarily of low quality or
recently ill fated. Moreover, the board might appoint the highest expected quality
candidate available to succeed a manager who departs voluntarily, but this does not
ensure that expected managerial quality increases. It could decrease. For these
reasons, changes in expected rm performance after a voluntary turnover should be
smaller than after a forced turnover.
2.3. Determinants of turnover-related performance changes
If quality does not vary across managers, and CEOs are scapegoats, then the only
performance changes we expect to see following turnover arise solely from mean
reversion. Expected performance improvements for rms experiencing turnover
should equal those for other rms with similar past performance. Under the
assumption that quality varies across managers, however, we can extend our
discussion to include determinants of the magnitude of the performance improvement that we might observe from voluntary and forced turnover. We rst consider
whether the new CEO is promoted from inside the rm or hired from outside the
rm. We then discuss the impact of monitoring mechanisms on performance
improvement.
2.3.1. Inside vs. outside succession
Most top management appointees are selected from among rm insiders, those
who are already senior ofcers of the rm. However, this is not always the case. The
decision to promote an insider or hire an outsider depends on the abilities of inside
ARTICLE IN PRESS
243
and outside candidates. Dalton and Kesner (1985) argue that outsiders will not be
appointed unless an incremental improvement relative to inside candidates is
expected, because it is more costly to appoint an outsider.
Chan (1996) discusses the effects on insider incentives of including external
candidates in competition for high positions within a rm. He explains that
increasing the number of candidates in this way can reduce the incentives of lowerlevel executives to exert effort. This is because the marginal effect of effort on the
probability of winning the competition is negatively related to the number of
contestants. Chan argues that, to mitigate the negative incentive effect of open
competition, the rm might award a competitive handicap to inside candidates.
Consequently, a greater increment to expected managerial quality would be required
to induce the appointment of an outsider than of an insider. This implies larger
increases in expected rm performance following the succession of an outsider than
the succession of an insider. Moreover, Borokhovich et al. (1996) report that
turnover announcement abnormal stock returns are signicantly positive for outside
successions, and signicantly negative for inside successions around forced turnover.
This suggests that the appointment of an executive from outside the rm is perceived
as more benecial to stockholders than an inside appointment.
2.3.2. Monitoring mechanisms and firm performance
Evidence on the relation between mechanisms to control agency problems and
rm performance is mixed. Demsetz and Lehn (1985) and Demsetz and Villalonga
(2001) focus on ownership structure as one such mechanism. They argue that
ownership structure is endogenously determined in equilibrium and reects
optimizing behavior on the part of investors and managers. Differences in ownership
structure arise because the optimal structure differs across rms, but there is no
systematic relation cross-sectionally between structure and rm performance. The
empirical evidence presented in these two studies is consistent with this prediction.
Agrawal and Knoeber (1996) study the relation between rm performance and
several different agency control mechanisms simultaneously. They distinguish between
mechanisms whose levels are decided internally, such as insider shareholdings and the
proportion of outside directors on the board, and those whose levels are decided by
outsiders, such as institutional shareholdings. In their view, value-maximizing rms
choose the levels of internally decided mechanisms to equate their marginal costs and
marginal benets. Mechanisms whose levels are decided by outsiders, however, need
not be set to maximize rm value. In this case, cross-sectional variation in rm
performance should be unrelated to the levels of internally decided mechanisms, but
could be related to the levels of mechanisms decided by outsiders. In their empirical
analysis, Agrawal and Knoeber report that rm performance is signicantly related to
board composition, but unrelated to the other control mechanisms that they examine.
The arguments of Demsetz and Lehn (1985), Demsetz and Villalonga (2001), and
Agrawal and Knoeber (1996) can be applied to our study as well. The implication is
that performance improvements following CEO turnovers should be unrelated to the
levels of internally decided control mechanisms, but could be related to the levels of
mechanisms set by outsiders.
ARTICLE IN PRESS
244
Aside from the equilibrium arguments of Demsetz and Lehn and the others cited
above, there may be no systematic relation between rm performance changes and
the levels of monitoring mechanisms. The characteristics that inuence the
effectiveness of monitoring mechanisms can operate in two ways. Better monitoring
can result in the appointment of superior replacement CEOs. Other things equal, this
would produce larger managerial quality increments and larger expected performance improvements. Better monitoring, however, also can permit more rapid and
accurate assessments of incumbent manager quality. As a consequence, relatively
small managerial deciencies can be detected and bring about turnover. In such
cases, there would be little room for quality increases, and expected performance
improvements would be small. Conversely, weak monitors could allow a bad
manager to perform very poorly before replacing him, leading to a large quality
increment and a positive expected performance change. On balance, therefore, the
net effect of better internal monitoring on expected performance improvements
following CEO turnover would be ambiguous.
There is a body of evidence, however, that suggests alternative hypotheses. For
example, Morck et al. (1988) nd a signicant (non-monotonic) relationship between
managerial stock ownership and Tobins Q: Yermack (1996) reports that board size is
negatively related to Tobins Q: Rosenstein and Wyatt (1990) nd that stock price
reactions to the appointment of outside directors are signicantly positive, on average.
Moreover, there is evidence that monitoring mechanism characteristics are related to
the likelihood that the CEO of a poorly performing rm is replaced.3 This evidence
suggests that the effectiveness of monitoring mechanisms, including those that are
decided internally, can affect changes in rm performance following turnover.
By examining the relations between the characteristics of monitoring mechanisms
and the change in rm performance following CEO turnover, we obtain direct
evidence on the impact of these governance characteristics on the quality of CEO
selection decisions. An advantage of focusing on performance changes around
turnover decisions, rather than on rm performance in general, is that we are able to
isolate a specic corporate event in which internal monitors make a decision that
potentially has important implications for the future performance of the rm. Our
tests do not address the general relation between monitoring mechanisms and rm
value, but rather the marginal impact of monitoring mechanisms in the selection of
new CEOs. When CEO turnover occurs, the board has the opportunity to appoint a
successor with greater ability than the departing CEO.
Internal monitoring can be affected by external factors through the market for
corporate control. For example, Mikkelson and Partch (1997) show that management turnover rates are more sensitive to rm performance during periods of
heightened takeover activity. While other researchers, such as Huson et al. (2001), do
not nd a relation between turnover rates and takeover activity, it remains possible
that outside takeover pressure motivated by potential performance improvements
puts pressure on board members to make changes or risk the loss of their board
positions and damage to their labor market reputations. The high cost of the
3
See, for example, Weisbach (1988), Perry (1998), or Huson et al. (2001).
ARTICLE IN PRESS
245
takeover market as a disciplinary mechanism suggests that takeover pressure will

only be observed when the potential gains are substantial. Therefore, if takeover
pressure inuences board decisions, we expect to observe larger improvements
following turnover associated with takeover pressure.
Evidence from prior studies indicates that internal monitoring is affected by board
composition. Weisbach (1988) and Borokhovich et al. (1996) report that outside
directors are more likely than inside directors to re a poorly performing CEO and
to replace him with an executive who will increase rm value. Consequently,
performance improvements following management turnover could be related to the
extent of outsider representation on the board of directors.
Post-turnover performance changes might also depend on the structure of
shareholdings. As Shleifer and Vishny (1986) point out, large stockholders are likely
to have greater incentives to monitor management than do small stockholders. This
is because the benets that large stockholders might receive from monitoring
activities are more likely to exceed the costs they bear. Evidence consistent with this
argument has been reported in several studies. For example, Agrawal and
Mandelker (1990, 1992) nd a positive overall relation between stock price reactions
to announcements of antitakeover charter amendment proposals and the fraction of
rm equity held by outside blockholders. Denis and Serano (1996) report evidence
suggesting that outside blockholders are instrumental in removing poorly performing managers subsequent to failed takeover bids. Denis et al. (1997) show that the
probability of top executive turnover in general is positively related to the presence
of such blockholders.
Many, including Black (1992) and Pound (1992), have contended that institutional
shareholders perform a monitoring function similar to that of blockholders.
Assuming that institutions increase the pressure on boards to make management
replacement decisions that serve stockholder interests, the quality of internal
monitoring would be positively related to institutional shareholdings.
3. Data
3.1. Sample
We begin by identifying all CEOs listed in the Forbes annual compensation
surveys over the 19711995 period who have held their position for one year or less.
This provides a list of CEOs who recently entered ofce. From this list, we construct
our turnover sample, which consists of 1,344 CEO successions at large public rms
from 1971 through 1994 that satisfy the following criteria:
(1) the incumbent and successor were both proled in the Forbes annual
compensation surveys;4
4
This criterion is not strictly met in all cases. In several instances a CEO was appointed and relinquished
the position before the next compensation survey was published. These CEOs were identied from the
succession announcements published in the Wall Street Journal.
ARTICLE IN PRESS
246
(2) the Wall Street Journal reported the succession announcement;

(3) accounting data for the rm are available on the Standard & Poors Compustat
database beginning in the year after the incumbent became CEO; and
(4) the succession was not directly related to a takeover.
Information on the CEOs age, tenure in ofce, and tenure with the rm was
obtained from the Forbes surveys and conrmed by the Wall Street Journal
announcements, various Marquis Whos Who publications, and Dun and
Bradstreets Reference Book of Corporate Managements. Because the Forbes annual
compensation surveys report the compensation of the highest paid ofcer, regardless
of titles held, the Wall Street Journal announcements were used to conrm the date
that a change in the individual holding the CEO title was announced. The reason for
each succession was obtained from the Wall Street Journal announcement and a
review of the business and trade press.
Each succession is classied as either forced or voluntary. If the Wall Street
Journal reported that the CEO was red, forced from the position, or departed due
to policy differences, the succession is classied as forced. If the departing CEO was
under the age of 60, we tentatively classied the succession as forced if the Wall
Street Journal announcement did not report the reason for the departure as death,
poor health, or the acceptance of another position (elsewhere or within the rm).
Similarly, we tentatively classied the succession as forced if the departing CEO was
under 60 and the Wall Street Journal reported that he was retiring, but did not
announce the retirement at least six months in advance. Those cases tentatively
classied as forced were further investigated by searching the business and trade
press for relevant articles to reduce the incidence of classication errors. These
successions are reclassied as voluntary if the incumbent took a comparable position
elsewhere or departed for personal or business reasons that were unrelated to the
rms activities.
We also designate successor CEOs as being either insiders or outsiders to their
rms. New CEOs who have been with their rms for one year or less at the time of
their appointments are classied as outsiders. All other CEOs are classied as
insiders.
In addition, successions are classied according to the extent of external takeover
pressure and corporate control market activity. We examined the Wall Street Journal
Index for evidence of events related to corporate control and takeovers at each rm
during the 12 months immediately preceding the turnover announcement. A
succession is classied as subject to takeover pressure if during this period there was
a proxy ght, takeover bid, or rumor of a bid involving the rm. If the rm adopted
a poison pill or other takeover defense, if there was a board shakeup or a stock
acquisition requiring a 13D ling, or if some other similar event occurred, the
succession is also classied as subject to takeover pressure.
The composition of the board of directors in the succession year was obtained
from the Dun and Bradstreet Million Dollar Directory, Moodys Industrial, Banking
and Financial, Transportation, and Utilities manuals, and proxy statements. All
directors who were employees of the rm are classied as insiders. Two different
ARTICLE IN PRESS
247
classication schemes were used to identify outside directors. In the rst scheme, the
proxy statement immediately preceding each succession was used to identify nonemployee directors who were former ofcers, consultants, commercial bankers,
investment bankers, lawyers, or insurance company executives, or were related to an
ofcer of the rm. All non-employee directors falling into one of these categories are
classied as greys (potentially afliated directors). All other non-employee directors
are classied as outsiders. This classication scheme is similar to those used in other
studies such as Weisbach (1988) and Byrd and Hickman (1992). Unfortunately, due
to the difculty of obtaining corporate proxy statements for the time period prior to
1978, we were only able to obtain proxy data for a subset of 876 turnovers, all of
which took place in 1978 or later. We employed a second, simpler classication
system, using data from the Million Dollar Directory, to obtain board composition
estimates for the full 19711994 sample period. Under this second approach,
directors are classied as insiders if they were ofcers of the rm and outsiders
otherwise.
Information on the stock ownership of CEOs, ofcers, and directors also was
taken from proxy statements for the 876 turnovers. In addition, the CDA/Spectrum
database was used to determine institutional ownership as of the end of the quarter
immediately preceding 737 of the turnovers that occurred after 1979.
Financial data for the seven-year period centered on the turnover year were
obtained from Standard & Poors Compustat les. We collected accounting numbers
for each full year that an incumbent CEO was in ofce. These data are used to
calculate ratios of operating income to book assets (OROA) and to sales (OROS).
For some of our tests we adjust the performance measures for industry effects. Both
of the accounting return measures also were analyzed using the performance-based
control group matching method described by Barber and Lyon (1996).
To control for industry effects, the accounting return measures are adjusted by
subtracting the median value of the corresponding measure for all rms in the
primary two-digit SIC industry in which the rm was active at the time of the
succession. A two-digit industry denition is used because Clarke (1989) has shown
that the two-digit denition captures similarities among rms as effectively as
industry denitions based on three- or four-digit SIC groupings.
For both OROA and OROS, the Barber and Lyon matching method is performed
as follows. Each sample rm is matched to comparison rms with the same two-digit
Compustat SIC code whose performance measures over the year before the turnover
are within 710% of the sample rms performance. If there are no such rms, we
match performance within the 710% lter using all rms with the same one-digit
SIC code. For rms without matches after this procedure, we use all rms with
performance within the lter bounds regardless of SIC code. In all but 1% of the
cases for OROA, the comparison rms and sample rm have the same one-digit SIC
code. Slightly over 91% of the cases are industry-matched at the two-digit level. For
OROS, 98% match at the one-digit level and 81% at the two-digit level. Each sample
rms performance is adjusted by subtracting the median performance of its control
group. Changes over time in adjusted performance are then calculated. This
procedure is intended to isolate the component of performance change due to
ARTICLE IN PRESS
248
management turnover from that attributable to mean reversion of industry and rmspecic factors.
We follow a similar procedure using groups of comparison rms approximately
matching the average performance of sample rms over the three years prior to the
turnover. We examine adjusted performance using these groups in some of the tests
discussed in Section 4. However, the results are similar to those obtained using oneyear performance matching so we do not report them in detail.
Our data span a longer time period and include a larger number of CEO turnover
events than the data used by Denis and Denis (1995). They study turnover events
occurring from 1985 through 1988. Our sample covers the 24 years from 1971
through 1994. Denis and Denis examine a sample of 908 management changes,
dened to include turnover involving the chairman, CEO, or president, and 353 top
management changes, where the top executive is dened to be the CEO if there is
one and the chairman otherwise. Due to data limitations, the numbers of
observations in the subsamples they actually use to estimate the changes in
operating income are smaller, including only 721 management changes and 296 top
management changes. In contrast, our sample includes 1,344 CEO changes and we
have sufcient data to estimate post-CEO turnover changes in OROA for 1,002 of
these turnovers (see Table 2).
One other distinction between our sample and that used by Denis and Denis is
that our sample consists of rms that are large enough to merit inclusion in the
Forbes compensation surveys. Each year the Forbes compensation survey covers the
800 largest rms in the United States. The Denis and Denis sample, on the other
hand, includes some smaller rms.
We focus solely on CEO turnover since it is likely to be more economically
signicant than other managerial turnover. Denis and Denis report insignicant
abnormal returns around the announcement of turnovers that do not involve the top
executive and nd no signicant change in operating performance following such
turnovers. In addition, Fee and Hadlock (2003) nd that, controlling for CEO
turnover, the sensitivity of turnover probability to rm performance is low for the
top ve corporate ofcers below the CEO level. We study larger rms because, as
suggested at the beginning of this paper, these rms represent the bulk of the market
value of public corporations in the United States. Turnover decisions at these rms
potentially have the greatest impact on aggregate wealth. It would, of course, also be
informative to examine small rms, but governance variables and turnover
characteristics are prohibitively costly to collect for a large sample of small rms.
3.2. Methodology
This study uses the limited information maximum likelihood method described by
Heckman (1979) to control for potential selection bias in an analysis of determinants
of post-turnover performance. A potential selectivity problem arises in the
estimation of an ordinary least squares (OLS) regression with post-turnover
operating performance as the dependent variable and CEO and other rm
characteristics as independent variables because the sample is censored. Data are
ARTICLE IN PRESS
249
not available to estimate this regression for all CEO turnover observations since
some rms do not survive through the end of the post-turnover performance
measurement period. The characteristics of the surviving rms and their CEOs
could differ systematically from rm and CEO characteristics at rms that do not
survive.
Heckman characterizes the sample selection problem as a special case of the
omitted variable problem in which the inverse Mills ratio IML is the omitted
variable in the OLS regression. Use of the two-step Heckman procedure allows us to
obtain consistent estimates for determinants of post-turnover rm performance. A
probit model, in which the dependent variable equals one if the rm survived the
entire post-turnover performance measurement period and zero otherwise, is rst
used to estimate the IML, where
IML
fx0i b=s
:
1 Fx0i b=s
In Eq. (1), f and F represent the density and cumulative density functions of the
standard normal distribution, respectively, x0i is a vector that contains observations
for the explanatory variables predicting whether a rm that experiences CEO
turnover will survive through the post-turnover measurement period, b is the
vector of coefcient estimates from the probit regression, and s is the standard
deviation for the residuals from the probit regression. The second step of the
Heckman procedure is to simply estimate the OLS regression with the IML as an
explanatory variable.
4. Empirical results
4.1. Sample characteristics
Table 1 contains summary statistics for the sample of CEO successions occurring
from 1971 through 1994. As the table shows, the median outgoing CEO is 63 years
old and has held that position for 7.5 years. Most often, the CEO leaves voluntarily.
Just 16% of the successions are classied under our procedures as forced, and in only
7% is takeover pressure apparent. The median successor CEO is 53 years old and
has been with the rm slightly over 19 years. However, 19% of new CEOs are
outsiders.
Table 1 also shows that sample rms are typically quite large, with the median rm
having $2.8 billion in annual sales, $4.1 billion in assets, and over 16,000 employees.
Outsider-dominated boards are prevalent in this sample. Three-quarters of the
directors on the median board are outsiders. Also, though not shown in the table,
approximately 17% of the sample rms have a board with fewer than 60% outsiders.
Institutional shareholders are prominently represented in the sample as well.
Institutions hold over 46% of the typical sample rms stock and at least 30% of the
stock in more than 75% of the rms. In most cases, however, individual institutions
ARTICLE IN PRESS
250
Table 1
Summary statistics
Statistics are for a sample of 1,344 CEO turnover events during the 19711994 period. Forced equals one
when the outgoing CEO is forced from ofce and zero otherwise. Takeover equals one if the rm was
subject to takeover pressure within 12 months preceding the succession and zero otherwise. Outsider is a
dummy variable that equals one if the successor CEO has been employed by the rm for one year or less at
the time of appointment and zero otherwise. Sales and asset values are restated in 1994 dollars, using the
Consumer Price Index, before statistics for these variables are computed.
Mean
Outgoing CEO
Age (years)
CEO tenure (years)
Share ownership
Succession characteristics
Forced
Takeover
Successor CEO
Age (years)
Years with rm when
appointed CEO
Outsider
61.60
8.82
1.02%
Median
63.00
7.50
0.20%
Std. dev.
5.85
6.27
3.37%
Minimum
37.00
0.08
0.00%
Maximum
91.00
1,343
46.92
1,343
45.32%
887
0.16
0.07
53.03
17.69
Observations
1,344
1,330
53.00
19.04
6.31
13.22
34.00
0.00
73.00
50.08
0.19
Firm
Sales (millions of
6,386.40
2,789.12
13,033.48
0.66
1994 dollars)
Assets (millions of
10,539.52
4,108.44
21,975.11
121.29
1994 dollars)
Employees
36,660
16,144
74,754
386
Outside directors
72.31%
75.00%
14.25%
0.00%
Institutional ownership
44.79%
46.68%
18.67%
0.31%
1,344
1,344
1,344
141,729.90
1,281
213,599.40
1,294
853,000
100.00%
87.34%
1,268
1,329
737
do not hold large blocks of stock. In more than half of the rms, no institution holds
as much as 5%.
4.2. Operating performance changes around turnover events
Figs. 13 plot sample median OROA over the period from three years before
to three years after the CEO turnover. Fig. 1 plots the median unadjusted OROA
series. Figs. 2 and 3 show the median industry-adjusted and control groupadjusted series. Separate plots are shown for forced and voluntary turnovers, and
for the combined sample. The gures suggest that top management turnover
follows a period of deteriorating rm performance and that performance tends
to improve subsequently. This is most clearly apparent for forced turnovers and
for the control group-adjusted measures. Plots of median OROS display similar
patterns.
ARTICLE IN PRESS
251
Fig. 1. Median unadjusted operating return on assets (OROA) around CEO turnover events.
Fig. 2. Median industry-adjusted operating return on assets (OROA) around CEO turnover events.
ARTICLE IN PRESS
252
Control group-adjusted OROA
0.6%
0.4%
0.2%
0.0%
-0.2%
-0.4%
All turnovers
Voluntary turnovers
Forced turnovers
-0.6%
-0.8%
-3
-2
-1
0
1
Year relative to turnover
Fig. 3. Median control group-adjusted operating return on assets (OROA) around CEO turnover events.
In Table 2 we report mean and median changes in OROA over the three-year
periods preceding and succeeding CEO turnover events.5 Panel A shows the
mean and median unadjusted changes while Panels B and C show mean and
median changes in industry-adjusted and control group-adjusted OROA, respectively. The results in the table conrm the impression gained from the gures.
Turnover follows declining performance. For all turnovers combined, mean and
median changes in OROA from year 3 to year 1 are negative and signicant at the 1% level. These changes are also negative and highly signicant in
all cases for forced turnovers. For voluntary turnovers, the pre-turnover mean
change in industry-adjusted OROA is signicant at only the 10% level. However,
the mean changes in unadjusted and control group-adjusted OROA for this
5
In calculating the averages and related test statistics, we exclude extreme observations that differ in
absolute value from the mean unadjusted, industry-adjusted, or control group-adjusted OROA by more
than three times the standard deviation of the respective distribution. The screens are implemented as
follows. First, raw, industry-adjusted, and control group-adjusted OROA are computed for each turnover
observation for years 3 to 3: All observations more than three standard deviations from the mean of its
respective OROA distribution (outliers) are then excluded from the sample and performance changes are
computed. Performance changes cannot be computed for all remaining observations because either (1)
data are not available to calculate the adjusted performance changes or (2) data are not available to
compute the unadjusted performance at both the beginning and end of the measurement period. The
screens result in the loss of 19 (5 positive, 14 negative), 24 (10 positive, 14 negative), and 39 (13 positive, 26
negative) unadjusted, industry-adjusted, and control group-adjusted OROA observations, respectively, for
the 3 to 1 period. The corresponding losses for the 1 to 3 period are 15 (8 positive, 7 negative), 23
(13 positive, 10 negative), and 22 (14 positive, 8 negative) observations. Without these screens, the
empirical evidence is qualitatively similar to that reported below.
ARTICLE IN PRESS
253
Table 2
Changes in operating return on assets (OROA)
Mean (median) changes in OROA from three years before to three years after CEO succession for a
sample of 1,344 CEO successions at large public rms during the 19711994 period. Results for three
performance measures are reported: changes in OROA (unadjusted OROA), changes in OROA adjusted
by subtracting SIC two-digit industry level median OROA (industry-adjusted OROA), and changes in
OROA adjusted by subtracting median OROA for a control group matched by two-digit industry and by
prior OROA performance (control group-adjusted OROA). Cases where the new CEOs tenure ended
before the end of year +3 are excluded. Extreme performance observations are also excluded (see footnote
5 for details). Sample sizes, t-statistics, and median sign-test statistics are reported in parentheses below the
mean (median) changes in the OROA measures.
Years
All turnovers
Voluntary turnovers
Forced turnovers
Panel A: unadjusted OROA

3 to 1
0.007 (0.003)
(1243, 4:95c ; 75:5c )
1 to 3
0:003 (0.000)
(1002, 1:45; 3:0)
0.005 (0.001)
(1051, 2:81c ; 38:5b )
0:005 0:001
(883, 2:25b ; 13:5)
0.023 (0.012)
(192, 6:10c ; 37:0c )
0.011 (0.007)
(119, 1:86a ; 10:5b )
Panel B: industry-adjusted OROA

3 to 1
0:005 (0:003)
(1238, 3:58c ; 71:0c )
1 to +3
0.003 (0.001)
(994, 1:82a ; 28:0b )
0:002 (0:001)
(1048, 1:67a ; 39:0b )
0.002 (0.001)
(876, 1.19, 14.0)
0:018 (0:012)
(190, 5:10c ; 32:0c )
0.011 (0.005)
(118, 2:03b ; 14:0b )
Panel C: control group-adjusted OROA

3 to 1
0:007 (0:002)
(1216, 4:88c ; 52:5c )
1 to +3
0.009 (0.003)
(990, 5:20c ; 49:5c )
0:005 (0:001)
(1027, 3:86c ; 38:5c )
0.009 (0.003)
(872, 4:70c ; 50:0c )
0:011 (0:002)
(189, 3:22c ; 14:0b )
0.013 0:001
(118, 2:23b ; 0.5)
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels, respectively, for twotailed tests.
group are signicantly negative at the 1% level. Furthermore, the pre-turnover

median changes are all negative and signicant at the 5% level for voluntary
turnovers.
The change in OROA from year 1 to year 3 measures the performance change
following the turnover. Looking at all turnovers, we nd no signicant increase in
mean or median unadjusted OROA, but mean industry-adjusted and control groupadjusted performance increase by 0.3% and 0.9% (t-statistics of 1.82 and 5.20),
respectively. The corresponding median increases are smaller, but are also positive
and signicant at the 5% level. After voluntary turnovers, the mean change in
unadjusted OROA from year 1 to year 3 is 0:5% (t-statistic 2:25). Industryadjusted operating performance over the same period does not change signicantly
but the increases in the mean and median control group-adjusted performance, 0.9%
and 0.3%, respectively, are both signicant at the 1% level. For forced turnovers, the
mean unadjusted and industry-adjusted changes in OROA from year 1 to year 3
are both 1.1% while the control group-adjusted performance change is 1.3%. All of
ARTICLE IN PRESS
254
these values are signicant at the 10% level (t-statistics of 1.86, 2.03, and 2.23,
respectively). The median unadjusted and industry-adjusted changes for forced
turnovers are also positive and signicant at the 5% level while the median control
group-adjusted change is insignicantly different from zero.
The signicant improvements in OROA from year 1 to year 3 do not
necessarily indicate more efcient operating performance. For example, OROA can
increase as a result of restructuring activities that reduce the overall capital intensity
of a rms businesses, eliminate businesses that are performing more poorly, or result
in writing down the book values of certain assets. To understand better the sources
of the increases in OROA, we examine data on restructuring activities and
alternative measures of operating performance during the period from year 1 to
year 3: Table 3, which presents some of these data, shows that unadjusted book
assets, capital expenditures, and gross property, plant, and equipment (gross PP&E)
increase over the measurement period for all turnovers and for the voluntary and
forced turnover samples. The change in total book assets from year 2 to year 2 is
also reported in Table 3 because the calculation of the change in OROA from year
1 to year 3 uses book assets from years 2 and 2 since OROA is computed
using book assets from the beginning of the period. Because there is no substantive
difference between the changes in the book asset values computed over the two
periods (year 2 to year 2 and year 1 to year 3), we focus on the changes from
year 1 to year 3: This is the period over which Denis and Denis (1995) measure
these changes.
The control group-adjusted gures for voluntary turnovers are all positive in
Table 3, suggesting that the signicant, positive, post-turnover changes in control
group-adjusted OROA for these turnovers in Table 2 do not simply reect the effects
of asset sales. In addition, consistent with the idea that operating efciency increases
following voluntary turnover, the control group-adjusted number of employees as
well as sales and operating prots per employee all increase from year 1 to year 3
for the voluntary turnover sample.
The control group-adjusted measures in Table 3 for the forced turnover sample
indicate that book assets, gross PP&E, and employment grow at a moderately slower
rate for the sample rms than for the control group. The substantial control groupadjusted increases in sales and operating prots per employee suggest that operating
efciency at these rms improved. These improvements could reect the disposition
of poorly performing assets as well as improved management. We would expect to
observe both in these generally nancially distressed rms.
Overall, the evidence is consistent with worsening performance preceding
turnovers and managerial quality improvements following turnovers. The change
in performance preceding voluntary turnovers is, as expected, smaller in absolute
value than that preceding forced turnover. After voluntary turnovers, the evidence
suggests further deterioration in unadjusted OROA, but control group-adjusted
OROA improves. For forced turnovers, the evidence is consistent with poor
performance preceding turnovers and performance improvements following turnovers for the three performance measures. These results tend to reject the scapegoat
hypothesis in favor of the improved management hypothesis.
ARTICLE IN PRESS
255
Table 3
Median percentage changes in book assets, capital expenditures, property, plant, and equipment (PP&E),
employment, and sales and operating income per employee following CEO turnover
Four different measures are reported: unadjusted changes, changes adjusted for ination (inationadjusted), changes adjusted by subtracting SIC two-digit industry median changes (industry-adjusted), and
changes adjusted by subtracting changes for a control group matched by two-digit industry and by prior
operating return on assets (OROA) performance (control group-adjusted). The number of observations
for which the median is estimated is reported in parentheses. Cases where the new CEOs tenure ended
before the various performance change measurement intervals are excluded.
All turnovers
Voluntary
turnovers
Forced
turnovers
p-Value for test of

voluntary vs. forced
%D Total book assets (1; +3)

Unadjusted
36.88c
Ination-adjusted
28.93c
Industry-adjusted
9.34c
Control group-adjusted
2.46b
(1198)
(1198)
(1198)
(1173)
40.26c
31.32c
6.97c
3.84c
(1029)
(1029)
(1029)
(1013)
17.80c
13.49c
28.99c
4.26
(169)
(169)
(169)
(160)
o0.001
o0.001
o0.001
0.018
%D Total book assets (2; +2)

Unadjusted
38.99c
Ination-adjusted
30.40c
Industry-adjusted
8.79c
0.73
(1225)
(1225)
(1224)
(1201)
42.48c
32.72c
5.70c
0.40
(1044)
(1044)
(1043)
(1028)
13.06c
10.95c
27.18c
9.68c
(181)
(181)
(181)
(173)
o0.001
o0.001
o0.001
0.002
%D Capital expenditures (1; +3)

Unadjusted
33.25c (978)
Ination-adjusted
25.26c (978)
Industry-adjusted
6.23c (978)
8.25c (971)
34.56c
26.77c
4.96b
8.98c
(853)
(853)
(853)
(847)
17.71
15.23
17.43b
6.65
(125)
(125)
(125)
(124)
0.069
0.104
0.151
0.573
%D Gross PP & E (1; +3)

Unadjusted
35.79c (990)
Ination-adjusted
28.31c (990)
Industry-adjusted
4.91c (990)
2.96c (982)
37.57c
30.10c
3.54b
3.68c
(866)
(866)
(866)
(858)
15.71c
13.57c
21.04c
4.50
(124)
(124)
(124)
(124)
o0.001
o0.001
o0.001
0.124
10.05c (161)
20.89c (161)
3.13 (153)
o0.001
o0.001
0.004
%D Number of employees (1; +3)

Unadjusted
2.56c (1163)
Industry-adjusted
5.19c (1163)
3.07c (1140)
%D Sales per employee (1;
Unadjusted
Ination-adjusted
Industry-adjusted
+3)
27.52c
22.03c
4.00c
2.96c
(1154)
(1154)
(1154)
(1139)
%D Operating income per employee (1; +3)

Unadjusted
4.12c (1138)
Ination-adjusted
3.38c (1138)
Industry-adjusted
1.51c (1138)
1.12c (1130)
4.03c (1002)
4.07c (1002)
4.07c (987)
27.03c
21.24c
3.58c
2.31c
(999)
(999)
(999)
(986)
31.66c
25.15c
7.35c
10.40c
(155)
(155)
(155)
(153)
0.262
0.343
0.070
0.007
3.83c
3.19c
1.39c
1.04c
(987)
(987)
(987)
(981)
6.09c
4.68c
3.01c
2.83c
(151)
(151)
(151)
(149)
0.004
0.018
0.011
0.065
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels, respectively, based
on the Brown-Mood test.
ARTICLE IN PRESS
256
4.3. Stock price performance around turnover events

We also examine evidence on abnormal stock returns around the turnovers
in our sample. Following Mitchell and Stafford (2000), we compute abnormal
returns using calendar-time portfolios. Specically, in each month t; all rms that
experience turnover within the next 36 months are included in that months
pre-turnover portfolio and all rms that have experienced a turnover during the
previous 36 months are included in that months post-turnover portfolio. Equally
weighted and value-weighted portfolio returns are calculated for each month. Like
Mitchell and Stafford, we exclude any month with fewer than ten event rms in its
portfolio. The median pre-turnover portfolio has 132 rms and the median postturnover portfolio has 119 rms. As in Fama and French (1993), the portfolio
returns, rpt ; are then used to estimate abnormal returns from the time-series
regression equation
rpt rft a b1 rmt rft b2 HMLt b3 SMBt et :
In Eq. (2) rft is the one-month Treasury bill rate; rmt rft ; the realized market risk
premium, is computed as the value-weighted return on all NYSE, AMEX, and
NASDAQ stocks in the Center for Research in Security Prices (CRSP) database, less
rft ; HMLt is the return difference between portfolios of high and low book-tomarket stocks; and SMBt is the return difference between portfolios of small
capitalization and big capitalization stocks.6 The subscript t refers to the calendar
month of the observations. et is assumed to be serially independent and normally
distributed with mean zero.
Under the joint null hypothesis that the model adequately describes returns and
that the expected abnormal performance around turnovers equals zero, the expected
value of the intercept is zero and estimates of the intercept measure the average
monthly abnormal return. Since Fama and French show that the three-factor model
does not completely describe returns, we calculate an adjusted intercept. As in
Mitchell and Stafford, we rst estimate an expected intercept, Ea; as the average
intercept from 1,000 calendar-time portfolio regressions for random portfolios
formed with the same size and book-to-market characteristics as the portfolios
constructed using the sample rms. The adjusted intercept is then computed as the
difference between the intercept estimated from Eq. (2), a# ; and Ea: The test statistic
for each adjusted intercept is estimated as
t
a# Ea
se
where se is the standard error of the estimate from the sample rm portfolios.
The abnormal returns estimated using this procedure are presented in Table 4.
Consistent with the evidence of a negative relation between stock performance and
6
We obtained the values for these factors from the website maintained by Ken French. More
information is available at this web site (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/) or
from the authors.
ARTICLE IN PRESS
257
Table 4
Long-run abnormal stock returns around CEO turnover events
Abnormal stock returns are estimated using calendar time portfolio regressions. In each month t; all rms
that experience turnover within the next 36 months are included in that months pre-turnover portfolio
and all rms that have experienced a turnover within the previous 36 months are included in that months
post-turnover portfolio. Equally weighted and value-weighted portfolio returns are calculated for each
month. The portfolio returns, rpt ; are then used to estimate abnormal returns using the following
regression:
rpt rft a b1 rmt rft b2 HMLt b3 SMBt et :
In this regression rft is the risk-free rate; rmt rft ; the realized market risk premium, is computed as the
value-weighted return on all NYSE, AMEX, and NASDAQ stocks in the Center for Research in Security
Prices (CRSP) database less rft ; HMLt is the return difference between portfolios of high and low book-tomarket stocks; and SMBt is the return difference between portfolios of small capitalization and big
capitalization stocks. The subscript t refers the calendar month of the observations. The intercept, a;
measures the monthly abnormal return conditioned on the model. The adjusted intercept, Adj:a; equals the
difference between the estimated intercept, a# ; and the average intercept from 1,000 calendar-time portfolio
regressions for portfolios formed from randomly selected rms from the same size and book-to-market
quintiles (using NYSE breakpoints) as the rms used to construct the sample portfolios. Observations for
calendar months with fewer than ten event rms in the portfolio are excluded. The Implied 3-year
Abnormal Return is calculated as 1 a36 1:
Pre-turnover
Post-turnover
Adj:a
Adj:a
Panel A: all turnovers

Equally weighted
t-statistic
Implied 3-year abnormal return (%)
0.16%
2:50b
5.67%
0.25%
3:87c
8.66%
0.02%
0:26
0.61%
0.03%
0:49
1.13%
Value-weighted
t-statistic
0.08%
1:41
2.83%
0.08%
1:33
2.67%
0.04%
0:57
1.35%
0.12%
1:78a
4.14%
Panel B: voluntary
Equally weighted
t-statistic
0.01%
(0.10)
0.23%
0.02%
0:32
0.68%
0.04%
0:63
1.45%
0.00%
(0.03)
0.06%
Value-weighted
t-statistic
0.01%
(0.14)
0.29%
0.01%
0:22
0.46%
0.07%
0:98
2.42%
0.08%
1:15
2.83%
Panel C: forced
Equally weighted
t-statistic
1.20%
8:05c
35.13%
1.25%
8:44c
36.49%
0.09%
1:53
3.14%
0.19%
3:21c
6.49%
Value-weighted
t-statistic
0.69%
4:52c
21:95%
0.73%
4:79c
23:07%
0.07%
(0.34)
2.54%
0.04%
(0.18)
1.32%
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels (for two-tailed tests),
respectively.
ARTICLE IN PRESS
258
the likelihood of CEO turnover reported by Warner et al. (1988), among others, we
nd evidence of negative abnormal returns during the three-year period prior to
turnover. These are largely attributable to substantial negative abnormal returns
preceding forced turnover. The average monthly abnormal return in the three-year
period prior to forced turnovers ranges from 1:25% to 0:69%; and all of the
estimates are signicant at the 1% level. Converting these monthly averages into
implied three-year abnormal returns indicates that forced turnovers in our sample
are preceded by abnormal returns ranging from approximately 22% to 36%;
depending on the measure selected. In contrast, abnormal stock price performance
preceding voluntary turnovers differs insignicantly from zero.
The evidence on post-turnover abnormal returns is mixed. The adjusted estimate
of abnormal return during the three-year period following a forced turnover is
signicantly negative for the equally weighted portfolio, though it is not very large in
absolute value. We obtain a similar result for the value-weighted portfolio of all
turnovers. None of the other estimates of post-turnover abnormal returns, however,
are signicantly different from zero.
We also examine daily returns around the date of the turnover announcement. In
this case the abnormal stock returns are estimated using the market model and
return data from the CRSP database. We regressed rm daily returns over the period
from 300 through 50 trading days before Wall Street Journal turnover announcements on the contemporaneous returns for the equally weighted CRSP index. The
resulting estimated market model was then used to generate stock return prediction
errors for the day of and day before the announcement of the management turnover.
We sum these two daily prediction errors to obtain the announcement period
abnormal return used in the regressions. The results reported are robust to using
windows of (+50, +300) and (50; 300: +50, +300).
The average announcement-period abnormal return for 1,302 cases in our sample
is 0.344%. The associated t-statistic is 2.55, which is signicant at approximately the
1% level. We also observe positive and statistically signicant average announcement-period abnormal returns for 1,097 voluntary turnovers (0.262% with a
t-statistic of 2.65) and for 111 forced turnovers where an outsider is appointed CEO
(2.146% with a t-statistic of 1.86). However, we nd that the average announcementperiod abnormal return for 94 forced turnovers where an insider is appointed CEO is
negative and insignicantly different from zero (0:832% with a t-statistic of 1:20).
A t-test indicates that the difference between the average abnormal returns for the
forced sample where an outsider is appointed and the forced sample where an
insider is appointed is signicant at approximately the 3% level (t-statistic of 2.21).
A Brown-Mood test indicates that the difference in the median for these two
samples is highly signicant, with a p-value of 0.003. The overall positive average
stock price reaction to turnover announcements indicates that investors anticipate
subsequent increases in operating returns along the lines reported in Table 2. The
evidence for outside and inside appointments around forced turnover, which is
consistent with evidence reported by Borokhovich et al. (1996), suggests that
anticipated performance improvements are greater when an outsider is appointed
CEO.
ARTICLE IN PRESS
259
4.4. Variation in performance changes over time

In contrast to Denis and Denis (1995), we nd only modest evidence that industryadjusted OROA increases after management turnover. We computed statistics for
post-turnover changes in unadjusted, industry-adjusted, and control group-adjusted
OROA for the 19851988 period examined by Denis and Denis and for all other
years in our sample period to investigate the differences between the evidence we
report and that reported by Denis and Denis. Panel A of Table 5 shows these
statistics for all turnovers, and for voluntary and forced turnover subsamples.
The evidence in Panel A of Table 5 shows that the changes in industry-adjusted
OROA after CEO turnover are signicantly higher from 1985 to 1988 than during
the other years in our sample period. For all turnovers, the mean and median
changes in industry-adjusted OROA are 1.2% and 0.6%, respectively, and
statistically signicant for the 19851988 period. In contrast, both the mean and
median changes in industry-adjusted OROA differ insignicantly from zero over the
rest of the sample period. Among forced turnovers, industry-adjusted performance
changes differ insignicantly between the two time periods. The results for voluntary
turnovers, however, are very similar to those for the sample as a whole. This evidence
suggests that the strong positive overall changes in industry-adjusted performance
reported by Denis and Denis are unique to the period they studied.
Examination of changes in the control group-adjusted performance measures
indicates that managerial quality improvements also were greater during the period
studied by Denis and Denis. However, our results here are not driven entirely by the
19851988 period. The changes in the mean control group-adjusted performance for
the other years covered by our sample period are all positive and statistically
signicantly greater than zero.
To obtain additional insight into how post-turnover changes in OROA varied
throughout our sample period, we compare post-turnover unadjusted, industryadjusted, and control group-adjusted OROA for the eight non-overlapping three-year
subperiods from 1971 through 1994. This analysis reveals that all three measures of
post-turnover changes in OROA tend to be negative or only slightly positive in each
of the three subperiods prior to 1980. Beginning in the early 1980s, changes in postturnover OROA tend to increase. This suggests that there was a change in the
nature of turnover decisions beginning in the early 1980s that coincided with the
changes in characteristics of corporate governance mechanisms, such as board
composition and institutional ownership, described by Huson et al. (2001). The
general magnitudes and statistical signicance of the post-turnover changes in OROA
are illustrated in Panel B of Table 5, which compares data for the rst half of the
sample period (19711982) to data for the second half (19831994). The magnitudes
and statistical signicance of the changes in control group-adjusted OROA in all
columns are particularly noteworthy, suggesting that the improvement in CEO
quality following turnovers was substantially greater in the latter half of the sample
period than in the rst half.
An alternative explanation for the larger post-turnover changes in OROA during
the latter part of our sample period is that governance mechanisms became weaker
ARTICLE IN PRESS
260
Table 5
Changes in operating return on assets (OROA) by subperiod
Changes in OROA from one year before to three years after CEO succession for a sample of 1,344 CEO
successions at large public rms during the 19711994 period. Results for three performance measures are
reported: changes in OROA (unadjusted OROA), changes in OROA adjusted by subtracting SIC twodigit industry level median OROA (industry-adjusted OROA), and changes in OROA adjusted by
subtracting median OROA for a control group matched by two-digit industry and by prior OROA
performance (control group-adjusted OROA). Cases where the new CEOs tenure ended before the end of
year +3 are excluded. Extreme performance observations are also excluded (see footnote 5 for details).
t-Statistics and z-approximations for Brown-Mood tests that the mean and median changes are equal in
the various subperiods are also reported.
Sample period
All turnovers
N
Mean
Median
Voluntary turnovers
N
Mean
Forced turnovers
Median N
Mean
Panel A: comparison of change in OROA from year 1 to

1. Unadjusted OROA
19851988
185 0.004 0.008
19711984 and 19891994
817 0.003
0.000
Statistic for test of difference
0.14 0.89
2. Industry-adjusted OROA
19851988
181 0.012b 0.006b
19711984 and 19891994
813 0.002
0.001
1.86a
1.89a
3. Control group-adjusted OROA
19851988
181 0.021c 0.012c
19711984 and 19891994
809 0.006c 0.001
3.23c
4.52c
year 3: 19851988 vs. other years
Panel B: comparison of change in OROA from year 1 to

1. Unadjusted OROA
19711982
515 0.006b 0.001
19831994
487 0.000
0.002
1.60 1.95b
2. Industry-adjusted OROA
19711982
518 0.001 0.001
19831994
476 0.008c 0.004c
2.59c 2.41b
3. Control group-adjusted OROA
19711982
513 0.003 0.000
19831994
477 0.016c 0.007c
3.85c 4.13c
year 3: 19711982 vs. 19831994
161 0.003 0.010

722 0.005b
0.000
0.23
0.77
24 0.004
0.002
95 0.015b 0.008a
1.21 0.86
157
719
0.012b
0.000
1.94a
0.006a 24 0.010
0.000 94 0.011a
1.67a
0.15
158
714
0.022c
0.006c
3.17c
0.012c 23
0.001 95
4.26c
463 0.008c 0.001

420 0.001
0.000
1.66a 1.51
Median
52
67
0.011
0.004b
0.46
0.018
0.007
0.012a 0.001
0.45
0.69
0.012
0.005
0.009
0.013a
1.21 0.86
466 0.003 0.000 52 0.011

0.004
410
0.008c
0.004b 66 0.011
0.011a
2.69c 2.16a
0.05 0.71
460
412
0.003
0.000 53 0.003 0.003
0.015c
0.007c 65 0.022b 0.007a
3.48c 3.39c
1.62a
1.66a
respectively.
over time and that pre-turnover declines in OROA also tended to be larger during
the latter part of our sample period. The bigger increases in post-turnover OROA
during the 19831994 period could simply reect a lower level of OROA in year 1:
We compare the pre-turnover change in unadjusted, industry-adjusted, and control
group-adjusted OROA across various subperiods and nd no support for this
alternative explanation. In fact, we nd weak evidence that the declines in
ARTICLE IN PRESS
261
industry-adjusted and control group-adjusted OROA from year 3 to year 1 were

greater during the 19711982 period than during the 19831994 period.
4.5. Cross-sectional determinants of performance changes
Table 6 presents evidence on the inuence of board composition, institutional
ownership, external takeover activity, and successor origin on changes in OROA
around CEO turnover events. Simple paired comparisons of average changes in
OROA from one year before to three years after turnover are presented. Specically,
we compare mean and median changes at rms with outsider-dominated boards to
those at other rms. We dene an outsider-dominated board as one with at least
60% outsiders.7 We also compare changes at rms with greater than the sample
median level of total institutional share ownership (46.68%) to those with less. In
addition, we report mean and median OROA changes for rms subject to takeover
pressure during the 12 months preceding turnover and contrast them with mean and
median changes for rms not subject to such pressure. Finally, we examine the
changes at rms where the successor CEO is an outsider and compare them to
changes where the successor is an insider.
Recall that the improved management hypothesis predicts that rm performance
changes should be greater when an outsider is appointed CEO and when takeover
pressure is present. The effects of board composition and institutional share
ownership, however, are ambiguous. Under the scapegoat hypothesis, performance
improvements are unrelated to board composition, ownership structure, takeover
pressure, or successor origin.
Panel A of Table 6 reports the comparisons for unadjusted performance changes.
Panels B and C contain the gures for industry-adjusted and control group-adjusted
changes, respectively. Mean and median unadjusted and industry-adjusted changes in
OROA are signicantly greater for CEO successions in which outsiders dominate the
board. Takeover pressure appears also to have a positive impact on mean and median
unadjusted and industry-adjusted changes in OROA following management turnover.
Moreover, contrary to the scapegoat hypothesis, changes in all three of the performance
measures are greater when an outsider is appointed CEO than when an insider is
appointed. For the mean and median unadjusted and industry-adjusted performance
measures, the differences are statistically signicant at the 1% level. For the control
group-adjusted measure, the difference in the median change is signicant at the 10%
level. These results generally are consistent with the improved management hypothesis.
Table 7 follows the same format as Table 6, but contains results for the subsample
of forced turnovers only. For this reason the sample sizes are much smaller than are
those for the comparisons in Table 6. None of the tests reported in Table 7 indicate
statistically signicant differences between the average rm performance changes for
7
The tables report results based on data from the Million Dollar Directory and the simple scheme for
classifying directors described in Section 3. We also performed tests using proxy statement data and the
alternative, more complex classication scheme described in Section 3 and reach similar conclusions. The
denition of an outsider-dominated board is the same as that used by Weisbach (1988).
Panel B: industry-adjusted
Sample size
Mean
t-stat for mean
t-stat for difference
Median
Sign stat for median test
Median test approximate Z-stat
Panel C: control group-adjusted
Sample size
Mean
t-stat for mean
Median
Institutional share
Takeover pressure
Successor
Outside
Inside
> Median
p Median
Yes
No
Outsider
Insider
1002
0.003
1.45
801
0.000
0.080
196
0.015
2.71c
291
0.001
0.24
276
0.003
0.80
63
0.011
1.37
938
0.004
1.86a
156
0.014
2.80c
846
0.006
2.85c
0.000
3.0
0.001
21.5
0.16
24.0c
0.002
7.5
0.000
3.0
0.007
2.5
0.000
5.0
0.007
23.0c
2.57c
3.81
994
0.003
1.82a
795
0.006
3.05c
0.001
28.0a
0.002
39.5c
0.80
194
0.007
1.34
283
0.007
1.91a
0.001
11.0
0.001
7.5
2.36b
2.07
990
0.009
5.20c
792
0.009
4.75c
0.003
49.5c
0.003
47.5c
275
0.010
2.88c
62
0.023
2.99c
0.005
26.5c
0.009
7.0a
286
0.021
6.10c
0.001
1.5
0.009
38.0c
0.28
1.84
272
0.014
4.13c
62
0.016
2.24b
0.006
31.5c
0.012
9.5b
1.54
1.18
:0002
26.0c
4.01
925
0.002
1.10
153
0.018
3.65c
0.001
21.5
0.008
20.5c
2.70c
1.27
193
0.010
2.15a
3.78c
0.92
0.62
0.39
1.82a
0.37
841
0.001
0.39
3.31c
0.000
7.5
3.07
919
0.009
4.80c
153
0.016
3.02
0.002
40.5c
0.005
12.5a
0.96
837
0.008
4.31c
1.39
1.58
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels, respectively, for two-tailed t- and median tests.
0.002
37.0b
1.67a
ARTICLE IN PRESS
Panel A: unadjusted
Sample size
Mean
t-stat for mean
Median
Board type
Total
Sample
262
Table 6
Changes in operating return on assets (OROA) from year 1 to year +3 for various CEO turnover groups
Changes in OROA are for a sample of 1,344 CEO successions at large public rms during the 19711994 period. Results for three performance measures are
reported: changes in OROA (unadjusted OROA), changes in OROA adjusted by subtracting SIC two-digit industry level median OROA (industry-adjusted
OROA), and changes in OROA adjusted by subtracting median OROA for a control group matched by two-digit industry and by prior OROA performance
(control group-adjusted OROA). Outside boards are those with at least 60% outsiders. All others are inside boards. Takeover pressure refers to events related
to corporate control and takeovers during the 12 months preceding the turnover announcement. Outside successors are those employed with their rms for one
year or less when appointed CEO. All others are inside successors. Cases where the new CEOs tenure ended before the end of year +3 are excluded. Extreme
performance observations are also excluded (see footnote 5 for details).
Total
Sample
Panel B: industry-adjusted
Sample size
Mean
t-stat for mean
Median
Panel C: control group-adjusted
Sample size
Mean
t-stat for mean
Median
Institutional share
Takeover pressure
Successor
Outside
Inside
> Median
p Median
Yes
No
Outsider
Insider
119
0.011
1:86a
104
0.007
1.31
14
0.038
1.32
31
0.017
1.69
39
0.013
1.39
15
0.015
0.69
104
0.010
1:74a
62
0.017
2:27b
56
0.004
0.42
0.007
10:5b
0.006
9:0a
0.047
1.0
0.030
6:5b
0.007
3.5
0.002
1.5
0.008
12:0b
0.008
9:5b
1.06
0.30
b
0.57
118
0.011
2:03b
103
0.006
1.22
0.005
14:0b
0.004
10:5b
14
0.043
1:90a
30
0.015
1.32
0.057
3.0
0.011
3.0
102
0.008
1.28
0.001
0.5
0.002
3.5
15
0.030
1.56
0.013
9:0c
0.007
0.5
15
0.048
2:04a
31
0.029
2:22b
0.038
2.5
0.023
4.5
15
0.014
0.62
0.001
0.5
0.000
0.0
0.44
1:79a
103
0.008
1.50
62
0.018
2:60b
0.05
13:5c
0.011
10:0b
56
0.003
0.40
1.36
0.28
38
0.021
1:70a
0.006
1.0
0.28
1.31
0.00
1.67
1.41
40
0.020
2:19b
0.42
1.16
1.14
0.79
2:15
1.58
118
0.013
2:23b
0.29
0.003
4.0
0.73
103
0.013
2:17b
63
0.011
1.38
0.001
0.5
0.001
0.5
55
0.016
1:75a
0.45
0.01
0.28
0.18
263
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels, respectively, for two-tailed t- and median tests.
0.000
0.0
ARTICLE IN PRESS
Panel A: unadjusted
Sample size
Mean
t-stat for mean
Median
Board type
Table 7
Changes in operating return on assets (OROA) from year 1 to year +3 for forced turnover groups
Changes in OROA are for a sample of 217 forced CEO successions at large public rms during the 19711994 period. Results for three performance measures
are reported: changes in OROA (unadjusted OROA), changes in OROA adjusted by subtracting SIC two-digit industry level median OROA (industryadjusted OROA), and changes in OROA adjusted by subtracting median OROA for a control group matched by two-digit industry and by prior OROA
performance (control group-adjusted OROA). Outside boards are those with at least 60% outsiders. All others are inside boards. Takeover pressure refers to
events related to corporate control and takeovers during the 12 months preceding the turnover announcement. Outside successors are those employed with
their rms for one year or less when appointed CEO. All others are inside successors. Cases where the new CEOs tenure ended before the end of year +3 are
excluded. Extreme performance observations are also excluded (see footnote 5 for details).
ARTICLE IN PRESS
264
the various groups based on board composition, takeover pressure, and successor
origin. However, the median differences for institutional ownership are signicant at
the 10% level for the unadjusted and control group-adjusted measures. This evidence
is consistent with the improved management hypothesis.
Overall, Tables 6 and 7 contain only relatively weak evidence that there are
reliable relations between performance changes and corporate governance characteristics (board composition and institutional share). One possible explanation is
that rms with superior governance do not let performance deteriorate too much
before taking corrective action. Under this scenario, the potential performance
improvements for such rms would be small. Firms with bad governance
characteristics would have relatively worse pre-turnover performance but might be
unable to select good replacement CEOs. Consequently, their performance
improvements would also be small.
To test this explanation we examine the pre-turnover performance of our sample
rms based on the groupings in Tables 6 and 7. Consistent with the explanation,
rms with institutional holdings above the median have signicantly higher
unadjusted and industry-adjusted OROA in the year prior to turnover. This holds
in general and for the forced turnover subsample. Firms with outsider-dominated
boards, however, have signicantly lower unadjusted and industry-adjusted OROA
prior to turnover. Hence, the evidence is mixed.
We also use a multivariate regression framework to investigate the cross-sectional
determinants of rm performance changes following CEO turnover events. The twostep method described by Heckman (1979) is used to obtain consistent estimates.
This method enables us to address the selectivity problem that arises because some
rms in the CEO turnover sample do not survive until the end of the year +3.
Among the sample rms, 83 merge or are acquired and 12 le for bankruptcy before
the end of year +3. Results for both the binomial probit and OLS regressions are
reported in Table 8.8 The probit regressions provide evidence on the predictors of
post-CEO turnover rm survival, but are estimated principally to obtain the IML
values used in the OLS regressions. In the OLS regressions, changes in OROA from
the year preceding turnover through three years after turnover OROA1; 3 are
regressed on dummy variables distinguishing forced successions, the presence of
prior takeover activity, outside successors, and outsider-dominated boards and on
the fraction of shares held by institutional investors and the two-day abnormal
return around the turnover announcement. Controls for contemporaneous industry
IOROA1; 3 and control rm COROA1; 3 performance, industry- and
control group-adjusted OROA during the year prior to turnover, and the natural log
of assets are included along with IML in various regressions.
8
Another potential source of bias exists because only observations for which the new CEO has survived
through the end of year +3 are included in our analysis. We examine the impact of this decision on the
regression evidence by reestimating the models in Tables 8 and 10 using data for all rms that survived
through year +3, regardless of the tenure of the CEO. The results are qualitatively similar to those
reported in Tables 8 and 10.
ARTICLE IN PRESS
265
Table 8
Sample selection models of post-turnover performance
This table reports the results of sample selection models, estimated as described by Heckman (1979), in
which the dependent variable for the rst regression equals one if the rm survives as an independent
entity for three years after the CEO turnover announcement and zero otherwise. The sample consists of
CEO turnovers announced over the 19801994 period. The dependent variable for the second regression
equals the change in operating performance over the period from one year before to three years after the
year of CEO turnover OROA1; 3: The second regression is estimated using only data for rms that
survived three years. The independent variables include dummy variables that equal one if the rm was
subject to takeover pressure during the 12 months preceding turnover (Takeover), if the old CEO is forced
from ofce (Forced), if the new CEO is employed with the rm for one year or less at the time of turnover
(Outsider), and if at least 60% of board members are outsiders (Outside board), and zero otherwise, as
well as the fraction of rm stock held by institutional investors (Institutional share), abnormal stock
returns over the day before and the day of the Wall Street Journal announcement of the turnover
(Abnormal return), changes in industry IOROA1; 3 or control COROA1; 3 rm OROA over
the period from one year before to three years after the year of CEO turnover, rm industry-adjusted and
control group-adjusted OROA in the year prior to turnover OROA1; and the natural log of assets at
the end of the scal year immediately preceding the turnover announcement (Ln(Assets)). IML is the
inverse Mills ratio. Cases where the new CEOs tenure was less than three years are excluded. Extreme
performance observations are also excluded (see footnote 5 for details). t-Statistics are reported in
parentheses.
Model 1
Intercept
Forced
Takeover
Outsider
Outside board
Institutional share
Abnormal return
Model 2
Model 3
Firm remains
independent
OROA
1; 3
Firm remains
independent
OROA
1; 3
Firm remains
independent
OROA
1; 3
1.267
2:32b
0.569
2:79c
0.725
1:91a
0.242
1:22
0.169
0:62
0.033
(0.08)
0.171
0:12
0.056
(1.54)
0.024
(0.89)
0.007
0:31
0.006
(0.40)
0.022
1:95a
0.046
2:43b
0.206
2:32b
1.334
5:52c
0.589
2:95c
0.729
1:93a
0.260
1:34
0.105
0:42
0.100
(0.24)
0.165
0:11
0.071
2:60c
0.006
(0.24)
0.004
(0.17)
0.003
0:26
0.014
1:78a
0.040
3:08c
0.167
2:65c
0.595
9:56c
1.412
2:65c
0.565
2:74c
0.669
1:80a
0.237
1:18
0.242
0:90
0.166
(0.42)
0.603
(0.40)
0.053
1:89a
0.008
(0.31)
0.007
(0.32)
0.007
(0.64)
0.011
(1.07)
0.025
1:84a
0.162
2:46b
IOROA 1; 3
COROA 1; 3
Industry-adjusted OROA 1
0.668
15:97c
1.802
(1.53)
0.507
6:81c
0.320
(0.30)
0.461
13:07c
Control group-adjusted OROA 1

Ln(Assets)
0.036
(0.61)
IML
N-total
N-remain independent
Regression p-value
0.009
2:98c
0.098
0:74
613
0.001
0.016
(0.27)
0.010
5:21c
0.038
0:32
613
556
0.000
0.006
0.006
(0.13)
0.021
(0.37)
0.743
1:40
0.008
3:98c
0.017
0:13
619
552
0.000
0.000
551
0.000
respectively.
ARTICLE IN PRESS
266
In all of the OLS regressions shown in Table 8 (columns 2, 4, and 6), the estimated
coefcients of the forced turnover and outside successor variables, though statistically
insignicant at conventional levels of condence, are positive. This is consistent with the
improved management hypothesis and with the results in Tables 2 and 6. The estimated
coefcients for the outsider-dominated board variable are also positive in all OLS
regressions and are signicant when we do not control for contemporaneous changes in
control rm performance. The estimated coefcient of fractional institutional share
ownership, however, is positive and signicant in all of the regressions. This, the only
truly robust result obtained for the governance variables in these regressions, tends to
support the improved management hypothesis.
We also nd strong evidence that the announcement-period abnormal stock
returns help predict performance changes. In each OLS regression, the estimated
coefcient for the abnormal return is positive and signicant at the 5% level or
better. These results suggest that positive abnormal returns around turnover events
reect rational anticipation by investors of later rm performance improvements.
Given this interpretation, positive stock price reactions to turnover news tend to
support the improved management hypothesis.
The regression results show that the estimated coefcients for abnormal return are
signicant in regressions that include board composition, fractional institutional
ownership, and the other structural variables derived from theory. This implies that
investors use information beyond that contained in these variables to forecast rm
performance following turnover.
One possible explanation for the statistically weak relations between governance
variables, other than institutional ownership, and OROA1; 3 in Table 8 is that
some of the governance variables are highly correlated with other regressors. To
explore the possibility that multicollinearity is inuencing the results, we examine the
correlations between the independent variables in Table 8. Table 9 shows that the
correlations between the governance variables and other regressors in Table 8 tend to
be relatively small. The notable exceptions are the correlations between institutional
share and lagged industry-adjusted OROA (industry-adjusted OROA1) r
0:2339 and between the forced turnover and outside succession dummy variables
r 0:4726: However, examination of the results from Model 2 (which excludes
industry-adjusted OROA) and Model 3 (which includes industry-adjusted OROA) in
Table 8 indicates that the relatively high correlation between institutional share and
industry-adjusted OROA1 does not have a substantial impact on the evidence.
Furthermore, re-estimation of all three models in Table 8, rst excluding the forced
turnover dummy, and then excluding only the outside succession dummy, reveals
that the high correlation between these two variables also has little impact on the
coefcient estimates.
Table 5 presents evidence that post-turnover changes in OROA are larger in the
second half of the sample period than in the rst half. As noted earlier, this increase
coincided with changes in the characteristics of governance mechanisms. It is also
possible that the larger changes in OROA in the second half of the sample period
reect shifts in the fundamental effectiveness of corporate governance mechanisms.
For example, outside directors or institutional investors might have become more
Takeover
Outsider
Outside
board
[619]
[619]
[619]
[619]
Forced
1.0000
Takeover
0.1166c
Outsider
0.4726c
Outside board
0.0791b
Institutional share
0.0880b
Abnormal return
0.0318
IOROA1; 3
0.0845b
COROA1; 3
0.0958b
Industry-adjusted OROA 1
0.1486c
Control group-adjusted OROA 1 0.0330
Ln(Assets)
0.0071
1.0000
0.0630
0.0010
0.1341c
0.0204
0.0146
0.0711a
0.1062c
0.0485
0.0866b
Institutional Abnormal IOROA1; 3 COROA1; 3 Industry-adjusted

Control
Ln(Assets)
share
return
OROA 1
group-adjusted
OROA1
[619]
[619]
[552]
[551]
[613]
[608]
[619]
1.0000
0.0689a
1.0000
0.0309 0.0266
1.0000
c
0.0331 0.0225
0.1176
c
0.0197
0.1269
0.0300
0.0699
0.2283c 0.0899b
c
c
0.1414 0.1529
0.2339c
0.0416
0.0246 0.0833b
0.0132
0.0985b
0.1324c
1.0000
0.0715a
0.0814a
0.0503
0.0326
0.0400
1.0000
0.4543c
0.1388c
0.0252
0.1350c
1.0000
0.6651c
0.1782c
0.1888c
1.0000
0.3187c
0.1854c
1.0000
0.1089c
1.0000
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels (for two-tailed tests), respectively.
ARTICLE IN PRESS
Forced
Table 9
Correlations between regressors
This table reports correlations between the independent variables used in the sample selection models in Tables 8 and 10. The sample used in these models
consists of 619 CEO turnovers announced over the 19801994 period. The variables include dummy variables that equal one if the old CEO is forced from
ofce (Forced), if the rm was subject to takeover pressure during the 12 months preceding turnover (Takeover), if the new CEO is employed with the rm for
one year or less at the time of turnover (Outsider), and if at least 60% of board members are outsiders (Outside board), and zero otherwise, as well as the
fraction of rm stock held by institutional investors (Institutional share), abnormal stock returns over the day before and the day of the Wall Street Journal
announcement of the turnover (Abnormal return), changes in industry IOROA1; 3 or control COROA1; 3 rm OROA over the period from one
year before to three years after the year of CEO turnover, rm industry-adjusted and control group-adjusted OROA in the year prior to turnover
OROA1; and the natural log of assets at the end of the scal year immediately preceding the turnover announcement (Ln(Assets)). Observations where the
new CEOs tenure was less than three years are excluded. Extreme performance observations are also excluded (see footnote 5 for details). The number of
turnovers for which the indicated variable is available is reported in brackets under the column header. The number of observations used to estimate each
correlation is the smaller of the number of observations available for the two variables. For example, 552 observations are used to estimate the correlation
between IOROA1; 3 and Forced.
267
268
Intercept
OROA1; 3
Firm remains
independent
OROA1; 3
Firm remains
independent
OROA1; 3
1.267
2:32b
0.068
1:91a
0.027
1:10
0.011
(0.39)
0.018
(1.11)
0.002
(0.09)
1.334
5:52b
0.082
2:90c
0.042
2:13b
0.002
0:06
0.010
(0.69)
0.009
(0.40)
1.412
2:65c
0.062
2:15b
0.052
2:65c
0.003
0:10
0.015
(0.98)
0.009
(0.42)
0.569
2:79c
Forced 19881994 dummy

Takeover
Model 3
Firm remains
independent
19881994 dummy
Forced
Model 2
0.725
1:91a
0.589
2:95a
0.729
1:93a
0.565
2:74c
0.669
1:80a
ARTICLE IN PRESS
Model 1
Table 10
Sample selection models of post-turnover performance: changes in sensitivity to governance characteristics over time
This table reports the results of sample selection models, estimated as described by Heckman (1979), in which the dependent variable for the rst regression
equals one if the rm survives as an independent entity for three years after the CEO turnover announcement and zero otherwise. The sample consists of CEO
turnovers announced over the 19801994 period. The dependent variable for the second regression equals the change in operating performance over the period
from one year before to three years after the year of CEO turnover OROA1; 3: The second regression is estimated using only data for rms that survived
three years. The independent variables include dummy variables that equal one if the turnover is announced during the 19881994 period (19881994
Dummy), if the rm was subject to takeover pressure during the 12 months preceding turnover (Takeover), if the old CEO is forced from ofce (Forced), if the
new CEO is employed with the rm for one year or less at the time of turnover (Outsider), and if at least 60% of board members are outsiders (Outside board),
and zero otherwise, as well as the fraction of rm stock held by institutional investors (Institutional share), abnormal stock returns over the day before and the
day of the Wall Street Journal announcement of the turnover (Abnormal return), changes in industry IOROA1; 3 or control COROA1; 3 rm
OROA over the period from one year before to three years after the year of CEO turnover, rm industry-adjusted and control group-adjusted OROA in the
year prior to turnover OROA1; and the natural log of assets at the end of the scal year immediately preceding the turnover announcement (Ln(Assets)).
IML is the inverse Mills ratio. Cases where the new CEOs tenure was less than three years are excluded. Extreme performance observations are also excluded
(see footnote 5 for details). t-Statistics are reported in parentheses.
Takeover 19881994 dummy

Outsider
0.242
1:22
Outsider 19881994 dummy

0.169
0:62
Outside board 19881994 dummy

Institutional share
0.033
(0.08)
Abnormal return
IOROA1; 3 (Model 2)/
COROA1; 3 (Model 3)
Industry-adjusted OROA 1 (Models 1 and 2)/
Control group-adjusted OROA1 (Model 3)
LnAssets
0.171
0:12
1.802
(1.53)
0.036
(0.61)
IML
N-total
N-remain independent
Regression p-value
0.497
6:96c
0.009
3:30c
0.087
0:68
613
0.000
0.105
0:42
0.100
(0.24)
0.165
0:11
0.320
(0.30)
0.016
(0.27)
613
556
0.000
0.006
0.237
1:18
0.242
0:90
0.166
(0.42)
0.603
(0.40)
0.006
(0.13)
0.021
(0.37)
0.002
0:12
0.016
(1.15)
0.016
1:15
0.004
0:37
0.055
3:66c
0.033
1:89a
0.010
0:38
0.153
2:32b
0.661
15:86c
0.848
1:62
0.008
3:94c
0.006
0:04
619
552
0.000
Superscripts a, b, and c denote statistical signicance at the 10%, 5%, and 1% levels (for two-tailed tests), respectively.
0.000
551
0.000
ARTICLE IN PRESS
Institutional share 19881994 dummy
0.260
1:34
0.010
0:60
0.007
(0.48)
0.008
0:59
0.001
(0.07)
0.053
3:37c
0.043
2:57b
0.011
0:43
0.151
2:42b
0.603
9:72c
0.452
12:98c
0.010
5:35c
0.032
0:27
Outside board
0:018
0:81
0.009
(0.60)
0.010
0:63
0.011
(0.90)
0.037
1:93a
0.039
1:79a
0.001
0:02
0.186
2:23b
269
ARTICLE IN PRESS
270
effective monitors. To examine this possibility, we added time-period dummy

variables and interaction terms for the forced turnover dummy variable and the
governance variables to the second regression in each of the three models in Table 8.
The time-period dummy variable equals one if the turnover is announced during or
after 1988 and zero otherwise.9 The coefcient estimates from these modied models
are presented in Table 10. The results show that the unconditional change in OROA
was actually lower, but that the change in OROA was more strongly inuenced by
the presence of an outsider-dominated board, during the 19881994 period than in
the previous eight years. The presence of an outsider-dominated board is associated
with a greater increase in the post-turnover change in OROA during the 19881994
period. This is consistent with the idea that outside directors became more effective.
It is also consistent with evidence, discussed by Huson et al. (2001), that stock-based
compensation for outside directors became more prevalent after the mid-1980s and
that outside directors that receive stock-based incentive compensation better
represent shareholder interests (Perry, 1998).
We estimate several regression models in addition to those reported in Tables 8
and 10. For example, in addition to investigating the potential impact of
multicollinearity on the reported evidence, we consider models containing other
subsets of the variables shown in Table 8 as well as models that do not contain the
logarithm of assets. We substitute the fraction of outside directors on the board for
the outsider-dominated board dummy variable. Unadjusted, industry-adjusted, and
control group-adjusted OROS measures were examined, as well as the OROA
measures. For the control group-adjusted measures, we try performance matching
over three years before management turnover instead of one year before turnover.
The results of these numerous experiments add little to those reported in Tables 8
and 10.
In addition, we estimate logistic models of the probability that performance
improves (i.e., that the change in OROA is positive) after turnover. We do not report
the results in detail because they are similar in most respects to those discussed
above.10 The evidence of a positive relation between performance improvement and
outsider-dominated boards, however, is somewhat stronger within the logistic
framework than with the ordinary linear model.
We analyze further the relations between abnormal return, board composition,
ownership structure, external takeover activity, successor origin, and rm
performance changes following CEO successions. To examine potentially nonlinear
relations between the variables, we rst extract residuals from regressions of changes
in OROA on rm size, lagged adjusted performance, and changes in industry or
control rm OROA. Firms are ranked by their residual performance changes and
grouped into quintiles. We then compare across quintiles the distributions of
abnormal returns and institutional shareholdings as well as of the dummy variables
9
Note that the regressions are estimated using only data for the 19801994 period because institutional
ownership data are only available beginning in 1980. The dummy variable allows us to compare the rst
half of this period with the latter half.
10
The results are available from the authors on request.
ARTICLE IN PRESS
271
Table 11
Characteristic comparisons for all CEO turnover observations grouped by mean changes in industryadjusted operating return on assets (OROA)
This table reports comparisons of rm characteristics for rms grouped by mean changes in OROA after
controlling for lagged performance, rm size, and changes in industry performance around all CEO
turnover events. Characteristics considered are the announcement-date abnormal stock returns (Abnormal
return), the fraction of rm stock held by institutional investors (Institutional share), and dummy variables
distinguishing successions at rms where at least 60% of board members are outsiders (Outside board),
successions where takeover activity occurred during the preceding 12 months (Takeover), forced
successions (Forced), and outside successors (Outsider). Abnormal stock returns are summed over the day
of and the day before the Wall Street Journal announcement of the turnover. Cases where the new CEOs
tenure ended before the end of year +3 are excluded. Extreme performance observations are also excluded
(see footnote 5 for details). p-Values for differences across groups in the means of OROA, abnormal
return, and institutional share are based on standard t-statistics. Median tests using median-sign (BrownMood) tests provide p-values similar to those reported in the table for the abnormal return and
institutional share variables. For the dummy variables (Outside board, Takeover, Forced, and Outsider),
the p-values are based on tests of the differences in proportions. (see Harnett, 1982, p. 401).
Industry-adjusted
OROA 1; 3
Panel A: performance change quintiles
Quintile 1
Mean (%)
6.21
(Lowest 20%)
N
{197}
Quintiles 24
Mean (%)
0.41
(Middle 60%)
N
{594}
Quintile 5
Mean (%)
7.45
(Highest 20%)
N
{197}
Abnormal
return
Outside
board
Takeover
Forced
Institutional
share
Outsider
0.01
{192}
0.03
{585}
1.08
{196}
74.36
{195}
84.43
{591}
75.13
{197}
6.09
{197}
5.56
{593}
8.63
{197}
12.18
{197}
11.78
{594}
11.17
{197}
45.49
{93}
44.35
{332}
46.94
{132}
14.21
{197}
14.81
{594}
18.27
{197}
0.755
0.593
0.275
0.880
0.526
0.836
0.817
0.177
0.247
Panel B: p-values for tests that performance change groups characteristics are equal
Tests that means are equal across quintiles 1 and 5
o0.001
0.861
0.334
0.917
0.002
0.781
0.001
0.004
0.125
indicating outsider-dominated boards, prior takeover activity, forced successions,

and outside successors.
In Table 11 we present the comparisons for groups of rms ranked by changes in
OROA after controlling for rm size, lagged industry-adjusted performance, and
industry changes in OROA. Consistent with the results in Tables 8 and 10,
announcement-period abnormal returns increase monotonically over the performance change groups. There is no apparent relation between performance ranking
and the incidences of takeover activity, forced succession, institutional holdings, and
outsider succession.
The evidence in Table 11 suggests that the relation between performance changes
and board composition is not monotonic. The proportion of rms with outsiderdominated boards is signicantly greater for the middle performance change group
than for either the highest or lowest performance change groups. Neither the
improved management hypothesis nor the scapegoat hypothesis predicts this result.
ARTICLE IN PRESS
272
Table 12
Characteristic comparisons for all CEO turnover observations grouped by changes in control groupadjusted operating return on assets (OROA)
This table reports comparisons of rm characteristics for rms grouped by changes in OROA after
controlling for lagged performance, rm size, and changes in control rm performance around all CEO
turnover events. Characteristics considered are the announcement-date abnormal stock returns (Abnormal
return), the fraction of rm stock held by institutional investors (Institutional share), and dummy variables
distinguishing successions at rms where at least 60% of board members are outsiders (Outside board),
successions where takeover activity occurred during the preceding 12 months (Takeover), forced
successions (Forced), and outside successors (Outsider). Abnormal stock returns are summed over the day
of and the day before the Wall Street Journal announcement of the turnover. Cases where the new CEOs
tenure ended before the end of year +3 are excluded. Extreme performance observations are also excluded
(see footnote 5 for details). p-Values for differences across groups in the means of abnormal return and
institutional share are based on standard t-statistics. Median tests using median-sign (Brown-Mood) tests
provide p-values similar to those reported in the table for the abnormal return and institutional share
variables. For the dummy variables (Outside board, Takeover, Forced, and Outsider), the p-values are
based on tests of the differences in proportions (see Harnett, 1982, p. 401).
Control group-adjusted Abnormal Outside Takeover Forced Institutional Outsider
OROA 1; 3
return
board
share
Panel A: performance change quintiles
Quintile 1
Mean (%)
6.14
(Lowest 20%)
N
{196}
Quintiles 24 Mean (%)
0.33
(Middle 60%)
N
{590}
Quintile 5
Mean (%)
7.14
(Highest 20%)
N
{196}
0.29
{195}
0.02
{575}
0.99
{196}
72.31
{195}
84.84
{587}
75.90
{195}
6.12
{196}
6.11
{589}
7.14
{196}
Panel B: p-values for tests that performance change groups characteristics are equal
0.086
0.418
0.685
0.337
0.000
0.999
0.004
0.004
0.609
13.27
{196}
10.51
{590}
14.29
{196}
43.59
{94}
44.33
{330}
48.85
{132}
14.80
{196}
14.24
{590}
19.39
{196}
0.770
0.032
0.227
0.722
0.730
0.941
0.150
0.017
0.085
Institutional ownership is also not monotonic across the performance change

groups, although institutional shareholdings are highest in the rms with the largest
performance changes.
Table 12 contains the results for comparisons of rms ranked by changes in
OROA after controlling for rm size, lagged control group-adjusted performance,
and control group changes in OROA. Again, we observe a positive relation between
performance rank and announcement-date abnormal returns.11 We also observe
signicantly higher concentrations of institutional ownership and outside succession
in the highest performance change quintile. This indicates that the turnover-related
change in managerial quality is positively related to these variables and tends to
11
It is worth noting that while the relation between performance rank and the mean abnormal returns in
Table 12 is not monotonic, the median abnormal returns do exhibit a positive monotonic relation.
ARTICLE IN PRESS
273
refute the scapegoat hypothesis. The relation between board composition and
control group-adjusted performance changes exhibits the same puzzling inverted U
pattern observed in Table 11.
5. Conclusions
We nd that rm nancial performance tends to deteriorate prior to top
management turnover. This result is statistically signicant and quite robust in our
sample. Hence, the evidence is consistent with that of Warner et al. (1988) and
Weisbach (1988). It indicates that boards of directors punish poor performance by
replacing CEOs. Like Bonnier and Bruner (1989) and Weisbach (1988), we nd also
that statistically signicantly positive average abnormal stock returns coincide with
management turnover announcements. Moreover, we show that turnover announcement abnormal stock returns are signicantly positively related to subsequent
changes in operating performance measured using accounting numbers. This
suggests that investors view turnover announcements as good news because they
expect that turnover will prompt performance improvements, on average.
To assess actual performance changes following turnover, we estimate unadjusted,
industry-adjusted, and control group-adjusted changes in operating performance.
Following turnover, we nd that the average change in unadjusted OROA for the
entire sample is actually negative, and unlike Denis and Denis (1995), we nd only
weak evidence that mean industry-adjusted OROA increases. Average changes in
control group-adjusted performance measures, however, are positive and highly
signicant. We believe that the control group-adjusted measures provide the most
reliable basis for assessing post-turnover changes in managerial quality because they
best control for mean reversion in performance. Hence, we conclude that managerial
quality and expected rm operating performance increase after CEO turnover. Our
evidence favors the improved management hypothesis over the scapegoat hypothesis. In addition, our analysis indicates that the increase in managerial quality
following turnover is signicantly greater in the latter half of our sample period
(19831994) than in the rst half (19711982). The quality of CEO replacement
decisions has improved over time.
We also investigate the cross-sectional determinants of post-turnover performance
changes. Simple pairwise comparisons suggest that board composition, institutional
shareholdings, takeover pressure, and outside successor CEOs positively affect
expected performance changes. Institutional shareholdings are signicantly positively related to post-turnover performance changes in our multivariate regression
analysis. In addition, though the presence of an outsider-dominated board does not
appear to be associated with an increase in post-turnover OROA during the
19801987 period, there is a strong positive relation in the 19881994 period.
Takeover pressure and the appointment of outside successors are not signicantly
related to post-turnover performance changes in our regression analysis. Overall, the
evidence from the cross-sectional analysis is consistent with the improved management hypothesis.
ARTICLE IN PRESS
274
One piece of unexpected evidence involves outside directors. Outsider-dominated

boards are less common among rms in the lowest and highest post-turnover
performance improvement groups than in the middle group. It is possible that
outside directors inuence rm behavior in ways that are more complex than
anticipated.
References
Agrawal, A., Knoeber, C.R., 1996. Firm performance and mechanisms to control agency problems
between managers and shareholders. Journal of Financial and Quantitative Analysis 31, 377397.
Agrawal, A., Mandelker, G., 1990. Large shareholders and the monitoring of managers: the case of
antitakeover charter amendments. Journal of Financial and Quantitative Analysis 25, 143161.
Agrawal, A., Mandelker, G., 1992. Shark repellents and the role of institutional investors in corporate
governance. Managerial and Decision Economics 13, 1522.
Barber, B.M., Lyon, J.D., 1996. Detecting abnormal operating performance: the empirical power and
specication of test statistics. Journal of Financial Economics 41, 359399.
Black, B.S., 1992. Institutional investors and corporate governance: the case for institutional voice.
Journal of Applied Corporate Finance 5, 1932.
Bonnier, K.-A., Bruner, R.F., 1989. An analysis of stock price reaction to management change in
distressed rms. Journal of Accounting and Economics 11, 95106.
Borokhovich, K.A., Parrino, R., Trapani, T., 1996. Outside directors and CEO selection. Journal of
Financial and Quantitative Analysis 31, 337355.
Byrd, J.W., Hickman, K.A., 1992. Do outside directors monitor managers? Evidence form tender offer
bids. Journal of Financial Economics 32, 195221.
Chan, W., 1996. External recruitment versus internal promotion. Journal of Labor Economics 14,
555570.
Clarke, R.N., 1989. SICs as delineators of economic markets. Journal of Business 62, 1731.
Comment, R., Schwert, G.W., 1995. Poison or placebo? Evidence on the deterrence and wealth effects of
modern antitakeover measures. Journal of Financial Economics 39, 343.
Coughlan, A.T., Schmidt, R.M., 1985. Executive compensation, management turnover, and rm
performance. Journal of Accounting and Economics 7, 4366.
Dalton, D.R., Kesner, I.F., 1985. Organizational performance as an antecedent of inside/outside chief
executive succession: an empirical assessment. Academy of Management Journal 28, 749762.
Demsetz, H., Lehn, K., 1985. The structure of corporate ownership: causes and consequences. Journal of
Political Economy 93, 11551177.
Demsetz, H., Villalonga, B., 2001. Ownership structure and corporate performance. Journal of Corporate
Finance 7, 209233.
Denis, D.J., Denis, D.K., 1995. Firm performance changes following top management dismissals. Journal
of Finance 50, 10291057.
Denis, D.J., Serano, J.M., 1996. Active investors and management turnover following unsuccessful control
contests. Journal of Financial Economics 40, 239266.
Denis, D.J., Denis, D.K., Sarin, A., 1997. Ownership structure and top executive turnover. Journal of
Financial Economics 45, 193221.
Fama, E., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of
Fee, C.E., Hadlock, C.J., 2003. Management turnover across the corporate hierarchy. Unpublished
working paper, Michigan State University.
Harnett, D.L., 1982. Statistical Methods, 3rd Edition. Addison-Wesley Publishing Company, Reading,
MA.
Heckman, J.J., 1979. Sample selection bias as a specication error. Econometrica 47, 153161.
.
Holmstrom,
B., 1979. Moral hazard and observability. The Bell Journal of Economics 10, 7491.
ARTICLE IN PRESS
275
Hotchkiss, E.S., 1995. Postbankruptcy performance and management turnover. Journal of Finance 50,
321.
Huson, M.R., Parrino, R., Starks, L.T., 2001. Internal monitoring mechanisms and CEO turnover: a long
term perspective. Journal of Finance 56, 22652297.
Khanna, N., Poulsen, A.B., 1995. Managers of nancially distressed rms: villains or scapegoats? Journal
of Finance 50, 919940.
Kim, Y., 1996. Long-term rm performance and chief executive turnover: an empirical study of the
dynamics. Journal of Law, Economics, and Organization 12, 480496.
Mikkelson, W.H., Partch, M.M., 1997. The decline of takeovers and disciplinary management turnover.
Journal of Financial Economics 44, 205228.
Mirrlees, J.A., 1976. The optimal structure of incentives and authority within an organization. The Bell
Journal of Economics 7, 105131.
Mitchell, M.L., Stafford, E., 2000. Managerial decisions and long-run stock price performance. Journal of
Business 73, 287320.
Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation: an empirical
analysis. Journal of Financial Economics 20, 293316.
Perry, T., 1998. Incentive compensation for outside directors and CEO turnover. Unpublished working
paper, Arizona State University.
Pound, J., 1992. Raiders, targets, and politics: the history and future of American corporate control.
Journal of Applied Corporate Finance 5, 618.
Reinganum, M.R., 1985. The effect of executive succession on stockholder wealth. Administrative Science
Quarterly 30, 4660.
Rosenstein, S., Wyatt, J.G., 1990. Outside directors, board independence, and shareholder wealth. Journal
of Financial Economics 26, 175191.
Shavell, S., 1979. Risk sharing and incentives in the principal and agent relationship. The Bell Journal of
Economics 10, 5573.
Shleifer, A., Vishny, R.W., 1986. Large shareholders and corporate control. Journal of Political Economy
94, 461488.
Warner, J.B., Watts, R.L., Wruck, K.H., 1988. Stock prices and top management changes. Journal of
Weisbach, M.S., 1988. Outside directors and CEO turnover. Journal of Financial Economics 20, 431460.
Weisbach, M.S., 1995. CEO turnover and the rms investment decisions. Journal of Financial Economics
37, 159188.
Yermack, D., 1996. Higher market valuation of companies with a small board of directors. Journal of

Articol HUSON MALATESTA SI PARRINO 2004 PDF

Enviado por

Dados do documento

Descrição original:

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Articol HUSON MALATESTA SI PARRINO 2004 PDF

Enviado por

Direitos autorais:

Formatos disponíveis

ARTICLE IN PRESS

Journal of Financial Economics 74 (2004) 237275

Managerial succession and rm performance$

Department of Finance and Management Science, School of Business, University of Alberta,

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

managements poor performance, rm value might fall on the news of management

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

no reliable difference between post-turnover performance changes for forced and

2. Theory and hypotheses

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

takeover market as a disciplinary mechanism suggests that takeover pressure will

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

(2) the Wall Street Journal reported the succession announcement;

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Control group-adjusted OROA

Panel A: unadjusted OROA

Panel B: industry-adjusted OROA

Panel C: control group-adjusted OROA

group are signicantly negative at the 1% level. Furthermore, the pre-turnover

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

p-Value for test of

%D Total book assets (1; +3)

%D Total book assets (2; +2)

%D Capital expenditures (1; +3)

%D Gross PP & E (1; +3)

%D Number of employees (1; +3)

%D Operating income per employee (1; +3)

4.3. Stock price performance around turnover events

Panel A: all turnovers

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

4.4. Variation in performance changes over time

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Panel A: comparison of change in OROA from year 1 to

year 3: 19851988 vs. other years

Panel B: comparison of change in OROA from year 1 to

year 3: 19711982 vs. 19831994

161 0.003 0.010

463 0.008c 0.001

466 0.003 0.000 52 0.011

industry-adjusted and control group-adjusted OROA from year 3 to year 1 were

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Control group-adjusted OROA 1

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Institutional Abnormal IOROA1; 3 COROA1; 3 Industry-adjusted

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Forced 19881994 dummy

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Takeover 19881994 dummy

Outsider 19881994 dummy

Outside board 19881994 dummy

Institutional share 19881994 dummy

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

effective monitors. To examine this possibility, we added time-period dummy

indicating outsider-dominated boards, prior takeover activity, forced successions,

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275

Institutional ownership is also not monotonic across the performance change

M.R. Huson et al. / Journal of Financial Economics 74 (2004) 237275