Harvard Law School
Empirical Law and Finance
Fall 2023
Lucian Bebchuk and Alma Cohen
Package #2
Session No. 1: Lucian Bebchuk (HLS),
and Scott Hirst (BU)
Lucian Bebchuk, “Dancing with Activists,” Journal of Financial Economics (2020), (co-authored with Alon
Brav, Wei Jiang, and Thomas Keusch).
Scott Hirst, “Oracles of the Vote: Predicting the Outcomes of Proxy Contests,” (March 2022), (co-authored
with Oğuzhan Karakaş, and Ting Yu).
NOTE: This session will take place on Monday, September 18, from 3:45 – 5:45 PM. Professor
Lucian Bebchuk (HLS) and Professor Scott Hirst (BU) will participate in the session and discuss
their respective papers with the class.
Students electing to submit one memo should focus on one of the papers. Students electing to
submit two memos should have one memo focused on one of the papers and the other memo focus
on the other paper.
Journal of Financial Economics 137 (2020) 1–41
Contents lists available at ScienceDirect
Journal of Financial Economics
journal homepage: www.elsevier.com/locate/jfec
Dancing with activists ✩
Lucian A. Bebchuk a, Alon Brav b, Wei Jiang c, Thomas Keusch d,∗
a
Harvard Law School, Harvard University, 1545 Mass. Ave., Cambridge, MA 02138, United States
Fuqua School of Business, Duke University, 100 Fuqua Dr., Durham, NC 27708, United States
Columbia Business School, Columbia University, 3022 Broadway, New York, NY 10027, United States
d
INSEAD, Boulevard de Constance, 77305 Fontainebleau Cedex, France
b
c
a r t i c l e
i n f o
Article history:
Received 27 November 2017
Revised 6 May 2019
Accepted 31 May 2019
Available online 29 January 2020
JEL Classification:
G12
G23
G32
G34
G35
G38
K22
Keywords:
Corporate governance
Hedge fund activism
Activist settlements
a b s t r a c t
An important milestone often reached in the life of an activist engagement is entering
into a “settlement” agreement between the activist and the target’s board. Using a comprehensive hand-collected data set, we analyze the drivers, nature, and consequences of
such settlement agreements. Settlements are more likely when the activist has a credible threat to win board seats in a proxy fight and when incumbents’ reputation concerns
are stronger. Consistent with incomplete contracting, face-saving benefits, and private information considerations, settlements commonly do not contract directly on operational
or leadership changes sought by the activist but rather on board composition changes.
Settlements are accompanied by positive stock price reactions, and they are subsequently
followed by changes of the type sought by activists, including CEO turnover, higher shareholder payouts, and improved operating performance. We find no evidence to support concerns that settlements enable activists to extract rents at the expense of other investors.
Our analysis provides a look into the “black box” of activist engagements and contributes
to understanding how activism brings about changes in target companies.
© 2020 Elsevier B.V. All rights reserved.
1. Introduction
✩
We are especially grateful to an anonymous referee for many valuable suggestions. We have also benefited from the comments of Manuel
Adelino, Amil Dasgupta, Veljko Fotak, Oliver Hart, Scott Hirst and participants in the 2015 FMA International Consortium on Activist Investors,
Corporate Governance and Hedge Funds in London, the 2015 FMA International meeting in Helsinki, and workshops at BlackRock, Harvard, INSEAD, NYU, Vanderbilt, and Maastricht University. We also wish to thank
Brady Baldwin, Seung Hwan Bang, Vlad Bouchouev, Hannah Clark, Ahmet Degerli, Timothy Goh, Robert Holowka, Wenyin Huang, Kirti Kamboj, Kobi Kastiel, Zheng Li, Cong Liu, Yaron Nili, John Rady, Qingrong
Ruan, June Rhee, Ruoxi Tian, Jun Xu, Jiaqi Yang, Zilan Yang, Kailei
Ye, and Lu Zheng for their excellent research assistance. Finally, we
are also grateful to many practitioners affiliated with the Harvard Law
School Program on Corporate Governance, including senior proxy solicitors, investment bankers, and hedge fund officers, for helpful discussions
about activist settlements. Financial support was provided by the INSEAD
https://doi.org/10.1016/j.jfineco.2020.01.001
0304-405X/© 2020 Elsevier B.V. All rights reserved.
In August 2013, Third Point, the hedge fund led by
Daniel Loeb, disclosed a significant stake in the auction
house Sotheby’s, criticized the company for its poor governance and its failure to take advantage of a booming market for luxury goods, and called for the ouster of the company’s CEO. Third Point launched a proxy fight for board
representation, and both sides were preparing for a con-
Corporate Governance center, Columbia Business School, Duke, and Harvard Law School.
∗
Corresponding author.
E-mail
addresses:
bebchuk@law.harvard.edu
(L.A.
Bebchuk),
brav@duke.edu
(A.
Brav),
wj2006@columbia.edu
(W.
Jiang),
thomas.keusch@insead.edu (T. Keusch).
2
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
tested election at the company’s upcoming annual meeting. However, the day before the scheduled annual shareholder meeting, the company and the activist fund entered
into a settlement agreement in which Sotheby’s agreed to
appoint three of the Third Point director candidates and
Third Point agreed to discontinue the proxy fight. The settlement terms did not require the company to make any of
the operational and executive changes that Third Point was
seeking. However, ten months later, Sotheby’s announced
the hiring of a new CEO, the appointment of a new board
chairman, and a plan to return capital to its investors.
While such settlements used to be rare, they now
occur with significant frequency, and they have been attracting a great deal of media and practitioner attention.
Understanding settlement agreements is important for obtaining a complete picture of the corporate governance
landscape and the role of activism within it. Using a
comprehensive, hand-collected data set of settlement
agreements, this paper provides the first systematic empirical investigation of settlements between activists and target companies. We study the drivers of settlements, their
growth over time, the terms of settlements, and their impact on board composition, and the stock market reaction
accompanying them. We further study the aftermath of
settlements in terms of CEO turnover, payouts to shareholders, mergers and acquisitions (M&A) activity, and operating performance.
With the growing recognition of the importance of
hedge fund activism, a large empirical literature on the
subject has emerged (see Brav et al., 2015b for a recent
survey). This literature has extensively studied the initiation of activist interventions—the time at which activists
announce their presence, usually by filing a Schedule 13(d)
with the Securities and Exchange Commission (SEC) after
passing the 5% ownership threshold, and the stock market
reactions accompanying such announcements.1 This literature has also studied extensively the changes in the value,
performance, and behavior of firms in the years following activist interventions; among other things, researchers
have studied the changes in Tobin’s Q, return on assets
(ROA), productivity, innovation, payouts to shareholders,
likelihood of an acquisition, divestitures, internal capital
markets, and accounting practices that ultimately follow
activist interventions.2 But there has been limited empirical work on the “black box” in between—the channels
1
Studies analyzing such initiations, including the stock market reactions accompanying them, and the stock accumulations preceding them,
include Clifford (2008); Klein and Zur (2009); Boyson and Mooradian
(2011); Brav et al. (2008); Bebchuk et al. (2013); Mietzner and Schweizer
(2014); Edmans et al. (2013); Collin-Dufresne and Fos (2015); Becht et al.
(2017); Norli et al. (2015); Gantchev and Jotikasthira (2017); Back et al.
(2018); and Boyson et al. (2016).
2
Studies of the subject include Brav et al. (2008); Clifford (2008); Klein
and Zur (2009, 2011); Greenwood and Schor (2009); Becht et al. (2009);
Boyson and Mooradian (2011); Cheng et al. (2012); Sunder et al. (2014);
Gow et al. (2014, 2016); Cheng et al. (2015); Bebchuk et al. (2015); Brav
et al. (2015a); Boyson and Pichler (2019); Brav et al. (2018); Aslan and
Kumar (2016); Boyson et al. (2017); Boyson et al. (2016); Corum and Levit
(2019); Gantchev et al. (2019a); Jiang et al. (2018); Krishnan et al. (2016);
Johnson and Swem (2017); Swanson and Young (2017); Bourveau and
Schoenfeld (2017); Hege and Zhang (2019); Guo et al. (2019); Kim (2018);
and Gantchev et al. (2019b).
through which activists’ influence is transmitted and is
reflected in targets’ economic outcomes.3 In particular, the
determinants, nature, and role of settlement agreements—
and the cooperation between activists and companies that
they target—have not been subject to a systematic empirical examination. Our study fills this gap.4
We begin by quantifying the upward trend in activist settlements. We show that the unconditional likelihood of an activism campaign leading to a settlement
increased sevenfold and almost linearly from 3% for campaigns launched in 20 0 0 to 21% for campaigns launched in
2013. This increase in the prevalence of settlement agreements is much stronger than the increase in contested
votes during the same period. While contested votes also
occurred in 3% of all activism campaigns started in 20 0 0,
this ratio only increases to 6% for campaigns started in
2013.
We build on the insights generated by the economics of
litigation and settlements to put forward hypotheses concerning the determinants of which cases will produce a
settlement rather than end up without either a contested
vote or a settlement. We hypothesize that settlements are
more likely in cases in which the activist has good chances
to win board seats should a contested vote take place, and
we find evidence consistent with this hypothesis. We also
hypothesize that settlements are more likely when a contested vote could impose larger reputational costs on incumbents, and we find evidence that is also consistent
with that hypothesis. As to timing, we hypothesize that
settlements are more likely to take place closer to the time
of the target’s shareholder meeting, as the arrival of information regarding the outcome of the expected vote narrows any divergence of expectations between the parties,
and we provide evidence that is consistent with this hypothesis.
Turning to the terms of settlements, we argue that
the terms of settlement agreements can commonly be expected to focus on board turnover rather than to stipulate directly the kind of operational and leadership
changes that activists ultimately seek and on which they
base their hopes for value appreciation. We put forward
three reasons—incomplete contracting that make some
direct action commitments infeasible, face-saving benefits, and informational asymmetry—that combine to preclude or discourage the incorporation of commitments
for direct action. We provide evidence that is consistent
with the hypothesis that settlement terms tend to focus
on board composition and avoid commitments of direct
action. We also find evidence that face-saving benefits,
3
Some studies have provided insights on the engagement process (see,
e.g., Appel et al., 2019; Becht et al., 2009; Boyson and Pichler, 2019;
Corum, 2018; Gantchev, 2013; He and Li, 2018; Levit, 2018; and McCahery,
et al., 2016). However, none of the empirical studies has paid significant
attention to investigating settlement agreements and their role.
4
Recent work by Schoenfeld (2018) studies the determinants (but not
the terms or consequences) of contracts between a wide variety of shareholders (not focusing on activists) and publicly traded companies. In contrast to the settlement agreements in our study, the contracts in his study
include a variety of provisions that pertain to financing terms, trading, directorships, payout policy, joint ventures, financial reporting, and selective
disclosure.
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
as well as informational asymmetry, at least partly contribute to the documented patterns.
We then turn to investigate systematically the effects
of settlements on board composition. We demonstrate that
settlements are a key channel through which activists
bring about board changes. In particular, we find that settlements are associated with an increase in the number of
activist-affiliated, activist-desired, and well-connected directors and decrease the number of old and long-tenured
directors.
Next, we show that settlements are accompanied by
positive abnormal stock returns on average. We find that
the positive market reaction is especially large when the
settlement is of “high impact” in terms of creating a high
volume of board turnover or providing for an immediate strategic transaction. We also find that activist engagements that produce settlement agreements are associated with higher abnormal stock returns at the time of
the activist’s initial 13(d) filing. These patterns are overall consistent with the view that the market views favorably the boardroom composition and other changes that
activist settlements produce and are inconsistent with the
view that such changes can be expected to be disruptive
and detrimental to target shareholders.5
Why do activists settle on changes in board composition if their ultimate goal is to bring about operational
or leadership changes? We argue that introducing individuals into the boardroom who are aligned with the activist, or at least open to the changes sought by the activist, is an intermediary step that often facilitates such
changes. Consistent with this view, we show that, while
settlements generally do not stipulate a replacement of
the CEO, settlements are followed by a considerable increase in CEO turnover and in the performance-sensitivity
of CEO turnover. Thus, the evidence is consistent with settlements planting the seeds for a subsequent CEO replacement. Similarly, while settlement agreements generally do
not require any specific operational changes, settlements
might facilitate such changes: consistent with the presence
of such an effect, we show that settlements are followed
by increased payouts to shareholders and improvements in
operating performance.
Finally, we also investigate concerns raised by some
practitioners and commentators that settlements between
activists and targets enable activists to extract rents at the
expense of other shareholders who are not “at the bargaining table.”6 We examine two suggested channels for such
rent extraction and find little evidence that settlements
distribute significant rents to activists at other shareholders’ expense. First, we find no evidence to support concerns that settlements enable activists to put on the board
5
We investigate the effects of activism settlements on a wide array of
dimensions, but we do not cover all possibly relevant dimensions. For example, in a recent paper, Coffee et al. (2019) focuses on the effect of such
settlements with activists on subsequent “information leakage” prior to
the issuance of 8-K filings and on bid-ask spreads.
6
For recent media accounts, see Reuters, 07/18/2016, “Big funds push
back against settlement agreements” and State Street Global Advisors,
10/10/2016, “Protecting long-term shareholder interests in activist engagements.”
3
directors that are not supported by other shareholders.7
In particular, we show that directors who enter the board
through settlements do not subsequently receive less voting support at the following annual general meetings than
other directors. Second, we also find little evidence that
settlements produce a significant incidence of “greenmail”
through the target’s purchasing of the activist’s shares at a
premium to the market price. Buybacks of activist shares
occur in a very small fraction of settlement agreements,
and when they do occur, they are typically executed at the
market price.
Our analysis is organized as follows. Section 2 discusses
the institutional background. Section 3 describes the data
collection and sample construction. Section 4 examines the
determinants and timing of settlements. Section 5 focuses
on the terms of settlements and, in particular, on the
prevalence of director turnover and the infrequent specification of direct actions. Section 6 examines the director
changes on which settlement terms focus and analyzes
their effects on board composition. Section 7 considers the
stock market reactions that accompany the announcements
of settlements as well as the initial stock market reactions
to the initiation of campaigns that ultimately end up in
settlements. Section 8 focuses on the economic aftermath
of settlements. In particular, we examine the changes in
CEO turnover, shareholder payouts, likelihood of a strategic transaction, and operating performance that follow
settlements. Section 8 also examines concerns about rent
extraction by activists, and we find no evidence for such
extraction. Finally, we present our conclusions in Section 9.
2. Institutional background
The first stage of an activist intervention occurs when
an activist reveals its presence and possibly the changes
it seeks to instigate at the company (e.g., to the target’s
board). This often happens when the activist files an SEC
Schedule 13(d) upon crossing a 5% ownership threshold of
company shares outstanding.8 Objectives often proclaimed
for the activism campaign include (i) maximizing shareholder value, which is the most common goal; (ii) changing capital structure, for example, by distributing excess
cash to shareholders; (iii) changing business strategy, for
example, by refocusing the company; (iv) sale of the target
company; and/or (v) effecting governance changes, for example, by replacing directors or the CEO. However, activists
often do not disclose at the 13(d) stage what changes they
will ultimately demand but rather keep their options open.
Most activists do not initially take an adversarial position,
hoping to be able to reach a cooperative outcome, but 25%
of the activists in our sample (similar to the frequency reported in Brav et al. (2008) for an earlier sample) express
an adversarial position when disclosing their presence.
7
Leading index fund manager State Street Global Advisors expressed
such a concern. See State St. Global Advisors, “Protecting long-term shareholder interests in activist engagements” (2016).
8
The Schedule 13(d) filing is a mandatory filing under Section 13(d) of
the Securities Exchange Act that requires investors to disclose within ten
days of acquisition of, or conversion into, more than 5% of any class of
securities of a publicly traded company if they have an interest in influencing the control or the management of the company.
4
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
How can an activist bring about changes it is seeking
given that the power to set the course of the company
is vested in the company’s board of directors? One important route that has received significant attention is to
win board seats in a proxy fight (see, e.g., Gantchev, 2013;
Fos and Tsoutsoura, 2014; Fos and Jiang, 2016; Fos, 2017).
To do so, the activist needs to nominate director candidates, file a proxy statement, campaign for shareholder
support, and win one or more board seats in the contested
vote.
A settlement agreement offers a “cooperative” alternative to a contested vote. The settlement ends, at least for
a specified time, the conflict between the two sides by
requiring concessions on both sides. Because reaching an
agreement with an activist is a material event, companies
are required by the securities laws to disclose the existence
and terms of a settlement agreement.
In such agreements, the most important and almost
universal concession made by the activist is a “standstill”
provision. Such a provision specifies a period during which
the activist agrees to refrain from activities designed to influence the control or policies of the company. The termination of the standstill can be event driven (i.e., it remains
in effect until certain agreed-upon conditions are in effect)
or time driven (i.e., until a definite expiration date). There
are two major types of standstill agreements: (i) a share
ownership standstill that prohibits the activist from acquiring additional voting securities of the target company and
(ii) a corporate governance standstill that prohibits further
activist activities such as seeking a merger, seeking additional board representation, or soliciting the company’s
proxies.
In consideration for the activist’s concessions, incumbent directors entering a settlement agreement can agree
to take certain steps that would move the operations or
strategy of the company in a direction favored by the activist. The incumbents have the power to agree to specified buybacks or special dividends, to explore a sale of the
company or a recapitalization, or to replace the CEO. Alternatively, the incumbents can agree to make changes to
board composition—adding specified new directors and/or
avoiding renomination of specified incumbent directors—
that could affect subsequent decisions concerning the company’s operations and strategy.
3. Data and sample overview
3.1. Data sources
We obtain stock returns from the Center for Research
in Security Prices (CRSP) and accounting data from Compustat. The remaining data used in this study come from
various sources. First, data on activism events come from
a data set that is an extension of the sample studied in
Brav et al. (2008). The events are identified mainly through
Schedule 13(d) filings submitted to the SEC (accessible via
the EDGAR system), which disclose beneficial ownership
of 5% or more of any class of publicly traded securities
of a company where the investors intend to influence corporate policy or control. In putting together this data set,
Schedule 13(d) filings aiming at bankruptcy reorganization
or distress investing, and those representing risk arbitrage
after the announcements of mergers and acquisitions were
manually excluded because such events are not related to
shareholder activism. To supplement Schedule 13(d) filings,
news searches were conducted for activists who launch
public activism with stock ownership below 5% at mid- to
large-cap companies. We remove target firms whose CRSP
common share codes differ from 10 or 11, the standard
codes for common stock.
Our event sample period begins in 20 0 0, because settlements were rare before 20 0 0, and ends at the end of
2013. We follow the settlement outcomes to 2018. For
each activist fund in our sample, we collect information
on the number of past campaigns, past settlements, past
proxy fights, past successful campaigns, and average market reaction to past campaign announcements, all measured over the five years prior to the start of the focal
campaign.
Second, information on the existence of settlement
agreements between activists and targets was obtained
from FactSet’s Shark Watch, Capital IQ Key Developments,
and via news archive searches. We use these sources to
determine, for each activism campaign launched between
20 0 0 and 2013, whether a settlement agreement was
reached by May 2018. Third, information on the contents
of each settlement agreement was hand-collected from the
public filings that targets and activists made and supplemented using a news archive search.9 Specifically, we collected information on the specific dates that settlement
agreements were reached and publicly disclosed; the number of directors added and whether these directors were
affiliated with the activist, unaffiliated but desired by the
activist, or neither affiliated nor desired; the number of director departures stipulated in the agreement; whether the
CEO has to leave the firm; whether the firm sells itself or
a division or whether the firm plans to engage in such actions; whether the board creates a “strategic transactions”
committee; and whether the firm initiates or increases the
scale of a buyback program.
Fourth, for each activism campaign, we also collected
data on whether the activist filed proxy material and
whether the proxy fight was settled (in which case we
would already have collected the settlement agreement),
withdrawn, or went to a vote. Fifth, information on director characteristics and turnover are collected from the
Directors Database Archive, a product of Corporate Board
Member Magazine, which provides quarterly snapshots of
executives and directors of firms listed on the NYSE, Nasdaq, and Amex going back to the year 20 0 0 until year
2013.10 We supplement this data with information gathered from proxy filings and news searches to determine
whether the new director is affiliated with the activist fund
or, in the case of a director independent of the fund, at
9
Appendix A.5 provides an example of a representative settlement
agreement.
10
This database has recently been used in Brav et al. (2018); Larcker
et al. (2013); and Stuart and Yim (2010). Since Board Magazine was acquired by NYSE, it does not update or sell the Directors Database any
longer, which means that the last year for which we have data on all the
directors of companies listed on NYSE, Nasdaq, and Amex is 2013.
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
5
Table 1
Incidence of settlements and proxy votes over time.
This table shows the distribution of all activism campaigns that were launched between calendar years 20 0 0 and 2013 and the percentage of those
interventions that led to settlement agreements between the activist and the target company, the percentage of campaigns that went to a proxy vote, the
percentage of campaigns that were settled or went to a proxy vote or both (over the life of a campaign, both outcomes can occur sequentially), and the
percentage of settled campaigns relative to all campaigns that were either settled or went to a proxy vote or both.
Panel A. Activism campaign outcomes over time
Year
Activism
campaigns
Settlements/
Activism campaigns
(in %)
Voted contests/
Activism campaigns
(in %)
(Settlements + Voted
contests)/
Activism campaigns (in %)
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
116
93
138
128
158
254
344
410
332
166
201
213
217
242
3
3
7
9
8
11
12
11
15
16
19
17
20
21
3
4
8
4
7
3
4
8
3
3
5
2
6
5
8
14
13
13
15
14
15
22
19
22
21
21
27
50
43
50
71
57
73
82
74
71
87
84
82
93
78
Total (Average)
3012
13
5
17
76
Settlements/
(Settlements + Voted
contests) (in %)
Panel B. Proxy contest outcomes over time
Year
Proxy contests
Proxy contests/Activism
campaigns (in %)
Voted contests/
Proxy contests (in %)
Settled contests/
Proxy contests (in %)
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
8
8
22
18
24
33
39
46
49
15
20
24
21
34
7
9
16
14
15
13
11
11
15
9
10
11
10
14
38
50
50
28
46
33
26
35
51
33
35
42
19
44
38
13
32
56
38
42
69
52
33
40
40
38
62
41
25
38
18
17
17
24
5
13
16
27
25
21
19
15
Total (Average)
361
12
38
45
17
least favored by the fund. Sixth, director election voting
outcomes were obtained from ISS Voting Analytics for the
time period 2001–2018. This database covers the Russell
30 0 0. We matched the names of the directors of our sample firms to the director names provided in ISS Voting Analytics by hand.
Seventh, information on target firm CEOs and their
characteristics and CEO turnover during activism campaigns is collected from the intersection of ExecuComp
and Equilar for years 20 0 0–2017. Eighth, for each of the
activism campaigns in the sample, we searched FactSet’s
Shark Watch, Capital IQ, Thomson Reuters’ SDC, and
news archives for privately negotiated share repurchase
transactions between hedge fund activists and their target companies. Last, we collected information on target
companies’ share-class structure, insider (board and management) ownership, and analyst coverage from Equilar,
ISS, Thomson, FactSet Shark Watch, and IBES.
Withdrawn contests/
Proxy contests (in %)
3.2. Summary statistics
Panel A of Table 1 reports the annual frequency of activist interventions, settlements, contested votes, and filings of proxy statements. Over the 14-year period 20 0 0–
2013, there were 3012 hedge fund activism campaigns.
Of these campaigns, 13% (399 campaigns) resulted in a
settlement agreement between the activist and the target company, while only 5% (137 campaigns) resulted in
a contested vote. The table indicates that settlements were
rare during the beginning years of the 20 0 0–2013 period,
but the incidence of settlements has gone up considerably
since then, increasing from 3% for campaigns starting in
20 0 0 to 21% for campaigns starting in 2013. By contrast,
the incidence of contested votes has been relatively stable
during the period we examine.
Panel B tabulates the frequency and outcome of proxy
contests—defined as campaigns in which the activist filed
6
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 2
Descriptive statistics on activist interventions.
This table reports descriptive statistics on activist characteristics and target characteristics separately for activism campaigns that lead to a settlement
agreement, a proxy vote, or none of the above. The sample covers campaigns launched between 20 0 0 and 2013. The number of observations on the
different variables differs depending on the availability of information on that variable.
Activist track record
Market reaction past campaigns
Ln (# settlements)
Ln (# proxy fights)
Ln (# campaigns)
Ln (# successful campaigns)
Ownership structure
Multiple share classes
High activist ownership
Insider ownership
Incumbents’ reputation concerns
# Directors older than 70
# Outside directors older than 70
Directors’ avg # directorships
Outside directors’ avg # directorships
CEO up for election
Board chair up for election
Ln (Analyst coverage)
Firm performance
ROA
Tobin’s Q
Abnormal returns pre-13D
Control variables
Market cap ($ million)
Settled campaigns (n = 399)
Campaigns w/ proxy vote (n = 137)
Other campaigns (n = 2490)
Mean
Median
StDev
Mean
Median
StDev
Mean
Median
StDev
0.04
0.69
0.66
1.75
1.56
0.04
0.69
0.69
1.79
1.61
0.06
0.76
0.76
1.09
0.78
0.06
0.45
0.68
1.52
1.60
0.06
0.00
0.69
1.61
1.61
0.06
0.68
0.79
1.18
0.74
0.03
0.26
0.42
1.74
1.27
0.02
0.00
0.00
1.61
1.10
0.07
0.51
0.62
1.27
0.70
0.03
0.58
0.11
0.00
1.00
0.07
0.16
0.49
0.13
0.02
0.46
0.10
0.00
0.00
0.06
0.14
0.50
0.14
0.07
0.49
0.15
0.00
0.00
0.08
0.26
0.50
0.18
0.88
0.83
1.62
1.68
0.85
0.87
1.30
0.00
0.00
1.56
1.60
1.00
1.00
1.39
1.25
1.19
0.51
0.57
0.36
0.33
1.07
1.03
0.97
1.63
1.67
0.75
0.77
1.09
1.00
1.00
1.57
1.60
1.00
1.00
1.10
1.28
1.22
0.54
0.58
0.43
0.42
1.00
0.73
0.68
1.64
1.71
0.79
0.78
0.86
0.00
0.00
1.50
1.60
1.00
1.00
0.69
1.10
1.04
0.57
0.63
0.41
0.41
0.94
0.06
1.72
−0.15
0.08
1.29
−0.20
0.17
1.66
0.46
0.04
2.00
−0.12
0.07
1.32
−0.15
0.20
2.35
0.36
0.05
1.90
−0.08
0.08
1.36
−0.18
0.21
1.92
0.90
2613
338
6503
1497
384
3833
1205
211
3804
a proxy statement—over time. Activists file proxy statements and thereby formally start a proxy contest in 12%
of activism campaigns. Of the proxy contests initiated, 38%
of them go all the way to a contested vote, 45% are settled, and 17% are withdrawn. In untabulated analyses, we
find that 44.4% of settlements are preceded by the activist filing proxy material. Thus, cases in which a proxy
fight is formally initiated have a higher incidence of settlement than cases without a formal filing of a proxy statement, though activists still obtain the majority of settlements without having to formally file proxy material. In
such cases, however, the background threat of pursuing a
proxy fight might still play a role in inducing incumbent
directors to agree to a settlement.
Appendices A.1 and A.2 provide information about notable targets and activists. Appendix A.1 lists settlements
related to activism campaigns launched between 20 0 0 and
2013 regarding firms with market capitalization of $7 billion and above, a total of 30 cases. The two largest target companies during the sample period were Microsoft,
which settled with ValueAct in 2013, and PepsiCo, which
settled with Trian Partners in 2015. Appendix A.2 lists the
30 activists that were involved in four or more settlements.
The three activists that entered into the largest number
of activist settlements were Carl Icahn (20 settlements),
Ramius Capital (14 settlements), and ValueAct (13 settlements).
In the next section, we compare the characteristics of
activist campaigns that result in a settlement, contested
vote, or neither a settlement nor a contested vote. Table 2
reports the mean, median, and standard deviation for all
independent variables in each of these three cases. For the
sake of brevity, we will not discuss these descriptive statistics of Table 2 here, but they will be informative when interpreting the results of the regressions analyses below.
4. The determinants and timing of settlements
4.1. Settlement theory and tested hypotheses
A settlement that averts a contested vote is akin to a
settlement of litigation prior to going to a trial in court.
The literature on the economics of litigation and settlement (see Spier, 2007, Wickelgren, 2013, Daugherty and
Reinganum, 2017 for surveys) provides insights about why
and when cases settle.
In particular, the litigation and settlement literature analyzes three potential outcomes of litigation situations:
such situations might go all the way to a court decision,
might produce a settlement agreement that avoids a court
decision, or might be “dropped” without either a settlement or a court decision. Similarly, we distinguish between
three possible outcomes of activist campaigns and accordingly three groups of campaigns.
The first group comprises all campaigns that result in a
settlement. The second group includes all campaigns that
go all the way to a contested vote. The third group consists
of all the campaigns that have no formalized outcome (e.g.,
result in neither a settlement agreement nor a vote). Thus,
the three outcomes we consider are settlement, contested
vote, and no settlement or contested vote.
The third group of campaigns—those without a formal
resolution that have no settlement or contested vote—is
likely heterogeneous. Some of these campaigns might be
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
ones that the activist drops or aborts without achieving
any of its objectives, say, because the activist concludes
that it does not have sufficient support among shareholders to have a credible threat to win board seats that would
enable the activist to extract a settlement or proceed to a
contested vote. In some of these cases, however, the activist might be able to achieve or expect some change despite the lack of a formal resolution. To begin, the activist might hope that the kick-in-the-pants provided by
the campaign will have a positive effect on future decisions of the incumbent directors and their responsiveness
to shareholder interests. Furthermore, the incumbent directors might, without any formal agreement, take some
action—say, announce an immediate buyback—that could
make the activist sufficiently content to drop its campaign
or, alternatively, could reduce the potential support that
the activist might obtain among other shareholders and
thus force the activist to abandon the campaign.
The making of such immediate concessions by the incumbent directors to avoid a contested vote or a formal
settlement can be regarded as “informal settlement.” However, such informal settlements are more limited in what
they can provide the activist relative to the formal settlements on which we focus. Because the securities laws
would require the incumbent directors to disclose any
commitment that incumbent directors make to undertake
some future action or any commitment that they obtain
from the activist in return for actions they take, any situation with such commitments would be classified as a formal settlement in our data set. Thus, informal settlements
can involve changes that the incumbents take right away
without accompanying commitments by the activist, but
such informal settlements cannot include binding commitments by incumbent directors to take any future action or
appointment of directors favored by the activist accompanied by a standstill or other commitment by the activist.
Although informal settlements and the incumbent steps
they consist of are thus more limited in the forms they can
take, Section 8 presents evidence that campaigns without
a settlement or a contested vote are still followed by an
elevated probability that certain outcomes favored by activists will occur, and this evidence is consistent with the
presence of informal settlements.
4.2. The determinants of reaching a settlement or contested
vote
In examining the determinants of campaign outcomes,
let us begin by discussing what will determine whether a
campaign will result in a formal outcome (e.g., a settlement agreement or a contested vote) or will be ended by
the activist without such a formal outcome of either a settlement or a contested vote.
The literature on the economics of litigation and settlement indicates that cases will end with either a settlement or a court decision only if the plaintiff has sufficiently good odds of winning at trial (Bebchuk, 1984; Spier,
2007; Bebchuk and Klement, 2012). Without such odds, a
rational plaintiff will not choose to bear the costs of pursuing the litigation all the way to a court decision, and
a rational defendant will not agree to a settlement if the
7
plaintiff’s low odds of winning deprive it of a credible
threat to pursue the litigation to a court decision. For activism campaigns, this economic logic implies that a campaign will be more likely to end with either a settlement
or a contested vote when the activist’s odds of winning
seats in a contested vote are higher. Without sufficiently
good odds for winning seats in a contested vote, a rational activist will be discouraged from bearing the costs of
pursuing the proxy fights all the way to a contested vote,
and rational incumbents will be less inclined to agree to a
settlement with an activist that does not have a credible
threat to win seats in a contested vote. Thus, we can put
forward the following hypothesis:
Hypothesis H1. The likelihood of a campaign leading either
to a settlement or a contested vote is positively related to
the odds that the activist would prevail in a contested vote.
4.3. The determinants of the choice between settlement and
a contested vote
We now turn to discuss, for the campaigns that result
in a formal outcome of either a settlement agreement or
a contested vote, what will determine whether there will
be a settlement or a contested vote. The economic literature on settlement decisions indicates that a driving force
for settlement are the efficiency gains that could result—
and could be shared by the two sides—from avoiding the
deadweight costs of trial. The larger these trial costs, the
stronger the incentive to settle (Spier, 2007; Prescott and
Spier, 2016).
Similarly, in the case of an activism campaign, settlement would save the costs that the two sides would have
to bear in the event of going all the way to a contested
vote. Accordingly, the larger these costs, the stronger the
incentive to have a settlement rather than a contested vote.
The costs of a contested vote include the out-of-pocket
costs of running a proxy fight all the way; a dissident’s
costs of running a proxy fight can easily rise to millions of
dollars (Gantchev, 2013).11 However, for the incumbent directors and CEO, the most important source of costs might
well be the expected reputational costs that going all the
way to a contested vote might impose. Resisting an activist
runs the risk of alienating other investors if the defense
is perceived as unnecessary, and attacks on incumbent directors by the activist in a contest can produce substantial reputation costs for these directors, which the directors
might have to bear irrespective of the outcome of the contest (Fos and Tsoutsoura, 2014; Gow et al., 2016). Indeed,
winning a heated proxy fight does not preserve the reputation of incumbent CEOs either. For example, after winning
11
In the Trian Partners versus DuPont proxy fight in 2015, the activist
spent $8 million launching the contest and DuPont spent $15 million to
defend. See USA Today, 05/19/2015, “DuPont spent $15 M to keep activist
investor off board.” In the more recent Trian Partners versus P&G proxy
fight in 2017 the two sides were estimated to have spent at least $60 million. See New York Times, 10/16/2017, “Procter & Gamble’s count shows
how close proxy vote was.” While the dissident might already have incurred part of these costs at the time when the settlement is negotiated,
many settlements specify that the activist is reimbursed part or all of its
contest-related expenses.
8
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
a proxy contest against Trian Partners in May 2018, DuPont
CEO Ellen Kullman resigned only five months later.
Thus, the prospect of avoiding the expected reputational damage of a bruising proxy fight going all the way
provides an incentive to have a settlement rather than a
contested vote. Accordingly, we put forward the following
hypothesis:
Hypothesis H2. The likelihood of an activism campaign
leading to a settlement rather than a contested vote is positively related to incumbents’ reputation concerns.
4.4. Analysis of the determinants of settlements
We test Hypotheses H1 and H2 using multinomial logit
regressions that include three groups of campaigns as parallel outcomes: a) settled campaigns, b) campaigns that
lead to a contested vote, and c) campaigns that lead neither to settlement nor to a contested vote. The relevant
sample includes all activism campaigns (with nonmissing information) launched by hedge funds over the period
from 20 0 0 to 2013.
We include the following explanatory variables listed
and discussed below. The first set of variables, from the
activist’s stake to the target firm’s past performance, serve
as proxies for the activist’s having good odds of winning
a contested vote. The subsequent set of variables, from director age to analyst coverage, serve as proxies for the reputation concerns of the incumbents.12
4.4.1. The activist’s stake
A larger stake for an activist is associated with an increased likelihood of success in a contested vote. The activist’s stake provides it with votes on which it can always count in a proxy contest. In addition, because a large
stake provides the activist with stronger incentives to increase value and can signal the activist’s confidence in the
prospects of value appreciation, having a large stake can
help the activist obtain votes from other shareholders.
4.4.2. Insider ownership
Conversely, when incumbents hold a larger stake in the
firm, their chances of winning a potential proxy fight increases. Accordingly, large insider ownership is associated,
other things equal, with lower odds for a victory by the
activist in a contested vote.
4.4.3. Target firm’s share-class structure
The activist’s chances in a contested vote are likely to
be diminished when the target has a share structure with
multiple classes. In such cases, the shares with superior
voting rights are likely to be disproportionately held by insiders or shareholders affiliated with them and thus improve the incumbents’ chances of prevailing in a contested
vote.
12
Even though untabulated analyses show that 44.4% of settlement
agreements are preceded by the activist formally filing proxy material,
we do not include such filings as determinants of the three groups (settlement, contested proxy vote, neither) because there is no variation in
this variable for those campaigns that reach a contested vote.
4.4.4. The average market reaction to past campaign
launches
The higher the stock price reaction to the initial announcement of activism, the higher the market’s expectation of the benefits that the activism campaign could bring
about and thus the greater the market’s approval of the
activism campaign. Thus, an activist track record of high
initial market reactions to the activist’s initiation of campaigns is likely to reflect the market’s viewing the activist
as capable of bringing about value increases. Such a market view is likely to increase the support that the activist
would be able to obtain from other shareholders in a contested vote and thus improve the activist’s chances in such
a vote.13
4.4.5. Success in past engagements
Similarly, when an activist has a track record of successful past campaigns, the activist is more likely to have good
relationships with, and enjoy a good reputation among, institutional investors. Thus, such a track record can be expected to be associated with improved chances in a contested vote.
4.4.6. Settlements in past engagements
Similarly, when an activist has a track record of obtaining settlements in past engagements, the activism is likely
to have the characteristics that enabled it to have been
viewed by the incumbents in such past engagements as
having had good chances of success in a contested vote.
Thus, such a track record is again likely to be associated,
other things equal, with improved chances in a contested
vote in the current engagement.
4.4.7. Proxy fights in past engagements
To have good chances in a contested vote, the activist
must be open to, and have the necessary know-how and
resources for, running a proxy fight. A track record of running proxy contests in past engagements implies that the
activist is open to running a proxy fight and has experience
in doing so.
4.4.8. Past firm performance
Poor past target firm performance (we measure performance using Tobin’s Q, ROA, and past stock returns) has
negative consequences for investors’ perception of the incumbents. Thus, poor past performance of the target bolsters the odds of activist success in a proxy contest.
4.4.9. Directors older than 70
Old incumbent directors—whether executives of the
firm or outside directors—can be expected to have lower
reputational concerns. Such directors are unlikely to be
present in the market for directorships for a long time. All
else equal, this shorter horizon of older directors thus decreases their reputation concerns.
13
We do not use the stock market reaction to the 13(d) filling in the
considered campaign, but rather such stock price reactions to the initiation of earlier campaigns by the activist, due to endogeneity concerns.
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
4.4.10. Directors’ average number of directorships
Incumbent directors with more other directorships enjoy a higher reputation stock and have more seats to lose
in case their reputation is tarnished.14 Fos and Tsoutsoura
(2014) show that directors who are targets in a contested
vote lose directorships they hold at other firms. Hence,
holding directorships at other firms can be expected to be
associated with higher reputational concerns for the directors.
4.4.11. CEO or board chair up for election
The CEO and the chair of the board (if different from
the CEO) are key decision-makers on the response to activists. When the company has a staggered board, these
individuals are not always up for election in the coming
shareholder meeting. When either of these two important
directors is up for election, success by the activist in a
contested vote in the coming shareholder meeting might
lead to the personal defeat and humiliation of these key
decision-makers on the incumbent team. In such a case,
these decision-makers will face the potential for a direct
personal reputational cost that they would not similarly
face if they were not up for reelection. Therefore, when the
CEO or the board chair are up for election, the expected
reputational costs pushing the key decision-makers toward
settlement are likely to be stronger.
4.4.12. Analyst coverage
Proxy contests that are more publicly visible are likely
to have a more significant effect on the incumbents’ reputation. Vega (2006) shows that the number of analysts following a company is strongly positively related to media
coverage. Therefore, we use this variable as a proxy for the
visibility—and hence potential reputational consequences—
of a contested vote.
In Panel A of Table 3, we report the results of the multinomial models that test Hypothesis H1, which predicts that
the credibility of the activist’s threat is positively related
to the likelihood of an activist campaign leading to a settlement or a contested vote (the first campaign subgroup)
or a proxy contest (second subgroup), relative to neither
of these two outcomes (third subgroup).15 The relative
propensity to likelihood of a settlement (contested vote)
relative to neither settlement nor proxy vote is shown in
the odd-numbered (even-numbered) columns. The coefficients represent the change in the odds ratio that a campaign leads to a settlement (or proxy vote) relative to the
base outcome, associated with a unit change in the independent variable.
Consistent with the hypothesis, we find that a number of proxies for the credibility of the activist’s threat
are related to the probabilities of a campaign leading either to settlement or proxy vote. Specifically, activists with
a better reputation, as measured by the market reaction
14
Jiang et al. (2016) show that directors with higher stock of external
reputation exhibit more reputation concerns.
15
We do not include all the different proxies of the credibility of the
activist’s threat and for incumbents’ reputation concerns in one regression
because of multicollinearity concerns and because of the significant drop
in sample size due to missing values.
9
to past campaign announcements and the number of past
campaign successes, are more likely to obtain a settlement
or to go to a proxy vote. When controlling for these two
proxies, activists with a large number of past campaigns
are less likely to reach a settlement or a contested election, implying that the mere number of past engagements
does not increase the credibility of the activist.16 A track
record of reaching settlements in past campaigns is associated with a higher propensity to settle the focal campaign,
and a track record of running proxy fights in the past is
related to a higher probability of reaching the proxy vote
stage again.
Target firm ownership structure is also an important
determinant of the credibility of the activist’s threat and,
in turn, whether campaigns lead to a settlement or a proxy
vote. For campaigns with above-median activist ownership
relative to campaigns with below-median activist stake
size, the coefficient of 1.85 indicates that the odds ratio for
settlement relative to the base outcome increases by 85%.
If a target firm has multiple share classes, the odds ratio
of settlement decreases by 54%. Furthermore, the extent of
insider ownership is negatively related to the likelihood of
settlement and proxy vote relative to the base state.
Depending on the specification, we find some support
for weaker past target firm performance increasing the
probabilities of settlement and vote. This finding is consistent with weaker performance increasing the credibility of
the activist’s threat, potentially because of the dissatisfaction of other shareholders.
Taken together, these results are consistent with Hypothesis H1 and imply that activist reputation, target firm
ownership structure, and, to a lesser extent, target firm
performance play a key role in determining the credibility
of the activist’s threat, which in turn affects the probabilities of the engagement ending without either a settlement
or a contested vote.
We also examine whether and how proxies for incumbents’ reputation concerns are related to the probabilities
of settlement and contested election vis-à-vis the base outcome. The results are presented in Panel B of Table 3. We
find that boards whose directors and independent directors
hold a larger amount of board positions are less likely to
settle, while boards with a larger number of old (independent) directors are more likely to participate in contested
elections. The latter result is consistent with reputation being less of a concern among older directors. In addition,
we find that the relative likelihood of settlement increases
when the board chair is up for election and when the campaign is more visible due to higher analyst coverage.
Table 4 completes the test of multinomial regressions to test Hypothesis H2, which predicts that the
16
One fund in our sample, GAMCO, has close to 100 campaigns. Even after dropping campaigns with more than 29 past campaigns (which is the
90th percentile), the number of past campaigns remain negatively and
significantly related to the probability that the focal campaign leads to
a settlement. When we exclude the other activist reputation proxies, the
number of past campaigns remains negatively related to the probability
that the current campaign leads to a settlement or proxy vote, albeit insignificantly so. Given the potential multicollinearity of the different activist reputation proxies, we do not include all of them in the same regression.
10
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 3
The determinants of settlement and proxy vote versus other campaign outcomes.
The analyses in this table predict whether a campaign leads to a settlement (columns 1, 3, 5, 7) or a proxy vote (columns 2, 4, 6, 8). For campaigns that
neither lead to a settlement nor to a proxy vote, these dependent variables are equal to zero. The sample includes hedge fund activism campaigns launched
between 20 0 0 and 2013. The sample size differs across models depending on the availability of the independent variables. The models are estimated using
multinomial logit regressions. The coefficient estimates represent relative risk ratios. Standard errors are robust to heteroskedasticity; z-statistics appear in
parentheses. ∗ ∗ ∗ , ∗ ∗ , and ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Panel A. Activist track record and ownership structure
Predicted outcome
Settlement
Proxy vote
Base outcome
Activist track record
Market reaction past campaigns
Ln (# campaigns)
Ln (# successful campaigns)
Settlement
Proxy vote
Settlement
(1)
(2)
1.052∗
(1.94)
0.354∗∗∗
(−8.56)
4.309∗∗∗
(8.58)
1.167∗∗∗
(2.93)
0.353∗∗∗
(−4.15)
4.070∗∗∗
(4.66)
Ln (# settlements)
Ln (# proxy fights)
(3)
(4)
2.624∗∗∗
(9.41)
0.990
(−0.10)
1.214
(1.10)
1.437∗∗
(2.44)
Ownership structure
Multiple share classes
High activist ownership
Insider ownership
Firm performance
ROA
Tobin’s Q
Abnormal returns pre-13D
Control variables
Ln (Market cap)
Observations
Pseudo R-squared
Proxy vote
Campaigns that neither lead to a settlement nor to a proxy vote
(5)
(6)
0.455∗∗
(−2.21)
1.850∗∗∗
(4.58)
0.272∗∗∗
(−2.71)
0.471
(−1.09)
1.144
(0.57)
0.060∗ ∗
(−1.99)
0.591
(−1.32)
0.879∗∗
(−2.26)
0.834
(−1.33)
0.350
(−1.63)
1.048
(0.68)
1.087
(0.59)
0.692
(−1.09)
0.942
(−1.20)
0.833
(−1.49)
0.554
(−1.12)
1.034
(0.66)
0.855
(−1.09)
1.223
(0.54)
0.893∗
(−1.78)
0.846
(−1.33)
0.355∗
(−1.74)
1.035
(0.56)
0.811
(−1.00)
1.208∗∗∗
(4.58)
1773
0.104
1.193∗∗ ∗
(2.67)
1773
0.104
1.124∗∗∗
(3.17)
2488
0.0624
1.203∗∗∗
(3.22)
2488
0.0624
1.215∗∗∗
(4.15)
1822
0.044
1.297∗∗∗
(3.88)
1822
0.044
Panel B. Incumbents’ reputation concerns
Predicted outcome
Settlement
Proxy vote
Base outcome
Incumbents’ reputation concerns
# Directors older than 70
Directors’ avg # directorships
Settlement
Proxy vote
Settlement
Proxy vote
(1)
(2)
1.067
(1.31)
0.714∗∗∗
(−2.75)
1.141∗
(1.84)
0.925
(−0.41)
# Outside directors older than 70
Outside directors’ avg # directorships
(3)
(4)
1.075
(1.38)
0.732∗∗∗
(−2.87)
1.157∗
(1.90)
0.878
(−0.79)
CEO up for election
Board chair up for election
(5)
(6)
0.713
(−0.85)
2.249∗
(1.94)
1.161
(0.37)
0.783
(−0.63)
Ln (Analyst coverage)
Firm performance
ROA
Tobin’s Q
Abnormal returns pre-13D
Control variables
Ln (Market cap)
Observations
Pseudo R-squared
Settlement
Proxy vote
Campaigns that neither lead to a settlement nor to a proxy vote
(7)
(8)
1.483∗∗∗
(4.75)
1.016
(0.12)
0.588∗
(−1.67)
0.922
(−1.63)
0.774∗
(−1.92)
0.498
(−1.33)
0.987
(−0.23)
0.868
(−1.00)
0.575∗
(−1.73)
0.923
(−1.63)
0.776∗
(−1.91)
0.487
(−1.35)
0.989
(−0.20)
0.856
(−1.06)
0.698
(−0.76)
0.951
(−0.61)
0.815
(−0.95)
1.507
(0.48)
1.043
(0.42)
0.974
(−0.36)
0.742
(−0.98)
0.919∗
(−1.68)
0.864
(−1.31)
0.603
(−0.99)
1.020
(0.39)
0.866
(−1.03)
1.297∗∗ ∗
(6.29)
2458
0.024
1.244∗∗∗
(3.47)
2458
0.024
1.290∗∗∗
(6.30)
2454
0.024
1.249∗∗∗
(3.62)
2454
0.024
1.291∗∗∗
(4.69)
946
0.036
1.359∗∗∗
(4.29)
946
0.036
1.046
(0.94)
2488
0.026
1.217∗∗
(2.54)
2488
0.026
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
11
Table 4
The determinants of settlement versus proxy vote.
This table documents the campaign, activist, and target characteristics that predict the occurrence of a settlement agreement versus a proxy vote among
the subset of campaigns that lead either to a settlement or a proxy vote. The sample includes hedge fund activism campaigns launched between 20 0 0 and
2013. The sample size differs across models depending on the availability of the independent variables. The models are estimated using multinomial logit
regressions. The coefficient estimates represent relative risk ratios. Standard errors are robust to heteroskedasticity; z-statistics appear in parentheses. ∗ ∗ ∗ ,
∗∗
, and ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Predicted outcome
Settlement
Base outcome
Proxy vote
(1)
Activist track record
Market reaction past campaigns
Ln (# campaigns)
Ln (# successful campaigns)
(2)
(3)
(4)
(5)
(6)
0.901∗
(−1.85)
1.006
(0.02)
1.059
(0.17)
2.161∗ ∗ ∗
(4.07)
0.689∗ ∗
(−2.23)
Ln (# settlements)
Ln (# proxy fights)
Ownership structure
Multiple share classes
0.965
(−0.05)
1.618∗
(1.86)
4.539
(1.03)
High activist ownership
Insider ownership
Incumbents’ reputation concerns
# Directors older than 70
0.935
(−0.82)
0.772
(−1.20)
Directors’ avg # directorships
# Outside directors older than 70
0.929
(−0.84)
0.834
(−0.97)
Outside directors’ avg # directorships
CEO up for election
0.614
(−0.98)
2.874∗ ∗
(2.07)
Board chair up for election
1.460∗ ∗
(2.51)
Ln (Analyst coverage)
Firm performance
ROA
Tobin’s Q
Abnormal returns pre-13D
Control variables
Ln (Market cap)
Observations
Pseudo R-squared
(7)
1.689
(0.74)
0.839∗ ∗
(−2.10)
0.767
(−1.46)
1.248
(0.37)
0.911
(−1.35)
0.974
(−0.14)
3.443∗
(1.82)
0.863
(−1.64)
1.044
(0.18)
1.181
(0.28)
0.934
(−0.93)
0.892
(−0.61)
1.182
(0.28)
0.933
(−0.96)
0.907
(−0.52)
0.463
(−0.84)
0.912
(−0.75)
0.837
(−0.80)
1.232
(0.36)
0.901
(−1.49)
0.997
(−0.02)
1.013
(0.18)
1773
0.104
0.935
(−1.05)
2488
0.062
0.936
(−0.85)
1822
0.044
1.042
(0.58)
2458
0.024
1.033
(0.47)
2454
0.024
0.950
(−0.62)
946
0.036
0.859∗
(−1.75)
2488
0.026
probability of an activism campaign leading to a settlement vis-à-vis a contested proxy vote increases in incumbents’ reputation concerns. Specifically, Table 4 compares
the likelihood of settlement to that of the base outcome of
a contested proxy vote. We find some support for Hypothesis H2, which predicts that incumbents’ reputation concerns are positively related to the relative likelihood of settlement.
In particular, the odds ratio of settlement vis-à-vis contested proxy vote is 2.9 times as high when the board chair
is up for election, consistent with reputation concerns being particularly strong for the chairs of target firm boards.
When the activism campaign is more visible, the relative
likelihood of settlement also increases, consistent with the
public visibility of a potential proxy fight increasing incumbents’ reputation concerns. While these findings support
Hypothesis H2, we find no evidence that the other proxies for incumbents’ reputation concerns, the number of old
directors, the average number of directors’ outside board
positions, and whether the CEO is up for election are associated with the more settlements versus contested votes.17
17
In addition to proxying for incumbents’ reputation concerns, these
measures might also capture the ability of these directors and boards to
12
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
In Table 4, we also examine whether the proxies for the
credibility of the activist’s threat are related to the relative
likelihood of settlement. We find that higher market reactions to the activist’s past campaigns and the number of
past proxy fights favor contested votes, while the number
of past settlements and above-median activist ownership
favor settlement. We find no consistent pattern for target
firm performance.
4.5. The timing of settlements
Furthermore, when we divide the period between campaign launch and the AGM into three equal-length subperiods, we find that about 47% of settlements fall during the
third subperiod, and about 84% fall during the second and
third equal-length subperiods, with only about 16% of settlements falling into the first equal-length period. We reject the uniform-distribution null hypothesis of one-third
of the settlements falling proportionately in each subperiod with a chi-square test (test statistic = 31.91, significant
at the 1% level).
Having examined the determinants of settlements, we
would like to conclude this section by considering the timing of settlements. The literature on the economics of settlement and litigation has shown that divergent expectations regarding the outcome of trial can be an impediment
to obtaining a settlement (see, e.g., Shavell, 1982). When
the parties have symmetric expectations regarding the expected outcome of trial, there would always be a range of
settlements that would be viewed by both parties as superior to trial. However, divergence of expectations might
sometimes narrow or eliminate the range of settlements
that would be viewed as superior to trial by both parties.
As a result, the discovery process and the arrival of information as trial nears can facilitate an out-of-court settlement.
Similarly, for the case of an activist campaign, divergence of expectations between the two sides about the expected effect of a contested vote can impede the reaching
of a settlement. However, as the two sides get closer to
the company’s shareholder meeting, the information the
two sides get from their proxy solicitors and from their
communications with institutional investors can be expected to provide them with better ability to predict the
expected outcome of a contested vote and to narrow any
widely divergent expectations regarding this expected outcome. Accordingly, with respect to the timing of settlements, we hypothesize that settlements are more likely to
occur closer to the shareholder meeting in which a contested vote would otherwise take place rather than earlier
on.
To test this hypothesis, we gather information about
the timing of settlement agreements within the period between the start of the activist campaign and the company’s first subsequent annual shareholder meeting (AGM)
in which a contested vote could take place. For the 211
settlements that occur within this period, we find that
the length of time between campaign start and the subsequent shareholder meeting has a mean (median) of 188
days (184 days), or about six months. Consistent with settlements being more likely to occur during the second half
of this period rather than the first part, about 70% of settlements take place after the mid-point of the period between campaign launch and AGM, and the ratio of (i) the
time between the settlement and the AGM to (ii) the time
between campaign launch and AGM has a mean (median)
value of 0.38 (0.35).
5. Settlement terms: the focus on board composition
monitor the CEO. For example, directors with a large number of directorships might be too “busy” to monitor properly (Fich and Shivdasani,
2006; Falato, et al., 2014).
18
We are grateful to Oliver Hart for a detailed discussion of the applicability of the incomplete contracting paradigm to our setting and the
consistency of our findings with an incomplete contracting view.
Activist hedge funds are ultimately interested in implementing operational or leadership changes that they believe would substantially increase the value of their shares.
Operational changes include a major transaction, such as a
divestiture of a peripheral division, and sale of the company or a major part of its assets. Leadership changes often
focus on replacing the CEO with another individual who
would be expected to perform better or change the company’s strategic direction.
However, rather than a settlement agreement that requires such specific changes (e.g., that specifies some direct action the board is committed to take), the parties can
make an agreement that focuses on changing some of the
individuals on the board of directors and leaves the possibility of making operational and leadership changes to decisions by the board down the road. For example, in the
Sotheby’s case, activist hedge fund Third Point campaigned
against the CEO and the strategic plan he was pursuing.
The settlement agreement signed in May 2014 provided
board seats for Third Point founder Dan Loeb and two of
his associates but kept Sotheby’s CEO William Ruprecht in
place. Subsequently, in November 2014, Ruprecht stepped
down and Dan Loeb handpicked Tad Smith from Madison
Square Garden for replacement.
In this section, we show that such cases should be
expected, and indeed are most common. We begin with
a theoretical discussion, putting forward three considerations that can be expected, individually or in combination,
to lead parties to focus on board turnover and to avoid
inclusion of direct action terms. We then proceed to look
at the evidence and show that it is consistent with the
theory.
5.1. Theory
5.1.1. Incomplete contracting
The first reason for why we do not expect settlement agreements to contract on operational or leadership
changes relates to the theory of incomplete contracting developed over the years by Oliver Hart and his co-authors,
which is discussed in detail in Hart’s Nobel Prize Lecture
on the subject (Hart, 2017).18 The incomplete contracting paradigm focuses on situations in which contracting
parties have primarily an interest in some subsequent ex
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
post actions or outcomes, but contracting on such ex post
actions is impossible due to the costs or limits on what
courts can or will verify and enforce ex post (see Aghion
and Bolton, 1992 for a well-known application of the insights of incomplete contracting to explain financial contracting choices that focus on corporate influence and control arrangements).
We argue that our setting is one in which directly contracting on the ex post operational changes sought by the
activists is often not feasible. Consider a hedge fund activist that would like to have a target sold. If a wellspecified offer from an outside buyer is not yet available,
the settlement agreement cannot require the acceptance of
a specific sale offer, and requiring the company to be sold
later on would be impractical. To begin, how and whether
a sale should be pursued would depend on contingencies
that would be difficult or impossible for a court to verify
and enforce. When and at what price it would be optimal
to sell would depend on what offers a search for buyers
would generate and what would take place in the market
in the meantime. Furthermore, and relatedly, corporate law
rules preclude directors from making a binding and irreversible commitment to recommend a sale without leaving
a “fiduciary out” that would allow them to avoid a sale if
doing so would in their judgment be best for shareholders.
To be sure, incumbent directors could commit to establishing a board committee to explore a sale. However, as
long as the team of incumbent directors does not change,
the activist could well be reasonably concerned that, not
having been interested in a sale earlier, the incumbents
could explore the option perfunctorily rather than wholeheartedly and a sale would remain unlikely. Thus, to increase the likelihood that the company will effectively pursue a sale, it would be best for the activist to add to the
board directors that the activist believes would be open to
or interested in a value-enhancing sale.
Similarly, consider a case in which the hedge fund activist is not merely interested in firing the current CEO but
also in the quality and fit of the CEO’s successor.19 Such a
successor might not be readily available at the time of the
settlement agreement and, given that courts would have
difficulty verifying the quality and fit of a chosen CEO versus alternatives, it would not be feasible for the activist
to secure the high-quality choice the activist seeks by requiring such a choice in the settlement agreement. Thus,
to increase the likelihood that the company will down the
road choose the successor CEO in the manner sought by
the hedge fund activist, it would be best for the activist to
add to the board directors that the activist believes would
be likely to push for such a CEO choice.
We note that while some changes sought by activists
are difficult or impossible to contract, others are not. In
particular, including the firing of the CEO in the settlement agreement is not precluded by incomplete contracting considerations. However, there are two other considerations that might keep such changes out of settlement
agreements.
19
Keusch (2020) examines activist involvement in the hiring of new
CEOs.
13
5.1.2. Face-saving benefits and reputational concerns
Another reason for keeping operational or leadership
changes out of settlement agreements relates to incumbents’ reputation concerns. Specifically, postponing operational and leadership changes until after the settlement
can provide face-saving benefits to the incumbent CEO and
directors, impose lower reputational costs on them, and
thus increase the parties’ joint surplus from making an
agreement. For example, if the settlement agreement were
to include a provision requiring the departure of a CEO, the
CEO’s being fired would be salient and the CEO would bear
substantial reputational costs. Furthermore, in this case,
the incumbent directors might be viewed as “throwing the
CEO under the bus” and thus would bear reputational costs
as well.
By contrast, if the CEO were to announce her departure several months down the road after the settlement
agreement, the departure can be attributed to changes in
the personal circumstances of the CEO and not be unambiguously viewed as a firing. For example, in the Sotheby’s
case, the hedge fund activist strongly campaigned against
the CEO but accepted a settlement agreement that did not
include the CEO’s removal. Six months after the settlement,
the company announced that the CEO was stepping down
“by mutual agreement,” and the lead independent director lavishly praised the departing CEO, stating that “[m]y
fellow Directors and I salute Bill’s unwavering dedication
and the many significant contributions he has made to
Sotheby’s for more than three decades.”
Similarly, if the incumbent directors accept a settlement
provision requiring the sale of a division that they have
been resisting during the campaign, they might be viewed
as capitulating under pressure to accept a strategy they do
not favor. By contrast, if such a change is made by the new
board several months down the road, it can be presented
as a consequence of discussions on the board, and the resulting development of views, or a change in the company’s circumstances.
Taken together, incumbents’ reputation concerns during activism campaigns represent another reason why settlements are unlikely to contract on CEO turnover or operational changes. In addition, the presence of reputation
concerns also suggests that incumbent directors should be
hesitant to agree to settlement terms that specify the immediate departure of some of them. Allowing incumbent
directors to retire several months or even years following
the settlement is more face-saving to them than their departure being stipulated under activist pressure in the settlement agreement.
5.1.3. Asymmetric information
Thus far we have discussed situations in which the activist is certain that it prefers a specific change. However,
as is emphasized and analyzed by the model of Corum
(2018), there might be situations in which the activist recognizes that the board of directors has private information that would be useful for evaluating whether a given
change would be value enhancing.
Suppose that the board of directors has private information that is useful for assessing whether the incumbent
14
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 5
The content of settlement agreements.
This table outlines the terms specified in settlement agreements between activist hedge funds and their target companies. Panel A tabulates the number of
settlement agreements that specify board changes or direct actions. Panel B provides descriptive statistics on individual settlement terms. The frequencies
represent the number and percentage of settlement agreements that led to a certain outcome. The averages represent the average number of director
additions or departures per settlement. Panel C compares the number of director departures to the number of director additions. ∗ ∗ ∗ , ∗ ∗ , and ∗ denote
statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed).
Panel A. Proxies for board turnover and direct actions
# Settlements
% Settlements
342
33
319
10
352
85.71
8.27
79.95
2.51
88.22
Director turnover
Sale | Merger | Asset sale | CEO departure
Director turnover only
Sale | Merger | Asset sale | CEO departure only
Director turnover & sale | Merger | Asset sale | CEO departure
Panel B. Individual settlement terms
# Settlements
% Settlements
Average
335
207
209
57
150
21
10
2
23
19
21
83.96
51.88
52.38
14.29
37.59
5.26
2.51
0.50
5.76
4.76
5.26
1.79
0.65
0.94
0.22
0.87
–
–
–
–
–
–
Addition of new director(s)
Addition of director(s) affiliated with activist
Addition of director(s) favored by but unaffiliated with activist
Addition of other director(s)
Director departure(s)
CEO departure
Sale or merger of target firm
Sale of a part of the firm’s assets
Formation of strategic transactions committee
Exploration of strategic alternatives w/o committee
Buyback program related announcement
Panel C. Director additions and departures
# Director additions per settlement
0
1
2
3
4
More than 4 additions (Average = 5.8)
# Settlements with 0, 1, 2, 3, 4,
>4 director additions
# Settlements
with director
departures
Average # of director
departures
64
120
127
49
19
20
7
21
53
37
13
19
0.22
0.30
0.59
1.80
2.26
4.50
CEO is still the best match for the target firm or whether
a subsidiary of the firm would be worth more if sold or
spun off. In this case, the best course of action for the
hedge fund activist might be to defer the decision on the
change, have new directors whose judgment the activist
respects join the board and obtain the private information
possessed by the board, and have these directors participate in making a decision down the road regarding the
suitabillity of the CEO or the spin-off.
5.2. Evidence
5.2.1. Board turnover versus direct action
The discussion in Section 5.1 suggests that due to incomplete contracting, reputation concerns of incumbent
directors, and informational asymmetry between incumbents and the activist, settlements commonly will not contract on direct actions such as CEO turnover or the sale
of the firm. Instead, settlements should commonly be expected to contract on the identity of the agents that may
choose among those actions down the road. The intermediate goal of the activist is to introduce into the boardroom individuals that are expected or likely to be open to
Average change in board size
−0.22∗ ∗
0.70∗ ∗ ∗
1.41∗ ∗ ∗
1.20∗ ∗ ∗
1.74∗ ∗ ∗
1.30∗ ∗ ∗
the type of changes that the activist seeks. Following the
theory in Section 5.1, we therefore test the following hypothesis regarding the frequency of settlements that specify board turnover and direct actions.
Hypothesis H3. The proportion of settlements specifying director turnover is higher than the proportion of settlements specifying direct actions such as CEO turnover or
strategic transactions.
Panel A of Table 5 reports the frequency of board
turnover and direct actions specified in 399 settlement
agreements reached in campaigns that were launched between 20 0 0 and 2013. Settlements, on average, are significantly more likely to stipulate board turnover (85.71% of
settlements) than direct actions such as the sale or merger
of the firm or of parts of the firm or CEO turnover (8.27%
of settlements). This difference is statistically significant at
the 0.1% level. Furthermore, the proportion of settlements
that only specify board turnover and no actions (79.95%) is
also significantly larger than the proportion of settlements
that only specify actions (2.51%).
Panel B of Table 5 shows that the actions that settlements do stipulate include CEO departure (5.3%), sale or
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
15
merger of the firm (2.5%), and sale of a part of the firm’s
assets (0.5%). Other settlement terms that are more difficult to classify as actions include the formation of strategic
transaction committees (5.76%), the “exploration of strategic alternatives” (4.76%), and the announcement of new
or increased buyback programs (5.26%).20 Taken together,
the evidence presented in Table 5 indicates that, consistent with Hypothesis H3, activists use settlement negotiations to contract for board changes rather than direct
actions.
director departures (relative to director additions) specified
in the settlement is not driven by an increase of the optimal board size of these firms.
While the above evidence indicates that reputational
considerations are present in the settlements sample, there
likely exists some variation in reputation concerns among
incumbent directors and CEOs. Specifically, reputational
concerns should be weaker if directors and CEOs have already reached customary retirement age and thus do not
expect to participate in the labor market much longer.
Therefore, we test the following hypothesis:
5.2.2. Face-saving benefits and reputational concerns
The
discussion
of
incomplete
contracting
in
Section 5.1 indicates that the nature of some outcomes desired by activists, such as selling the firm down the road to
a buyer yet to be identified, is not practically contractible.
However, some other outcomes desired by activists, such
as replacement of the CEO, can be contracted upon. We
therefore now turn to examine whether the two other
factors discussed in Section 5.1, face-saving benefits and
informational asymmetry, contribute to the infrequent
incorporation of direct actions in settlement terms.
Beginning with the role of face-saving considerations,
we begin by observing patterns suggesting that incumbent
directors are sensitive to face-saving considerations. Panel
B of Table 5 shows that 84% of settlements result in the appointment of new directors, while director departures only
occur in 37.6% of cases. Similarly, while the average number of new directors added in a settlement is 1.79, the average number of departures is only 0.87. This difference is
statistically significant at the 0.1% level.
Fig. 1 plots the number of director additions and director departures in the 11 years around settlement agreements and provides additional evidence that the rate of
director addition is significantly higher than the rate of
director departure in the year of settlement. In addition, this figure shows that while the addition rate reverts back to normal levels in the year following the settlement, the departure rate remains high in the ensuing
years, which suggests that many incumbent directors are
allowed to depart quietly in the years following the settlement rather than being publicly ousted via the settlement
agreement.
Panel C of Table 5 shows that for settlements that specify one, two, three, four, or more director additions, the
average number of director departures is consistently below the number of additions. Consistent with the patterns
in Fig. 1, this leads to statistically significant temporary
increases in the board size of target firms. In unreported
analyses, we find consistent evidence showing that target
firm board size increases in the year of the settlement and
then gradually reverts back to normal levels in the ensuing years. These findings suggest that the low rate of
Hypothesis H4. The probability of a settlement specifying
director departure or CEO departure is negatively related to
the strength of incumbent directors’ and CEOs’ reputation
concerns.
20
The creation of a strategic transaction committee and the claim to
“explore strategic alternatives” do not represent actions, as they may or
may not result in a strategic transaction. Similarly, announcements about
buyback programs are difficult to classify because firms have significant
discretion about whether and when to execute and complete buyback
programs.
Panel A of Table 6 tests this hypothesis. The dependent
variables identify settlement terms that are most costly to
incumbents’ reputation, namely CEO departure and director departure. The independent variables are proxies for directors’ and CEOs’ reputation concerns related to whether
they have reached customary retirement age. We find that
the higher the number of septuagenarian directors or outside directors on the board, the greater the probability
of settlement terms specifying director departure and the
higher the number of director departures. For example, for
each director on the board who is older than 70 years, the
likelihood of director departure increases by 8.3 percentage points. We also find that if the CEO is close to retirement (above 62 years of age), the probability of settlement
terms specifying CEO departure increases by 7.7 percentage points, an effect that is economically large though not
statistically significant.21
These results are consistent with the parties being more
likely to incorporate departures directly in the settlement
terms when reputational concerns, and thus the facesaving benefits from delaying departures, are greater.22
Thus, the results are consistent with Hypothesis H4. However, we acknowledge that these results are also subject to
the alternative explanation that the identified positive relation between director/CEO age and immediate (i.e., stipulated in settlement) director/CEO departures arises because
the presence of old directors/CEOs implies stronger entrenchment that the activist wishes to break up right away
in the settlement rather than later. These two different explanations are difficult to disentangle.
5.2.3. Asymmetric information
Finally, we turn to examine whether the informational
asymmetry between incumbents and activists also contributes to the infrequent incorporation of direct actions
21
The CEO age cutoff is similar to Kaplan and Minton (2012) and Jenter
and Lewellen (2015).
22
We do find evidence that other variables that could be regarded as
proxies for incumbents’ reputation concerns, such as the aggregate number of incumbents’ directorships and analyst coverage, are not associated
with settlement terms. Also, whereas a board with its chair up for reelection is more likely to enter a settlement with the activist (Table 4), having
the chair up for reelection is not associated with the nature of settlement
terms.
16
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Fig. 1. Director additions and departures around settlement agreements.
This figure plots the average number of new directors joining the board (top) and incumbent directors leaving the board (bottom) around the year of
a settlement agreement and around a placebo year for matched control firms. The sample of settlements consists of all settlement agreements reached
between hedge fund activists and target companies during campaigns that were launched between 20 0 0 and 2013. We select untargeted control firms by
propensity score matching on the year that the settlement agreement is reached. See Section 6.1. Year 0 on the x-axis represents the year during which
the settlement agreement was reached or the placebo year.
among settlement terms. As is suggested by the discussion in Section 5.1.3 and the model of Corum (2018), informational asymmetry can prevent the activist from fully
understanding whether certain outcomes would indeed be
value enhancing and thus to prefer that a decision regarding them be deferred. Below we investigate whether
the infrequent incorporation of direct actions in settlement terms is at least partly due to such informational
asymmetry.
To this end, we exploit the fact that information asymmetry between incumbents and the activist likely differs
in the cross-section of activism campaigns. Given this
variation, the informational asymmetry
suggests the following hypothesis:
explanation
Hypothesis H5. The probability of settlements specifying direct actions is negatively related to the extent of information asymmetry between incumbents and the activist.
We follow prior literature (see, e.g., Minton and
Schrand, 1999, and Zhang, 2006) in using cash flow volatility as a proxy for informational asymmetry. In Panel B of
Table 6, the independent variable of interest, serving as
a proxy for information asymmetry between incumbents
and the activist, is the firm’s cash flow volatility over the
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
17
Table 6
Cross-sectional variation in settlement terms.
This table examines cross-sectional differences in settlement terms. Panel A examines whether incumbents’ reputation concerns are related to whether
director departures and CEO departures are stipulated in the settlement. Panel B tests the relation between information asymmetry between firm insiders
and outsiders and the stipulation of actions such as the sale or merger of the firm, the sale of part of the firm’s assets, and CEO turnover in the settlement
agreement. The sample includes settlements for activism campaigns that are launched between 20 0 0 and 2013, but the number of observations differs
with the availability of data on the independent variables. All models are estimated using OLS or linear probability models. Standard errors are robust to
heteroskedasticity; t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , and ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All
variables are defined in Appendix A.3.
Panel A. Incumbents’ reputation concerns and settlement terms
Dependent variable
# Directors older than 70
Directors’ average # directorships
Director
departure
# Director
departures
Director
departure
# Director
departures
Director
departure
# Director
departures
CEO
departure
(1)
(2)
(3)
(4)
(5)
(6)
(7)
0.086∗ ∗ ∗
(3.32)
−0.023
0.179∗ ∗
(2.28)
−0.188
(−0.35)
(−1.10)
−0.005
(−0.10)
0.020
(0.14)
−0.034
(−1.40)
∗∗∗
0.083
(3.47)
−0.007
(−0.09)
(8)
∗∗
0.170
(2.42)
−0.196
(−1.09)
# Outside directors older than 70
Outside directors’ avg #
directorships
Ln (Analyst coverage)
CEO age > 62
ROA
Tobin’s Q
Abnormal returns pre-13D
Ln (Market cap)
Industry & year FE
Observations
R-squared
Adjusted R-squared
−0.269
(−1.33)
0.013
(0.70)
−0.010
(−0.15)
−0.008
(−0.42)
Yes
335
0.170
−0.001
CEO
departure
−1.012
(−1.40)
−0.035
(−0.90)
−0.004
(−0.02)
0.015
(0.27)
Yes
335
0.239
0.082
−0.260
(−1.28)
0.012
(0.67)
−0.013
(−0.19)
−0.007
(−0.36)
Yes
335
0.169
−0.002
−0.996
(−1.38)
−0.036
(−0.90)
−0.008
(−0.05)
0.013
(0.24)
Yes
335
0.239
0.082
−0.187
(−0.93)
0.012
(0.74)
0.014
(0.22)
−0.010
(−0.36)
Yes
339
0.139
−0.032
−0.810
(−1.14)
−0.033
(−0.96)
0.105
(0.64)
−0.026
(−0.33)
Yes
339
0.218
0.063
0.077
(1.28)
−0.373∗ ∗
(−2.36)
0.015
(1.48)
0.009
(0.33)
0.010
(1.26)
Yes
300
0.279
0.127
−0.158
(−1.34)
0.015
(1.35)
−0.002
(−0.09)
0.024∗ ∗
(1.98)
Yes
339
0.252
0.104
Panel B. Information asymmetry and settlement terms
Dependent variable
Prob (Sale | Merger | Asset sale | CEO departure)
(1)
Cash flow volatility
−0.00004
(−2.41)
(2)
∗∗
−0.00004
(−2.07)
Ln (Market cap)
(3)
∗∗
−0.00006
(−1.93)
0.00859
(0.93)
(4)
∗
−0.00006
(−1.79)
0.00934
(0.88)
ROA
Tobin’s Q
Abnormal returns pre-13D
Industry & year FE
Observations
R-squared
Adjusted R-squared
No
310
0.003
0.000
Yes
290
0.255
0.117
No
310
0.006
−0.001
past five years. The dependent variable is an indicator for
whether the settlement terms include direct actions whose
value consequences can be better assessed using private
information possessed by the board, such as the sale or
merger of the firm, the sale of parts of the firm’s assets,
and CEO departure. Across several specifications, we find a
negative association between our proxy for informational
asymmetry and the likelihood that settlement terms specify such direct actions. These finding are consistent with
Hypothesis H5 and thus the predictions suggested by the
Yes
290
0.257
0.117
(5)
∗
−0.00004
(−2.50)
(6)
∗∗
−0.05306
(−0.35)
0.02325∗
(1.67)
−0.00517
(−0.19)
No
306
0.027
0.014
−0.00004
(−1.85)
(7)
∗
−0.13359
(−0.88)
0.02278∗
(1.77)
−0.00342
(−0.12)
Yes
286
0.282
0.136
(8)
∗
−0.00006
(−1.94)
0.00670
(0.64)
−0.07566
(−0.47)
0.02170
(1.57)
−0.00417
(−0.15)
No
306
0.028
0.012
−0.00005
(−1.65)
0.00904
(0.79)
−0.16228
(−1.04)
0.02092
(1.62)
−0.00196
(−0.07)
Yes
286
0.284
0.135
informational asymmetry factor and the model of Corum
(2018).
6. Effects on board composition
In the preceding section, we argued that settlement
terms are likely to focus commonly on director turnover
rather than the type of leadership and operational changes
that activists ultimately seek to obtain to enhance share
value, and we provided evidence consistent with this
18
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
hypothesis. In this section, we investigate the board composition changes brought about by settlements. Among
other things, we analyze the extent to which the board
turnover produced by settlements is abnormal; the type of
directors that settlements introduce; the post-settlement
board turnover changes that follow to reverse the temporary increase in board size that settlements often produce;
and whether settlements lead to an improvement in board
monitoring capabilities as evidenced by director characteristics and shareholder votes in director elections.
6.1. Methodology
To verify that the extent of board turnover specified
in the settlement agreement is indeed abnormal, and to
examine whether settlement agreements are followed by
changes in board composition, CEO turnover, and operational changes, we need to find control firms with similar
ex ante characteristics as settlement firms prior to the settlement. In addition, we want to control for any potential
relation between the initial activist intervention and board
turnover, as this relation is distinct from the effect of the
settlement. Finally, we also want to filter out the relation
between contested proxy votes and board turnover. Thus,
our analyses of the consequences of settlement include six
groups, three treatment groups and three matched control
groups, as described below.
The first treatment group is the subsample of campaigns that leads to a settlement and the year of treatment is the year in which the settlement agreement is
reached. The corresponding control group consists of untargeted firms and is matched based on the year of the settlement. This implies that treated and control firms have
similar observable characteristics during the years leading
up to the settlement year. The second treatment group
consists of the subsample of campaigns that neither experience a settlement nor a contested proxy vote. The treatment year is the year of the initial intervention, and untargeted control firms are matched based on that year. The
third treatment group includes campaigns that lead to a
contested vote. The treatment year is the year in which the
proxy fight is announced, and we find a third control group
of untargeted firms based on that year.
We use three propensity score matching iterations to
find control firms for the three treatment groups. The variables that we include in the propensity score matching regressions are Ln(Market Cap)t − 1 (the natural logarithm of
market capitalization lagged by one year), ROAt − 1 , ROAt − 3 ,
Tobin’s Qt − 1, and Tobin’s Qt − 3 , based on the earlier findings that small and underperforming firms are more likely
to be targeted (Brav et al., 2008) while, among the population of target firms, large and underperforming firms are
more likely to settle or be subject to a contested vote (see
Table 3). Then we match treated and control firms on the
resulting propensity score. For each firm in the three control groups, we require the placebo year to be the same
year as the treatment year, and we require the matched
firm to be in the same industry by the two-digit Standard Industrial Classification (SIC) classification to control
for unobserved industry trends and macroeconomic conditions.
For each treated firm, we select at most five matched
control firms (without replacement), but we delete
treatment-control pairs that belong to the worst 5% of
matches with respect to the absolute distance between the
two propensity scores. Appendix A.4 shows that covariate
balance is satisfied in all 21 mean comparison tests and
pre-event parallel trends tests in ROA and Tobin’s Q between treated and control firms. To build the final data set,
we collect data on dependent and independent variables
for the 11 years around the treatment year for treated
firms, or the matched placebo year for control firms.
For each outcome variable, we present three differencein-differences regression specifications. We cluster standard errors by firm across all specifications. The first
estimates the difference in the outcome variable from preto post- settlement year relative to the difference in the
outcome variable from pre- to post- placebo year.
Yi,t =
αi + αt + β · d[s + k]i,t
+ γ · Set t lement · d[s + k]i,t + εi,t ,
(1)
where α i and α t represent firm and year fixed effects,
respectively. d[s + k] with –5 ≤ k ≤ 5 is a set of indicator
variables for the year relative to the settlement or placebo
year. Year s−1 is the omitted base year. Thus, d[s + k]
capture the difference in the level of the dependent variable among control firms between years s-1 and s + k. In
this model, the level effect associated with the Settlement
firms is subsumed by the firm fixed effects. The dummies
Settlement · d[s + k] capture the change from s−1 to s + k
in the difference in the dependent variable between firms
that settle and control firms.
The approach summarized in Eq. (1) only relies on campaigns that were settled and matched control firms and
therefore does not control for the potential association between the initial activist intervention and the dependent
variable. To mitigate concerns that such association confounds the results, we use a second regression specification in which we include the second treatment group (i.e.,
target firms that neither reach a settlement nor a proxy
vote and their matched control firms). These firms were
matched on the year of initial intervention. To further control for the potential confounding effect of the initial intervention among settling firms and their matched control
firms, we identify the year of the initial intervention for
firms that reach a settlement and assign that year to their
matched control firms as a placebo intervention year. Then
we estimate Eq. (2).
Yi,t =
αi + αt + δ · Set t lementi + β · d[s + k]i,t
+ γ · Set t lement · d[s + k]i,t
+ ζ · d[t + k]i,t + η · Activism · d[t + k]i,t + εi,t , (2)
where d[t + k] with –5 ≤ k ≤ 5 is a set of indicator variables
for the year relative to the initial intervention year or relative to the placebo year. As before, year t−1 is the omitted
base year. The coefficient of Settlement captures the difference in the dependent variable between target firms that
settle and matched control firms in year s−1. The level effect of firms treated by Activism is subsumed by the firm
fixed effects. The dummies Activism · d[t + k] capture the
change from t−1 to t + k in the difference in the dependent
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
variable between target firms and control firms. The advantages of this matching approach are that we can control
for any association between the initial intervention and the
outcome variables such that the difference-in-differences
estimates around the settlement year capture the effect of
settlement over and above the effect of the initial intervention. In the final model Eq. (3) we further add firms that
experience a voted contest, as well as their matched control firms, to be able to compare the changes in the dependent variables around settlement to the changes around
voted contests.
Yi,t =
αi + αt + δ · Set t lementi + β · d[s + k]i,t
+ γ · Set t lement · d[s + k]i,t
+ ζ · d[t + k]i,t + η · Activismi · d[t + k]i,t
+ θ · V ot edCont esti + λ · d[c + k]i,t
+ μ · V ot edCont est · d[c + k]i,t + εi,t ,
(3)
where d[c + k] with –5 ≤ k ≤ 5 is a set of indicator variables
for the year relative to the announcement year of a proxy
fight that went to a vote or relative to the placebo year.
Year c−1 is the omitted base year. The coefficient of VotedContest captures the difference in the dependent variable between target firms that experience a proxy fight
and matched control firms in year c−1. The dummies VotedContest · d[c + k] capture the change from c−1 to c + k
in the difference in the dependent variable between firms
that experience a contested vote and control firms.
6.2. Abnormal board turnover
Table 5, discussed above, shows that 84% of settlements
stipulate director additions and 37.6% director departures.
Fig. 1, which plots the annual number of director additions
and departures for settlement firms and matched control
firms, provides a first indication that the amount of board
turnover specified in these settlement agreements is indeed abnormal. We observe no difference between treatment and control firms before the settlement but a peak in
both the addition and departure of directors among treated
firms in the year of the settlements. The average abnormal
numbers of director additions and departures (in excess of
control firms) are 1.43 and 0.85 during the year of the settlement.
This asymmetry in the addition versus departure implies that the boards of settlement firms expand around
the year of settlement. In untabulated analyses we find
that the average board size increases from about 8.3 directors prior to the settlement to a peak of 9.1 the year after
the settlement is reached. Following the year of the settlement or placebo, treated and control firms show similar trends in board additions, but settlement firms experience significantly higher board departures than control
firms so that treated firms’ board size gradually adjusts to
pre-settlement levels. These patterns are consistent with
incumbent directors being allowed to “retire” from target
firm boards in a face-saving way in the years following settlement agreements.
Table 7 maps the pattern shown in Fig. 1 into a linear
regression framework. In this table, the dependent variables are Number of director additions and number of direc-
19
tor departures. The observations are recorded at the firmyear level for all target and control firms covered in the
Board Magazine Directors Database universe between 20 0 0
and 2013. The coefficients on Settlement d[s + k] represent
the difference-in-differences estimates and clearly confirm
the patterns in Fig. 1: both additions and departures spike
in the year of the settlement relative to the year before,
with an average of 1.3–1.4 new directors and 0.74–0.84 departing ones. These estimates represent an “abnormal” increase in board turnover relative to matched control firm
years and are statistically significant. We continue to find
a strong, abnormal increase in board turnover among settlement firms when we use year s−2 as the benchmark
period instead of year s−1 as reported at the bottom of
Table 7. Also consistent with Fig. 1, firms that eventually
settle do not experience a higher rate of board turnover
prior to the settlement compared to control firms. We find
some director additions in the year following settlement
and continue to observe abnormally high numbers of director departures until two years after settlement. This result further supports the claim that settlement agreements
allow incumbent directors to exit in a face-saving way in
the months or even years following the settlement.
The regression framework allows us to explore related
effects. In columns 2, 3, 5, and 6, we include indicators
capturing abnormal changes in board turnover around the
year of the initial activist intervention, which typically
takes place one year prior to the year of settlement if there
is one. Table 7 indicates that the initial intervention per
se is associated with abnormal increases in board turnover
but again in an asymmetric way: while we observe abnormal increases in new director additions in the year of campaign launch, t + 1, and t + 2, we find evidence of abnormal
director departures only in year t + 2. By controlling for
the time window surrounding the initial activist intervention, we effectively disaggregate total board turnover following settlements into turnover associated with the initial activist intervention and the incremental turnover that
is directly related to the settlement. Importantly, estimates
of the increase in board turnover around the year of the
settlement remain statistically and economically significant
after controlling for the changes that are associated with
the initial activist engagement.
Finally, columns 3 and 6 show evidence for abnormal
board turnover around contested proxy votes. We find 1.4
new director additions above the normal level in the first
year following the proxy fight announcement and between
0.3 and 1.1 abnormal director departures in years c, c + 1,
and c + 2. These magnitudes suggest that settlements potentially provide a less costly avenue to achieve a similar
level of board turnover as proxy fights do. Unlike settlements and initial activism campaign launches, proxy fights
are preceded by an abnormal amount of director departures three years prior to contest announcement.
6.3. Activist directors
Having established the magnitude of board turnover
following activist engagements in general, and settlement
agreements in particular, we now turn to the characteristics of the directors that are added and removed.
20
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 7
Board turnover around settlement agreements.
This table presents difference-in-differences analyses of board turnover around settlement agreements. Columns 1–3 examine the number of new directors
on the board and models 4–6 the number of director departures. For target firms that settle (“Settled campaigns”; all columns), we select untargeted control
firms by propensity score matching on the year that the settlement agreement is reached. For target firms that experience a proxy contest vote (columns
3 and 6), we select untargeted control firms by propensity score matching on the proxy contest announcement year. For target firms that experience
neither settlement nor proxy contest vote (“Other campaigns”; columns 2, 3, 5, 6), we select untargeted control firms by propensity score matching on
the year of the initial activism announcement that, in most cases, is the year of the first 13D filing. The tabulated coefficients measure how the difference
between treatment firm board turnover and control firm board turnover has changed from the year before treatment/placebo (the omitted base year) to
year s ± k, t ± k, or c ± k. Treatment can either be the settlement year (for settled campaigns) or the proxy fight announcement year (for those proxy fights
that went to a vote) or the year of the initial activism announcement (for “Other campaigns” and “All campaigns”). Columns 2, 3, 5, and 6 (columns 3
and 6) include dummies for the Settlement (Voted proxy) main effect. All columns include a set of 11 d[s ± k] dummies, one for each year relative to the
settlement year (for treatment firms) and matched placebo year (for control firms). In addition, columns 2, 3, 5, 6 (columns 3 and 6) include a set of
11 d[t ± k] (d[c ± k]) dummies, one for each year relative to the initial campaign launch year (proxy fight announcement year) and matched placebo year.
At the bottom of the table, p-values of partial F-tests indicate the statistical significance of difference-in-differences estimations of board turnover around
settlement agreements using year t-2 as the base year. Coefficients and standard errors clustered by firm are estimated using OLS. t-statistics appear in
parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent variable
Sample
Settlement · d[s-5]
Settlement · d[s-4]
Settlement · d[s-3]
Settlement · d[s-2]
Settlement · d[s]
Settlement · d[s + 1]
Settlement · d[s + 2]
Settlement · d[s + 3]
Settlement · d[s + 4]
Settlement · d[s + 5]
13D · d[t-5]
13D · d[t-4]
13D · d[t-3]
13D · d[t-2]
13D · d[t]
13D · d[t + 1]
13D · d[t + 2]
13D · d[t + 3]
13D · d[t + 4]
13D · d[t + 5]
Voted contest · d[c-5]
Voted contest · d[c-4]
Voted contest · d[c-3]
Voted contest · d[c-2]
# Director additions
Settled campaigns
& control firms
Settled campaigns,
other campaigns &
control firms
# Director departures
All campaigns &
control firms
Settled campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
(3)
(4)
(5)
−0.207∗ ∗
(−1.99)
−0.003
(−0.02)
−0.158
(−1.55)
−0.116
(−0.95)
1.399∗ ∗ ∗
(9.51)
0.328∗ ∗
(2.48)
0.147
(1.00)
0.008
(0.06)
0.055
(0.26)
0.000
(0.00)
−0.110
(−1.06)
0.082
(0.70)
−0.102
(−1.01)
−0.098
(−0.82)
1.299∗ ∗ ∗
(9.00)
0.237∗
(1.87)
0.119
(0.84)
0.023
(0.16)
0.064
(0.32)
−0.008
(−0.05)
−0.095∗
(−1.73)
−0.054
(−1.04)
−0.065
(−1.38)
−0.015
(−0.31)
0.101∗ ∗
(2.01)
0.171∗ ∗ ∗
(3.23)
0.125∗ ∗
(2.23)
−0.004
(−0.06)
−0.009
(−0.15)
0.038
(0.61)
−0.135
(−1.34)
0.057
(0.52)
−0.124
(−1.29)
−0.117
(−1.02)
1.270∗ ∗ ∗
(9.02)
0.215∗
(1.74)
0.100
(0.73)
−0.002
(−0.02)
0.039
(0.19)
−0.030
(−0.17)
−0.083
(−1.54)
−0.037
(−0.73)
−0.055
(−1.18)
−0.003
(−0.06)
0.112∗ ∗
(2.29)
0.201∗ ∗ ∗
(3.86)
0.114∗ ∗
(2.08)
0.006
(0.12)
0.008
(0.13)
0.054
(0.88)
0.435∗ ∗
(2.20)
−0.125
(−0.90)
0.005
(0.03)
0.120
(0.78)
−0.087
(−0.81)
−0.110
(−1.05)
−0.033
(−0.32)
−0.071
(−0.64)
0.840∗ ∗ ∗
(6.49)
0.375∗ ∗ ∗
(2.78)
0.493∗ ∗ ∗
(2.90)
0.185
(1.29)
0.191
(0.82)
0.082
(0.44)
−0.058
(−0.55)
−0.087
(−0.85)
−0.022
(−0.22)
−0.090
(−0.85)
0.742∗ ∗ ∗
(5.88)
0.256∗
(1.94)
0.416∗ ∗
(2.58)
0.165
(1.20)
0.169
(0.76)
0.043
(0.24)
−0.068
(−1.28)
−0.047
(−0.93)
−0.050
(−1.05)
−0.040
(−0.84)
0.078
(1.63)
0.068
(1.32)
0.150∗ ∗
(2.36)
−0.038
(−0.68)
−0.012
(−0.18)
−0.028
(−0.46)
All campaigns &
control firms
(6)
−0.063
(−0.62)
−0.092
(−0.94)
−0.021
(−0.21)
−0.087
(−0.86)
0.743∗ ∗ ∗
(6.04)
0.253∗ ∗
(1.97)
0.414∗ ∗ ∗
(2.63)
0.160
(1.17)
0.160
(0.73)
0.035
(0.20)
−0.065
(−1.23)
−0.035
(−0.71)
−0.046
(−0.97)
−0.043
(−0.93)
0.070
(1.49)
0.076
(1.51)
0.149∗ ∗
(2.39)
−0.042
(−0.74)
0.001
(0.02)
−0.018
(−0.30)
0.286
(1.41)
−0.025
(−0.17)
0.307∗
(1.74)
0.101
(0.78)
(continued on next page)
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
21
Table 7 (continued)
Dependent variable
Sample
# Director additions
# Director departures
Settled campaigns
& control firms
Settled campaigns,
other campaigns &
control firms
All campaigns &
control firms
Settled campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
(3)
(4)
(5)
Yes
80,864
0.183
0.191
(1.14)
1.421∗ ∗ ∗
(5.85)
0.226
(1.31)
−0.073
(−0.49)
0.254
(1.15)
−0.043
(−0.19)
Yes
85,867
0.184
Yes
12,922
0.157
Yes
80,864
0.197
0.339∗ ∗
(2.18)
1.076∗ ∗ ∗
(4.44)
0.807∗ ∗ ∗
(3.03)
0.192
(1.09)
0.230
(1.02)
0.008
(0.04)
Yes
85,867
0.197
0.000
0.011
0.127
0.407
0.448
0.627
0.000
0.012
0.126
0.431
0.467
0.638
0.000
0.002
0.000
0.114
0.267
0.402
0.000
0.016
0.001
0.105
0.248
0.449
0.000
0.018
0.001
0.117
0.270
0.483
Voted contest · d[c]
Voted contest · d[c + 1]
Voted contest · d[c + 2]
Voted contest · d[c + 3]
Voted contest · d[c + 4]
Voted contest · d[c + 5]
Firm & year FE
Yes
Observations
12,922
Adjusted R-squared
0.150
p-values for partial F-tests comparing coefficients
S · d[s] vs. S · d[s-2]
0.000
S · d[s + 1] vs. S · d[s-2]
0.001
0.071
S · d[s + 2] vs. S · d[s-2]
S · d[s + 3] vs. S · d[s-2]
0.407
S · d[s + 4] vs. S · d[s-2]
0.445
S · d[s + 5] vs. S · d[s-2]
0.546
We provide evidence that activism campaigns are indeed
associated with the addition of new directors and settlements are associated with the appointment of more proactivist directors than campaigns not leading to a settlement. To examine whether activists indeed influence the
appointment of directors, we proxy for the relationship between activists and directors by separating newly added
directors into three mutually exclusive categories: (i) directors that are affiliated with the activist fund; (ii) directors
that are not affiliated with but (publicly) favored by the
activist; and (iii) other directors.
The status of “favored but not affiliated” is based
on manual classification from proxy materials and news
searches. This dummy variable is coded as one for directors
who are not employees of the fund but for whom there is
evidence of any of the following: a) they are described in
media accounts as being selected or desired by the fund;
b) they appear, on the fund’s proxy material, as directors
that the fund would like to choose for replacements in
case some directors are unable to continue serving on the
board; c) they have to resign if the fund falls below a certain ownership threshold; or d) they have previously been
appointed by the fund during other campaigns.
Panel B of Table 5 already provides some indication that
hedge fund activists are indeed involved in the selection
of the new directors who join target firm boards following
settlement negotiations. In 51.9% of the settlements, directors affiliated with the activist fund join the board, and in
52.4% of cases the added directors are favored by, though
not affiliated with, the activists. The average number of
activist-affiliated (activist-desired) directors added is 0.65
(0.94).
We next examine the activist’s influence on the choice
of new directors more thoroughly. Table 8 shows the like-
All campaigns &
control firms
(6)
lihood that a new director who is added to the board following the initial intervention, settlement or proxy contest is activist-affiliated, activist favored but unaffiliated,
or neither of these two. We include new directors who
are added within five years following the initial intervention, up to 2013, which is the year in which we lose access to the Board Magazine data base. Activists are unlikely
to place directors on the boards of untargeted companies,
so Table 8 only includes target firm-years post intervention and no matched control firms. We treat any of the
years three, four, and five following the initial intervention
as the omitted base years if they do not overlap with the
three-year periods following settlements or proxy fight announcements.
The results provide strong support for activists’ role in
the selection of new directors. In the year of settlement,
the likelihood that a new director is affiliated with the
activist is between 18.6 and 24 percentage points higher
than in years t + 3 through t + 5 following the initial intervention. In addition, the probability of new directors being
unaffiliated with, but desired by, the activist increases by
19 and 30 percentage points, while the likelihood that a
new director is not desired by the activist is between 44
and 50 percentage points lower. We find some weak evidence that the number of activist-affiliated new directors
is abnormally low in year s + 2 following the settlement,
likely driven by the observation in Table 7 that by s + 2, the
settlement has no effect anymore on the number of new
board additions. We find that the abnormally high number of newly added unaffiliated, but desired, directors and
other directors continues until year s + 1 but with lower
magnitude.
These results hold after controlling for dummies that
capture the type of director additions following the
22
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 8
Activist-favored directors following settlement agreements.
This table shows the annual probability of activist-affiliated (column 1), activist favored but unaffiliated (column 2), and other directors (column 3) being
added to the boards of activism target companies following settlement agreements. Since hedge fund activists unlikely place directors on the boards of
untargeted companies, this analysis excludes control firms and targets firm-years prior to the intervention. The time period identifiers measure the annual
level of the dependent variable relative to the average level among activism targets in years three to five following the activist intervention. Coefficients and
standard errors clustered by firm are estimated using linear probability models. t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at
the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent variable
Sample
Settlement · d[s]
Settlement · d[s + 1]
Settlement · d[s + 2]
Prob (Favored and affiliated director)
Prob (Other new director)
Settled
campaigns
Settled
campaigns
& other
campaigns
All
campaigns
Settled
campaigns
Settled
campaigns
& other
campaigns
All
campaigns
Settled
campaigns
Settled
campaigns
& other
campaigns
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
∗∗∗
∗∗∗
0.240
(3.10)
0.076
(1.00)
−0.023
(−0.31)
0.186
(6.11)
0.037
(0.84)
−0.070∗ ∗
(−2.52)
0.095∗ ∗ ∗
(4.20)
0.069∗ ∗ ∗
(3.94)
0.043∗ ∗
(2.45)
Yes
792
0.058
Yes
3059
0.104
13D · d[t]
13D · d[t + 1]
13D · d[t + 2]
Voted contest · d[c]
Voted contest · d[c + 1]
Voted contest · d[c + 2]
Firm & year FE
Observations
Adjusted R-squared
Prob (Favored but unaffiliated director)
∗∗∗
0.187
(6.41)
0.046
(1.07)
−0.067∗ ∗
(−2.50)
0.092∗ ∗ ∗
(4.22)
0.070∗ ∗ ∗
(4.04)
0.036∗ ∗
(2.08)
0.118∗ ∗
(2.08)
0.119∗ ∗ ∗
(2.72)
0.055
(1.41)
Yes
3369
0.102
∗∗∗
0.194
(1.63)
0.079
(0.70)
0.004
(0.03)
0.302
(6.73)
0.155∗ ∗ ∗
(3.02)
0.035
(0.97)
0.057∗ ∗
(2.15)
0.026
(1.64)
0.030∗ ∗
(2.13)
Yes
792
0.170
Yes
3059
0.259
initial intervention and following proxy contest announcements. For the first three years following the initial intervention, we find increased probabilities of new
activist-affiliated directors and activist-desired directors
and fewer other directors added to the board, although
the magnitudes are lower than those observed during the
settlement year. Similarly, we find a higher likelihood of
activist-affiliated and activist-desired directors, and lower
other directors, added to target firm boards in the year of
and the year following proxy contest announcement.
Though the board turnover around initial intervention,
during proxy fights, and following settlement is statistically significant and economically meaningful, it reflects
turnovers that, in most cases, involve a strict minority relative to the full size of a board (typically 8–9 directors).
The activist’s strategy is thus about influence rather than
control of the board. The activists, even if successful, are
not able to dominate the boardroom or to dictate corporate
policy. Instead, they aim at persuading other shareholders
to support their candidates and then influencing decisionmaking inside the boardroom. Such a strategy differentiates hedge fund activists from the corporate raiders of the
1980s, who tended to seek outright control as well as limits the scope for extraction of private benefits.
6.4. Personal characteristics of new and departing directors
Some institutional investors, such as BlackRock and
State Street, have publicly expressed concerns about the
∗∗∗
0.306
(7.00)
0.165∗ ∗ ∗
(3.27)
0.028
(0.74)
0.021
(0.76)
0.000
(0.00)
0.003
(0.17)
0.256∗ ∗ ∗
(4.29)
0.210∗ ∗ ∗
(4.04)
0.084
(1.39)
Yes
3369
0.257
∗∗
All
campaigns
(9)
∗∗∗
−0.443
(−2.43)
−0.160
(−1.01)
0.016
(0.09)
−0.492
(−8.34)
−0.194∗ ∗ ∗
(−3.17)
0.034
(0.84)
−0.148∗ ∗ ∗
(−3.93)
−0.091∗ ∗ ∗
(−3.67)
−0.069∗ ∗ ∗
(−2.93)
Yes
792
0.223
Yes
3059
0.324
−0.500∗ ∗ ∗
(−8.59)
−0.213∗ ∗ ∗
(−3.56)
0.038
(0.92)
−0.109∗ ∗ ∗
(−2.82)
−0.065∗ ∗
(−2.36)
−0.035
(−1.35)
−0.360∗ ∗ ∗
(−3.74)
−0.324∗ ∗ ∗
(−4.26)
−0.120
(−1.39)
Yes
3369
0.329
directors who are added to corporate boards as a result of
settlements rather than victory in a proxy fight, where institutional investors would have a say. We aim to assess
these concerns about the quality of the new directors in
two ways. First, we examine the personal characteristics
of the directors who enter (and leave) the board following settlement agreements. Second, we test, in the next
section, whether activist-affiliated and activist-desired directors receive lower voting support at the ensuing annual
general meeting elections.
To examine the personal characteristics of directors
who enter and leave the boardroom following settlement
agreements, we focus on three director traits that are likely
to proxy for their effectiveness as overseers. The first is a
dummy variable Age > 70 for directors who are at least in
their 70s. The second is Connectedness, which is the number of different listed companies a director served on as
a board member over the five preceding years. This measure assumes that directors preserve personal connections
they made in the recent past (Stuart and Yim, 2010). For
the departure analysis, we also use the variable Tenure, defined as years of service on the current board, which might
reflect both experience and entrenchment.
Table 9 shows the results of linear probability regressions on all director-year observations for treatment and
control firms covered in the Board Magazine Directors
Database. To increase statistical power by decreasing the
number of interaction terms to estimate, we aggregate
the years around the settlement, around proxy fight
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
23
Table 9
Board renewal around settlement agreements.
This table presents triple-difference analyses of the relation between director characteristics and the probability of director turnover around settlement
agreements. Columns 1 to 3 examine additions of new directors and columns 4 to 6 departures of incumbent directors. For target firms that settle (“Settled campaigns”; all columns), we select untargeted control firms by propensity score matching on the year that the settlement agreement is reached.
For target firms that experience a proxy contest vote (columns 3 and 6), we select untargeted control firms by propensity score matching on the proxy
contest announcement year. For target firms that experience neither settlement nor proxy contest vote (“Other campaigns”; columns 2, 3, 5, 6), we select
untargeted control firms by propensity score matching on the year of the initial activism announcement (in most cases the year of the first 13D filing). The
tabulated coefficients measure how the difference between treatment firms and control firms in the relation between director characteristics and director
turnover changes from the three years before treatment/placebo (POST=0) to the year of and the two years following treatment/placebo (POST=1). Treatment can either be the settlement year (for settled campaigns) or the proxy fight announcement year (for those proxy fights that went to a vote) or the
year of the initial activism announcement (for “Other campaigns” and “All campaigns”). Columns 2, 3, 5, and 6 (columns 3 and 6) include dummies for the
Settlement (Voted proxy) main effect. All columns include a dummy variable d[POSTS] that is equal to one in the year of and the two years following the
settlement year (for treatment firms) or placebo year (for control firms) and equal to zero in the three years prior to the settlement or placebo. In addition,
columns 2 and 3 (column 3) include another d[POSTD] (d[POSTC]) dummy that is equal to one in the year of and the two years following the initial intervention year or matched placebo year (proxy fight announcement year or matched placebo year) and equal to zero in the three years prior to the initial
intervention year or placebo year (proxy fight announcement year or placebo year). All columns include the main effects for the different director characteristics. All relevant two-way interactions are included. Coefficients and standard errors clustered by firm are estimated using linear probability models.
t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in
Appendix A.3.
Dependent variable
Sample
Settlement · d[POSTS] · Age >70
Settlement · d[POSTS] · Connectedness
Prob (Director addition)
Settled
campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
Prob (Director departure)
All campaigns
& control
firms
Settled
campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
(3)
(4)
(5)
−0.047∗ ∗ ∗
(−2.87)
0.011∗
(1.85)
−0.037∗ ∗
(−2.38)
0.010∗
(1.78)
−0.037∗ ∗
(−2.40)
0.010∗
(1.83)
−0.031
(−0.98)
−0.002
(−0.37)
0.003∗ ∗
(2.34)
−0.009
(−1.56)
0.004∗ ∗
(2.27)
−0.010∗
(−1.85)
0.004∗ ∗
(2.08)
−0.044
(−1.47)
−0.002
(−0.44)
0.003∗ ∗
(2.29)
0.009
(0.78)
−0.000
(−0.10)
0.000
(0.55)
Settlement · d[POSTS] · Tenure
13D · d[POSTD] · Age >70
13D · d[POSTD] · Connectedness
13D · d[POSTD] · Tenure
−0.029∗
(−1.69)
0.002
Voted Contest · d[POSTC] · Age >70
Voted Contest · d[POSTC] ·
Connectedness
(0.40)
Voted Contest · d[POSTC] · Tenure
Firm & year FE
Observations
Adjusted R-squared
Yes
66,054
0.033
Yes
362,992
0.040
announcement, and around the initial activist intervention
into three-year periods. d[POSTS], d[POSTD], and d[POSTC]
are equal to zero in the three years prior to the settlement
year, initial intervention year, or contest announcement
year, respectively, and are equal to one in the three years
afterwards. The incremental probability of the addition
or departure of directors with certain characteristics
post-settlement is captured by the three-way interaction terms Settlement · d[POSTS] · Age>70, Settlement
· d[POSTS] · Connectedness, and Settlement · d[POSTS] ·
Tenure. Similarly, we define three-way interactions for the
initial intervention and the announcement of voted proxy
contests.
What is noteworthy about this analysis is how the characteristics of the directors involved in board turnover differ when turnover is associated with the initial activist intervention versus with a settlement. We find that in the
[s, s + 2] period following a settlement, target firms show
Yes
386,402
0.040
Yes
66,054
0.041
Yes
362,992
0.043
All campaigns
& control
firms
(6)
−0.046
(−1.58)
−0.002
(−0.35)
0.003∗ ∗
(2.48)
0.012
(1.00)
−0.001
(−0.46)
0.000
(0.19)
−0.006
(−0.14)
−0.010∗
(−1.65)
0.007∗ ∗ ∗
(3.26)
Yes
386,402
0.044
an abnormal increase in the appointment of more wellconnected directors and an abnormal decrease in the appointment of older directors relative to control firms. We
observe a similar, but weaker, pattern of director appointments around the time of initial intervention and following the announcement of proxy fights that go to a vote.
With respect to director departures, the likelihood of incumbent directors with longer tenure to resign increases
abnormally following settlement agreements and following
proxy fight announcements, but not following the initial
intervention.
Taken together, these results imply that following settlements, boards end up being composed of fewer old
directors, and more shorter-tenured and better-connected
directors than prior to the settlement, thereby arguably increasing board effectiveness according to the general consensus from the literature (Core et al., 1999; Larcker et al.,
2013; Nili, 2016; Huang and Hilary, 2018).
24
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 10
Market reaction to settlement agreements.
This table shows the stock market reaction to settlement agreements. Panel A shows the average market reaction, and Panel B examines the determinants of
the market reaction. The stock market reaction is measured over a five-day window and a seven-day window centered on the date on which the settlement
agreement is filed. We measure the buy and hold return on the target firm over these windows relative to the buy and hold return on the value-weighted
market index. The sample includes settlements for activism campaigns that are launched between 20 0 0 and 2013 and for which we have information
on the market reaction and control variables. Coefficients and heteroskedasticity-robust standard errors are estimated using OLS. t-statistics appear in
parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Panel A. Average market reaction
Buy-and-hold abnormal return
Event window
(t−2, t + 2)
(t−3, t + 3)
Average
Observations
Standard error
t-statistic
1.16%
1.28%
374
374
0.0043
0.0049
2.71
2.60
Panel B. Determinants of the market reaction
Buy-and-hold abnormal return (t−2, t + 2)
(1)
High board turnover
(2)
(3)
0.026
(1.20)
Observations
R-squared
Adjusted R-squared
(6)
0.081∗ ∗ ∗
(2.64)
−0.012
(−1.33)
0.000
(0.03)
−0.004
(−0.48)
360
0.015
0.004
(7)
(8)
0.027
(1.16)
High-impact settlement
Book / Market
(5)
0.021∗
(1.65)
Strategic transaction
Ln (Market cap)
(4)
0.020∗
(1.87)
CEO departure
Proxy fight
Buy-and-hold abnormal return (t−3, t + 3)
−0.011
(−1.24)
0.000
(0.17)
−0.003
(−0.42)
360
0.009
−0.002
−0.013
(−1.50)
0.001
(0.31)
−0.001
(−0.16)
360
0.032
0.021
7. Stock market reactions
In examining stock market reactions, the literature has
paid close attention to the stock price reactions accompanying the initial campaign launch disclosures (usually the
filing of SEC 13(d) schedules) that alert the market to the
purchase of a significant stake by an activist. The existing
literature to date has provided consistent evidence indicating that hedge fund activism announcements lead to a
price increase on the order of about 4.5% during the 20day window, which is not reversed in the longer term (for
up to five years) for the time period of 1994 to 2016.23
Moreover, this positive reaction from the stock market is
justified by the ex post improvement in both operating
performance and corporate governance, e.g., increases in
the ROA and in CEO pay-for-performance. (e.g., Brav et al.,
2008; Bebchuk et al., 2015).
We first examine the market reaction to the announcement of a settlement agreement. Panel A of Table 10
shows that there is a statistically significant, 1.16% average buy-and-hold abnormal return in excess of the
value-weighted market benchmark during the [t−2, t + 2]
day window around the settlement date; the results are
slightly stronger if the time window is expanded to [t−3,
t + 3].
23
See the updated analysis provided by Barry, Brav, and Jiang (2019),
available at: http://people.duke.edu/∼brav/HFactivism_March_2019.pdf.
0.089∗ ∗ ∗
(2.91)
0.029∗ ∗ ∗
(2.79)
−0.014
(−1.63)
−0.000
(−0.05)
−0.004
(−0.48)
360
0.028
0.017
−0.010
(−0.97)
0.002
(0.68)
0.003
(0.28)
360
0.012
0.001
−0.009
(−0.87)
0.002
(0.81)
0.003
(0.32)
360
0.007
−0.004
−0.011
(−1.11)
0.002
(0.96)
0.005
(0.55)
360
0.028
0.017
0.031∗ ∗ ∗
(2.61)
−0.013
(−1.25)
0.002
(0.60)
0.003
(0.28)
360
0.024
0.013
When we use the estimates of abnormal returns as a
dependent variable in Panel B of Table 10, high-impact settlements (those that stipulate at least three director additions or departures, or a strategic transaction, or CEO
turnover) earn an extra 3 percentage points compared to
nonhigh impact settlements. This result is consistent with
prior literature documenting higher announcement and
holding-period returns for activism campaigns that result
in board turnover, takeovers, and other forms of corporate
restructurings (Greenwood and Schor, 2009; Becht et al.,
2017).
Among the components of high-impact settlements,
strategic transactions (sale or merger of the firm or of
parts of the firm’s assets) generate the highest short-term
returns, specifically between 8.1% and 8.9% above the average abnormal return of settlements in which no strategic
transaction is specified. CEO turnover is associated with an
insignificant abnormal price increase of about 2.6%. These
results are consistent with Corum (2018), who predicts
that settlements, which specify the ultimate outcomes that
activists seek, will generate larger market reactions.
Finally, in untabulated analysis, looking back in time,
we compare the abnormal returns around the initial campaign announcement for the sample of campaigns that
subsequently lead to a settlement with campaigns that do
not. We find that the former have a significantly greater
market reaction at campaign announcement than the latter. Specifically, for a 5 (7) day window centered on the
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
campaign announcement date, we find market-adjusted
buy-and-hold returns of 3.74% (4.2%) for campaigns that
were eventually settled and only 2.59% (2.89%) for campaigns that were not settled. This pattern is consistent
with markets viewing settlements favorably and being able
to identify at the outset that some campaigns are more
likely to results in settlement than others.
Overall, the results presented in this section are consistent with the market reacting favorably to settlements,
viewing their direct and indirect effects as overall value increasing. The results are also inconsistent with shareholders viewing privately negotiated settlement agreements as
devices for rent extraction by activists that are overall
value reducing for other shareholders.
8. The aftermath of settlements
We have seen that, although hedge fund activists
are ultimately interested in leadership and operational
changes, settlement agreements often focus on the introduction of new directors. We have argued that activists
seek to obtain such new directors to facilitate the subsequent occurrence of leadership and operational changes. In
this section, we test the hypothesis that, in line with the
incomplete contracting, asymmetric information, and reputational concerns theories, predicts the occurrence of CEO
turnover and operational changes in the years following
settlement agreements:
Hypothesis H6. Settlement agreements are positively associated with the occurrence of future CEO turnover and operational changes.
To test this hypothesis, we employ the same empirical
methodology that we developed in Section 6.1, which relies on propensity score matched samples and differencein-differences estimation. Below we analyze in turn how
settlements are associated with subsequent CEO turnover,
shareholder payouts, strategic transactions, and operating
performance. We conclude by examining whether there is
post-settlement evidence of rent extraction by the activist
at the expense of other shareholders.
8.1. CEO turnover
While the high portion of settlement agreements that
contractually specify director additions (84%) suggest
that the above-documented board renewal is directly attributable to the settlement agreement, only very few settlements stipulate the ouster of the incumbent CEO (5.3%).
While this result is consistent with Hypothesis H3, it might
still seem puzzling given that Becht et al. (2009) find that
replacing target firms’ CEOs is an important mechanism
through which activists can implement their objectives. In
line with this claim, untabulated analyses show that CEO
turnover spikes to an annual rate of 28%, in the year after
settlement, up from a pre-settlement average of about 12%
annually, and significantly higher than the rate of turnover
among matched control firms of 11%. CEO turnover reverts back to 14% by the third year following settlement
and lingers at about that rate until the fifth year. Taken
together, these results suggest that the CEO is allowed to
25
resign “quietly” after the settlement as opposed to
her ouster being publicly announced in the settlement
agreement.
Table 11 examines CEO turnover around initial activist
interventions, the announcement of proxy fights that go to
a vote, and settlement agreements in a regression framework using a similar difference-in-differences specification
as in Table 7. In the year of the settlement agreements, the
CEO turnover rate at target firms increases by between 6.3
and 7.7 percentage points in excess of the change in CEO
turnover at control firms during the same time. This abnormal increase grows even further in the first year following the settlement with an abnormal increase (relative
to the year prior to the settlement or placebo year) of between 13.2 and 14 percentage points. Importantly, we do
not observe abnormal turnover prior to the settlement, and
the difference-in-differences estimates revert back to zero
by year s + 2 or s + 3. Taken together, these results provide
strong support for Hypothesis H6 that settlements facilitate leadership changes down the road.
In the years following the announcement of voted
proxy contests, we find increases in the CEO turnover rate
that are of similar magnitude compared to CEO turnover
following settlements, suggesting again that settlements
might be a less confrontational, and hence cheaper, means
to achieve similar leadership changes as going all the way
to a vote. We also show a significant increase in abnormal
CEO turnover in the year of the initial intervention and the
two subsequent years. This result is in line with the notion
that many activism campaigns are successful in replacing
the CEO even in the absence of an explicit settlement or
proxy vote, potentially signaling the presence of informal
settlements.
To the extent that activists attempt to replace CEOs at
target companies as a way to restore governance and performance, this effort should be reflected in an increased
sensitivity of CEO turnover to performance. To this end, we
test whether activism, settlement agreements, and proxy
fights affect the relation between prior firm performance
and CEO turnover. Specifically, we run a linear probability model for CEO turnover at the firm-year level that is
similar to the model in Table 9. The key independent variables are three-way interaction terms between indicator
variables for the time periods around the launching of activism campaigns, settlement agreements, and proxy fight
announcements and a measure of prior-year stock return
performance (e.g., Settlement · d[POSTS] · Stock returns for
settled campaigns). Table 12 reports the results, which suggest that CEO turnover following settlements is different,
namely more performance based, than CEO turnover following initial activist interventions or following the announcement of voted proxy contests.
The negative and statistically significant main effect on
Stock returns shows that past stock return performance is
negatively related to the likelihood of CEO turnover among
control firms in the pre-placebo period, which is a standard finding in the governance literature (e.g., Kaplan and
Minton, 2012). Most important in our finding is that the
negative and significant coefficient on Settlement · d[POSTS]
· Stock returns is around three times as large as the Stock
returns baseline effect, which means that the abnormal
26
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 11
CEO turnover around settlement agreements.
This table presents difference-in-differences analyses of CEO turnover around settlement agreements. For target firms that settle (“Settled campaigns”; all
columns), we select untargeted control firms by propensity score matching on the year that the settlement agreement is reached. For target firms that
experience a proxy contest vote (column 3), we select untargeted control firms by propensity score matching on the proxy contest announcement year. For
target firms that experience neither settlement nor proxy contest vote (“Other campaigns”; columns 2, 3), we select untargeted control firms by propensity
score matching on the year of the initial activism announcement (in most cases the year of the first 13D filing). The tabulated coefficients measure how
the difference between treatment firm CEO turnover and control firm CEO turnover has changed from the year before treatment/placebo (the omitted
base year) to year s ± k, t ± k, or c ± k. Treatment can either be the settlement year (for settled campaigns) or the proxy fight announcement year (for
those proxy fights that went to a vote) or the year of the initial activism announcement (for “Other campaigns” and “All campaigns”). Columns 2 and 3
(column 3) include dummies for the Settlement (Voted proxy) main effect. All columns include a set of 11 d[s ± k] dummies, one for each year relative to
the settlement year (for treatment firms) and matched placebo year (for control firms). In addition, columns 2 and 3 (column 3) include another set of
11 d[t ± k] (d[c ± k]) dummies, one for each year relative to the initial campaign launch year (proxy fight announcement year) and matched placebo year.
At the bottom of the table, p-values of partial F-tests indicate the statistical significance of difference-in-differences estimations of CEO turnover around
settlement agreements using year t-2 as the base year. Coefficients and standard errors clustered by firm are estimated using OLS. t-statistics appear in
parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent variable
Sample
Settlement · d[s−5]
Settlement · d[s−4]
Settlement · d[s−3]
Settlement · d[s−2]
Settlement · d[s]
Settlement · d[s + 1]
Settlement · d[s + 2]
Settlement · d[s + 3]
Settlement · d[s + 4]
Settlement · d[s + 5]
13D · d[t−5]
13D · d[t−4]
13D · d[t−3]
13D · d[t−2]
13D · d[t]
13D · d[t + 1]
13D · d[t + 2]
13D · d[t + 3]
13D · d[t + 4]
13D · d[t + 5]
Voted contest · d[c-5]
Voted contest · d[c−4]
Prob (CEO turnover)
Settled campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
(3)
−0.029
(−0.94)
0.010
(0.32)
−0.003
(−0.10)
−0.010
(−0.31)
0.077∗ ∗
(2.18)
0.140∗ ∗ ∗
(3.43)
0.056
(1.44)
0.007
(0.19)
0.003
(0.07)
−0.021
(−0.44)
−0.004
(−0.14)
0.034
(1.09)
0.011
(0.36)
−0.003
(−0.09)
0.063∗
(1.84)
0.132∗ ∗ ∗
(3.33)
0.059
(1.54)
0.027
(0.76)
0.031
(0.81)
0.003
(0.07)
−0.021
(−1.37)
−0.007
(−0.50)
−0.015
(−1.14)
0.007
(0.49)
0.035∗ ∗
(2.56)
0.028∗
(1.86)
0.032∗
(1.94)
0.009
(0.57)
−0.014
(−0.95)
−0.016
(−0.95)
−0.003
(−0.11)
0.036
(1.21)
0.013
(0.44)
−0.002
(−0.05)
0.064∗
(1.94)
0.134∗ ∗ ∗
(3.48)
0.061∗
(1.66)
0.027
(0.79)
0.031
(0.84)
0.004
(0.10)
−0.019
(−1.33)
−0.009
(−0.65)
−0.016
(−1.21)
0.004
(0.32)
0.035∗ ∗ ∗
(2.62)
0.029∗ ∗
(2.01)
0.029∗
(1.80)
0.007
(0.47)
−0.011
(−0.78)
−0.018
(−1.12)
0.048
(1.16)
−0.004
(−0.10)
All campaigns &
control firms
(continued on next page)
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
27
Table 11 (continued)
Dependent variable
Sample
Prob (CEO turnover)
Settled campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
(3)
Voted contest · d[c−3]
Yes
12,162
0.037
Yes
67,304
0.071
0.057
(1.23)
0.063
(1.30)
0.018
(0.37)
0.178∗ ∗ ∗
(2.75)
0.147∗ ∗
(2.53)
−0.017
(−0.36)
0.156∗ ∗
(2.47)
0.006
(0.10)
Yes
71,603
0.073
0.015
0.000
0.077
0.648
0.750
0.819
0.065
0.001
0.100
0.425
0.404
0.893
0.068
0.001
0.097
0.445
0.420
0.893
Voted contest · d[c−2]
Voted contest · d[c]
Voted contest · d[c + 1]
Voted contest · d[c + 2]
Voted contest · d[c + 3]
Voted contest · d[c + 4]
Voted contest · d[c + 5]
Firm & year FE
Observations
Adjusted R-squared
p-values for partial F-tests comparing coefficients
S · d[s] vs. S · d[s−2]
S · d[s + 1] vs. S · d[s−2]
S · d[s + 2] vs. S · d[s−2]
S · d[s + 3] vs. S · d[s−2]
S · d[s + 4] vs. S · d[s−2]
S · d[s + 5] vs. S · d[s−2]
All campaigns &
control firms
Table 12
The CEO turnover-performance sensitivity around settlement agreements.
This table presents analyses of the CEO turnover-performance sensitivity around settlement agreements. For target firms that settle (“Settled campaigns”;
all columns), we select untargeted control firms by propensity score matching on the year that the settlement agreement is reached. For target firms
that experience a proxy contest vote (column 3), we select untargeted control firms by propensity score matching on the proxy contest announcement
year. For target firms that experience neither settlement nor proxy contest vote (“Other campaigns”; columns 2, 3), we select untargeted control firms
by propensity score matching on the year of the initial activism announcement (in most cases the year of the first 13D filing). The tabulated coefficients
on the three-way interactions measure how the difference between treatment firms’ CEO turnover-sensitivity and control firms’ CEO turnover sensitivity
changes from the three years before treatment/placebo (POST=0) to the year of and the two years following treatment/placebo (POST=1). Treatment can
either be the settlement year (for settled campaigns) or the proxy fight announcement year (for those proxy fights that went to a vote) or the year of
the initial activism announcement (for “Other campaigns” and “All campaigns”). Columns 1 to 3 (column 3) include dummies for the 13D and Settlement
(Voted proxy) main effects. All columns include a dummy variable d[POSTS] that is equal to one in the year of and the two years following the settlement
year (for treatment firms) or placebo year (for control firms) and equal to zero in the three years prior to the settlement or placebo. In addition, columns
2 and 3 (column 3) include d[POSTD] (d[POSTC]) dummy that is equal to one in the year of and the two years following the initial intervention year or
matched placebo year (proxy fight announcement year or matched placebo year) and equal to zero in the three years prior to the initial intervention year
or placebo year (proxy fight announcement year or placebo year). All relevant two-way interactions are included. Coefficients and standard errors clustered
by firm are estimated using linear probability models. t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at the 1%, 5%, and 10% level,
respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent variable
Sample
Stock returns
Settlement · d[POSTS] · Stock returns
Prob (CEO turnover)
Settled campaigns &
control firms
Settled campaigns, other
campaigns & control firms
(1)
(2)
∗∗∗
(3)
∗∗∗
−0.038
(−3.54)
−0.109∗ ∗ ∗
(−3.42)
−0.029
(−6.22)
−0.094∗ ∗ ∗
(−3.02)
−0.015
(−1.14)
Yes
7499
0.017
Yes
41,076
0.012
13D · d[POSTD] · Stock returns
Voted contest · d[POSTC] · Stock returns
Year FE
Observations
Adjusted R-squared
All campaigns &
control firms
−0.030∗ ∗ ∗
(−6.63)
−0.090∗ ∗ ∗
(−2.88)
−0.015
(−1.20)
0.007
(0.12)
Yes
43,691
0.013
28
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
(i.e., control firm adjusted) increase in the CEO turnoverperformance sensitivity among target firms from pre- to
post- settlement is economically significant. For the initial intervention or the year following the announcement
of proxy fights that go to a vote, we find no significant
increase in the relation between CEO turnover and firm
performance. Overall, the empirical evidence implies that
activists oust CEOs more selectively following settlement
agreements.
8.2. Shareholder payouts
Activists often seek an increase in shareholder payouts,
and Brav et al. (2008) show increases in payout among
target firms following activist interventions. As we show
in Table 5, however, a share buyback is directly stipulated
in only 5.3% of the settlement agreements in our sample.
When we go beyond the contractual stipulations in the
settlement and examine the annual level of payout in
the years surrounding the settlement date, we find that
payout yield, which is the sum of dividends and share
repurchases scaled by the firm’s market capitalization,
spikes in the year that the settlement agreement is filed.
This can be seen in Table 13. In the settlement year, target
firm payout yield increases by 1.1–1.2 percentage points,
in excess of the change in payout among control firms,
compared to the year prior to settlement or placebo. These
magnitudes are economically significant given the average payout yield of around 4% observed in the 11 years
surrounding the settlement agreement. However, we find
some evidence that payout yield was lower among firms
that ultimately settle compared to matched control firms
in the pre-settlement years. This suggests that activists do
not demand unreasonable payout from target firms but
rather revert target firm payout back to normal levels.
Similar to the CEO turnover-performance analyses, settlements are different from the initial intervention or
proxy fights also with respect to payout. We find no evidence of abnormal changes in payout following activism
campaign launch or the announcement of proxy contests
that go to a vote. Overall these results support Hypothesis
H6 that settlement agreements facilitate the distribution of
free cash flow to shareholders in the future.
8.3. Strategic transactions
Prior research shows activists’ ability to put target
firms into play (Greenwood and Schor, 2009; Burkart and
Lee, 2015; Boyson et al., 2017; Corum and Levit, 2019).
However, only about 8% of settlement agreements signal
the possibility for a strategic transaction.24 As discussed in
the development of Hypothesis H3, strategic transactions,
such as selling the target firm, are difficult to contract on
in the settlement due to incomplete contracting, asymmetric information, and potential reputation concerns, but
as Hypothesis H6 predicts, settlements can be used to
24
Two point five percent of agreements specify the sale of the firm or
a merger, another 0.5% the sale of parts of the firm’s assets, and 5.8%
stipulate the formation of a strategic transaction committee on the board
of directors.
facilitate a transaction later on. Thus, in this section we
examine whether the likelihood of target firms being sold
increases in the aftermath of settlements.
In Table 14, we include only those years following the
initial intervention or placebo year and find an increase in
the probability of target firm stock market delisting following settlements. Importantly, this increase in overall delisting propensity is driven by those delistings that are related to M&A or going private transactions, while we find
no increase in distress-related delistings post settlement.
Estimates of the abnormal probability of M&A or going
private transactions is about 6.7 percentage points higher
than controls in the year of the settlement. After controlling for the dynamics of delistings around the initial intervention, the abnormal likelihood of an acquisition or going
private transaction among settlement targets decreases to
2.6 percentage points in the year of settlement relative to
control firms. While this is an economically large magnitude, given an average in-sample likelihood of M&A or going private transactions of 5.4%, the estimate is statistically
insignificant at conventional levels.
For the years following initial activist interventions,
we find that the likelihood of stock market delistings
is abnormally high, which is almost entirely driven by
M&A and going private transactions. The rate of distressrelated delistings is marginally higher among target firms
compared to control firms only in the third year following the campaign launch year. For proxy contests, we find
some evidence that delistings occur less frequently in the
year of and the year following the announcement year,
which could be a partly mechanical relation, however, if
those are the years during which contests are actually
voted on. In other words, we cannot observe a contested
vote if the company has delisted prior to the shareholder
meeting. In the third year following proxy fight announcement, we find that the probability of delisting increases,
which is driven by M&A and going private transactions.
While the results for settlement agreements alone are
weak, collectively, these findings imply that target firms
are more likely to delist for favorable reasons (M&A and
going private) than control firms, refuting the claim that
hedge fund activists destabilize target firms.25
8.4. Operating performance
Several prior studies on hedge fund activism show
deterioration in ROA and Tobin’s Q among target firms
prior to the intervention and a full recovery afterwards (for recent evidence, see Bebchuk et al., 2015 and
Gantchev et al., 2019a). In this section, we examine
whether post-intervention performance increases are similar among activism campaigns that lead to a settlement or
to a proxy vote. The first three columns of Table 15 examine ROA. In columns 2 and 3, we find target firm performance indeed deteriorates from year t−3 through the year
of the initial intervention. From year t + 1 through t + 5,
ROA improves but this improvement turns significant only
25
For an example of such claims, see Reuters, 11/16/2015, “The cannibalized company.”
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
29
Table 13
Payout policy around settlement agreements.
This table presents difference-in-differences analyses of payout yield around settlement agreements. For target firms that settle (“Settled campaigns”; all
columns), we select untargeted control firms by propensity score matching on the year that the settlement agreement is reached. For target firms that
experience a proxy contest vote (column 3), we select untargeted control firms by propensity score matching on the proxy contest announcement year. For
target firms that experience neither settlement nor proxy contest vote (“Other campaigns”; columns 2, 3), we select untargeted control firms by propensity
score matching on the year of the initial activism announcement (in most cases the year of the first 13D filing). The tabulated coefficients measure how
the difference between treatment firm and control firm payout policy has changed from the year before treatment/placebo (the omitted base year) to year
s ± k, t ± k, or c ± k. Treatment can either be the settlement year (for settled campaigns) or the proxy fight announcement year (for those proxy fights
that went to a vote) or the year of the initial activism announcement (for “Other campaigns” and “All campaigns”). Columns 2 and 3 (column 3) include
dummies for the Settlement (Voted proxy) main effect. All columns include a set of 11 d[s ± k] dummies, one for each year relative to the settlement year
(for treatment firms) and matched placebo year (for control firms). In addition, columns 2 and 3 (column 3) include another set of 11 d[t ± k] (d[c ± k])
dummies, one for each year relative to the initial campaign launch year (proxy fight announcement year) and matched placebo year. At the bottom of the
table, p-values of partial F-tests indicate the statistical significance of difference-in-differences estimations of payout yield around settlement agreements
using year t-2 as the base year. Coefficients and standard errors clustered by firm are estimated using OLS. t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗
denote statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent variable
Sample
Settlement · d[s−5]
Settlement · d[s−4]
Settlement · d[s−3]
Settlement · d[s−2]
Settlement · d[s]
Settlement · d[s + 1]
Settlement · d[s + 2]
Settlement · d[s + 3]
Settlement · d[s + 4]
Settlement · d[s + 5]
13D · d[t−5]
13D · d[t−4]
13D · d[t−3]
13D · d[t−2]
13D · d[t]
13D · d[t + 1]
13D · d[t + 2]
13D · d[t + 3]
13D · d[t + 4]
13D · d[t + 5]
Voted contest · d[c−5]
Voted contest · d[c−4]
Payout yield
Settled campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
∗∗∗
−0.015
(−3.18)
−0.008
(−1.52)
−0.010∗ ∗
(−2.04)
−0.011∗ ∗
(−2.27)
0.011
(1.60)
−0.002
(−0.32)
0.005
(0.76)
−0.005
(−0.67)
−0.002
(−0.32)
−0.007
(−0.91)
All campaigns &
control firms
(3)
∗∗
−0.010
(−2.14)
−0.003
(−0.47)
−0.005
(−1.07)
−0.008
(−1.63)
0.011∗
(1.73)
−0.001
(−0.21)
0.008
(1.11)
0.000
(0.04)
0.005
(0.68)
−0.001
(−0.16)
−0.001
(−0.24)
−0.003
(−1.12)
−0.003
(−1.47)
−0.001
(−0.50)
0.001
(0.51)
0.004
(1.58)
0.002
(0.80)
0.001
(0.31)
−0.004
(−1.15)
−0.005
(−1.53)
−0.010∗ ∗
(−2.13)
−0.002
(−0.38)
−0.005
(−1.06)
−0.007∗
(−1.65)
0.012∗
(1.90)
−0.000
(−0.08)
0.009
(1.30)
0.001
(0.16)
0.005
(0.79)
−0.001
(−0.11)
−0.001
(−0.32)
−0.003
(−1.38)
−0.003
(−1.45)
−0.001
(−0.49)
0.001
(0.50)
0.004
(1.53)
0.002
(0.57)
0.000
(0.15)
−0.004
(−1.24)
−0.005
(−1.55)
0.025∗
(1.82)
0.010
(1.30)
(continued on next page)
30
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 13 (continued)
Dependent variable
Sample
Payout yield
Settled campaigns &
control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
Voted contest · d[c−3]
(3)
Yes
14,674
0.280
Yes
88,487
0.283
0.005
(0.77)
0.005
(0.56)
−0.002
(−0.35)
0.005
(0.59)
−0.005
(−0.63)
−0.001
(−0.09)
0.002
(0.22)
−0.001
(−0.11)
Yes
93,674
0.283
0.000
0.108
0.016
0.318
0.186
0.517
0.002
0.247
0.022
0.191
0.049
0.316
0.001
0.218
0.018
0.180
0.048
0.320
Voted contest · d[c−2]
Voted contest · d[c]
Voted contest · d[c + 1]
Voted contest · d[c + 2]
Voted contest · d[c + 3]
Voted contest · d[c + 4]
Voted contest · d[c + 5]
Firm & year FE
Observations
Adjusted R-squared
p-values for Partial F-tests comparing coefficients
S · d[s] vs. S · d[s−2]
S · d[s + 1] vs. S · d[s−2]
S · d[s + 2] vs. S · d[s−2]
S · d[s + 3] vs. S · d[s−2]
S · d[s + 4] vs. S · d[s−2]
S · d[s + 5] vs. S · d[s−2]
All campaigns &
control firms
by t + 5. In columns 1–3, we find little evidence that settlements are followed by improvements in profitability relative to controls. Only in the first year following the settlement do we observe an abnormal improvement in ROA,
but this is not sustained in later years. For voted proxy
contests, however, we observe significant incremental improvements in profitability in four out of five years following the contest announcement.
Columns 4–6 present the results for Tobin’s Q. We find
significant abnormal (i.e., in excess of control firms) improvements in Tobin’s Q relative to the year before settlement. These improvements start materializing in the year
of the settlement, which is consistent with the market reaction to settlement disclosure presented in Table 10. For
campaigns that lead to a proxy vote, the coefficients on
Voted contest · d[c + 1] through Voted contest · d[c + 5] also
show an increasing pattern, but these coefficients are not
statistically significant, potentially due to the small number
of voted proxy contests. Four out of the five coefficients on
the interaction terms identifying the years following the
initial intervention are positive, but none of them attain
statistical significance at conventional levels.
These results suggest that more confrontational campaigns, such as those that lead to a settlement or a proxy
vote, are associated with stronger improvements in firm
valuations, which is inconsistent with the claim that performance improvements associated with activism are entirely due to stock picking. These findings also provide
some support for Hypothesis H6 and the claim that settlements facilitate operational changes down the road and
that these changes are value enhancing.
8.5. Rent extraction?
Commentators and some institutional shareholders
have expressed the concern that settlements, which activist hedge funds and incumbents make without other
shareholders being present at the bargaining table, enable
hedge fund activists to push for a short-term agenda to extract rents from other shareholders with a longer investment horizons. We find, however, that the average length
of engagement is 979 days for campaigns that are settled
and 660 days for campaigns that are not settled, a statistically significant difference. This long period for settled
campaigns is natural to expect since a settlement takes
time to obtain and takes additional time to bear fruit. This
pattern is inconsistent with a concern that settlements
push companies in directions favored by agents with very
short time horizons.
In this concluding section we further examine this concern. In particular, we investigate whether settlements enable activists to pressure incumbents to a) accept the appointment of new directors who are not favored by other
shareholders and/or b) buy out the activist in privately negotiated transactions that might resemble the greenmail
practices of the 1980s. We find no evidence to support the
rent-extraction concern.
8.5.1. Self-serving director appointments
First, we test whether directors added to the board
during settlement negotiations (i.e., without the vote of
other shareholders) are perceived by investors as less qualified or committed. Specifically, we examine director vot-
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
31
Table 14
Stock market delisting following settlement agreements.
This table presents analyses of the probability of different types of stock market delisting following settlement agreements. For target firms that settle
(all columns), we select untargeted control firms by propensity score matching on the year that the settlement agreement is reached. For target firms
that experience a proxy contest vote (columns 3, 6, 9), we selected untargeted control firms by matching on the proxy contest announcement year. For
target firms that experience neither settlement nor proxy contest vote (columns 2, 3, 5, 6, 8, 9), we select untargeted control firms by matching on the
year of the initial activism announcement (in most cases the year of the first 13D filing). We exclude firm-years prior to the event years. To identify
the coefficients in the absence of pre-periods, we include as a base group firms that were neither targeted nor selected as matched control firms. The
tabulated coefficients measure the difference in the probability of delisting between treatment firms and matched control firms in the years following
the treatment or placebo event. Treatment event can be the settlement announcement (all columns), the proxy fight announcement (columns 3, 6, 9), or
the initial activism campaign announcement (columns 2, 3, 5, 6, 8, 9). All columns include a set of six d[s + k] dummies, one for each year relative to the
settlement year (for treatment firms) and matched placebo year (for control firms). In addition, columns 2, 3, 5, 6, 8, 9 (columns 3, 6, 9) include another
set of six d[t + k] (d[c + k]) dummies, one for each year relative to the initial campaign launch year (proxy fight announcement year) and matched placebo
year. Coefficients and standard errors clustered by firm are estimated using linear probability models. t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote
statistical significance at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent Variable
Sample
Settlement · d[s]
Settlement · d[s + 1]
Settlement · d[s + 2]
Settlement · d[s + 3]
Settlement · d[s + 4]
Settlement · d[s + 5]
Prob
(Delisting)
Prob
(Acquisition
/ going
private
delisting)
Prob
(Distress
delisting)
Settled campaigns & control firms
Prob
(Delisting)
Prob
(Acquisition
/ going
private
delisting)
Prob
(Distress
delisting)
Settled campaigns, other campaigns &
control firms
Prob
(Acquisition
/ going
private
delisting)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
0.067∗ ∗ ∗
(3.43)
0.022
(1.23)
0.003
(0.19)
0.009
(0.61)
0.010
(0.57)
0.043∗
(1.85)
0.011
(1.18)
−0.002
(−0.18)
0.000
(0.02)
0.009
(0.77)
0.003
(0.41)
0.007
(0.59)
0.028
(1.27)
−0.005
(−0.24)
−0.004
(−0.21)
−0.006
(−0.30)
0.009
(0.43)
0.041
(1.47)
0.066∗ ∗ ∗
(6.17)
0.027∗ ∗ ∗
(2.85)
0.005
(0.57)
0.021∗ ∗
(2.20)
0.013
(1.32)
0.015
(1.31)
0.026
(1.32)
0.002
(0.11)
−0.011
(−0.70)
−0.006
(−0.40)
−0.007
(−0.43)
0.033
(1.43)
0.067∗ ∗ ∗
(7.11)
0.035∗ ∗ ∗
(4.25)
0.012∗
(1.73)
0.013∗
(1.79)
0.017∗ ∗
(2.18)
0.023∗ ∗
(2.41)
0.010
(1.07)
−0.003
(−0.29)
−0.004
(−0.47)
0.006
(0.54)
0.005
(0.69)
0.009
(0.75)
0.004
(0.90)
−0.001
(−0.24)
0.001
(0.28)
0.009∗
(1.87)
−0.000
(−0.07)
−0.007
(−1.60)
Yes
84,178
0.044
Yes
84,178
0.028
Yes
84,178
0.027
Yes
123,663
0.036
Yes
123,663
0.023
Yes
123,663
0.021
0.028
(1.27)
−0.005
(−0.24)
−0.004
(−0.20)
−0.006
(−0.30)
0.009
(0.43)
0.041
(1.48)
0.066∗ ∗ ∗
(6.25)
0.027∗ ∗ ∗
(2.82)
0.006
(0.63)
0.021∗ ∗
(2.16)
0.013
(1.38)
0.014
(1.30)
−0.008
(−0.23)
−0.006
(−0.17)
0.014
(0.40)
0.115∗ ∗
(2.20)
0.028
(0.73)
−0.022
(−0.63)
Yes
126,565
0.036
0.026
(1.33)
0.002
(0.10)
−0.011
(−0.71)
−0.006
(−0.41)
−0.007
(−0.41)
0.034
(1.45)
0.067∗ ∗ ∗
(7.17)
0.034∗ ∗ ∗
(4.23)
0.013∗
(1.83)
0.013∗
(1.80)
0.018∗ ∗
(2.21)
0.021∗ ∗
(2.26)
0.014
(0.46)
−0.047∗ ∗
(−2.49)
0.004
(0.12)
0.097∗ ∗
(2.25)
−0.001
(−0.03)
−0.022
(−0.88)
Yes
126,565
0.023
13D · d[t + 1]
13D · d[t + 2]
13D · d[t + 3]
13D · d[t + 4]
13D · d[t + 5]
Voted contest · d[c]
Voted contest · d[c + 1]
Voted contest · d[c + 2]
Voted contest · d[c + 3]
Voted contest · d[c + 4]
Voted contest · d[c + 5]
ing outcomes at target firms’ annual general meetings for
those new directors whose names we can match to director names in the ISS Voting Analytics data. Table 16 reports the results of analyses where the dependent variables are the fraction of votes supporting (column 1) or
against (column 2) a particular director in a given year
Prob
(Distress
delisting)
All campaigns & control firms
0.064∗ ∗ ∗
(2.93)
0.010
(0.46)
0.009
(0.46)
0.008
(0.45)
0.021
(1.00)
0.048∗
(1.72)
13D · d[t]
Industry & year FE
Observations
Adjusted R-squared
Prob
(Delisting)
(9)
0.010
(1.07)
−0.003
(−0.36)
−0.004
(−0.49)
0.006
(0.57)
0.005
(0.67)
0.009
(0.74)
0.004
(0.85)
−0.001
(−0.28)
0.003
(0.60)
0.009∗
(1.79)
−0.000
(−0.10)
−0.006
(−1.44)
−0.026∗ ∗ ∗
(−3.47)
−0.011
(−0.81)
0.015
(0.70)
0.015
(0.50)
0.012
(0.78)
0.010
(0.40)
Yes
126,565
0.021
or the fraction of votes withheld (column 3). The directors whose voting results are considered are classified into
four mutually exclusive categories based on whether they
were added following a settlement and, among directors
added afterwards, whether they are affiliated with, desired
by, or unrelated to (“Other”) the activist. These classifica-
32
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Table 15
Operating performance around settlement agreements.
This table presents difference-in-differences analyses of firm performance around settlement agreements. Columns 1 to 3 examine return on assets and
models 4 to 6 Tobin’s Q. For target firms that settle (“Settled campaigns”; all columns), we select untargeted control firms by propensity score matching on
the year that the settlement agreement is reached. For target firms that experience a proxy contest vote (columns 3 and 6), we select untargeted control
firms by propensity score matching on the proxy contest announcement year. For target firms that experience neither settlement nor proxy contest vote
(“Other campaigns”; columns 2, 3, 5, 6), we select untargeted control firms by propensity score matching on the year of the initial activism announcement
(in most cases the year of the first 13D filing). The tabulated coefficients measure how the difference between treatment firm and control firm operating
performance has changed from the year before treatment/placebo (the omitted base year) to year s ± k, t ± k, or c ± k. Treatment can either be the settlement year (for settled campaigns) or the proxy fight announcement year (for those proxy fights that went to a vote) or the year of the initial activism
announcement (for “Other campaigns” and “All campaigns”). Columns 2, 3, 5, and 6 (columns 3 and 6) include dummies for the Settlement (Voted proxy)
main effect. All columns include a set of 11 d[s ± k] dummies, one for each year relative to the settlement year (for treatment firms) and matched placebo
year (for control firms). In addition, columns 2, 3, 5, 6 (columns 3 and 6) include another set of 11 d[t ± k] (d[c ± k]) dummies, one for each year relative
to the initial campaign launch year (proxy fight announcement year) and matched placebo year. At the bottom of the table, p-values of partial F-tests
indicate the statistical significance of difference-in-differences estimations of firm performance around settlement agreements using year t-2 as the base
year. Coefficients and standard errors clustered by firm are estimated using OLS. t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance
at the 1%, 5%, and 10% level, respectively (two-tailed). All variables are defined in Appendix A.3.
Dependent variable
Sample
Settlement · d[s−5]
Settlement · d[s−4]
Settlement · d[s−3]
Settlement · d[s−2]
Settlement · d[s]
Settlement · d[s + 1]
Settlement · d[s + 2]
Settlement · d[s + 3]
Settlement · d[s + 4]
Settlement · d[s + 5]
13D · d[t−5]
13D · d[t−4]
13D · d[t−3]
13D · d[t−2]
13D · d[t]
13D · d[t + 1]
13D · d[t + 2]
13D · d[t + 3]
13D · d[t + 4]
13D · d[t + 5]
Voted contest · d[c−5]
Voted contest · d[c−4]
Voted contest · d[c−3]
Voted contest · d[c−2]
Return on assets (ROA)
Settled campaigns
& control firms
Settled campaigns,
other campaigns &
control firms
Tobin’s Q
All campaigns &
control firms
Settled campaigns
& control firms
Settled campaigns,
other campaigns &
control firms
All Campaigns &
control firms
(1)
(2)
(3)
(4)
(5)
(6)
0.025∗ ∗
(2.23)
0.014
(1.59)
0.001
(0.18)
0.002
(0.35)
−0.001
(−0.24)
0.013
(1.60)
0.004
(0.43)
0.007
(0.54)
0.004
(0.36)
0.008
(0.76)
0.018∗
(1.70)
0.008
(0.93)
−0.003
(−0.35)
0.002
(0.28)
0.003
(0.53)
0.013
(1.62)
0.000
(0.04)
0.002
(0.14)
−0.003
(−0.22)
0.003
(0.28)
0.007
(1.25)
0.007
(1.59)
0.008∗
(1.89)
0.007∗ ∗
(2.28)
−0.009∗ ∗ ∗
(−3.07)
−0.003
(−0.61)
0.001
(0.21)
0.005
(0.83)
0.006
(1.14)
0.010∗
(1.66)
0.020∗
(1.93)
0.010
(1.24)
−0.001
(−0.08)
0.003
(0.55)
0.005
(0.79)
0.015∗
(1.89)
0.003
(0.33)
0.004
(0.35)
−0.000
(−0.00)
0.005
(0.49)
0.006
(1.16)
0.007
(1.43)
0.006
(1.54)
0.005∗
(1.88)
−0.009∗ ∗ ∗
(−2.93)
−0.003
(−0.67)
0.000
(0.06)
0.003
(0.57)
0.005
(0.94)
0.008
(1.45)
−0.001
(−0.04)
0.006
(0.30)
0.009
(0.53)
0.013
(0.99)
−0.027
(−0.23)
−0.029
(−0.33)
−0.109
(−1.20)
−0.021
(−0.24)
0.050
(0.60)
0.117
(0.89)
0.251∗ ∗
(2.07)
0.153
(1.10)
0.234
(1.45)
0.317
(1.45)
0.143
(1.15)
0.142
(1.53)
0.059
(0.64)
0.147∗
(1.69)
0.223∗ ∗ ∗
(2.66)
0.264∗ ∗
(1.99)
0.357∗ ∗ ∗
(3.06)
0.228∗
(1.73)
0.337∗ ∗
(2.05)
0.431∗ ∗
(2.10)
−0.009
(−0.13)
−0.024
(−0.37)
−0.004
(−0.08)
−0.007
(−0.14)
−0.066
(−1.42)
0.014
(0.24)
−0.039
(−0.66)
0.096
(1.40)
0.071
(0.94)
0.023
(0.26)
0.119
(0.96)
0.116
(1.25)
0.028
(0.30)
0.117
(1.35)
0.199∗ ∗
(2.36)
0.239∗
(1.82)
0.333∗ ∗ ∗
(2.87)
0.207
(1.58)
0.307∗
(1.88)
0.397∗
(1.94)
−0.012
(−0.17)
−0.034
(−0.54)
−0.003
(−0.06)
−0.002
(−0.05)
−0.067
(−1.49)
0.006
(0.12)
−0.040
(−0.68)
0.090
(1.34)
0.057
(0.79)
0.038
(0.46)
−0.210
(−0.80)
−0.103
(−0.42)
0.018
(0.06)
−0.045
(−0.19)
(continued on next page)
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
33
Table 15 (continued)
Dependent variable
Sample
Return on assets (ROA)
Tobin’s Q
Settled campaigns
& control firms
Settled campaigns,
other campaigns &
control firms
All campaigns &
control firms
Settled campaigns
& control firms
Settled campaigns,
other campaigns &
control firms
(1)
(2)
(3)
(4)
(5)
Yes
95,711
0.704
0.010
(1.07)
0.039∗ ∗
(2.15)
0.028
(1.56)
0.046∗ ∗
(2.03)
0.062∗ ∗
(2.43)
0.060∗ ∗
(2.16)
Yes
101,574
0.71
Yes
15,805
0.536
Yes
96,957
0.515
0.058
(0.52)
−0.073
(−0.45)
0.088
(0.50)
0.154
(0.48)
0.324
(1.44)
0.219
(0.86)
Yes
102,880
0.514
0.842
0.195
0.898
0.994
0.730
0.906
0.138
0.767
0.911
0.630
0.987
0.864
0.182
0.961
0.931
0.788
0.872
0.118
0.851
0.992
0.701
0.967
0.552
0.281
0.065
0.217
0.137
0.140
0.605
0.103
0.476
0.226
0.254
0.530
0.373
0.152
0.566
0.292
0.197
0.756
0.266
0.972
0.470
0.362
0.502
0.354
0.141
0.525
0.292
0.206
0.760
0.265
0.954
0.491
0.387
Voted contest · d[c]
Voted contest · d[c + 1]
Voted contest · d[c + 2]
Voted contest · d[c + 3]
Voted contest · d[c + 4]
Voted contest · d[c + 5]
Firm & year FE
Yes
Observations
15,642
Adjusted R-squared
0.752
p-values for partial F-tests comparing coefficients
S · d[s] vs. S · d[s-2]
0.621
0.212
S · d[s + 1] vs. S · d[s−2]
S · d[s + 2] vs. S · d[s−2]
0.859
S · d[s + 3] vs. S · d[s-2]
0.708
0.856
S · d[s + 4] vs. S · d[s−2]
S · d[s + 5] vs. S · d[s−2]
0.576
S · d[s + 1] vs. S · d[s]
0.027
S · d[s + 2] vs. S · d[s]
0.562
S · d[s + 3] vs. S · d[s]
0.489
S · d[s + 4] vs. S · d[s]
0.618
S · d[s + 5] vs. S · d[s]
0.388
tions are captured by three indicator variables with the
fourth group, the omitted base case, being incumbent directors who were already on the target firm’s board prior
to campaign launch.
Columns 1 and 2 of Table 16 show that directors who
are added to the board following a settlement (irrespec-
All Campaigns &
control firms
(6)
tive of whether affiliated with or desired by the activist)
generally receive significantly more votes in their favor
and fewer withheld votes compared to incumbent directors. Specifically, new activist-affiliated directors receive
1.7 percentage points more in their favor (1.8 percentage
points less against them) than incumbent directors. For
Table 16
Settlement agreements and annual director elections.
This table examines whether directors who are affiliated with or desired by the activists and appointed to the board following a settlement agreement
receive lower voting support at future annual shareholder meetings than incumbent directors who were already on the board before the activism campaign
launch. The sample for this analysis is restricted to settled activism campaigns. The independent variables of interest are indicator variables for the different
director classifications, the omitted base case being incumbent directors who had been on the board prior to the activist intervention. Coefficients and
standard errors clustered by firm are estimated using OLS. t-statistics appear in parentheses. ∗ ∗ ∗ , ∗ ∗ , ∗ denote statistical significance at the 1%, 5%, and 10%
level, respectively (two-tailed). All variables are defined in Appendix A.3.
Activist-affiliated directors added following settlement
Activist-desired directors added following settlement
Other directors added following settlement
Firm & year FE
Observations
Adjusted R-squared
p-values for partial F-tests comparing coefficients
Affiliated directors = Other directors
Affiliated directors = Desired directors
Desired directors = Other directors
Fraction of votes
supporting director
Fraction of votes
against director
(1)
(2)
∗∗
Fraction of votes
withheld
(3)
∗∗
0.017
(2.44)
0.038∗ ∗ ∗
(3.69)
0.039∗ ∗
(2.58)
Yes
4027
0.135
−0.018
(−2.59)
−0.039∗ ∗ ∗
(−3.71)
−0.040∗ ∗
(−2.59)
Yes
4027
0.135
0.001
(0.82)
0.000
(0.76)
0.001∗
(1.74)
Yes
4027
0.118
0.155
0.045
0.943
0.158
0.046
0.927
0.997
0.791
0.615
34
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
activist-desired directors, these differences increase to 3.8
percentage points for supporting votes and 3.9 percentage points for disapproving votes, respectively, which is
similar to the third category of other new directors. At
the bottom of the table, partial F-tests show that activistdesired directors receive more voting support (and less
dissent) than activist-affiliated directors, but there is no
significant difference between activist-affiliated directors
and other directors. Taken together, our results provide no
evidence that directors who join boards following settlements receive lower voting support than incumbent directors, and we find no evidence that activist-affiliated or
activist-desired directors are less popular than other new
directors.
8.5.2. Greenmail
Some commentators have expressed concern that target
companies might buy out the activist in some settlements
and settlements might thus enable activists to extract rents
at the expense of other shareholders similar to economic
rents corporate raiders might have extracted during the
1980s through greenmail.26 We investigate whether such
a concern is empirically justified.
To this end, we search for information about privately
negotiated buybacks in the settlement agreements and
other company disclosures, Capital IQ, SDC Platinum, and
FactSet Shark Watch. We find that only 19 settlements out
of 399 during our period were followed by a privately negotiated buyback that enabled the activist to sell shares to
the issuer. Furthermore, in a large majority of these cases,
the record indicates that the buyback was executed without a premium to the market price, and in the remaining
26
See, e.g., Wall Street Journal, 6/11/2014, “Activist funds dust off ’greenmail’ playbook” and Forbes, 2/19/2016, “Greenmail lives! Activist RiverNorth folds at Fifth Street after a Tannebaum-led buyout.”
cases we are unable to determine the price of the transaction. Thus, we find no evidence supporting the greenmail
concern.
9. Conclusion
Settlement agreements between the activist and the
target’s board have been playing an increasingly important role in the corporate governance landscape. Using
a comprehensive hand-collected data set, we provide the
first systematic analysis of the drivers, nature, and consequences of such settlement agreements.
We identify the factors that determine the likelihood of
a settlement, showing that the evidence is consistent with
settlements being more likely when the activist has a credible threat to win board seats in a proxy fight and when
incumbent directors face significant reputation concerns.
We then show that settlements bring about major changes
in board composition, rather than provide directly the operational changes that activists seek, but that settlements
are followed by such changes, including increases in CEO
turnover, increased payout to shareholders, and improvements in valuation. We find no evidence that settlements
allow activists to appoint directors who are not supported
by other shareholders or that they enable the extraction of
economically meaningful rents through greenmail. Indeed,
the disclosure of settlements is accompanied by positive
stock price reactions. Our analysis highlights the importance of settlements, and it also provides an empirical
foundation on which future examination of settlements
and activist engagements can more generally build
upon.
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
35
Appendix
A.1. List of settlement agreements involving large companies
Table A.1
This table reports sample settlement agreements reached during campaigns that were launched between 20 0 0 and 2013 involving target companies with at
least $7 billion market capitalization. Market capitalization is expressed in millions of dollars. Time refers to the month and year the settlement agreement
was reached.
Company
Market cap
Activist
Time
Microsoft
PepsiCo
Home Depot
Time Warner
228,220
140,680
81,635
76,432
ValueAct
Trian
Relational
Icahn
Aug 2013
Jan 2015
Feb 2007
Feb 2006
Mondelez
Hewlett-Packard
Motorola Inc
55,724
55,635
36,300
Trian
Relational
Icahn
Jan 2014
Nov 2012
April 2008
Illinois Tool Works
Williams Companies
Motorola Solutions
22,570
22,069
21,309
Relational
Corvex
ValueAct
Jan 2012
Feb 2014
Oct 2012
-
Yahoo
19,195
Third Point
May 2012
-
Hess Corp
18,088
Elliott
May 2013
Genzyme
17,967
Relational
Apr 2010
Air Products & Chemicals
17,508
Pershing
Square
Sep 2013
-
May 2008
- Three new directors are added, two favored by the activist but unaffiliated,
one not identified to be favored by the activist
- The CEO resigns
- Two incumbent directors resign
Clear Channel Com.
17,188
Highfields
Capital
Yahoo
16,977
Icahn
July 2008
Transocean
Biogen Idec
16,052
14,705
Icahn
Icahn
Nov 2013
Mar 2010
Chesapeake Energy
14,695
Icahn
July 2012
Adobe Systems
Genzyme
Sara Lee
Intuit
13,460
13,042
12,782
9586
ValueAct
Icahn
ValueAct
Relational
Dec 2012
June 2010
Aug 2008
Oct 2009
ITT Corp
Ingersoll Rand
Forest Laboratories
Kerr McGee
9567
9512
9241
8766
Relational
Trian
Icahn
Jana Partners
Jan 2012
Aug 2012
June 2013
April 2005
Sun Microsystems
8182
Sovereign Bancorp
7784
Southeastern Dec 2008
Asset
Mgmt.
Relational
Mar 2006
Nuance Communications
7767
Icahn
Oct 2013
Main changes obtained by activists
-
One new director added who is affiliated with the activist
One new director added who is affiliated with the activist
One new director added who is affiliated with the activist
Two new directors, both not identified as favored by activist
Increase of share buyback activity
Intention to explore strategic alternatives
One new director added who is affiliated with the activist
One new director added who is affiliated with the activist
One incumbent director resigns
Two new directors, one affiliated with the activist, one favored by the
activist but unaffiliated
One new director added who is affiliated with the activist
One new director added who is affiliated with the activist
Two new directors added, one affiliated with the activist, one not identified
to be favored by the activist
Six incumbent directors resign
Four new directors are added, three affiliated with the activist, one favored
by the activist but unaffiliated
Three incumbent directors resign
Three new directors, all favored by the activist but unaffiliated
Two new directors added, one favored by the activist but unaffiliated
Establishment of board committee to evaluate strategic alternatives
One incumbent director resigns
- Two new directors are added, one favored by the activist but unaffiliated,
one not identified to be favored by the activist
- One incumbent director resigns
- Three new directors are added, one affiliated with the activist, two favored
by the activist but unaffiliated
- Two new directors are added, both affiliated with the activist
- One incumbent director resigns
- Two new directors added, one favored by the activist but unaffiliated, one
not identified to be favored by the activist
- Five incumbent directors resign
- Five new directors are added, one affiliated with the activist, four not
identified to be favored by the activist
- One new director added who is affiliated with the activist
- Two new directors, who are both favored by the activist but unaffiliated
- One new director added who is affiliated with the activist
- One incumbent director resigns
- One new director added who is affiliated with the activist
- The CEO resigns
- One new director added who is affiliated with the activist
- One new director added who is affiliated with the activist
- One new director added who is affiliated with the activist
- Divestiture of part of the firm’s assets
- Increase in share buyback volume
- Two new directors, who are both favored by the activist but unaffiliated
- Two new directors are added, one affiliated with the activist, one favored
by the activist but unaffiliated
- Two new directors are added, both affiliated with the activist
36
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
A.2. Activists with four or more settlements
Table A.2
This table reports the names of activists who reached four or more settlements for interventions launched between 20 0 0 and 2013. Number of settlements
refers to the number of settlements the activist reached during the sample period. Number of Interventions refers to the number of interventions launched
by the activist during the sample period. % Settlements refers to the percentage of interventions that resulted in a settlement agreement.
Activist name
Carl Icahn
Ramius Capital Group
ValueAct
Barington Capital Group
SRB Management / Becker Drapkin
Relational Investors
Clinton Group
Steel Partners
Financial Edge Fund / PL Capital
Elliott Associates
Starboard Value
SACC Partners
Crescendo Partners
Red Oak Partners
Stillwell Partners
Third Point
Jana Partners
MMI Investments
Pershing Square
Pirate Capital
Trian Partners
Bulldog Capital Management
Seth Hamot
Lawrence Seidman
Oliver Press Partners
Raging Capital Management
RMCP GP
Sandell Asset Management
Southeastern Capital Management
Wynnefield Capital
Number of
settlements
Number of
interventions
20
14
13
12
12
12
11
10
9
8
8
7
6
6
6
6
5
5
5
5
5
4
4
4
4
4
4
4
4
4
61
31
74
26
19
28
31
50
29
29
36
32
12
11
40
40
30
24
16
19
12
17
12
27
7
11
12
15
15
36
% Settlements
33
45
18
46
63
43
35
20
31
28
22
22
50
55
15
15
17
21
31
26
42
24
33
15
57
36
33
27
27
11
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
37
A.3. Variable definitions
Table A.3
Settlement characteristics
High board turnover is an indicator variable equal to one if the settlement agreement stipulates that at least three new directors join the board or at least three
incumbent directors leave the board and equal to zero otherwise.
Strategic transaction is an indicator variable equal to one if the target firm merges or sells itself or sells part of its assets and equal to zero otherwise.
CEO departure is equal to one if the settlement agreement stipulates that the CEO will leave the company and equal to zero otherwise.
High-impact settlement is an indicator variable equal to one if “High board turnover” is equal to one or “CEO departure” is equal to one or “Strategic transaction” is equal
to one and equal to zero otherwise.
Buy & hold industry-adjusted return is the continuously compounded industry-adjusted share price performance around the date the settlement agreement is reached.
Identifiers for time relative to initial activist intervention, proxy fight announcement, or settlement
13D is equal to one for hedge fund activism target firms and equal to zero otherwise.
Settlement is equal to one for target firms that settle with the activist and equal to zero otherwise.
Voted contest is equal to one for target firms that experience a contested proxy vote and equal to zero otherwise.
d[t ± k] with −5 ≤ k ≤ 5 is a set of indicator variables equal to one in year k relative to the initial intervention year or placebo year and equal to zero otherwise.
d[s ± k] with −5 ≤ k ≤ 5 is a set of indicator variables equal to one in year k relative to the settlement year or placebo year and equal to zero otherwise.
d[c ± k] with −5 ≤ k ≤ 5 is a set of indicator variables equal to one in year k relative to the proxy fight announcement year (for voted proxy contests) or placebo year and
equal to zero otherwise.
d[POSTD] is equal to zero in years t-1 through t-3 around the initial activist intervention or placebo year and equal to one in years t through t + 2.
d[POSTS] is equal to zero in years s-1 through s-3 around the settlement year or placebo year and equal to one in years s through s + 2.
d[POSTC] is equal to zero in years c-1 through c-3 around the proxy fight announcement year (for voted proxy contests) or placebo year and equal to one in years c
through c + 2.
CEO, board, and director characteristics
CEO age ≥ 62 is equal to one if the CEO is older than 62 and equal to zero otherwise.
# Director additions is the number of new directors on the board.
# Director departures is the number of directors that leave the board.
Favored and affiliated director is equal to one if a new director is affiliated with the activist and equal to zero otherwise.
Favored but unaffiliated director is equal to one if a new director is favored by but unaffiliated with the activist and equal to zero otherwise.
Other new director is equal to one if a new director is neither desired by nor affiliated with the activist.
Favored and affiliated director added following settlement is equal to one for activist-affiliated directors who were added to the board following a settlement and equal to
zero otherwise.
Favored but unaffiliated director added following settlement is equal to one for activist-desired but unaffiliated directors who were added to the board following a
settlement and equal to zero otherwise.
Other new director added following settlement is equal to one for directors who were added to the board following a settlement but are neither activist-affiliated nor
activist-desired and equal to zero otherwise.
Age ≥ 70 is an indicator variable equal to one if the director is older than 70 and equal to zero otherwise.
# Directors older than 70 is the number of directors who are older than 70.
# Outside directors older than 70 is the number of outside directors who are older than 70.
Connectedness is the number of other boards the director served on over the past five years.
Tenure is the number of years the director has been on the board of the company.
Directors’ average # directorships is the average number of directorships held by incumbent board members.
Outside directors’ avg # directorships is the average number of directorships held by incumbent outside directors.
Board chair up for election is equal to one if the board chair is up for election at the next shareholder meeting.
CEO up for election is equal to one if the CEO is up for election at the next shareholder meeting.
Fraction of votes supporting director is the proportion of votes at the annual meeting supporting the director.
Fraction of votes against director is the proportion of votes at the annual general meeting against the director.
Fraction of votes withheld is the proportion of votes at the annual general meeting withheld in a director election.
Activism campaign characteristics
# Campaigns is the number of campaigns launched by the activist in the past five years.
# Settlements is the number of settlements reached by the activist in the past five years.
# Successful campaigns is the number of campaigns launched by the activist in the past five years during which the activist achieved its campaign goals.
Market reaction past campaigns is the average buy-and-hold market-adjusted return over a seven-day window centered on the campaign announcement date for all
campaigns launched by the activist in the past five years.
# Proxy fights is the number of proxy fights launched by the activist in the past five years.
Proxy fight is an indicator variable equal to one if the activist has threatened a proxy fight and equal to zero otherwise.
High activist ownership is an indicator variable equal to one for campaigns with activist ownership above the sample median and equal to zero otherwise.
Company characteristics and outcomes
Market cap is the target firm’s market capitalization in millions.
Multiple share classes is equal to one if the target firm’s share structure includes multiple classes.
Analyst coverage is the number of sell-side analysts covering the firm.
ROA is earnings before interest, taxes, and depreciation over lagged total assets.
Tobin’s Q is the sum of the book value of debt and the market value of equity scaled by the sum of the book value of debt and the book value of equity.
Book / Market is the book value of equity scaled by the market value of equity.
Insider ownership is the percentage of shares held by incumbent executives and directors.
Cash flow volatility is the standard deviation of operating cash flows over the past five years.
Abnormal returns pre-13D is the market-adjusted share price performance over the 12 months leading up to the 13(d) filing.
Stock returns is equal to the continuously compounded stock return over the calendar year.
Delisting is an indicator variable equal to one the year of a stock market delisting and equal to zero otherwise.
Acquisition / Going private delisting is an indicator variable equal to one in the year of a stock market delisting that is related to merger and acquisition or going private
transactions and equal to zero otherwise.
Distress delisting is an indicator variable equal to one in the year of a stock market delisting that is related to financial distress based on the CRSP delisting codes and
equal to zero otherwise.
Payout yield is the sum of dividends and repurchases divided by market capitalization.
CEO turnover is an indicator variable equal to one if the CEO leaves the firms and equal to zero otherwise.
38
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
A.4. Covariate balance
Table A.4
This table presents descriptive statistics on selected firm size and firm performance for target firms that reached a settlement (Panel A), target firms that
experiences a proxy fight that went to a vote (Panel B), all other target firms (Panel C), as well as the matched control groups of these three types of target
firms. The firm characteristics are measured prior to the year of settlement (Panel A), prior to the year of proxy fight announcement (Panel B), or prior
to the year of the initial intervention (Panel C), respectively. p-values for differences in means between treated and control firms evaluates whether the
matching procedure was successful in selecting control firms that are similar to target firms on observable firm characteristics. All variables are defined in
Appendix A.3.
Panel A: Propensity score matching on the year of the settlement (for settled campaigns)
Target firms (n = 285)
Market cap t − 1
ROA t − 1
ROA t − 2
ROA t − 3
Tobin’s Q t − 1
Tobin’s Q t − 2
Tobin’s Q t − 3
Control firms (n = 1382)
Mean
Median
SD
Mean
Median
SD
p-value for difference in means
2750.061
0.056
0.074
0.081
1.645
1.859
1.898
294.960
0.073
0.096
0.093
1.270
1.364
1.491
6791.195
0.159
0.168
0.172
1.330
1.676
1.658
3209.980
0.072
0.087
0.095
1.691
1.926
2.046
454.699
0.100
0.107
0.109
1.362
1.440
1.552
7072.831
0.182
0.177
0.179
1.401
1.947
1.986
0.482
0.181
0.269
0.260
0.576
0.564
0.171
Panel B: Propensity score matching on the year of proxy contest announcement (for voted contests)
Target firms (n = 106)
Market cap t − 1
ROA t − 1
ROA t − 2
ROA t − 3
Tobin’s Q t − 1
Tobin’s Q t − 2
Tobin’s Q t − 3
Control firms (n = 509)
Mean
Median
SD
Mean
Median
SD
p-value for difference in means
1555.371
0.029
0.058
0.066
1.941
2.078
2.440
394.575
0.064
0.074
0.081
1.361
1.561
1.474
4093.265
0.189
0.192
0.168
2.061
2.116
2.962
1667.273
0.045
0.058
0.068
1.995
2.379
2.679
317.185
0.096
0.098
0.098
1.550
1.563
1.626
4274.404
0.227
0.228
0.211
1.543
2.740
3.211
0.380
0.508
0.996
0.920
0.846
0.302
0.484
Panel C: Propensity score matching on the year of the initial activist intervention (for other campaigns)
Target firms (n = 1461)
Market cap t − 1
ROA t − 1
ROA t − 2
ROA t − 3
Tobin’s Q t − 1
Tobin’s Q t − 2
Tobin’s Q t − 3
Control firms (n = 6996)
Mean
Median
SD
Mean
Median
SD
p-value for difference in means
1282.688
0.064
0.076
0.083
1.866
1.994
2.030
198.613
0.084
0.091
0.093
1.339
1.426
1.466
3986.801
0.188
0.183
0.179
1.860
1.949
1.864
1539.532
0.072
0.078
0.083
1.924
2.046
2.073
240.943
0.098
0.100
0.102
1.427
1.467
1.506
4303.394
0.189
0.186
0.185
1.726
1.992
1.869
0.395
0.261
0.699
0.943
0.344
0.399
0.472
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
A.5. Example of a settlement agreement
The text below provides the central parts of the settlement agreement between Charming Shoppes, Inc. and
Crescendo Partners, dated as of May 8, 2008 (The entire
text with more details of the agreement is available online at: https://www.sec.gov/Archives/edgar/data/19353/
0 0 0 0921895080 01394/ex991to13da407148002_051208.
htm).
This Settlement Agreement, dated as of May 8, 2008
(the “Agreement”), is by and among Charming Shoppes,
Inc., a Pennsylvania corporation (the “Company”), and the
other parties signatory hereto (collectively, the “Committee,” and individually, a “member of the Committee”).
WHEREAS, the Committee beneficially owns (as defined
below) shares of Common Stock, $0.10 par value, of the
Company (the “Common Stock”) as specified on Schedule A
of this Agreement (the “Shares”);
WHEREAS, prior to the date hereof the Committee
(i) delivered a letter (the “Nomination Letter”) to the Company, dated as of January 14, 2008, stating its intention
to nominate (the “Committee Nomination”) three individuals for election to the Board of Directors of the Company (the “Board”) by the shareholders of the Company
(the “Shareholders”) and (ii) filed a definitive proxy statement on Schedule 14A with the Securities and Exchange
Commission (the “SEC”) related to the matters set forth in
the Nomination Letter;
WHEREAS, the Company and the Committee have
agreed that it is in their mutual interests to enter into
this Agreement, which, among other things, terminates the
pending proxy contest for the election of directors at the
2008 Annual Meeting (as defined below);
WHEREAS, the Company has agreed that the size of
the Board will be increased from eight to eleven members as permitted by the Company’s Articles of Incorporation, such increase to be effective as of the 2008 Annual
Meeting;
WHEREAS, the Company has agreed that, in connection
with the Company’s 2008 Annual Meeting of Shareholders
(including any adjournment or postponement thereof in
accordance with this Agreement, the “2008 Annual Meeting”), the Board will include in its nominations for election
as members of the Board, and recommend that the shareholders vote to elect as directors of the Company, Michael
Appel and Arnaud Ajdler (each, a “Committee Nominee”);
and
WHEREAS, the Company has agreed to submit a proposal at the 2008 Annual Meeting for the declassification
of the Company’s Board.
NOW, THEREFORE, in consideration of the mutual
covenants and agreements contained herein, and for
other good and valuable consideration, the receipt and
sufficiency of which is hereby acknowledged, the parties
hereto agree as follows:
Board of Directors, Annual Meeting and Related Matters.
(a) Board Expansion. As promptly as practicable following the date of this Agreement, the Company shall
increase the size of the Board from eight to eleven
directors, such increase to be effective as of the date
39
hereafter that the Company’s Proxy Statement and
proxy card are first sent to shareholders.
(b) 2008 Annual Meeting. The Company shall adjourn the
2008 Annual Meeting until June 26, 2008 for purposes of the election of directors and the declassification of the Board as contemplated herein.
(c) Nomination of New Directors. The Company agrees
that at the 2008 Annual Meeting, the Board will:
(1) nominate each of Michael Appel, Arnaud Ajdler, Dorrit J. Bern and Alan Rosskamm for election as a director of the Company at the 2008 Annual Meeting
to serve as Class C directors with terms scheduled to
end in 2011;
(2) nominate each of Michael Goldstein and Richard W.
Bennet III for election as directors of the Company
at the 2008 Annual Meeting to serve as Class B directors with terms schedule to end in 2010; and
(3) cause all proxies received by the Company to be
voted in the manner specified by such proxies.
(d) Board Declassification. In accordance with the Company’s Restated Articles of Incorporation, Amended and
Restated Bylaws and applicable state law, the Company
will submit, recommend and actively solicit proxies in
favor of a resolution for approval by its shareholders at
the 2008 Annual Meeting to declassify the Company’s
Board to provide for the annual election of all directors (the “Declassification Proposal”). The Company will
seek to have such Declassification Proposal classified
as a “routine matter” under New York Stock Exchange
rules. Under such proposal, if approved by the Company’s shareholders, the first of such annual elections
would take place at the Company’s 2009 Annual Meeting. The members of the Board will vote all of their
shares in favor of the Declassification Proposal.
(e) Proxy Solicitation Materials. The Company and the Board
agree that the Company’s Proxy Statement and proxy
cards for the 2008 Annual Meeting and all other solicitation materials to be delivered to shareholders in
connection with the 2008 Annual Meeting (in each
case excepting any materials delivered prior to the
date hereof) shall be prepared in accordance with, and
in furtherance of, this Agreement. The Company will
provide the Committee with copies of any portion of
proxy materials or other solicitation materials that contain statements relating to the Committee, the Committee Nominees or this Agreement a reasonable period in advance of filing such materials with the SEC
or disseminating the same in order to permit the Committee a reasonable opportunity to review and comment on such materials. The Committee will provide,
as promptly as reasonably practicable, all information
relating to the Committee Nominees (and other information, if any) to the extent required under applicable law to be included in the Company’s Proxy Statement and any other solicitation materials to be delivered to shareholders in connection with the 2008 Annual Meeting.
(f) Committees. At the first meeting of the Board following
the 2008 Annual Meeting, the Company shall cause at
40
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
least one Committee Nominee, such Committee Nominee to be selected by the Company, to be a member
of each committee of the Board and each committee of
the Board which is created after the date of this Agreement.
(g) Expenses. Within fifteen business days from the date
of this Agreement, the Company shall reimburse the
Committee an amount equal to the Committee’s actual
out-of-pocket expenses incurred in connection with
the Committee Nomination (the Committee shall provide reasonable documentation with respect to such
expenses), including the preparation of related filings
with the SEC, the fees and disbursements of counsel and other advisors, and expenses incurred in connection with the litigation between the Company and
the Committee, up to a maximum reimbursement of
$10 0 0,0 0 0, and the Committee hereby agrees that such
payment shall be in full satisfaction of any claims or
rights it may have as of the date hereof for reimbursement of fees, expenses or costs in connection with the
Committee Nomination.
Voting Provisions. The Committee, together with its Affiliates, will cause all shares of Common Stock for which they
have the right to vote as of the record date for the 2008
Annual Meeting to be present for quorum purposes and to
be voted at such meeting or at any adjournments or postponements thereof, (a) in favor of each director nominated
and recommended by the Board for election at such meeting, (b) in favor of the Declassification Proposal and each
other matter recommended by the Board at such meeting,
and (c) against any shareholder nominations for director
which are not approved and recommended by the Board
for election at such meeting.
Additional Undertakings by the Committee. By executing
this Agreement and in consideration of the agreements
contained herein, the Committee hereby irrevocably withdraws its Nomination Letter and any nominations to the
Board made prior to the date hereof and agrees to terminate the pending proxy contest with respect to the election
of directors at the 2008 Annual Meeting. Within two business days of the date of this Agreement, the Committee
shall file, or cause to be filed on its behalf, with the SEC an
amendment to its Schedule 13D with respect to the Company disclosing the material contents of this Agreement.
Publicity. Promptly after the execution of this Agreement, the Company and the Committee will issue a press
release in the form attached hereto as Schedule B. Any
press release to be issued by the Committee relating to the
matter covered by this Agreement shall be provided prior
to issuance to the Company for the Company’s review and
approval, such approval not to be unreasonably withheld.
Dismissal of Claims. Promptly after the execution of this
Agreement, the Company and the Committee shall stipulate to the voluntary dismissal with prejudice and without costs the action entitled Charming Shoppes v. Crescendo
Partners II, L.P., et al., No. 08-CV-1156-AB, pending in the
United States District Court for the Eastern District of
Pennsylvania, and file with the Court a stipulation of dismissal.
Mutual Releases. Promptly after the execution of this
Agreement, the Company and the Committee shall provide
mutual releases in the forms annexed hereto as Schedule
C.
References
Aghion, P., Bolton, P., 1992. An incomplete contracts approach to financial
contracting. Rev. Financ. Stud. 59, 473–494.
Appel, I.R., Gormley, T.A., Keim, D.B., 2019. Standing on the shoulders of
giants: the effect of passive investors on activism. Rev. Financ. Stud.
32, 2720–2774.
Aslan, H., Kumar, P., 2016. The product market effects of hedge fund activism. J. Financ. Econ. 119, 226–248.
Back, K., Collin-Dufresne, P., Fos, V., Li, T., Ljungqvist, A., 2018. Activism,
strategic trading, and liquidity. Econometrica 86, 1431–1463.
Bebchuk, L.A., 1984. Litigation and settlement under imperfect information. RAND J. Econ. 15 (3), 404–415.
Bebchuk, L.A., Brav, A., Jackson, R.J., Jiang, W., 2013. Pre-disclosure accumulations by activist investors: evidence and policy. J. Corp. Law 39,
1–34.
Bebchuk, L., Brav, A., Jiang, W., et al., 2015. The long-term effects of hedge
fund activism. Columbia Law Rev. 115, 1085–1156.
Bebchuk, L., Klement, A., 2012. Negative-expected-value suits. In:
Sanchirico, C. (Ed.), Procedural Law and Economics. Edward Elgar Publishing, Cheltenham, pp. 341–349.
Becht, M., Franks, J.R., Grant, J., Wagner, H.F., 2017. The returns to
hedge fund activism: an international study. Rev. Financ. Stud. 30,
2933–2971.
Becht, M., Franks, J., Mayer, C., Rossi, S., 2009. Returns to shareholder activism: evidence from a clinical study of the Hermes UK focus fund.
Rev. Financ. Stud. 22, 3093–3129.
Bourveau, T., Schoenfeld, J., 2017. Shareholder activism and voluntary disclosure. Rev. Account. Stud. 22, 1307–1339.
Boyson, N.M., Gantchev, N., Shivdasani, A., 2017. Activism mergers. J. Financ. Econ. 126, 54–73.
Boyson, N.M., Ma, L., Mooradian, R.M., 2016. Serial activists. Unpublished
working paper. Northeastern University.
Boyson, N.M., Mooradian, R.M., 2011. Corporate governance and hedge
fund activism. Rev. Deriv. Res. 14, 169–204.
Boyson, N.M., Pichler, P., 2019. Hostile resistance to hedge fund activism.
Rev. Financ. Stud. 32, 771–817.
Brav, A., Jiang, W., Kim, H., 2015a. The real effects of hedge fund activism.
Rev. Financ. Stud. 28, 2723–2769.
Brav, A., Jiang, W., Kim, H., 2015b. Recent advances in research on hedge
fund activism: value creation and identification. Annu. Rev. Financ.
Econ. 7, 579–595.
Brav, A., Jiang, W., Ma, S., Tian, X., 2018. How does hedge fund activism
reshape corporate innovation? J. Financ. Econ. 130, 237–264.
Brav, A., Jiang, W., Partnoy, F., Thomas, R., 2008. Hedge fund activism, corporate governance, and firm performance. J. Financ. 63, 1729–1775.
Burkart, M., Lee, S., 2015. Hedge fund activism vs. hostile takeover bids.
Unpublished working paper. NYU Stern School of Business and Stockholm School of Economics.
Cheng, C.A., Huang, H.H., Li, Y., Stanfield, J., 2012. The effect of hedge fund
activism on corporate tax avoidance. Account. Rev. 87, 1493–1526.
Cheng, C.S., Huang, H.H., Li, Y., 2015. Hedge fund intervention and accounting conservatism. Contemp. Account. Res. 32, 392–421.
Clifford, C.P., 2008. Value creation or destruction? Hedge funds as shareholder activists. J. Corp. Financ. 14, 323–336.
Coffee, J., Jackson, R.J., Mitts, J.R., Bishop, R.E., 2019. Activist directors and
agency costs: what happens when an activist director goes on the
board? Cornell Law Rev. 104, 381–466.
Collin-Dufresne, P., Fos, V., 2015. Do prices reveal the presence of informed trading? J. Financ. 70, 1555–1582.
Core, J.E., Holthausen, R.W., Larcker, D.F., 1999. Corporate governance,
chief executive officer compensation, and firm performance. J. Financ.
Econ. 51, 371–406.
Corum, A.A., Levit, D., 2019. Corporate control activism. J. Financ. Econ.
133 (1), 1–17.
Corum, A.A., 2018. Activist Settlements. Cornell University Unpublished
working paper.
Daugherty, A.J., Reinganum, J.F., 2017. Settlement and trial: selected
analyses of the bargaining environment. In: Parisi, F. (Ed.), Oxford
Handbook of Law and Economics. Oxford University Press, Oxford,
pp. 229–246.
Edmans, A., Fang, V.W., Zur, E., 2013. The effect of liquidity on governance.
Rev. Financ. Stud. 26, 1443–1482.
L.A. Bebchuk, A. Brav and W. Jiang et al. / Journal of Financial Economics 137 (2020) 1–41
Falato, A., Kadyrzhanova, D., Lel, U., 2014. Distracted directors: does
board busyness hurt shareholder value? J. Financ. Econ. 113,
404–426.
Fich, E.M., Shivdasani, A., 2006. Are busy boards effective monitors? J. Financ. 61, 689–724.
Fos, V., 2017. The disciplinary effects of proxy contests. Manag. Sci. 63,
655–671.
Fos, V., Jiang, W., 2016. Out-of-the-money CEOs: private control premium
and option exercises. Rev. Financ. Stud. 29, 1549–1585.
Fos, V., Tsoutsoura, M., 2014. Shareholder democracy in play: career consequences of proxy contests. J. Financ. Econ. 114, 316–340.
Gantchev, N., 2013. The costs of shareholder activism: evidence from a
sequential decision model. J. Financ. Econ. 107, 610–631.
Gantchev, N., Gredil, O.R., Jotikasthira, C., 2019a. Governance under the
gun: spillover effects of hedge fund activism. Rev. Financ. 23 (6),
1031–1068.
Gantchev, N., Jotikasthira, C., 2017. Institutional trading and hedge fund
activism. Manag. Sci. 64, 2930–2950.
Gantchev, N., Sevilir, M., Shivdasani, A., 2019b. Activism and empire building. J. Financ. Econ. forthcoming.
Gow, I.D., Shin, S.P., Srinivasan, S., 2014. Activist directors: determinants and consequences. Unpublished working paper. Harvard Business School.
Gow, I.D., Shin, S.P., Srinivasan, S., 2016. Consequences to directors of
shareholder activism. Unpublished working paper. Harvard Business
School.
Greenwood, R., Schor, M., 2009. Investor activism and takeovers. J. Financ.
Econ. 92, 362–375.
Guo, J., Utham, V., Wang, G.J., 2019. Shepherds in wolves’ clothing? Hedge
fund activism using corporate divestitures. Unpublished working paper. Durham University Business School.
Hart, O., 2017. Incomplete contracts and control. American Economic Review 107, 1731–1752.
He, Y.E., Li, T., 2018. The benefits of friendship in hedge fund activism.
Unpublished working paper. University of Warwick.
Hege, U., Zhang, Y., 2019. Activism pressure and the market for corporate
assets. Unpublished working paper. University of Toulouse-Capitole.
Huang, S., Hilary, G., 2018. Zombie board: board tenure and firm performance. J. Account. Res. 56, 1285–1329.
Jenter, D., Lewellen, K., 2015. CEO preferences and acquisitions. J. Financ.
70, 2813–2852.
Jiang, W., Li, T., Mei, D., 2018. Influencing control: jawboning in risk arbitrage. J. Financ. 73, 2635–2675.
Jiang, W., Wan, H., Hao, S., 2016. Reputation concerns of independent directors: evidence from individual director voting. Rev. Financ. Stud.
29, 655–696.
Johnson, T., Swem, N., 2017. Reputation and investor activism. Unpublished working paper. The University of Texas at Austin and Board of
Governors of the Federal Reserve System.
Kaplan, S.N., Minton, B.A., 2012. How has CEO turnover changed? Int. Rev.
Financ. 12, 57–87.
41
Keusch, T., 2020. Activist involvement in CEO recruiting. Working paper.
Kim, S., 2018. Hedge fund activism and internal capital markets. Unpublished working paper. University of Florida.
Klein, A., Zur, E., 2009. Entrepreneurial shareholder activism: hedge funds
and other private investors. J. Financ. 64, 187–229.
Klein, A., Zur, E., 2011. The impact of hedge fund activism on the target
firm’s existing bondholders. Rev. Financ. Stud. 24, 1735–1771.
Krishnan, C.N.V., Partnoy, F., Thomas, R., 2016. The second wave of hedge
fund activism: the importance of reputation, clout, and expertise. J.
Corp. Financ. 40, 296–314.
Larcker, D.F., So, E.C., Wang, C.C.Y., 2013. Boardroom centrality and firm
performance. J. Account. Econ. 55, 225–250.
Levit, D., 2018. Soft shareholder activism. Rev. Financ. Stud. 32,
2775–2808.
McCahery, J.A., Sautner, Z., Starks, L.T. , 2016. Behind the scenes: the corporate governance preferences of institutional investors. J. Financ. 71,
2905–2932.
Mietzner, M., Schweizer, D., 2014. Hedge funds versus private equity funds
as shareholder activists in Germany—differences in value creation. J.
Econ. Financ. 38, 181–208.
Minton, B.A., Schrand, C., 1999. The impact of cash flow volatility on discretionary investment and the costs of debt and equity financing. J.
Financ. Econ. 54, 423–460.
Nili, Y., 2016. The “new insiders”: rethinking independent directors’
tenure. 68 Hastings Law J. 67, 97–158.
Norli, Ø., Ostergaard, C., Schindele, I., 2015. Liquidity and shareholder activism. Rev. Financ. Stud. 28, 486–520.
Prescott, J.J., Spier, K.E., 2016. A Comprehensive Theory of Civil Settlement.
91 New York University Law Review, p. 59.
Schoenfeld, J., 2018. Action-based contracts between firms and shareholders. Unpublished working paper. Dartmouth College.
Shavell, S., 1982. Suit, settlement, and trial: a theoretical analysis under
alternative methods for the allocation of legal costs. J. Legal Stud. 11,
55–82.
Spier, K., 2007. Litigation. In: Polinsky, M., Shavell, S. (Eds.), The Handbook
of Law and Economics. North Holland, Amsterdam, pp. 259–342.
Stuart, T.E., Yim, S., 2010. Board interlocks and the propensity to be targeted in private equity transactions. J. Financ. Econ. 97, 174–189.
Sunder, J., Sunder, S.V., Wongsunwai, W., 2014. Debtholder responses to
shareholder activism: evidence from hedge fund interventions. Rev.
Financ. Stud. 27, 3318–3342.
Swanson, E., Young, G., 2017. Are activist investors good or bad for business? Evidence from capital market prices, informed traders, and firm
fundamentals. Unpublished working paper. Texas A&M University.
Vega, C., 2006. Stock price reaction to public and private information. J.
Financ. Econ. 82, 103–133.
Wickelgren, A.L., 2013. Law and economics of settlement. In: Arlen, J.
(Ed.), Research Handbook on the Economics of Tort Law. Edward Elgar
Publishing, Cheltenham, pp. 330–359.
Zhang, X.F., 2006. Information uncertainty and stock returns. J. Financ. 61,
105–137.
Oracles of the Vote:
Predicting the Outcomes of Proxy Contests
Scott Hirst,a Oğuzhan Karakaş,b and Ting Yuc
March 2023
[Preliminary Draft – Please Do Not Distribute or Cite without Authors’ Permission]
Abstract: This paper examines proxy contests and the value of shareholder voting rights (i.e., the
voting premium) estimated using option prices. Our sample consists of 873 proxy contests for
board seats at U.S. publicly listed firms from 1994 to 2017. We find that voting premium can help
predict the outcomes of the proxy contests. Specifically, increased voting premiums around contest
announcements is associated with higher likelihood of the contest being subsequently settled or
going to a vote, and to a lower likelihood of the contest being withdrawn. Further, the likelihood
of the dissidents being elected if the contest goes to a vote increases with the voting premium
around the record date.
JEL classification: G13, G30, G34, K22
Keywords: Proxy Contests; Market for Corporate Control; Voting Premium; Value of Voting
Rights; Corporate Governance
a
b
c
∗
Boston University School of Law, 765 Commonwealth Avenue, Boston, MA 02215, USA. Phone: +1 (617) 353
5753. Email: hirst@bu.edu.
Centre for Endowment Asset Management, Cambridge Judge Business School, University of Cambridge,
Trumpington Street, Cambridge CB2 1AG, UK. Phone: +44 (0)1223 766449. Email: o.karakas@jbs.cam.ac.uk.
Cambridge Judge Business School, University of Cambridge, Trumpington Street, Cambridge CB2 1AG, UK.
Email: ty302@jbs.cam.ac.uk.
We are grateful for helpful feedback from John Matsusaka, as well as seminar participants at Corporate Finance
Webinar, and CERF Webinar. Karakaş gratefully acknowledges the financial support of the Cambridge Endowment
for Research in Finance (CERF) and the J M Keynes Fellowship. Jonas Bahceci, Garret Podolan, Harry Quinn,
Jaclyn Rothenberg, and Ji Xi provided excellent research assistance. All errors are ours.
Introduction
What is the role of the value of shareholder votes during the proxy contests? Does the
market predict the outcomes of proxy contests for board seats? In this paper, we seek to answer
these questions, applying an options-based approach to calculate the market value of shareholder
voting rights.
Proxy contests attract extensive interest from investors. With the advent of the poison pill,
proxy contests seeking board control became a critical element of a hostile takeover. Since that
time, proxy contests seeking minority positions on boards have also become an important tool for
hedge fund activists seeking to influence directors and executives (Fos (2017)).
Most corporate elections are uncontested—the only directors eligible for election are those
nominated by the incumbent directors, usually themselves. A proxy contest occurs when a
“dissident” shareholder nominates their own nominees for election to the board, who then compete
with the “management nominees” put forward by the incumbent board. In many cases, the
dissident withdraws their nominees, or the dissident and the directors reach a settlement, either
agreeing to implement some of the requests of the dissident, or appointing some dissident
nominees to the board of directors, or both. If the dissident does not withdraw their nominees, and
a settlement is not reached, the contest goes to a vote of the shareholders. As part of this process,
both the incumbents and the dissident will actively solicit “proxies” from shareholders. A proxy,
usually in the form of a “proxy card” distributed by the incumbents or the dissident, gives the party
soliciting it the right to vote on behalf of the shareholders, in the manner indicated by the
shareholder on the card. In the period covered by our study, the proxy cards distributed by each
party would give them the right to vote (on behalf of the shareholder) for some or all of that party’s
nominees, as the shareholder decided. Proxy contests can be categorized into those seeking at least
2
a majority of the board (“control contests”) and those seeking only a minority of the board (“noncontrol contests”). In general, proxy contests are sponsored by activist hedge funds, groups of
shareholders, former directors or executives, and other firms.
Proxy contests are generally considered to involve significant expense, both to the dissident
and to the company. Gantchev (2013) estimates the cost of a proxy fight to the dissident to be
$10.71 million. Rational dissidents can only be expected to outlay such significant expense if they
expect the proxy contest to have some effect. Consistent with this theory, studies by Dodd and
Warner (1983), Mulherin and Poulsen (1998), and Fos (2017) find positive and significant
abnormal returns around proxy contests for target firms.
Early work regarding proxy contests raised the important question of the extent to which
the share price increase observed around proxy contests relates to expected increases in the value
of the firm. Dodd and Warner (1983) observe that a portion of share price increase around proxy
contests is not permanent. They conjecture that these increases may therefore not be associated
with expected changes in the value of the company, but could instead be partially explained by a
temporary change in the market value of the voting rights of shares, echoing an earlier suggestion
by Manne (1962). Ghosh, Owers, and Rogers (1992) revisit this “value of the vote” hypothesis,
and find evidence consistent with increased demand for and value of votes during the proxy
contests.2
2
The value of vote is a function of the probability of going for a control contest and the expected benefits from the
contest (Zingales (1995)). A contest can be expected to go to a vote where both the dissident and the incumbents
expect to win the contest, as there will be no settlement or withdrawal. Proxy contests that are expected to go to a
vote can therefore be expected to be more contentious, and consequently the voting premiums will be higher. When
the incumbents face a credible threat of losing to the dissident, they can be expected to reach a settlement involving
concessions, in order to avoid the (potential) adverse effects of losing the proxy contest (Chen et al. (2020)). In this
case, the likelihood of the contest going to a vote is low, but the expected benefits of winning a contest could still
be high. In contrast, if the dissident does not believe they have a significant likelihood of winning the contest, the
probability of the contest going to a vote is close to zero. The dissident can be expected to withdraw the contest in
order to avoid the substantial costs of conducting an expensive proxy contest (Bhattacharya (1997)).
3
This paper sets out to examine the value of the vote hypothesis. We focus on how the value
of the vote behaves around proxy contests, and whether it predicts the conduct or outcome of a
proxy contest. Our sample consists of 873 proxy contests for board seats that occurred during the
period 1994-2017 at “target” firms that were publicly listed on U.S. exchanges. 3 Of the 873
contests, 159 proxy contests (18%) were withdrawn by the dissidents before the annual or special
meeting; 325 contests (37%) were settled between the incumbents and dissidents before the
meeting date; and 389 contests (45%) “went the distance” to a vote. Of those that went the distance,
vote results are available for 292 contests (33% of all contests).4 We find a positive and significant
share price performance around the proxy contest announcement in our sample, which is consistent
with previous studies, including Dodd and Warner (1983), Mulherin and Poulsen (1998), and Fos
(2017).5
We take as our main measure the value of the vote the voting premium—the premium of
the share price over the underlying value of the cash flow rights represented by a share. To isolate
the voting premium we use the method introduced by Kalay, Karakaş, and Pant (2014). Kalay,
Karakaş, and Pant calculate the value of a synthetic stock, constructed using options, which mimics
the cash flow rights from a share of stock, but which does not incorporate any voting rights. They
subtract the value of the synthetic stock from the actual stock price, which values both the voting
rights and cash flow rights, and calculate the difference as proportion of the stock price. We apply
this method to the voting premium of three separate subsamples of firms targeted by proxy contests:
3
4
5
We restrict our sample to proxy contests for board seats since they are more contentious compared to “issue-only”
contests, and can therefore be expected to involve larger voting premiums.
Our sample is comparable with Brav et al. (2020), where 17% of proxy contests are withdrawn, 45% are settled,
and 38% voted.
Interestingly, to the best of our knowledge, our study is the first in the literature to observe that target firms with
proxy contests that are subsequently withdrawn by the dissidents experience higher abnormal returns around the
contest announcement date than contests that are settled or that go to a vote.
4
those where proxy contests were withdrawn by dissidents; those where proxy contests were settled
between the dissidents and the incumbents; and those where proxy contests went to a vote of
shareholders.
Of the 873 contests in our sample, 314 contests have options available. Based on our
analysis of this subsample, we find a positive voting premium around the contest announcement
date for proxy contests. Analyzing the proxy contests by their outcomes, we observe a higher
voting premium for proxy contests that subsequently go to a vote, or are settled, compared to those
that are withdrawn before the shareholder meeting. These findings suggest that a portion of
positive stock market reaction around proxy contest announcement results from the change in the
shareholder voting premium, especially for contests that are later settled or that go to a vote.
We next conduct exploratory analysis to determine whether the outcomes of proxy contests
are predictable from the behavior of the voting premiums. Our findings suggest that it is possible
to predict whether a proxy contest will subsequently be withdrawn (as opposed to being settled or
going to a vote) from the voting premium around the contest announcement date. For proxy
contests that are withdrawn by dissidents, we do not observe an increase in the shareholder voting
premium around the contest announcement date. In contrast, there is an increase in the voting
premium for firms in contests that were subsequently settled or that went to a vote. Further, for
proxy contests that go to a vote, we are able to predict the level of success of the dissidents in the
proxy contest from the shareholder voting premium around the record date. 6 Specifically, the
proportion of board seats sought by the dissident that the dissident wins increases with increases
in the shareholder voting premium around the record date.
6
“The ‘record date’ refers to the date on which the record holders of securities entitled to vote at the meeting is
determined.” Rule 14a-1(h), 17 CFR § 240.14a-1.
5
Our paper contributes to the corporate governance literature on proxy contests in two main
ways. First, this is the first paper to predict the outcomes of the proxy contests, based on the
behavior of the voting premium. This finding holds important implications for proxy contests,
since dissidents and incumbents could benefit from observing the market reaction during the proxy
contests, such as by using it to re-assess their strategy and tactics during proxy contests. Second,
our paper is the first to systematically analyze the voting premium around the proxy contests. Our
findings provide initial evidence suggesting that the shareholder voting premium can explain the
share price performance around proxy contests. It also supports the hypothesis that the value of
voting rights is a meaningful component of target firms’ stock price reactions around proxy
contests.
The remainder of the paper is structured as follows: Part 1 outlines related literature and
sets out our empirical predictions. Part 2 introduces our data, describes our sample, and provides
summary statistics. Methodology and empirical results are presented in Part 3.
1. Institutional Details and Empirical Predictions
1.1 How Do Proxy Contests Work?
The board of directors serves to appoint executives, to advise them, and to monitor and
oversee their decision-making. When the shareholders are dissatisfied with incumbent directors’
decisions, or believe that the firm is being poorly managed, it is common practice to seek to replace
executives, and/or directors. The proxy contest is the corporate governance mechanism by which
a “dissident” shareholder aims to replace one or more directors.7 The public firms that are the
7
Proxy contests are governed by the general corporation laws of the company’s state of incorporation, and by the
federal proxy rules, which specify the information required to be disclosed by the parties soliciting proxies.
6
targets of proxy contests vary widely in their size and past performance. However, previous
research has demonstrated that firms that are undervalued or have poor share price performance
are more likely to be targeted (Duvall and Austin (1965) and Fos (2017)).8
During the period we consider, proxy contests were most commonly associated with
shareholder activism. As part of an activism campaign, the activist generally selects a potential
target company and acquires a stake in that company. If the stake is more than 5% of the target
firm, the activist must disclose it to the SEC on Schedule 13D. After acquiring a stake in the
company, the activist generally engages in correspondence or discussions with the target company
seeking changes in the company’s strategy, operations, management, and/or governance. If the
activist regards the engagement as unsuccessful, they may commence a proxy contest.
A proxy contest is commenced when the “dissident” shareholder informs the company that
they are nominating directors for election at a subsequent shareholder meeting. However, the
nomination does not mean that the proxy contest will necessarily go to a vote.9 The dissident is
entitled to withdraw its nominations at any time before the annual meeting.10 In the meantime, the
incumbents and dissidents may engage in dialogue. In some contests, the parties will reach a
settlement before either is required to file and distribute its proxy statement to investors.11 If there
Pound (1988) questions the efficiency of proxy contests, arguing that (i) proxy vote solicitation favors management,
due to inefficiencies in the system, (ii) institutional investors may support management, due to conflict-of -interest,
and (iii) dissidents may have to incur costs to signal that their bids are not ‘crank’ bids.
9
It is not required for the dissidents to disclose their nominations and/or proposals publicly, and the incumbents have
no obligation to announce the receipt of nominations and/or proposals to the general public prior to filing the proxy
statement.
10
The dissidents seldom give a formal reason for the withdrawal. But it can be reckoned that the dissidents withdraw
proxy solicitations when their chance of winning the contest is low (Chen et al. (2020)). Brav et al. (2020) also
confirm this proposition by studying the early votes of mutual funds.
11
The incumbents choose to make concessions to the dissident group when facing credible threats (Chen et al. (2020)
and Bebchuk et al. (2020)). Settled contests allow incumbents to avoid potential losses from losing a proxy contest
(Huang and Yen (1996)). Bebchuk et al. (2020) also show that settlements are more likely when incumbents’
reputation concerns are severer. Also, the dissident could avoid the costs of initiating an expensive proxy contest
(Bhattacharya (1997)).
8
7
is no settlement, the dissidents and incumbents are required to file and distribute to shareholders
their preliminary proxy statement and a definitive proxy statement (labelled DEFC14A).12
In most proxy contests, the dissident and incumbents undertake vigorous communications
with investors seeking to win their proxies.13 The term “solicitation” includes any request for a
proxy. Any activities that relate to a solicitation are required to be filed with the SEC.14 The
solicitation of proxies normally begins several months before the annual or special meeting, where
the dissident and incumbent parties start soliciting proxies from shareholders. The record date
determines the shares that are eligible to vote at the meeting--the shareholders as of the record date
are entitled to sign and return a form of proxy (referred to as “proxy card”, though it is now
generally completed electronically). Separate proxy cards are distributed by each party. Each party
employs a proxy solicitor that accumulates and tracks completed proxies. The proxies give the
party receiving the proxy the right to vote on behalf of the shareholder submitting the proxy at the
meeting of shareholders. The company is required to disclose the results of the election, and in
many contests there is a formal or informal announcement of the results at or shortly after the
meeting. For the number of board seats to be elected at the meeting, that same number of nominees
that receive the most votes will be elected. Generally, the party that obtains the most proxies will
have more of their nominees elected.15
The preliminary proxy statement with regard to contested solicitations required by SEC is labelled PREC14A on
the SEC’s EDGAR database. It includes information about the details of the meeting and proxies. The definitive
proxy statement is labelled DEFC14A in the SEC’s EDGAR database, and is required to be filed if the contest is
not settled or withdrawn before filing.
13
Parties in a proxy contest normally have a team that includes professional advisors consisting of executives,
directors, and nominees in target firms, public relations personnel, outside legal counsel, proxy solicitors, and
investment bankers (De Wied (2018)).
14
Rule 14a-1(l), 17 CFR § 240.14a-1. See Black (1990) for comments.
15
A small number of companies have “cumulative voting,” which allows shareholders to “cumulate” votes that could
otherwise only be used to vote for different nominees, such that they can cast votes equal to the number of shares
they hold, multiplied by the number of directors up for election. This feature ensures that shareholders with a
sufficiently-sized minority interest in a company can elect a minority of board seats. See Dodd and Warner (1983);
Aranow and Einhorn (1968, p. 331).
12
8
Figure 1 illustrates a typical timeline for a proxy contest in our sample.
[ ~Insert Figure 1 about here~ ]
1.2 The Characteristics and Efficiency of Proxy Contests
Proxy contests can be divided into three basic categories based on the goals of the dissident
and the proportion of the board that the dissident seeks to elect. Control contests are filed by
dissidents who seek control of the firm by voting out incumbents, and involve the dissident putting
forward nominations for at least a majority of the board of directors. Short-slate contests are proxy
contests where the dissident seeks to elect nominees that would constitute less than a majority of
the board, and so would not have control of the company. In an issue contest the dissident is
focused on making a particular change the corporation, and does so by making a proposal in their
proxy materials to amend the company’s bylaws or policies, without nominating any directors for
election.17 This study focuses on proxy contests for the election of directors, and thus includes both
control contests and short-slate contests, but not contests that are only issue contests.
The motivations of dissident investors to wage a proxy contest may be driven by several
different factors, depending on the type of dissident.18 Activist hedge funds have become the most
common sponsors of proxy contests (Fos (2017)).19 Activist hedge funds normally acquire a stake
in a potential target of between 5% and 10% (Brav, Jiang, and Kim (2010)). They then pressure
the board of directors to undertake certain changes by (explicitly or implicitly) threatening a proxy
Common topics of issue contests include opposition to mergers and acquisitions, or to executive compensation plans.
Hedge fund activists typically own less than 10% of the target company’s shares (Fos and Jiang (2016)). As a
consequence, the hedge fund activist will only be successful in a proxy contest if they attract the support of other
shareholders (Brav et al. (2020)).
19
See Christie (2019) for a detailed discussion of the role of activist hedge funds in seeking and securing board
representation in public firms, and Hancock (1992) for a review of proxy contests.
17
18
9
contest to elect their nominees to the board (Brav et al. (2008)). The prevalence of activist investors
generally using short-slate contests is consistent with the fact that the majority of proxy contests
are non-control contests (Fos (2017); Goodwin (2017)). Other types of dissidents include current
or former directors or executives that are in conflict with all or part of the incumbent board and
seek to increase their power by having their nominees elected to the board. A third important
category of dissidents in our sample (though less prevalent in recent times) are hostile acquirers
seeking control of the board approve the acquisition (Bertrand and Mullainathan (2003)). This
summary of the major types of investors that initiate proxy contests is also notable for the kinds of
investors that it leaves out. In particular, despite the large blocks of corporate stock that they hold,
large investment managers and pension funds rarely if ever initiate proxy contests (Bebchuk and
Hirst (2019)).
The benefits of proxy contests, and their effects on the value of the firms they target, have
been the subject of considerable debate. This debate seeks to explain the widely-observed evidence
that the announcement of a proxy contest—or of an activist engagement that may lead to a proxy
contest—is associated with an increase in the share price of the target company (e.g., Fos (2017);
Brav et al. (2008)). One view is that the firm value is likely to increase at the time of proxy contests
since proxy contests are corporate governance mechanisms that aim to replace managers
responsible for the underperformance of firms. There are two main reasons why we such an
increase could be observed. First, in a short-slate contest, even though the dissident would only
obtain a minority of board seats if successful, that may be sufficient for enhancing the performance
of the company. Second, proxy contests may have disciplinary effects on targeted firms even
where there are no dissidents elected. Specifically, Dodd and Warner (1983) show evidence of
positive and statistically significant share price performance around proxy contests regardless of
10
the proxy contest outcome based on a sample consisting of 96 proxy contests for board seats from
1962 to 1978. Moreover, Mulherin and Poulsen (1998) find that proxy contests create value for
target firms with a sample of 270 proxy contests for board seats between 1979 and 1994. Fos (2017)
also shows that the average abnormal returns of targets reach 6.5% around proxy contest
announcements using a dataset of all proxy contests from 1994 to 2012.
However, Dodd and Warner (1983) find that a portion of the share price increase around
proxy contests is not permanent in the first part of their 1962-1978 sample period. Specifically,
they observe that share price falls between the contest announcement and the announcement of the
outcome of the contest. They argue that this systematic phenomenon can be partially explained by
the ‘market value of a vote’ as suggested by Manne (1962). During a proxy contest, the market
value of voting rights will increase, and the price of shares will therefore also increase. According
to this hypothesis, the share price should fall when the voting rights of shares cease to be vital in
the contest. Dodd and Warner (1983) test this hypothesis by examining the abnormal returns
around the record dates for contests. They find statistically significant negative abnormal returns
around record date for contests where the record date follows the contest announcement, but
insignificant negative abnormal returns around record dates where the record date precedes or
coincides with the announcement of the contest itself.
Ghosh, Owers, and Rogers (1992) reexamine the value of the vote hypothesis by studying
proxy contests between 1974 and 1985. They include both issue contests and contests for board
seats. Specifically, they show that those types of contests are not only distinct in the level of control
sought, but also in the changes in voting premium. For contests for board seats, the dissidents’
vote solicitation can be expected to be more intense, due to the potential for control and
participation in firms’ future decision making, which could be expected to be reflected in increased
11
demand for the voting rights, and thus a higher voting premium. They find that the positive returns
around announcement and the negative returns around the record date are more significant for
contests for board seat than for issue contests. Their results are consistent with the value of the
vote hypothesis. However, their study design only allows indirect inferences regarding the value
of the vote. Our study directly uses voting premium data from 1996 to 2019 calculated based on
the method introduced by Kalay, Karakaş, and Pant (2014), and aims to test the hypothesis that
the voting premium of shares can explain part of the change in share prices around proxy contests.
1.3 Voting Premium
Voting is one of the core rights of shareholders in corporations. Shareholders exercise their
voting rights to elect directors, approve mergers, approve changes to the firm’s charter, and many
other matters. Previous work has estimated the value of shareholder voting rights to be negatively
associated with the level of investor protection across countries (Nenova (2003) and Dyck and
Zingales (2004)).21
This paper uses insights and methods developed by Kalay, Karakaş, and Pant (2014) to
estimate the voting premium of shares. Kalay, Karakaş, and Pant (2014) estimate the voting
premium as the difference in price between shares of a company—which provide both cash flow
rights and voting rights—and a synthetic share of the same company constructed using put and
call options—which provides the same cash flow rights as a share of the company, but not voting
rights—normalized by the stock price.22
21
22
See Levit, Malenko, and Maug (2021) for a discussion of the voting premium.
There exist three other methods in previous literature to measure the values of voting rights in stocks. First, the
difference between multiple classes of stocks with diversified voting rights is a proxy for the value of voting rights
(see, for example, Levy (1983), Zingales (1994, 1995), Rydqvist (1996), Nenova (2003)). Second, the premium of
selling controlling blocks is a measurement of voting premium (see Barclay and Holderness (1989) and Dyck and
Zingales (2004)). Third, Christoffersen et al. (2007) and Aggarwal, Saffi, and Sturgess (2015) use the incremental
cost of borrowing stock, which is the equity lending fee to measure the voting premium. An important advantage of
12
Analyzing U.S. public firms, Kalay, Karakaş, and Pant (2014) find that the voting premium
is positive (about 1.5% of the share price, in annualized terms) and varies over time at the firm
level. Their study analyses types of control events (shareholder meetings, hedge fund activism,
and mergers and acquisitions) and concludes that voting rights become more valuable as the firm
approaches key events affecting control of the company. First, the annualized voting premium
increases from 0.86% to 2.28% during special shareholder meetings. Second, the annualized
voting premium increases when hedge fund activism is announced (from 1.05% to 1.72%), with
engagements considered to be more hostile associated with higher voting premiums. Finally, the
announcement of merger and acquisition events results in an increase in the annualized voting
premium by 2.38% on average. The magnitude of the increase in voting premium is much higher
for contentious deals, particularly in contests that are likely to involve pivotal votes. This finding
is consistent with the hypothesis of Zingales (1995) that the value of a vote is “a function of the
probability that a vote is pivotal in a control contest and the magnitude of the private benefits
obtainable by controlling the company.”
Two other recent papers have applied the methodology in Kalay, Karakaş, and Pant (2014)
to determine voting premiums in other situations. Mohseni and Karakaş (2021) find that firms with
staggered boards have higher voting premium, which is consistent with the entrenchment view of
the staggered boards. Gurun and Karakaş (2021) find that negative earnings surprises lead to
increases in the voting premiums, consistent with the increased chances of control contests
following poor firm performance.
the Kalay, Karakaş, and Pant (2014) approach is that one could calculate the voting premium for a large sample of
firms at any point in time, which suffer less from selection biases. Please see further discussion of the pros and cons
of each method in Kalay, Karakaş, and Pant (2014).
13
1.4 Empirical Predictions
Based on this body of prior work, we predict that there is likely to be an increase in the
voting premium at the announcement of proxy contests, and around the record dates for contests
that are expected to go to a vote, and that the increase in the voting premium is likely to be strongest
for contentious proxy contests where votes are more likely to be pivotal.23 We predict that this will
explain a portion of the increase in the share price of target firms around the announcement of
proxy contests which has been reported in the prior literature and that we replicate in this paper.
We also expect a decrease in the voting premium after the conclusion of proxy contests, following
the shareholder meetings for those contests that go to a vote.
2. Data and Sample Description
2.1 Sample Description
Our sample consists of proxy contests at target companies that are publicly listed on the
New York Stock Exchange (NYSE), the American Stock Exchange (AMEX), or NASDAQ. Proxy
contest data from 2000 to 2017 are retrieved from the SharkRepellent dataset of FactSet Data
Systems. 24 Separately, we collected all preliminary and definitive proxy statements filed on
EDGAR from its inception in 1994 to 2017. We then manually added missing records, and
corrected missing or erroneous data, including contests that occurred before SharkRepellent began
The probability that market votes are pivotal is equal to the “Shapley value” of the market votes in a simple majority
game. This is a model of a shareholder meeting with a small number of major shareholders that own large blocks
of shares, and an infinite number of small shareholders with tiny shareholdings. The Shapley value of each
shareholder reflects the ability of the respective shareholder to affect the outcome of a vote (Rydqvist (1987)). If
one group controls 50% of the votes, then a marginal vote is unvalued. However, if the group could not gather 50%
of votes even if the marginal vote were included, then the marginal vote makes no contribution. A vote is pivotal if
it would success in the proxy contest.
24
The SharkRepellent dataset was previously part of SharkRepellent, which was later acquired by FactSet Data
Systems, which incorporated it into its Corporate Governance data product.
23
14
collecting proxy contest data in 2000. We collect voting data and recommendations for proxy
contests from ISS Voting Analytics. We also manually check the ownership by institutional
investors and hedge funds for each company proxy contest from FactSet Ownership.
The above data collecting process gives us a sample of 1,799 proxy contests for a short
slate or full slate of board seats. We exclude targets that are fund or trust. We further remove
contests that have missing data to ensure we have data available for important dates including
record date, contest announcement date, campaign end date, and the outcome of the contest. The
final sample we identified is 873 proxy contests for board seats from 1994 to 2017. Of the 873
contests, 159 contests (18%) were withdrawn by the dissident before the annual or special meeting;
325 contests (37%) were settled among the incumbents and dissidents prior to the meeting; and
389 contests went the distance to a vote (45%). Of those contests that went the distance, vote
results are available for 292 contests (33% of our sample).
We collect options data from the OptionMetrics dataset on the Wharton Research Data
Service (WRDS). OptionMetrics contains data for options on all US exchange-listed and
NASDAQ equities and market indices since 1996. We winsorize voting premiums at the 1% and
99% level, to eliminate outliers with options very close to maturity. We then combine our proxy
contest data with our voting premium data and retain only those contests where voting premium
data are available during the contest period.25 This leaves us with a sample of 314 proxy contests,
between 1998 and 2017. Of the 314 contests, 100 proxy contests have vote results available. To
analyze share price performance, we merge our proxy contest data with stock returns on the target
firms from the Center for Research in Security Prices (CRSP).
25
The contest period is illustrated in Section 3.1.
15
2.2 Summary Statistics
2.2.1
The Time Distribution of Proxy Contests
Figure 2 illustrates the time distribution of proxy contests for board seats announced each
year in the 1994 to 2017 sample period, and the number of contests at companies for which options
are available. The total number of proxy contests for board seats from 1994 to 2017 is 873 for the
full sample and 314 for the subsample for which options are available. The frequency of proxy
contests is generally increasing over the sample period, though it fell slightly in 2004. Overall, we
observe an increase in the number of proxy contests occurrence each year in our sample period,
which is consistent with the analysis of Fos (2017). The number of proxy contests for board seats
for companies with options follows a pattern similar to the full sample. Since firms with option
data available are typically large and liquid firms, our evidence suggests that firms targeted by
proxy contests are not limited to small or illiquid firms, particularly in the later part of our sample
period.
[ ~Insert Figure 2 about here~ ]
2.2.2
Classification by Type and Outcome of Proxy Contests
Our study focuses on contests for board seats. We classify these contests as either control
contests or short-slate contests, based on whether the dissidents nominate directors for 50% or
more of the available board seats. We divide the outcome of proxy contests into three categories.
The first category is proxy contests that went to a vote. The second category is proxy contests that
were withdrawn by dissidents. A withdrawal by a dissident of its nominees for election generally
occurs when the dissident believes there to be a low likelihood of their nominees being elected.
16
The third category is proxy contests that were withdrawn following a settlement among the
incumbents and the dissident. If the incumbents believe they are unlikely to be successful, they
have a greater incentive to reach a settlement with the dissident. If the concessions offered by the
incumbents are satisfactory to the dissident, the parties will reach a settlement and the dissident
will withdraw their nomination for election. Often, the concessions will include the addition of
one or more dissident nominees to the board of directors by the incumbents.
Table 1 sets out the details of our sample of contests. Of the 873 proxy contests for board
seats, 151 are control contests (17%), and 722 are short-slate contests (83%). The prevalence of
short-slate contests that do not seek control of the target firm is consistent with the argument of
Fos (2017), that the current corporate governance environment is closer to a “market for corporate
influence” rather than a market for corporate control.
Of the control contests in our sample, 62 contests (41%) went the distance to a vote, of
which we have vote results for 43 contest (28%). In addition, 55 contests (36%) were settled and
34 contests (23%) were withdrawn. Of the 722 short-slate contests in our sample, 338 (45%) went
the distance (of which we have vote results for 249, or 34%); 270 (37%) settled; and 125 (17%)
were withdrawn by the dissidents before the meeting.
[ ~Insert Table 1 about here~ ]
3. Methodology and Empirical Results
3.1 Excess Stock Returns around Proxy Contests
This paper uses the Carhart (1997) four-factor model (consisting of factors for the return
on the market portfolio, the size of the company, the book-to-market ratio of the company, and
17
stock price momentum) to calculate excess returns (abnormal returns) for securities.27 The contest
period is defined as the period from the 90th day prior to the initial campaign announcement date
through the 90th day after the campaign end date. The campaign end date is the meeting date if
the contest went to a vote, otherwise it is the date on which the dissident withdraws its nominees,
or the date when a settlement is announced. The estimation period is the 255-trading day period
prior to the contest period. For an individual security i, the abnormal returns (AR) between two
dates d1i and d2i is given by the cumulative abnormal returns (CAR), calculated as:
𝑡𝑡=𝑑𝑑
𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖 = ∑𝑡𝑡=𝑑𝑑2𝑖𝑖
𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 .
1𝑖𝑖
(1)
For a sample of N securities, the abnormal stock performance is measured by the mean of
cumulative abnormal returns, calculated as:
������ = 1 ∑𝑁𝑁
𝐶𝐶𝐶𝐶𝐶𝐶
𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖
𝑁𝑁 𝑖𝑖=1
3.1.1
(2)
Share Price Performance around Contest Announcement
To analyze the share price performance around proxy contests, we calculate cumulative
abnormal returns (CARs) for both the full-sample and the sample for which we have option data,
based on monthly data from the 24 months prior to the contest announcement and the 24 months
following the contest announcement. Figure 3 illustrates these results. It shows two important
features with respect to the full sample returns. First, the CARs decline by about 20% during the
24 months prior to the contest announcement, with the largest decreases taking place starting about
27
In unreported results we also use the market model and the Fama-French Three-Factor model to calculate the excess
returns. We find that the results are not sensitive to the model we chose. Consistent with Brown and Warner (1980),
if an event has a clear effect, the choice of the model only has an impact on abnormal return quantitatively, rather
than qualitatively.
18
one year prior to the contest announcement. This feature is consistent with the finding of Fos (2017)
that target companies experience poor share price performance before proxy contests, and which
could also explain dissidents’ decisions to initiate proxy contests at those companies.
Second, the CARs increase in the two-month period prior to the contest announcement, by
about 5%. This is generally the period after a campaign has been rumored or announced, and
suggests that the market reacts favorably to the campaign, or to the prospect of a proxy contest.
The share price performance after the announcement appears to be more complex. It decreases
slightly within the 12 month period following the contest announcement, and then increases after
12 months following the contest. The sample for which we have option data exhibits a similar but
amplified pattern. For this subsample, CARs decline to around -32% prior to the contest date,
which indicates that targets with options have worse share price performance compared to the full
sample. The subsample of companies with options also shows a more significant decline in
performance following the contest announcement.
[ ~Insert Figure 3 about here~ ]
Monthly data can only give us a broad overview of how share price behaves around contest
announcement. We therefore use daily data to focus on the period from 60 days before and the
contest announcement date to 60 days after that date. Figure 4 illustrates the share price
performance around the contest announcement date for the full sample and the subsample with
options. For both samples, CARs increase 60 days prior to the contest announcement date, which
is consistent with the results illustrated in Figure 3. The period prior to the contest announcement
date is nonetheless generally after the campaign announcement date. Increases in this period are
likely due to the market responding positively to news of the campaign, and anticipating increases
19
in the performance of the company. There is a jump in returns around the contest announcement
date, followed by further increases in returns after the contest announcement. After the contest
announcement, the full sample experiences a higher CAR compared to the subsample with options.
[ ~Insert Figure 4 about here~ ]
3.1.2
Share Price Performance around Proxy Contests by Category and Outcome
We next split our analysis of share price performance around proxy contest announcements
by the different categories of proxy contests, and different contest outcomes. In our full sample of
contests, 389 contests went the distance, of which 292 proxy contests have vote results. We use
daily data to analyze share performance for those contests for which vote results are available.
Table 2 summarizes the daily excess (abnormal) returns around proxy contests for all voted
contests by types and outcomes, for different event windows relative to the announcement date,
the meeting date, and the date after which the shares lose the right to vote, which we refer to as the
“ex-vote date.”28
Consistent with the results from monthly data, there is strong share price performance
around proxy contest announcements for all contests. The cumulative mean residual for the period
from 60 days prior to and including contest announcement through the meeting date (the [-60, 100]
event window around the contest announcement) is 14.71%, with a Z-statistic of 4.614. This
28
In our analysis, we refer to the last day a share could be traded before it loses the right to vote at the meeting as the
“ex-vote” date (analogous to an “ex-rights” date). Shareholders of the company as of the record date are entitled to
vote at the meeting. If a share is sold prior to the record date, it will lose the right to vote at the meeting. The date
on which this would occur depends on the length of time it takes to settle the trade. Prior to September 5, 2017, the
standard settlement period for most broker-dealer transactions was three business days (“T+3”). Hence, the ex-vote
date was the date three days before the record date. On September 5, 2017, the SEC adopted Rule 15c6-1(a), which
amended the settlement cycle to two days (“T+2”) for most broker-dealer transactions. As a result, after that date,
the ex-vote date is the date that is two days before the record date.
20
positive share price performance is consistent across all contests, regardless of their type or
outcome, which is consistent with the finding of Dodd and Warner (1983). In addition, contests
where dissidents win seats exhibit better share price performance than those where they do not—
the cumulative mean residual for contests where the dissidents win seats is 16.21%, with a Zstatistic of 4.286, 3% higher than the cumulative mean residual for contests where the dissidents
fails to win any seats.
During the [–1,0] event window around the meeting date, the share price moves only
slightly. Although we only observe small share price change, it shows that the movement is related
to the contest outcome. For contests where the dissidents win seats, the cumulative mean residual
is positive, while for those contests where the dissident fails to win seats, the cumulative mean
residual is negative, however none of these changes is statistically significant. One explanation for
the relatively minor adjustment of the share price around the meeting date is that the information
of whether the dissidents are likely to win seats is largely anticipated.
During the [1, 20] event windows following the meeting date, we observe small negative
cumulative mean residuals for all contests except for those where dissidents win seats. Normally,
we would expect some non-negative share price movement where the market expects the proxy
contest outcome to benefit shareholders. One possible explanation for the negative movement is
that the market value of vote decreases, since the votes cease to be important after the election.
More importantly, for most contests, we observe an increase in cumulative mean residuals
in the [-21, 0] event window, from 21 days prior to the ex-vote date, to the day after that date. This
is economically and statistically significant for all contests except for those where the dissidents
fail to win seats (model (5)). One possible explanation for this increase is that the shareholder
21
voting premium increases around the record date, since that is the date on which the company
determines which shares are eligible to vote at the meeting.
We next analyze the share price performance around contests at companies with options
available (and hence, for which voting premium data is available). Table 3 provides the summary
of excess returns around contests that went to a vote at companies with options data available. The
results reported in this subsample are consistent with those in Table 2. There are positive share
price performances prior to and including the contest announcement though the meeting date for
all contests regardless of types and outcomes. There is a substantial increase in CARs during the
[-21, 0] period, from 21 days prior to and including the ex-vote date, followed by modest declines
after the meeting date (which are not statistically significant). 30 Overall, the share price
performance around proxy contests for all voted contests and voted contests with options are
similar. We can observe a positive share price performance around the proxy contest. Moreover,
the cumulative mean residual jumps during the period leading up to and including the ex-vote date,
and decreases slightly after the meeting date.
The above analysis is based on proxy contests that went to a vote. However, more than half
of our sample consists of contests that were settled or withdrawn prior to the meeting. To avoid
the possibility of sample selection bias, we further analyze the full sample size and plot the CAR
around proxy contests based on their outcomes.
30
We note two minor differences between the results for our full sample in Table 2 and those for the subsample in
Table 3. First, for the subsample, the cumulative mean in the [-60, 100] event window around the contest
announcement date (event window (1)), the returns for control contests are greater than those for short-slate contests,
whereas the opposite is the case for the full sample. Second, the for the subsample, the returns in the [-21, 0] window
around the ex-vote date are less significant (both economically and statistically) than those for the full sample. We
do not believe there are conclusions that can be drawn from these differences, other than the effects of a smaller
sample size.
22
Figure 5 illustrates the share price performance around the contest announcement for the
subsample with options, by the outcome of the contest (that is, whether the contest was withdrawn,
settled, or went to a vote). Similar to our prior analyses, we observe positive share price
performance around the contest announcement for all groups of contests. However, there are
substantial differences in the nature of these increases. Target firms where proxy contests were
withdrawn by the dissident experience the highest abnormal returns around the contest
announcement date. Contests that settled, and those that went to a vote, experienced much more
moderate market responses.
[ ~Insert Figure 5 about here~ ]
This result suggests that market reactions around announcement dates do vary by proxy
contest outcomes—that is, that markets incorporate information that reflects the likely outcome of
the contest. To better understand the source of these abnormal returns, and the extent to which
extent shareholder voting premiums can explain the share price performance, we next examine the
behavior of voting premium around proxy contests.
3.2 Voting Premium
We calculate voting premiums using option pairs. An option pair consists of a call option
on the underlying stock, matched with a put option, both with the same strike price X and time to
expiration T. The option pairs we used in our sample are subject to four constraints. First, we only
use data for options that have between 10 and 90 days to expiration.31 Second, we discard option
31
We limit the maturity of our options for calculating the voting premium to those greater than 10 days because we
the prices of options with maturity less than 10 days to be too noisy to include in our analysis.
23
pairs where the quotes for either the call or the put option are locked or crossed. Third, we drop
option pairs for which the volume for either the call or put is less than zero, or for which the implied
volatility (calculated using the binomial option pricing model) for the call or the put is undefined.
Fourth, among possible option pairs for a particular stock, we select the most liquid options—
those that have the highest volume, that are closest-to-the-money, and that have the shortest
maturity.32
The main function we use for calculating the value of the vote is the following:
𝑆𝑆−𝑆𝑆̂(𝑇𝑇)
𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 =
𝑆𝑆
(3)
where S is the stock price (the closing price on the particular date, or if the closing price is not
available, the average of the bid and ask), and 𝑆𝑆̂(𝑇𝑇) is calculated as follows:
𝑑𝑑𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑𝑑𝑑
𝑆𝑆̂(𝑇𝑇) = 𝐶𝐶 − 𝐸𝐸𝐸𝐸𝐸𝐸𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
− 𝑃𝑃 + 𝐸𝐸𝐸𝐸𝐸𝐸𝑝𝑝𝑝𝑝𝑝𝑝
+ 𝑃𝑃𝑃𝑃(𝑋𝑋) + 𝑃𝑃𝑃𝑃(𝐷𝐷𝐷𝐷𝐷𝐷)
(4)
where C and P are the American option prices for the call and put options, calculated (in each case)
as the midpoint of the bid and ask quotes; PV(X) is the present value of the strike price of each
option in the pair (with present value calculated using time to expiration and the risk-free rate);
PV(Div) is the present value of any dividend payment amounts paid before exercise; and
𝑑𝑑𝑑𝑑𝑑𝑑
𝐸𝐸𝐸𝐸𝐸𝐸[𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜]
is the early excise premium of each of the options due to dividends.
𝑑𝑑𝑑𝑑𝑑𝑑
𝐸𝐸𝐸𝐸𝐸𝐸[𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜]
is calculated as follows:
𝑑𝑑𝑑𝑑𝑑𝑑
𝐸𝐸𝐸𝐸𝐸𝐸[𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜]
= 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴) − 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸)
32
(5)
For each firm, there exist options with multiple strike price and maturity on each date. The moneyness is defined as
ln(S/X), where S is the stock price on the date and X is the strike price of the option.
24
The EEP is calculated using the 1,000-step binomial option pricing model. Up and down factors
are calculated as 𝑢𝑢 = 𝑒𝑒 𝜎𝜎√∆𝑡𝑡 and 𝑑𝑑 = 𝑒𝑒 −𝜎𝜎√∆𝑡𝑡 . The implied volatility, time to expiration, strike price,
price of the underlying share, dividends, and ex-dates are obtained from the OptionMetrics
database.
3.2.1
Event and Benchmark Periods
Figure 6 shows an illustrative timeline of a hypothetical voting premium. We take the
meeting date as “date 0” in our event period, and thus the campaign end date. Our key variable of
interest is the average voting premium during this period, compared to the average voting premium
during a benchmark period. The benchmark period starts 90 days before the campaign
announcement date, and finishes on the day before the campaign announcement date. To
investigate the various results presented in Table 2 and Table 3, we choose corresponding event
windows for different critical periods of a contest. To capture any campaign announcement effect,
we choose a [–1, 15] window around the campaign announcement date. To capture contest
announcement effects, we use a [–15,1] window around the contest announcement date. To capture
any record date effects, we use a [–15,0] window around the ex-vote date. To capture meeting
effects, we use a [0,15] window around the meeting date.
[ ~Insert Figure 6 about here~ ]
25
3.3 Predictability of Voting Premium on the Proxy Contest Outcome
3.3.1
Predictability in the Full Sample
In this Part, we explore whether we can predict the outcomes of contests by observing
shareholder voting premiums around proxy contests. To examine whether voting premiums can
predict contest outcomes, we create three dummy variables: withdrawn, settled, and voted, which
take the value of 1 for contests that were withdrawn, contests that settled, and contests that went
to a vote, respectively, and otherwise take the value of 0.33
Table 4 reports univariate analyses of voting premium behavior for event windows around
the most important dates during proxy contests—the campaign announcement date, the contest
announcement date, the record date, and the meeting date. For each event window, Table 4 shows
the voting premium, the annualized voting premium, and the change of annualized voting premium
from the benchmark window. We report the mean, median, and standard deviation for each of
these variables during the various event windows. For robustness, for each relevant date we
consider two different event windows: we use [–1,10] and [–1,15] event windows around the
campaign announcement date; [–10,1] and [–15,1] event windows around the contest
announcement date. We use [–10,0] and [–15,0] event windows around the record date, as trading
to buy or sell shares around the record date will be effective up to and including the record date
itself. We use [0,10] and [0,15] event windows around the meeting date, as we are most interested
in the behavior of the voting premium after the meeting.
33
We note that there is a fourth potential case in our sample—contests that went to a vote, but for which vote results
are not available. This means that that we can safely perform regressions that include all three dummy variables.
26
[ ~Insert Table 4 about here~ ]
Table 4 shows that the contests that were withdrawn exhibit much lower voting premiums
than those that settled or that went to a vote, around both the record date and the contest
announcement date. This suggests that we could predict whether a proxy contest will be withdrawn
from observing voting premiums at an early stage of the proxy contest.
To test whether this is actually the case, we conduct multinomial logit regressions of proxy
contests outcomes or voting premium. We conduct separate regressions comparing voting
premiums for (a) contests that are go to a vote versus those that are withdrawn, and (b) contests
that are settled versus those that are withdrawn. For each of these pairs we run regressions using
three separate specifications, each including progressively more independent variables. Table 5
reports the regression results. Consistent with the univariate analysis, the difference in the
annualized voting premium around contest announcement date from that in the benchmark window
has predictive power in determining whether a contest will be withdrawn. In particular, an increase
in the voting premium around the contest announcement is associated with a greater likelihood
that the contest will go to a vote, and a lower likelihood that the contest will be withdrawn. While
an increase in the voting premium also results in a greater likelihood of settlement, the effect is
not statistically significant. Interestingly, the involvement of hedge funds increases the likelihood
of settlement (Table 5, Regression 3), which is consistent with results observed by Bebchuk et al.
(2020).
[ ~Insert Table 5 about here~ ]
Table 6 reports the results of a multinomial logit regression of vote outcome (settled or
voted) on the same independent variables as in Table 5. We find that the voting premium around
27
the contest announcement date is a weak (though statistically insignificant) predictor of whether
the contest will go to a vote or be settled. The higher voting premium around contest
announcements for contests that go to a vote could reflect the fact that both the dissidents and the
incumbents will participate in the electoral process only if they both expect to be successful in the
contest. If that is the case, the probability of the contest going to a vote is likely to be higher, and
the value of the vote will increase accordingly. Similar to Table 5, we find that involvement of
hedge funds decreases the chances of the contest going to a vote, in favor of settlement. However,
having a 5% blockholders, and activist ownership, increases the likelihood of a contest going to a
vote, and reduces the likelihood of settlement.
[ ~Insert Table 6 about here~ ]
3.3.2
Predictability in the Voted Contest Subsample
We next examine whether the voting premium can predict which side is likely to prevail in
the contest. We limit our sample to those contests that go to a vote, and for which voting results
are available (i.e., where voted equals 1). We create a variable, seats won/sought, which is the
number of seats won by the dissident as a proportion of the number of dissident nominees in the
contest. The mean of seats won/sought is around 45% in the sample of contests for which we have
vote results.
Table 7 shows the results of ordinary least squares regression of seats won/sought on the
voting premium around the record date, and various controls. These regression show that it is
possible to predict the proportion of dissidents that are elected in a contest. Specifically, during
the [-15, 0] window from 15 days before the record date through the record date itself, seats
won/sought increases by 21% for one standard deviation in change of annualized voting premium,
28
ceteris paribus. This corresponds to a 50% increase in the mean of seats won/sought. We find that
the increase in the voting premium around the record date strongly predicts the fraction of board
seats won by dissidents.
[ ~Insert Table 7 about here~ ]
Table 8 shows that there is a negative relationship between seats won/sought and the
change in annualized voting premium after the meeting date. Specifically, Table 8 shows the
results of ordinary least squares regressions of seats won/sought on the change in annualized voting
premium after the meeting date, and various controls. In each specification the coefficient on the
change of annualized voting premium after the meeting date is negative and significant. These
results can be understood as the converse of the comparable results for the change in the annualized
voting premium around the record dates in Table 7. After the contest, the voting premium
decreases. The decrease in the voting premium is negatively related to the fraction of the board
members won by the dissidents, consistent with our findings in Table 7. These results reinforce
the possibility that shareholder voting premium explains share price performance around proxy
contests, especially for those contests that go to a vote.
[ ~Insert Table 8 about here~ ]
Overall, the empirical results described in this Part reinforce our confidence in the
hypothesis that the voting premium of shares can partially explain the share price performance
around proxy contests, as well as some outcomes of proxy contests. We show that a portion of the
positive market reaction around proxy contests announcements is related to the change in the
shareholder voting premium, especially for contests that go to a vote and those that are settled. In
29
addition, for contests that go to a vote, we observe an increase in voting premium around the record
date, and a decrease after the meeting date, which also reflects the share price performance of
target firms in those circumstances. Finally, we find that we can predict whether the proxy contest
will be withdrawn by the dissidents from observation of the shareholder voting premium around
the contest announcement date, and that we can predict how successful the dissidents will be in
the contest from the voting premium around the record date.
Conclusion
This paper investigates how share price behaves around proxy contests, and in particular,
the part of the share price attributable to the voting premium. We explore the extent to which
shareholder voting premium can explain contest outcomes and target share price performance.
Consistent with prior literature, we find a positive and statistically significant increase in share
price performance around the announcement of proxy contests. However, the positive market
reaction differs among proxy contests with different outcomes. Contests that are withdrawn
experience higher abnormal returns around contest announcement than contests that are settled, or
than contests that go to a vote.
We estimate the shareholder voting premium, and find that a portion of the positive market
reaction around proxy contests announcement results from the change in the shareholder voting
premium, especially for contests that were settled, and those that go to a vote. Moreover, for
contests that go to a vote, we find that the voting premium increases around the record date, and
declines after the campaign end date, which reflects similar movements in the share price
performance. These results suggest that shareholder voting premium can partially explain the share
price performance around proxy contests.
30
We also find that we can predict the certain aspects of the outcome of a proxy contest based
on the behavior of the voting premium. In particular, an increased voting premium around the
campaign announcement predicts a decrease in the likelihood of the contest being withdrawn by
the dissident. A higher voting premium around the record date also predicts a greater proportion
of dissidents nominated in the contest being elected. This finding has important implications for
the proxy contest market, as dissidents and incumbents could observe the market reaction to the
campaign announcement as an indicate of the likely outcome of the contest, and use that
information to inform their decision whether to pursue the contest, and which tactics and strategies
to adopt.
The observation that withdrawn contests exhibit the highest CAR around the proxy contest
announcement date is puzzling. As highlighted by Jensen and Warner (1988), the wealth effect
shown by the share price increase contains both a real effect and an information effect. If a proxy
contest has a disciplining effects on incumbents, we would expect a positive abnormal share price
reaction (the real effect). However, it is difficult to understand how a contest that is ultimately
withdrawn has a disciplining effect. That leaves two potential information effects that could
explain the extremely high CAR for withdrawn contests. One possibility is that activist hedge
funds—the most common dissidents in proxy contests—are promoting short-term goals at the
expense of long-term value, as many of their critics have argued. If that is the case, if the market
expects them to withdraw their contest, that would be construed as having a positive effect on the
company’s share price. Another potential explanation is that, because proxy contests impose
considerable costs on companies, the likely withdrawal of a contest is expected to increase the
value of the company. A final possibility is that the causal relationship actually runs in the reverse
direction—that is, that it is not that the market predicts withdrawn contests and trades in such a
31
way as to increase share price performance, but rather that better share price performance makes
the dissident more likely to withdraw the contest. For example, incumbents could cause the
company to repurchase stock, which would increasing stock price performance, and might also
meet or obviate the demands of the dissident. These findings—and the open questions they
reveal—suggest that shareholder vote premiums around proxy contests merit further investigation.
32
References
Aggarwal, R., P.A.C. Saffi, and J. Sturgess, 2015, The Role of Institutional Investors in Voting:
Evidence from the Securities Lending Market, Journal of Finance 70, 2309–2346.
Aranow, E.R., and H.A. Einhorn, 1968, Proxy Contests for Corporate Control, Columbia
University Press, New York.
Barclay, M.J., and C.G. Holderness, 1989, Private Benefits from Control of Public Corporations,
Journal of Financial Economics 25, 371–395.
Bebchuk, L.A., and S. Hirst, 2019, Index Funds and the Future of Corporate Governance: Theory,
Evidence, and Policy, Columbia Law Review 119, 2019-2146.
Bebchuk, L.A., and M. Kahan, 1990, A Framework for Analyzing Legal Policy Towards Proxy
Contests, California Law Review 78, 1071–1136.
Bebchuk, L.A., A. Brav, W. Jiang, and T. Keusch, 2020, Dancing with Activists, Journal of
Financial Economics 137, 1–41.
Bertrand, M., and S. Mullainathan, 2003, Enjoying the Quiet Life? Corporate Governance and
Managerial Preferences, Journal of Political Economy 111, 1043–1075.
Bhattacharya, U., 1997, Communication Costs, Information Acquisition, and Voting Decisions in
Proxy Contests, Review of Financial Studies 10, 1065–1097.
Black, B.S., 1990, Shareholder Passivity Reexamined, Michigan Law Review 89, 520–608.
Brav, A., W. Jiang, F. Partnoy, and R. Thomas, 2008, Hedge Fund Activism, Corporate
Governance, and Firm Performance, Journal of Finance 63, 1729–1775.
Brav, A., W. Jiang, and H. Kim, 2010, Hedge Fund Activism: A Review, Foundations and Trends
in Finance 4, 185–246.
Brav, A., W. Jiang, T. Li, and J. Pinnington, 2020, Picking Friends Before Picking (Proxy) Fights:
How Mutual Fund Voting Shapes Proxy Contests, Working Paper, Duke University.
Brown, S.J., and J.B. Warner, 1980, Measuring Security Price Performance, Journal of Financial
Economics 8, 205–258.
Carhart, M.M., 1997, On Persistence in Mutual Fund Performance, Journal of Finance 52, 57–82.
Chen, F., L. Du, S.M.V. Flaherty, J. Huang, and G. Torna, 2020, Not All Threats are Taken Equally:
Evidence from Proxy Fights, Financial Review 55, 145–168.
33
Christie, A.L., 2019, The New Hedge Fund Activism - Activist Directors and the Market for
Corporate Quasi-Control, Journal of Corporate Law Studies 19, 1–41.
Christoffersen, S.E.K., C.C. Geczy, D.K. Musto, and A.V. Reed, 2007, Vote Trading and
Information Aggregation, Journal of Finance 62, 2897–2929.
Corum, A.A., and D. Levit, Corporate Control Activism, Journal of Financial Economics 133, 1–
17.
De Wied, W.S., 2018, Proxy Contests, Practical Law, Resource ID: 6-503-6878.
Dodd, P., and J.B. Warner, 1983, On Corporate Governance: A Study of Proxy Contests, Journal
of Financial Economics 11, 401–438.
Duvall, R.M., and D.V. Austin, 1965, Predicting the Results of Proxy Contests, Journal of Finance
20, 464–471.
Dyck, A., and L. Zingales, 2004, Private Benefits of Control: An International Comparison,
Journal of Finance 59, 537–600.
Fos, V., 2017, The Discipline Effect of Proxy Contests, Management Science 63, 655–671.
Fos, V., and W. Jiang, 2016, Out-of-the-Money CEOs: Private Control Premium and Option
Exercises, Review of Financial Studies 29, 1549–1585.
Gantchev, N., 2013, The Costs of Shareholder Activism - Evidence from a Sequential Decision
Model, Journal of Financial Economics 107, 610–631.
Ghosh, C., J.E. Owers, and R.C. Rogers, 1992, Proxy Contests: A Re-Examination of the Value
of Vote Hypothesis, Managerial Finance 18, 3–18.
Goodwin, S., 2017, Management Practices in an Age of Engaged Investors, Columbia Business
School Research Paper No. 17-97.
Gurun, U., and O. Karakaş, 2021, Earnings and the Value of Voting Rights, Working Paper,
University of Texas at Dallas.
Hancock, G.D., 1992, Battles for Corporate Control: An Overview of Proxy Contests, Managerial
Finance 18, 59–76.
Hirst, S., 2018, Universal Proxies, Yale Journal on Regulation 35, 437–511.
Huang, W.-C., and G. Yen, 1996, The Impact of Proxy Contests on Managerial Turnover: A Test
of the Job Security Hypothesis. Managerial and Decision Economics 17, 551–558.
Ikenberry, D., and J. Lakonishok, 1993, Corporate Governance through the Proxy Contests:
Evidence and Implications, Journal of Business 66, 405-435.
34
Jha, M., 2021, Catching the Conscience of Kings: How Activists Pander Mutual Funds, Working
Paper, Washington University in St. Louis.
Kalay, A., O. Karakaş, and S. Pant, 2014, The Market Value of Corporate Votes: Theory and
Evidence from Option Prices, Journal of Finance 69, 1235–1271.
Karakaş, O., and M. Mohseni, 2021, Staggered Boards and the Value of Voting Rights, Review of
Corporate Finance Studies 10, 513–550.
Levit, D., N. Malenko, and E.G. Maug, 2021, The Voting Premium, Working Paper, University
of Washington.
Levy, H., 1983, Economic Evaluation of Voting Power of Common Stock, Journal of Finance 38,
79–93.
Manne, H.G., 1962, The ‘Higher Criticism’ of the Modern Corporations, Columbia Law Review
62, 399–432.
Mulherin, J.H., and A.B. Poulsen, 1998, Proxy Contests and Corporate Change: Implications for
Shareholder Wealth, Journal of Financial Economics 47, 279–313.
Nenova, T., 2003, The Value of Corporate Voting Rights and Control: A Cross-Country Analysis,
Journal of Financial Economics 68, 325–351.
Pound, J., 1988, Proxy Contests and the Efficiency of Shareholder Oversight, Journal of Financial
Economics 20, 237–265.
Rydqvist, K., 1987, Empirical Investigation of the Voting Premium, Northwestern University
Working Paper No. 35.
Rydqvist, K., 1996, Takeover Bids and the Relative Prices of Shares That Differ in Their Voting
Rights, Journal of Banking and Finance 20, 1407–1425.
SEC (U.S. Securities and Exchange Commission), 1992, Regulation of Communication Among
Shareholders, Securities Exchange Act of 1934, Release no. 31,326, 57 Federal Register
48,276, Oct. 16.
Sharara, N.M., and A.E. Hoke-Witherspoon, 1993, The Evolution of the 1992 Shareholder
Communication Proxy Rules and Their Impact on Corporate Governance, Business
Lawyer 49, 327–358.
Zingales, L., 1994, The Value of the Voting Right: A Study of the Milan Stock Exchange
Experience, Review of Financial Studies 7, 125–148.
35
Zingales, L., 1995, What Determines the Value of Corporate Votes? Quarterly Journal of
Economics 110, 1047–1073.
36
Figure 1. The Timeline of a Campaign
The figure characterizes the timeline of proxy contests in our sample, the key events during contests, and the median
and mean dates between certain key events for contests in our sample. The campaign start date is the date on which
the dissident announce the campaign. The dissident generally engages with the target firm during the campaign. If the
firm and the dissident are unable to reach agreement, the dissident commences a proxy contest (the contest start date),
by nominating directors for election. It is possible that the dissident goes directly to a proxy contest, without first
beginning a campaign and engaging with the target directors, in which case the campaign announcement date and the
contest announcement date are the same. The record date is set by the incumbent directors. Only the shareholders of
the company as of that date are eligible to vote at the meeting. The record date could be prior to the campaign
announcement date or proxy contest announcement, but is usually after both dates. The meeting date is the date of the
annual or special meeting at which the director election takes place. If the contest goes to a vote then the campaign
end date is the same as the meeting date. If the parties reach a settlement, or the dissident withdraws its nominees,
then the campaign end date is the date on which the contest was settled or withdrawn.
37
Figure 2. Time Distribution of Proxy Contests
The figure shows the number of proxy contests announced each year in the 1994-2017 period, and the number of those
contests for which sufficient option data exists for inclusion in our sample. The total number of proxy contests for
board seats from 1994 to 2017 is 873 for the full sample and 314 for the subsample with option data.
38
Figure 3. Cumulative Abnormal Returns (CAR) around Proxy Contest Using Monthly
Data for the Full Sample and Subsample with Options
The figure illustrates the share price performance, by cumulative abnormal returns, for proxy contests in both the full
sample and the subsample with for which options data is available. The sample sizes of the full sample and the
subsample with options are 873 and 314, respectively.
39
Figure 4. CARs around Proxy Contest Announcement Using Daily Data for Full Sample
and Subsample with Options
The figure uses daily data to report the share price performance around the proxy contest announcement date for the
full sample and the subsample with options. The sample sizes of the full sample and the subsample with options are
873 and 314, respectively.
40
Figure 5. CAR around Proxy Contest Announcements Using Daily Data for the Subsample
with Options By Contest Outcome
The figure reports the share price performance around proxy contests for the subsample with options using daily data,
by the outcome of the contest. The sample size of the subsample with options is 314, with 100 contests that went to a
vote, 61 that were withdrawn, and 134 that settled.37
37
We exclude from consideration cases that went the distance but for which vote results were not available.
41
Figure 6. Event and Benchmark Windows
The figure depicts the event and benchmark windows used in our analyses.
42
Table 1. Proxy Contests from 1994 to 2017 by Type and Outcome
This table reports the sample of proxy contests for board seats from 1994 to 2017, by type and outcome. The total
sample size is 873 proxy contests, of which 151 are control contests and 722 are short-slate contests. Control contests
are contests where dissidents seek a majority of board seats available, i.e., for corporate control. Short-slate contests
are contests where dissidents seek less than half of the board seats available. Went distance indicates that the contest
“went the distance” to a vote at a shareholder meeting. Settled / concessions made indicates that the contest was settled
between the incumbent and the dissident before the annual or special meeting, with concessions by the incumbent.
Withdrawn indicates that the contest was withdrawn by dissidents before the meeting.
Panel A: Control Contests
Outcomes
Went distance, vote
Went
Settled/ Concessions
results
distance
made
151
43
19
55
34
100%
28.48%
12.58%
36.42%
22.52%
Total
No. of contests
Percentage
Withdrawn
Panel B: Short-slate Contests
Outcomes
Went distance, vote
Went
Settled/ Concessions
results
distance
made
722
249
78
270
125
100%
34.49%
10.80%
37.40%
17.31%
Total
No. of contests
Percentage
43
Withdrawn
Table 2. Summary of Excess Returns around Proxy Contests for All Voted Contests
This table reports the summary of daily excess returns around all contests with vote results by types and outcomes.
The sample consists of 292 proxy contests for board seats with vote results from 1994 to 2017. The table provides
cumulative abnormal returns (in percentages) using different event windows around the contest announcement date,
the ex-vote date, and the meeting date. Two-sided Z-statistics are displayed in parentheses. The symbols *, **, and
*** denote statistical significance at the 0.10, 0.05, and 0.01 levels, respectively.
Mean Cumulative Residuals in Percentages
All contests where
dissidents
All contests
(1)
[292]
(2)
[43]
Short
slate
(3)
[249]
14.71***
(4.614)
13.90
(1.276)
14.86***
(4.474)
16.21***
(4.286)
13.24**
(2.249)
(2) Days –1 through 0 prior to and including
meeting date
0.19
(0.759)
1.27
(0.039)
–0.02
(0.811)
0.80
(1.509)
–0.41
(–0.424)
(3) Days 60 through 0 prior to and including
contest announcement date
7.97***
(4.486)
6.15
(0.627)
8.30***
(4.613)
9.61***
(3.558)
6.35***
(2.790)
(4) Days 1 through 100 following contest
announcement date
6.75
(1.412)
7.75
(–0.023)
6.56
(1.546)
6.60**
(1.739)
6.90
(0.264)
(5) Days 1 through 20 following meeting date
–1.42
(–1.295)
4.69***
(3.315)
–1.01
(–0.922)
7.78**
(2.191)
–1.50
(–1.011)
4.12***
(2.666)
0.47
(0.778)
4.79***
(3.460)
–3.26***
(–2.589)
4.60
(1.223)
–0.20
(–0.665)
–0.21
(–0.406)
–0.20
(–0.549)
–0.58
(–0.525)
0.18
(–0.415)
(1) Days –60 through 100 before and after contest
announcement date38
(6) Days -21 to 0 prior to and including ex-vote
date
(7) Days 1 through 2 following ex-vote date
38
Control
The median days between contest announcement date and meeting date is 100 in our sample.
44
Wins
seats
(4)
[142]
Wins no
seats
(5)
[150]
Table 3. Summary of Excess Returns around Proxy Contests for Voted Contests with
Options
This table reports the summary of daily excess returns around contests with options available with vote results by
types and outcomes. The sample consists of 292 proxy contests for board seats with vote results from 1994 to 2017.
The table provides cumulative abnormal returns (in percentages) using different event windows around the contest
announcement date, the ex-vote date, and the meeting date. Two-sided Z-statistics are displayed in parentheses. The
symbols *, **, and *** denote statistical significance at the 0.10, 0.05, and 0.01 levels, respectively.
Mean Cumulative Residuals in Percentages
All contests where
dissidents
Wins
All
seats
contests
(4)
(5)
[53]
[47]
All
contests
(1)
[100]
Control
(2)
[17]
Short
slate
(3)
[83]
(8) Days –60 through 100 before and after contest
announcement date
9.58***
(2.982)
13.05*
(1.823)
8.87**
(2.449)
7.71**
(2.255)
11.69*
(1.956)
(9) Days –1 through 0 prior to and including meeting
date
0.18
(1.619)
2.14
(0.805)
–0.22
(1.413)
1.17**
(2.252)
–0.94
(–0.030)
(10) Days 60 through 0 prior to and including contest
announcement date
5.36***
(2.582)
3.96
(0.367)
5.65***
(2.668)
5.08
(0.880)
5.69***
(2.832)
(11) Days 1 through 100 following contest
announcement date
4.22
(0.982)
9.09
(0.853)
3.22
(0.692)
2.64
(0.880)
6.01
(0.498)
(12) Days 1 through 20 following meeting date
–1.27
(–0.381)
3.61**
(2.023)
–0.07
(0.023)
–1.91
(–0.651)
6.66
(0.799)
0.50
(1.284)
–1.13
(–0.124)
2.98*
(1.859)
–0.19
(–0.556)
–1.13
(–1.045)
3.51**
(1.984)
–0.17
(0.610)
–1.42
(0.554)
3.72
(0.844)
0.03
(–0.614)
(13) Days -21 to 0 prior to and including ex-vote date
(14) Days 1 through 2 following ex-vote date
45
Table 4. Univariate Analysis of Voting Premium, Annualized Voting Premium, and
Change of Annualized Voting Premium Around Proxy Contests
This table reports the mean, median, and standard deviation of voting premium, annualized voting premium, and
change in annualized voting premium (each in percentages) for our sample of contests at companies with options. The
sample consists of 314 proxy contests for board seats from 1999 to 2017. Each scenario displays the number of cases
of proxy contests used for the calculation of the relevant statistic as n. Voting premium are winsorized at 1%. The
365
annualized normalized value of voting rights is estimated as follows: 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑢𝑢𝑚𝑚 = 1 − (1 − 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉) 𝑇𝑇 . 39
Change in annualized voting premiums are calculated as the annualized voting premium, less the comparable statistics
for the annualized voting premium during a benchmark window of [-90, -1] relative to the campaign announcement
date. Event windows are in trading days.
Variables
Voting Premium
Event
Window
Withdrawn Settled
Annualized Voting Premium
Voted
Withdrawn Settled
Voted
∆ Annualized Voting Premium
Withdrawn Settled
Voted
Around
[–1,10]
Campaign
Announce
Date
Mean
–0.00
0.07
0.14
–0.07
0.84
1.49
–1.16
0.21
0.98
Median
0.01
0.03
0.04
0.18
0.28
0.47
0.05
–0.24
0.36
0.0020
0.0024
0.0038
0.0331
0.0250
0.0377
0.0418
0.0289
0.0380
30
54
53
30
54
53
21
39
38
[–1,15]
Mean
–0.00
0.11
0.12
0.04
1.12
1.48
–1.04
0.59
0.66
Median
0.00
0.04
0.04
0.04
0.35
0.41
–0.35
–0.11
0.07
0.0020
0.0029
0.0033
0.0319
0.0296
0.0407
0.0421
0.0436
0.0307
32
63
56
32
63
56
23
45
39
S.D.
n
S.D.
n
Around
[–10,1]
Contest
Announce
Date
Mean
0.02
0.08
0.08
0.37
0.87
0.85
–0.39
0.20
0.51
Median
0.00
0.03
0.03
0.01
0.43
0.29
–0.51
0.39
0.37
0.0018
0.0017
0.0017
0.0240
0.0215
0.0190
0.0177
0.0258
0.0232
35
61
57
35
61
57
21
44
41
[–15,1]
Mean
0.02
0.07
0.10
0.31
0.78
1.05
–0.29
0.24
0.68
S.D.
n
Median
S.D.
n
Around
Record
Date
[–10,0]
Mean
Median
S.D.
n
[–15,0]
Mean
[0,10]
0.06
0.42
0.35
–0.02
0.29
0.22
0.0203
0.0169
0.0221
0.0193
0.0257
0.0268
38
74
63
38
74
63
24
52
45
–0.00
0.15
0.21
–0.49
1.67
1.93
–0.47
0.41
1.75
0.00
0.01
0.04
0.03
0.17
0.52
0.11
0.00
0.40
0.0020
0.0044
0.0054
0.0261
0.0465
0.0454
0.0336
0.0275
0.0510
24
52
47
24
52
47
17
34
32
–0.04
0.14
0.19
–0.06
1.48
1.79
–0.76
0.75
1.70
–0.02
0.02
0.04
–0.03
0.20
0.71
–0.35
0.10
0.20
0.0027
0.0041
0.0049
0.0249
0.0405
0.0438
0.0395
0.0381
0.0527
28
60
50
28
60
50
19
39
34
Mean
0.03
0.20
0.06
0.20
1.63
0.32
0.30
0.15
–1.66
Median
–0.04
0.02
0.02
–0.49
0.29
0.24
–0.22
0.19
–0.29
S.D.
0.0035
0.0068
0.0041
0.0423
0.0523
0.0506
0.0468
0.0563
0.0550
16
45
50
16
45
50
15
39
44
Mean
0.10
0.14
0.08
1.13
1.32
0.45
1.04
0.14
–1.40
Median
–0.00
0.03
0.02
–0.01
0.38
0.24
0.05
0.26
–0.00
S.D.
0.0031
0.0054
0.0043
0.0396
0.0486
0.0495
0.0371
0.0420
0.0533
22
51
53
22
51
53
20
45
47
n
39
0.03
0.0019
S.D.
n
[0,15]
0.04
0.0015
Median
n
Around
Meeting
Date
0.00
0.0017
For an explanation and further details of regarding this approach, see the Internet Appendix of Kalay et al. (2014).
46
Table 5. Multinomial Logit Regression of Proxy Contest Outcomes and Voting Premium:
Voted vs. Withdrawn and Settled vs. Withdrawn
This table reports the voting premium that predict the occurrence of a voted or settled contest versus a withdrawn
contest. The base outcome is withdrawn. The predicted outcome of column (1), (3), and (5) is voted. The predicted
outcome of column (2), (4), and (6) is settled. The models are estimated using multinomial logit regressions. Our key
explanatory variable is ΔAVote, the change in annualized voting premium, which is the Annualized Voting Premium
(AVote) for the [-15, 1] window around the contest announcement, less AVote for the benchmark window [-90, -1]
relative to the campaign announcement date. Size is the size of target firm. Return on Assets is net income divided by
total assets. Tobin’s Q is the market value of assets divided by the book value of assets. Excess return is equal to stock
return less market return over last financial year. Ownership by 5% Holders is total ownership of 5% holders. Hedge
Fund is a dummy which equals 1 if the dissident was a hedge fund dissident, otherwise 0. Activist Ownership is the
activist group ownership percentage. The sample is proxy contests for board seats with voting premium data available.
The number of observations in each specification is reduced by the number of contests for which independent variables
are not available. The symbols *, **, and *** denote statistical significance at the 0.10, 0.05, and 0.01 levels,
respectively.
Voted
(1)
Settled
(2)
Voted
(3)
Settled
(4)
Voted
(5)
Settled
(6)
17.07*
(1.821)
10.34
(1.201)
24.97**
(2.487)
15.06
(1.640)
28.30**
(2.485)
18.67
(1.351)
0.000220
0.000352
0.000365
6.93e–05
Return on Assets
–0.202
0.450
0.468
2.478
Tobin’s Q
0.554
0.348
1.011*
0.916*
–0.0665
–0.934
0.564
–0.659
0.0551***
–0.0503
0.0990***
–0.0411
0.779
1.968***
–0.0127
2.526***
0.0543
–0.136
ΔAVote (Contest
Announcement [-15, 1])
(T-statistic)
Size
Excess Return
Ownership by
Holders
Hedge Fund
5%
Activist Ownership
Constant
Observations
0.502*
0.620**
–1.839**
–1.528
–3.244**
–2.226
99
99
95
95
70
70
47
Table 6. Multinomial Logit Regression of Proxy Contest Outcomes and Voting Premium:
Voted vs. Settled
This table reports the voting premium that predict the occurrence of a voted contest versus a settled contest. The base
outcome is settled. The predicted outcome of column (1), (2), and (3) is voted. The models are estimated using
multinomial logit regressions. Our key explanatory variable is ΔAVote, the change in annualized voting premium,
which is the Annualized Voting Premium (AVote) for the [-15, 1] window around the contest announcement, less
AVote for the benchmark window [-90, -1] relative to the campaign announcement date. Return on Assets is net income
divided by total assets. Tobin’s Q is the market value of assets divided by the book value of assets. Excess return is
equal to stock return less market return over last financial year. Ownership by 5% Holders is total ownership of 5%
holders. Hedge Fund is a dummy which equals 1 if the dissident was a hedge fund dissident, otherwise 0. Activist
Ownership is the activist group ownership percentage. The sample is proxy contests for board seats with voting
premium data available. The number of observations in each specification is reduced by the number of contests for
which independent variables are not available. The symbols *, **, and *** denote statistical significance at the 0.10,
0.05, and 0.01 levels, respectively.
Voted
(1)
Voted
(2)
Voted
(3)
6.224
(0.914)
9.277
(1.286)
22.39*
(1.949)
–6.64e–05
0.000595
Return on Assets
–1.003
–0.332
Tobin’s Q
0.167
0.583
Excess Return
0.790
1.458
0.108***
0.151***
–1.033
–2.785***
ΔAVote (Contest Announcement [-15, 1])
(T-statistic)
Size
Ownership by 5% Holders
Hedge Fund
Activist Ownership
Constant
Observations
0.301***
–0.113
–0.431
–2.927**
78
74
54
48
Table 7. OLS Regression of Success in Obtaining Board Seats: Around Record Date
This table reports the results of ordinary least squares regressions of the success in obtaining board seats on ΔAVote,
the change in annualized voting premium around the ex-vote date. The dependent variable is seats won divided by
seats sought. ΔAVote is calculated as the Annualized Voting Premium (AVote) for the [0,15] window around the exvote date, less AVote for the benchmark window [-90, -1] relative to the campaign announcement date. Return on
Assets is net income divided by total assets. Tobin’s Q is the market value of assets divided by the book value of assets.
Excess return is equal to stock return less market return over last financial year. Ownership by 5% Holders is total
ownership of 5% holders. Hedge Fund is a dummy which equals 1 if the dissident was a hedge fund dissident,
otherwise 0. Activist Ownership is the activist group ownership percentage. Regressions are clustered on campaignid,
which is the arbitrary index number of each proxy contest. The sample is proxy contests for board seats with voting
premium data available. The number of observations in each specification is reduced by the number of contests for
which independent variables are not available. The symbols *, **, and *** denote statistical significance at the 0.10,
0.05, and 0.01 levels, respectively.
Seats Won ÷ Sought
(1)
Seats Won ÷ Sought
(2)
Seats Won ÷ Sought
(3)
3.442***
3.755***
3.657**
(3.414)
(2.976)
(2.507)
–0.000137**
–0.000150
Return on Assets
–0.665***
–0.628*
Tobin’s Q
–0.0822*
–0.0480
Excess Return
0.670**
0.568
Ownership by 5% Holders
–0.00159
0.000229
Hedge Fund Dummy
–0.326*
–0.368
ΔAVote (Ex-vote date [–15,0])
T-statistic
Size
Activist Ownership
Constant
Observations
R-squared
–0.0132
0.385***
1.021***
1.054**
34
34
26
0.151
0.392
0.418
49
Table 8. OLS Regression of Success in Obtaining Board Seats: Around Meeting Date
This table reports the results of ordinary least squares regressions of the success in obtaining board seats on ΔAVote,
the change in annualized voting premium around the meeting date. The dependent variable is seats won divided by
seats sought. ΔAVote is calculated as the Annualized Voting Premium (AVote) for the [-15,0] window around the
meeting date, less AVote for the benchmark window [-90, -1] relative to the campaign announcement date. Return on
Assets is net income divided by total assets. Tobin’s Q is the market value of assets divided by the book value of assets.
Excess return is equal to stock return less market return over last financial year. Ownership by 5% Holders is total
ownership of 5% holders. Hedge Fund is a dummy which equals 1 if the dissident was a hedge fund dissident,
otherwise 0. Activist Ownership is the activist group ownership percentage. Regressions are clustered on campaignid,
which is the arbitrary index number of each proxy contest. The sample is proxy contests for board seats with voting
premium data available. The number of observations in each specification is reduced by the number of contests for
which independent variables are not available. The symbols *, **, and *** denote statistical significance at the 0.10,
0.05, and 0.01 levels, respectively.
Seats Won ÷
Sought
(1)
Seats Won ÷ Sought
(2)
Seats Won ÷ Sought
(3)
–2.301***
(–3.343)
–3.152***
(–2.960)
–2.831**
(–2.667)
–0.000137**
–0.000111
–0.708***
–0.734***
–0.0364
–0.0211
0.319
0.171
Ownership by 5% Holders
–0.00294
0.000660
Hedge Fund Dummy
–0.238*
–0.232
ΔAVote (Meeting Date [0,15])
T-statistic
Size
Return on Assets
Tobin’s Q
Excess Return
Activist Ownership
Constant
Observations
R-squared
0.00176
0.444***
0.813***
0.667**
47
46
35
0.072
0.286
0.291
50