Communication and Decision-Making in

Corporate Boards
Nadya Malenko
Boston College, Carroll School of Management

Time constraints, managerial power, and reputational concerns can impede board
communication. This paper develops a model where board decisions depend on directors’
effort in communicating their information to others. I show that directors communicate
more effectively when pressure for conformity is stronger—that is, when directors are

Downloaded from [Link] at Boston College on February 15, 2015

more reluctant to disagree with each other. Hence, open ballot voting can be optimal, even
though it induces directors to disregard their information and conform their votes to others.
I also show that communication can be more efficient when directors’ preferences are more
diverse. The analysis has implications for executive sessions, transparency, and committees.
(JEL D71, D72, D82, D83, G34)

Introduction
The board of directors is a collective body, whose members have
diverse expertise in various aspects of the company’s business. Therefore,
communication between directors is critical to successful board functioning.
In recent years, regulators, shareholders, and directors themselves have been
paying increased attention to decision-making policies that could increase the
quality of board discussions. Executive sessions that exclude the management,
separation of the CEO and chairman positions, board retreats, and separate
committees on specific topics have been put in place to promote more effective
communication. As governance experts Carter and Lorsch (2004) emphasize,
“If we could offer only one piece of advice, it would be to strive for open
communication among board members.”

I am grateful to my dissertation committee, Anat Admati, Paul Pfleiderer, and Jeffrey Zwiebel, for valuable
comments and suggestions. I would also like to thank the anonymous referees, David Hirshleifer (the editor),
Francesco D’Acunto (discussant), David Chapman, Peter DeMarzo, Darrell Duffie, Joseph Grundfest, Sebastian
Infante, Dirk Jenter, Jens Josephson (discussant), Ron Kaniel, Arthur Korteweg, Doron Levit, Andrey Malenko,
George Parker, Francisco Perez-Gonzalez, Jeffrey Pontiff, Jun “QJ” Qian, Francisco Santos, Kenneth Shotts, Eric
So, Philip Strahan, Luke Taylor (discussant), James Van Horne, Bilge Yilmaz, and seminar participants at Boston
College, the University of British Columbia, University of California, Berkeley, University of Chicago, Columbia
University, Harvard University, University of Houston, London Business School, London School of Economics,
New Economic School, New York University, University of North Carolina at Chapel Hill, Northwestern
University, Stanford University, Washington University in St. Louis, the 2011 WFA Meeting, the 2011 SED
Meeting, the 2011 EFA Meeting, and the 2011 ESEM Meeting for helpful comments and discussions. The online
Appendix can be found on the Review of Financial Studies web site. Send correspondence to Nadya Malenko,
Boston College, Carroll School of Management, Finance Department, Fulton Hall 332, 140 Commonwealth
Avenue, Chestnut Hill, MA 02467; telephone: (617) 552-2178. E-mail: [Link]@[Link].

© The Author 2013. Published by Oxford University Press on behalf of The Society for Financial Studies.
All rights reserved. For Permissions, please e-mail: [Link]@[Link].
doi:10.1093/rfs/hht075 Advance Access publication December 2, 2013

[18:17 8/4/2014 [Link]] Page: 1486 1486–1532

 Communication and Decision-Making in Corporate Boards

In practice, communication in boards is impeded by several factors. In

many board meetings, time allocated for discussion of important issues is very
limited, and managers use their power to prevent any contentious discussions.1
Because of that, directors have been increasingly engaging in discussions
outside board meetings (Hart 2003), which is difficult given directors’ busy
schedules. According to a former board chairman, “If you want to be influential
on a board it takes time, so you have to make the commitment of time. Cause
you’re gonna have a lot of discussions outside of board meetings” (Stevenson
and Radin 2009). Moreover, discussions outside board meetings can result
in retaliation from the manager, who may want to prevent directors from
communicating behind his back. Another obstacle to communication is that
directors often refrain from openly expressing their views due to reputational

Downloaded from [Link] at Boston College on February 15, 2015

concerns, such as the fear to appear incompetent by voicing a controversial
opinion, or to be perceived as a troublemaker.2 Finally, given directors’ diverse
preferences, backgrounds, and areas of expertise, conveying one’s knowledge
effectively and persuasively requires preparation and effort. All these factors
make it personally costly for directors to communicate their position openly
and effectively.3
The goal of this paper is to examine whether existing board policies mitigate
or exacerbate the problem of ineffective communication, and to understand
which policies are optimal. To study this question, I develop a theory of
board decision-making whose key element is that the quality of discussion
is endogenous and determined by the efforts directors put into communicating
their position to others. The model yields novel implications for the choice
between open and secret ballot voting, the role of board diversity, executive
sessions, transparency, and board committees. These implications follow from
two main results, which show that when communication is personally costly,
certain biases in directors’ preferences can improve the quality of board
discussions. In particular, both directors’ concerns about conformity and
diversity in directors’ private interests can give directors stronger incentives
to incur the costs and communicate their position to others.
Anecdotal and survey evidence suggests that conflicts of interest and pressure
for conformity play an important role in directors’ decisions. Conflicts of
interest arise due to private benefits directors receive from certain board

1 According to Lipton and Lorsch (1992), “Too much of this limited time is occupied with reports from
managements and various formalities. In essence, the limited time outside directors have together is not used
in a meaningful exchange of ideas.” The manager’s ability to prevent debate is especially high if the CEO and
chairman roles are combined. For example, Michael Eisner, the former CEO and chairman of Disney, was known
to allow very little discussion and no heated disagreement during board meetings. See “Are Disney’s Directors
Only Eisner’s Puppets?” Los Angeles Times, February 16, 2004.
2 Sonnenfeld (2002) describes a story of a CEO who was invited to join the board of a well-known company. He
was told that new directors were expected to say nothing for the first 12 months. The candidate answered “Fine,
I’ll see you in a year,” and never got the appointment.
3 The fact that communication can be privately costly for the sender is well recognized in the literature. Dewatripont
(2006) provides a review of the relevant literature. See Section 2.1 for a more detailed discussion.

1487

[18:17 8/4/2014 [Link]] Page: 1487 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

decisions. For example, such conflicts are likely to arise in corporate control
transactions due to directors’ ownership, affiliation, or relation with the CEO.
Desire for conformity, which makes directors reluctant to deviate from other
board members, is perceived by many as equally important.4 This reluctance
to take a minority stand is due to several reasons, including the influence of
the CEO and directors’ reputational concerns. For example, anecdotal evidence
suggests that directors who oppose the CEO without support from other board
members face retaliation and pressure to resign.5
The paper develops a model that incorporates the key features of board
decision-making described above—costly communication, conflicts of interest,
and pressure for conformity. In the model, the board contemplates a decision
whose value is uncertain—for example, an acquisition. Each director has

Downloaded from [Link] at Boston College on February 15, 2015

private information relevant to the decision. The board’s decision process
takes place in two stages—communication, followed by decision-making.
At the communication stage, each director decides whether to incur a cost
to communicate his information to others. At the decision-making stage,
all directors take actions (e.g., vote) based on their private information
and the information inferred from the discussion, and the board’s collective
decision is made. Directors can have conflicts of interest and thus prefer
a decision that is not optimal for shareholders. Directors can also have
a preference for conformity and thus incur a loss if their action deviates
from other directors’ actions—for example, if they vote differently from the
majority.
As a benchmark, I show that if decisions are made without prior
communication—that is, if directors vote just on the basis of their private
signals, then both conflicts of interest and conformity are detrimental for board
decisions. However, the conclusions are different when prior to making the
decision, directors can choose to communicate their information to others at
a cost.
The first result shows that pressure for conformity at the decision-making
(voting) stage gives directors stronger incentives to incur the costs and
communicate their information to others prior to the vote. By encouraging better
communication, conformity at the decision-making stage can have an overall
positive effect on board decisions. As an illustration of this result, consider the

4 According to one director, “Groupthink is one of the greatest problems boards face” (Leblanc and Gillies 2005).
Schwartz-Ziv and Weisbach (2013) examine board minutes of Israeli companies and find that dissension is very
rare: in 97.5% of the 1,422 votes they observe, the vote was unanimous. Survey evidence in Lorsch and MacIver
(1989) suggests that 49% of directors feel inhibited in taking a minority stand. See also Sonnenfeld (2002) and
“What boards should know about groupthink” (Compliance Week, May 19, 2009) for case studies of conformity
in the boardroom.
5 Mace (1986) describes a case study in which an outside director was excluded from the company’s proxy
statement after openly criticizing the manager’s press releases during a board meeting. “Don’t raise questions
with the president unless you can, for sure, count on the support of others on the board,” commented the director
afterward. The first section in the Appendix discusses several microfoundations for conformity and explains why
the results are robust to different microfoundations.

1488

[18:17 8/4/2014 [Link]] Page: 1488 1486–1532

 Communication and Decision-Making in Corporate Boards

difference between open and secret ballot voting. Anecdotal evidence suggests
that in the majority of cases, directors vote by open ballot.6 At first glance, this
seems puzzling: pressure for conformity is stronger when voting is by open
rather than secret ballot because the vote of each director is observable to other
directors and the CEO. Thus, the open ballot system is likely to prevent directors
from using their information and honestly voting their opinions. However, my
first result implies that open ballot voting can nevertheless be optimal because
it improves communication prior to the vote.
To see the intuition, suppose that the board contemplates a merger supported
by the manager. If voting is by secret ballot, a director with negative information
about the prospects of the merger will vote against it. Even if he is the only
one to vote negatively, he will not be identified by the manager and hence will

Downloaded from [Link] at Boston College on February 15, 2015

not be penalized. However, given potential costs of sharing his concerns with
others, the director may be reluctant to speak up during the discussion and will
simply vote against. Conversely, suppose that voting is by open ballot. Unless
the director shares his negative information prior to the vote, he faces the risk
that he will be the only one to vote against and hence will be penalized by the
manager. Realizing this, the director has strong incentives to share his concerns
with others, for example, outside the board meeting. By communicating his
information, he ensures that other directors also become pessimistic about
the merger and will support him in his opposition to the CEO. Thus, board
communication is more effective under open ballot voting.
In general, the choice between open and secret ballot (stronger and
weaker pressure for conformity) depends on several factors, including the
decision under consideration, the nature of directors’ information, and their
private interests. Section 4 presents the implications of the model for voting
procedures, transparency, and executive sessions, and discusses anecdotal
evidence consistent with these implications. Pressure for conformity may
exist at the communication stage as well. In particular, it may be costly
for directors to voice a dissenting opinion, not only to vote in dissent.
However, such decision-making policies as the use of open versus secret ballot
voting affect pressure for conformity at the voting stage without affecting
conformity at the communication stage. Thus, the first result shows that various
inefficiencies during the discussion (due to conformity, fear of the CEO, or
direct communication costs) can be overcome by increasing conformity at the
voting stage, for example, by using open ballot voting. Section 5.2 discusses
these issues in more detail.
The second result of the paper shows that a stronger conflict of interest
relative to other board members gives a director stronger incentives to
communicate his information to them. Thus, board discussions can be more

6 For example, in the sample of Schwartz-Ziv and Weisbach (2013), all board votes were taken by open ballot.
See also the discussion in Rock (2004). Swartz (2007) documents that while open ballot voting is most common,
some firms adopt secret ballot voting due to shareholder pressure or by management initiative.

1489

[18:17 8/4/2014 [Link]] Page: 1489 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

effective when directors’ private interests are more diverse. For example, this
implies that in the context of a takeover, a board that includes both insiders,
who are biased against the takeover, and bidder representatives, who are biased
in its favor, can be more effective than a fully independent board.
The intuition is the following. Consider a firm that receives a takeover
bid and an inside director who is biased against being taken over because of
private benefits of control. Other board members expect the director to actively
participate in the discussion and present evidence that the bid undervalues
the target whenever he has such evidence. Hence, if the director does not
speak up, other directors infer that he privately knows that the target’s value
is low. This negative inference increases the probability that the board will
accept the bidder’s offer. The more biased is the director relative to other

Downloaded from [Link] at Boston College on February 15, 2015

board members, the more harmful for him is their negative inference when
he does not speak up. At the margin, this gives a more biased director stronger
incentives to incur the costs of communication and try to credibly convey his
information.
Finally, the paper has implications for the allocation of decision-making
rights among directors. I show that even if directors are identical, it can be
optimal to allocate full decision-making power to one director, as opposed to
giving voting rights to all directors. In the presence of communication costs,
such allocation of control leads to the most efficient use of directors’ diverse
information, with the director in control serving as a communication link.
Directors without decision-making power communicate their information to
the director in control, who makes the decision by incorporating all available
information. Section 4 discusses the implications of this result for the advisory
role of the board and board committees. I also show that when directors
are heterogeneous, it is optimal to allocate control to directors who care
the least about conforming to others because such directors distort their
decisions the least. This is consistent with the observed practices, where
more experienced directors (who are likely to have lower costs of disagreeing
with others) are more likely to be appointed to the chairman or lead director
positions.
The paper proceeds as follows. The remainder of this section reviews
the related literature. Section 1 presents the benchmark case without
communication, and Section 2 analyzes the general setup. Section 3 considers
an extension of the model to differential decision-making rights. Section 4
presents the policy and empirical implications. Section 5 discusses the
robustness of the results, and Section 6 concludes. All proofs are given in
the Appendix.

Related literature
The paper contributes to the theoretical literature on corporate boards. Most
papers in this literature focus on the board-CEO interaction and consider the

1490

[18:17 8/4/2014 [Link]] Page: 1490 1486–1532

 Communication and Decision-Making in Corporate Boards

board as a single agent.7 In contrast, this paper considers the board as a collective
decision-making body and focuses on the interaction between directors.8 In
this respect, it is most closely related to Warther (1998) and Chemmanur and
Fedaseyeu (2011), who analyze directors’ decisions whether to fire the CEO
in the presence of costs of dissent: directors who vote against the CEO incur a
loss, but only if their vote deviates from the majority. Hence, costs of dissent
are similar to pressure for conformity. Warther (1998) and Chemmanur and
Fedaseyeu (2011) show that costs of dissent introduce a coordination problem:
directors may not vote against the CEO even if each of them has negative private
information about him. My paper contributes to these papers by allowing for
communication prior to voting. It emphasizes that communication can alleviate
the coordination problem and also shows that when communication is costly,

Downloaded from [Link] at Boston College on February 15, 2015

higher costs of dissent improve communication. Communication within the
board is also studied in Harris and Raviv (2008), who compare insider- and
outsider-controlled boards when the party without control can communicate its
information to the party in control by sending a costless message. My paper
is different from Harris and Raviv (2008) in that it studies collective decision-
making by directors and focuses on costly communication.9
The theoretical contribution of this paper is to extend the literature on
voluntary information disclosure, started by Verrecchia (1983), to the case of
multiple senders and multiple receivers of information. The framework allows
directors to be different in their preferences, information quality, and other
characteristics, but remains very tractable.
The literature on costly communication acknowledges that communication
between team members is costly and studies the effects of specialization and
decentralization (see Dewatripont 2006 for a review). The closest paper in
this literature is Dewatripont and Tirole (2005), where both the sender and
the receiver need to incur costly effort to improve communication between
them. The focus of their paper is on the “moral hazard in teams” problem in
communication.
The paper is also related to the literature on voting in committees,
and specifically, to several papers in this literature that allow for pre-vote
communication.10 Unlike my paper, these papers study a cheap talk setting.
My paper is also different in the way it models the decision-making stage.

7 See, e.g., Hirshleifer and Thakor (1994), Hermalin and Weisbach (1998), Almazan and Suarez (2003), Song
and Thakor (2006), Adams and Ferreira (2007), Kumar and Sivaramakrishnan (2008), Burkart and Raff
(2012), Chakraborty and Yilmaz (2011), and Levit (2012). Adams, Hermalin, and Weisbach (2010) provide
a comprehensive review of the literature.
8 Baranchuk and Dybvig (2009) also study individual directors’ preferences and information, but use a cooperative
solution concept, while my paper models directors’ behavior explicitly.
9 Communication between directors is also considered in Raheja (2005), where informed insiders, who compete
with each other to become the CEO’s successor, can reveal their information to uninformed outsiders.
10 E.g., Coughlan (2000), Gillette, Noe, and Rebello (2003), and Austen-Smith and Feddersen (2005). Gerling et al.
(2005) provide a review of the literature on decision-making in committees.

1491

[18:17 8/4/2014 [Link]] Page: 1491 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

Because the analysis of voting with communication is technically involved

and gives rise to multiple equilibria, the above papers focus on two- or three-
member committees or assume a degenerate information structure. My paper
abstracts from voting, which allows a closed-form solution for any board size
and a general information structure.
Finally, the paper is related to the literature on agents’ concerns about
conformity. Scharfstein and Stein (1990), Banerjee (1992), Bikhchandani,
Hirshleifer, and Welch (1992), and Zwiebel (1995) show how incentives to
conform arise endogenously, when agents have reputational concerns or infer
information from other agents’ actions. Morris and Shin (2002) take concerns
about conformity as given and compare the effects of public versus private
information on social welfare. In their paper, the separation of information

Downloaded from [Link] at Boston College on February 15, 2015

into public and private is exogenous, while in my paper it arises endogenously
through directors’communication decisions. Alonso, Dessein, and Matouschek
(2008) study communication between agents who benefit from coordinating
their actions but focus on costless communication.

1. Benchmark Case: No Communication

The analysis starts with the benchmark case without communication. I show
that in this case, conflicts of interest and pressure for conformity reduce board
effectiveness. Section 2 shows that these biases can have the opposite effect
when costly communication is allowed.

1.1 Setup
1.1.1 Information structure. The board, which consists of N directors, is
contemplating a decision. I adopt the information structure and firm value
specification of Harris and Raviv (2005, 2008) and assume that the value of the
firm is equal to
V (a,θ ) = V0 −(a −θ)2 , (1)
where a is the decision made by the board, and θ is the unknown state of the
world, which is equal to the sum of N independent signals xi and an independent
noise term ε:
 N
θ= xi +ε. (2)
i=1

Signal xi is distributed on [−ki ,ki ], where ki can be infinite, with a density

function fi (·), which is symmetric around zero. A director with a higher signal
variance can be interpreted as having more expertise. Director i perfectly
observes xi but has no information about other signals. The term ε is distributed
with a density function fε (·) with E[ε] = 0 and corresponds to the information
that is not known to any director. As discussed in Sections 2 and 5.4, the results
would remain unchanged if directors’ signals were correlated or if directors
observed their signals only with some probability. Expressions (1) and (2)

1492

[18:17 8/4/2014 [Link]] Page: 1492 1486–1532

 Communication and Decision-Making in Corporate Boards

imply that a social planner

 who has access to all directors’ private signals will
choose decision a = N i=1 i to maximize the value of the firm.
x
To illustrate this setup, suppose that the firm is a potential target, and its
board decides on the minimum price a it is willing to accept (other examples
include choosing the scale of investment or the maximum price to pay in an
acquisition). For example, in the well-known Air Products’ takeover fight for
Airgas, directors ofAirgas strongly disagreed with each other over the minimum
price they would accept.11 While the minimum price set by the board should
not be too low, it should also not be too high since it can deter the bidder
completely. Hence, there exists some optimal minimum price a ∗ = θ . This price
is determined by several factors, including the value of the target as a stand-
alone firm, the number and quality of potential bidders, and operational and

Downloaded from [Link] at Boston College on February 15, 2015

financial synergies. These factors are represented by signals xi and ε. Directors
have different areas of expertise and thus have information that is relevant to
different aspects of the decision.

1.1.2 Decision-making stage. Since the board is a collective body, its final
decision is the result of collective decision-making by directors. Any modeling
of one-shot collective decision-making implies that the board’s decision is
some, potentially probabilistic, function of individual directors’ actions: a =
h(a1 ,...,aN ).
In the main part of the paper, I focus on the “random dictator” rule, which
has been discussed in the political economy literature as an alternative to
majority voting.12 According to the “random dictator” rule, if directors’ actions
are a1 ,...,aN , then the board’s decision equals ai with probability N1 . For
example, in the context of a target’s board deciding on the minimum takeover
price, each director proposes the price ai that he thinks should be set. If K
out of N directors propose price p1 and N −K directors propose price p2 ,
then p1 is chosen with probability K N
. The advantage of this specification is
that unlike most models of collective decision-making with communication,
it allows a tractable solution. In particular, regardless of signal distributions,
there is a unique linear equilibrium at the decision-making stage and closed-
form expressions for communication strategies and firm value. Combined with
(1), this decision rule implies that firm value is equal to:

1
N
V (a1 ,...,aN ,θ) = V0 − (ai −θ )2 . (3)
N i=1

11 The Airgas chairman claimed that the board had unanimously agreed on the minimum acceptable price of $78
a share. However, 3 out of 10 Airgas directors disputed that they agreed with this price. See “Airgas directors
disagree over minimum price while fending off hostile bid,” Bloomberg, December 13, 2010.
12 See, e.g., Zeckhauser (1973). Zhang and Casari (2012) present experimental evidence that the actual decision-
making process in groups lies between the “random dictator” and the majority rule processes.

1493

[18:17 8/4/2014 [Link]] Page: 1493 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

The results of the paper hold for a wider set of decision rules. Section 3
generalizes the model to the case in which some directors have more decision-
making power than others, while the online Appendix analyzes a general
linear specification of h(a1 ,...,aN ), as well as a model with voting.13 More
generally, the intuition behind the results holds for decision rules such that in
the absence of communication, the decision-making stage alone does not allow
efficient aggregation of directors’ signals (i.e., it does not result in the social
planner’s choice), and hence communication has value. Majority voting is the
simplest example of an inefficient decision rule because directors are restricted
to binary actions, which cannot convey continuous and multidimensional
information.14 The “random dictator” rule, analyzed in the basic setup, does
not efficiently aggregate directors’ information either. The formal definition of

Downloaded from [Link] at Boston College on February 15, 2015

efficient decision rules is as follows.

Definition 1. Let ai∗ (xi ) be the equilibrium action of director i in the absence
of communication given decision rule h(a1 ,...,aN ) when the director has
information xi . This decision rule is said
 to efficiently aggregate
 information
if for any realization of x1 ,...,xN , h a1∗ (x1 ),...,aN∗ (xN ) equals the social

planner’s choice N i=1 xi with probability 1.

The motivation for the assumption of an inefficient decision rule is that

efficient rules are difficult to implement in practice. First, directors’information
is complex and multidimensional, and its distribution is unknown. Thus, to be
efficient, the decision rule would need to be tailored to the specific company
(based on its board structure and directors’expertise) and to the specific decision
under consideration. Even if the information structure is known, an ex post
efficient decision rule may not exist15 or may be costly to implement. For
example, a large literature studies how communicational constraints make
implementation of mechanisms prohibitively costly (see Segal 2006 for a
review). For these reasons, many papers on decision-making in organizations
restrict attention to specific decision rules, which are easy to implement but are
not efficient. The delegation literature focuses on delegating authority to a single
agent (Dessein 2002; Harris and Raviv 2008; Chakraborty and Yilmaz 2011),
and the literature on committees focuses on voting (see Gerling et al. 2005 for a

13 The general linear specification of h(a ,...,a ) is characterized by m vectors (γ

1 N 1(j ) ,...,γN (j ) ), γi(j ) ≥ 0,
N m
i=1 γi(j ) = 1, and probabilities qj > 0, j =1 qj = 1, such that the board’s decision equals the linear combination
N
i=1 γi(j ) ai with probability qj .
14 For example, if all but one director have moderately positive signals but the remaining director has an extremely
negative signal about the value of a proposal, the proposal is detrimental to firm value. Nevertheless, in the
absence of communication, such a proposal will be approved by a majority of the votes.
15 Committing to an ex post inefficient decision rule can be optimal to give agents incentives to report truthfully.
For example, Li, Rosen, and Suen (2001) show that when committee members have conflicting preferences,
there is no mechanism that efficiently pools members’ private information, and that decisions have to be made
through voting (or other procedures that partition continuous information into ranks).

1494

[18:17 8/4/2014 [Link]] Page: 1494 1486–1532

 Communication and Decision-Making in Corporate Boards

review). I adopt the approach in these literatures and consider decision rules that
inefficiently aggregate information. This assumption makes communication
valuable.

1.1.3 Preferences. To capture conflicts of interest and conformity, I assume

that the utility of director i is:
ri 
Ui (a,a1 ,...,aN ,θ) = −(a −(bi +θ))2 − (ai −ak )2 , (4)
N −1 k=i

where a is the board’s decision. These preferences are common knowledge. As

discussed in Section 5.4, the results would continue to hold if preferences were

Downloaded from [Link] at Boston College on February 15, 2015

unknown.
The director’s utility has two components. The first component reflects
conflicts of interest: the director’s preferred decision is bi +θ , rather than θ.
For example, due to private benefits, an inside director may prefer the target
to remain independent and would like to set a higher minimum price than is
optimal for shareholders. I will refer to bi as a “directional bias” in the rest
of the paper, capturing the idea that directors may be biased in one or the
other direction. The second component reflects pressure for conformity: the
director suffers a loss if his action deviates from the actions of other directors.
Since the board’s decision equals ak with probability N1 , the director effectively
minimizes the expected squared distance of his action from the final board’s
decision. I will refer to ri as the director’s “conformity bias” in the rest of the
paper. This specification is similar to Morris and Shin (2002) and captures,
in a reduced-form way, different reasons for conformity, such as directors’
reputational concerns and desire to avoid confrontation. The first part of the
Appendix discusses several explanations for the conformity bias and explains
why the results of the paper are robust to different explanations.
Specification (4) does not explicitly incorporate the fear of the CEO as
one of the reasons for the conformity bias. The online Appendix presents two
extensions of the model that capture directors’ desire to support the CEO and
shows the robustness of the results. Section 5.1 contains a discussion of these
extensions.
This setup is sufficiently flexible to describe different board structures. For
example, the presence of the CEO on the board could be captured by assuming
that one of the directors has a large directional bias, has a small conformity bias,
is particularly informed (Exi2 is high relative to Exj2 ), and is more influential
than other directors (pi is high in the extension of Section 3).16

16 When r < 0, this framework captures a preference for disagreeing, when a director derives utility from taking a
i
contrarian position. The effect of such preferences is discussed in note 24. In addition, in Section 5.6, I analyze a
generalized specification with weight w(ri ) on the firm value term. Finally, the online Appendix analyzes more
general preferences, where directors have reputational concerns and derive utility from taking the correct action
even if the ultimate decision of the board is different. I show that such reputational concerns make directors’
actions more efficient but reduce their incentives to share information.

1495

[18:17 8/4/2014 [Link]] Page: 1495 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

Combined with the above specification for h, (4) implies that the director’s
utility is

1 ri 
N
Ui (a1 ,...,aN ,θ) = − (ak −(bi +θ ))2 − (ai −ak )2 . (5)
N k=1 N −1 k=i

1.2 Analysis of the benchmark case

Throughout the paper, the solution concept is perfect Bayesian equilibrium
(PBE).
Let ai (Ii ) be the action of director i given his information set Ii at the
beginning of the decision-making stage. By (5), this action is determined by

Downloaded from [Link] at Boston College on February 15, 2015

the first-order condition
 
1 1
ai (Ii ) = (E[θ |Ii ]+bi )+ 1− E[ā−i |Ii ], (6)
1+N ri 1+N ri

where ā−i = N 1−1 j =i aj . If there are no biases in the director’s preferences
(ri = 0 and bi = 0), his action is ai (Ii ) = E[θ |Ii ], which is his best estimate of
the state of the world. The presence of biases introduces distortions into the
director’s behavior. A conformity bias gives him an incentive to mimic the
actions of other directors. In particular, when ri > 0 and bi = 0, the director tries
to balance his desire to maximize firm value and his desire to conform to other
1
directors. The weight 1+N ri
, attached to his best estimate of the state, decreases
as the director’s conformity bias becomes larger. The presence of a directional
bias introduces an additional distortion, inducing the director to pursue the
decision E[θ|Ii ]+bi instead of E[θ |Ii ].
Because there is no communication in the benchmark case, director i only
knows his own signal xi . Hence, E[xj |Ii ] = 0 for j  = i by the independence
assumption, and E[θ|Ii ] = xi .
I now show that there is a unique linear equilibrium of the game. Suppose
that a linear equilibrium exists—that is, ai is linear in the director’s signal:
ai = γi xi +gi . Plugging in the conjectured strategies of other directors in the
first-order condition (6), we get:
 
1 1 1
ai = xi + bi + 1− ḡ−i , (7)
1+N ri 1+N ri 1+N ri

where ḡ−i = N 1−1 j =i gj . Thus, the best response strategy of director i is indeed
1
linear in his signal. Comparing the coefficients, we see that γi is equal to 1+N ri
,
and gi is defined by a system of linear equations:
 
1 1
gi = bi + 1− ḡ−i . (8)
1+N ri 1+N ri
Lemma A.2 in the Appendix shows that (8) has a unique solution given by (15)
below. In particular, gi = bi if ri = 0, and gi = 0 if bi = 0 for all i. We conclude

1496

[18:17 8/4/2014 [Link]] Page: 1496 1486–1532

 Communication and Decision-Making in Corporate Boards

that there is a unique linear equilibrium:

1
ai∗ = gi + xi . (9)
1+N ri
If there are no biases in directors’ preferences (ri = bi = 0 for all i), then
ai∗ = xi . However, even if information about the state is known to the board as a
whole (ε ≡ 0), the board does not make the first-best decision xi because each
director only knows his own signal xi . This inefficiency captures the assumption
that the decision-making stage alone does not efficiently aggregate directors’
signals and hence communication improves board decisions. In particular, (3)
implies that for ε ≡ 0, the first-best is achieved if and only if each director takes
action xi , a necessary condition for which is that all directors have shared

Downloaded from [Link] at Boston College on February 15, 2015

their information with others.17
The presence of biases introduces additional inefficiencies:
1. Only conformity biases. If directors have conformity biases but no
directional biases (ri > 0,bi = 0), then ai∗ = 1+N 1
x . Thus, a conformity bias
ri i
induces a director to put less than optimal weight on his private signal. In
the extreme case, when ri is infinitely large, the director does not care about
the correct decision being made and wishes only to deviate as little as possible
from other directors. Because the expected value of other directors’ signals is
zero, the director takes action ai∗ = 0. Hence, in this case, his private information
never affects the board’s decision.
2. Only directional biases. If directors have directional biases but no
conformity biases (bi  = 0,ri = 0), then ai∗ = bi +xi . The presence of a directional
bias induces the director to push the board’s decision in the direction of his
bias, moving it farther away from the optimal decision from shareholders’
perspective.
To summarize, in the absence of communication, both conformity and
directional biases impede decision-making by distorting directors’ actions at
the decision-making stage.

2. Communication Prior to Decision-Making

This section introduces a communication stage prior to the decision-making
stage. It starts with the setup of the communication stage and then presents
the analysis. To illustrate the results more clearly, I first analyze the
setting where directors have only conformity biases (Section 2.2), and then
generalize the results to the setting where both types of biases are present
(Section 2.3).

17 Note that directors’ actions reveal information about their signals. However, the decision-making rule h is
interpreted as a binding rule that matches directors’ actions to the final decision, i.e., as an analog to majority
voting. Since the rule is binding, I do not allow the social planner to use the information revealed through
directors’ actions to change the board’s decision after the decision-making stage.

1497

[18:17 8/4/2014 [Link]] Page: 1497 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

2.1 Communication stage

Factors such as CEO power, time constraints, and reputational concerns make
it personally costly for directors to communicate their position openly and
effectively. In addition, communication may require time and effort due to
directors’ diverse preferences and backgrounds. If a director’s preferences
are not aligned with those of other board members, he has incentives to
misrepresent information. This inhibits communication (Crawford and Sobel
1982) and requires the director to expend effort to make his report more
credible (see Section 5.3 for an additional discussion). Even if preferences
are aligned, directors’ diverse backgrounds and expertise, together with their
complex and multidimensional information, may require additional effort to
make communication effective. For example, it can be difficult for a director

Downloaded from [Link] at Boston College on February 15, 2015

with a marketing background to convey his knowledge of the product market
to directors with financial expertise. Dewatripont and Tirole (2005), who
also study costly communication, point out that the sender needs to address
receiver-specific knowledge and make his report both detailed enough to convey
all relevant information and concise enough so as not to distract attention.
They illustrate communication costs in the context of academic presentations
and journal submissions and interpret them as the costs of transforming soft
information into hard.
I model these different types of communication costs in the simplest possible
way. Specifically, I assume that at the communication stage, each director
simultaneously decides whether to incur a cost ci to convey his signal to others.
If the director incurs the cost, other directors learn xi with certainty.18 Directors
can withhold their information but cannot lie—that is, information is verifiable.
Generally, the literature uses two approaches to model communication. One
approach, started by Grossman (1981) and Milgrom (1981), assumes that agents
cannot lie. The other, started by Crawford and Sobel (1982), assumes that
agents can lie without incurring any penalties. Actual board communication lies
somewhere in between these two extremes: some misrepresentation is possible,
but significant misrepresentation is likely to result in reputational penalties or
other costs.19 I adopt the first approach because it is more tractable and allows
me to demonstrate the intuition in the simplest way, and discuss the robustness
of the results to allowing for information manipulation in Section 5.3.
A fixed cost ci is a reasonable way to capture direct costs of communication—
that is, the time and effort needed to present information effectively. Other
communication costs, coming from the director’s reputational concerns and
his reluctance to alienate the CEO, are likely to depend on the director’s
signal: these costs are likely to be higher if the signal is extreme and deviates

18 The benchmark case corresponds to c = ∞ for all i . The assumption that directors communicate simultaneously
i
and disclose their signals to the whole board is made for simplicity. This assumption is standard in models of
committee decision-making (Coughlan 2000; Austen-Smith and Feddersen 2005).
19 Cornelli, Kominek, and Ljungqvist (2013) discuss the use of hard vs. soft information in board decisions.

1498

[18:17 8/4/2014 [Link]] Page: 1498 1486–1532

 Communication and Decision-Making in Corporate Boards

significantly from the position of the CEO. For simplicity, the basic model
captures these costs in a reduced-form way, together with direct costs of
communication. Section 5.2 discusses a more general modeling of these costs
and why the results are robust to this assumption.20
As will become clear below, the main results of the paper rely on the
assumption that directors’ communication costs are strictly positive. Only in
this case can conformity and directional biases have a positive effect on board
decisions by encouraging communication. In contrast, as the following lemma
shows, when ci ≤ 0, directors always disclose their information, and hence the
quality of board discussions is not affected by these biases.

Lemma 1. If ci ≤ 0, director i always reveals his signal.

Downloaded from [Link] at Boston College on February 15, 2015

From now on, I assume ci > 0 and discuss the case ci ≤ 0 at the end of Sections
2.2 and 2.3.

2.2 Conformity biases

The analysis begins with the case in which directors have only conformity
biases: ri ≥ 0, bi = 0. To find the PBE, the model is solved by backwards
induction. Let JC ⊆ {1,...,N} be the set of signals that were communicated
at the first stage and JN C = {1,...,N}\JC be the set of signals that were not
communicated. Either of the sets could be empty. Suppose also that given
the equilibrium strategies at the communication stage, the expected value of
director i’s signal conditional on it not being communicated is yi .

2.2.1 Equilibrium at the decision-making stage. Let us search for linear

equilibria at the decision-making stage. In a linear equilibrium, the action of
director i is some linear function of signals xj ,j ∈ JC that were communicated,
and his private signal xi . Similarly to the benchmark case, using (6), it can be
shown that there is a unique linear equilibrium, characterized by the following
proposition.

Proposition 1. Suppose that at the communication stage signals xi ,i ∈ JC

were communicated, signals xi ,i ∈ JN C were not communicated, and yi is
the expected value of xi conditional on no communication. Then there is the
following linear equilibrium at the decision-making stage:
1. If director i communicated his signal, i ∈ JC , his action is given by
 
ai∗ = xj + yj . (10)
j ∈JC j ∈JN C

20 When a director takes the time to explain his position to others, he not only incurs a cost himself, but may also
impose a cost on other directors. Introducing the costs of receiving information would affect directors’equilibrium
utility, but would not change the equilibrium strategies and firm value. This is because communication decisions
are made simultaneously and directors care only about other directors’ actions, not about other directors’ utility.

1499

[18:17 8/4/2014 [Link]] Page: 1499 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

2. If director i did not communicate his signal, i ∈ JN C , his action is

given by
   
1 1
ai∗ = xj + yj + xi + 1− yi . (11)
j ∈J j ∈J ,j  =i
1+N ri 1+N ri
C NC

The intuition is the following. First, if director i did not communicate his
signal (i ∈ JN C ), other directors’ actions will be based on the expectation of his
signal, yi . Indeed, according to (10) and (11), all directors except i put weight
1 on yi . As in the benchmark case, a conformity bias then induces the director
to put less than optimal weight on his private signal to move his action closer to
other directors’ actions. This is captured by the last two terms in (11): instead

Downloaded from [Link] at Boston College on February 15, 2015

of using xi with weight 1, as would be optimal from shareholders’ perspective,
the director uses a weighted average of his signal, xi , and the expectation
of his signal by others, yi . Second, consider a director who communicated his
signal (i ∈ JC ). By communicating, the director made other board members fully
internalize his information and use it in their decisions. Indeed, as (10) and (11)
show, all other directors now put weight 1 on xi .As a result, even though director
i still has a conformity bias, conformity no longer induces him to underreact
to his information. By using his signal xi with optimal weight 1, the director
minimizes the expected difference of his action from other directors’ actions.
Thus, signals that were communicated are used efficiently by all directors.

2.2.2 Equilibrium at the communication stage. At the communication

stage, each director takes into account the equilibrium strategies at the
decision-making stage and compares his expected payoff from paying the
cost ci to communicate his signal to others and the expected payoff from
not communicating. The following proposition characterizes the equilibrium
communication strategies.

Proposition 2. There exists an equilibrium in which director i communicates

his signal xi if and only if |xi | > di , where
 1/2
ci
di = . (12)
1− N1 1+N1
ri

Due to the symmetry of density functions fi (·) around zero and because
directors do not have directional biases, the noncommunication intervals
[−di ,di ] are symmetric around zero, and hence yi = 0. In the Appendix, I prove
that if the distribution of signals is single-peaked at zero (e.g., normal), the
equilibrium of Proposition 2 is unique.21

21 For a general distribution, there may be multiple equilibria. Importantly, the result that conformity at the decision-
making stage encourages communication holds in any chosen equilibrium. Multiple equilibria can arise if the

1500

[18:17 8/4/2014 [Link]] Page: 1500 1486–1532

 Communication and Decision-Making in Corporate Boards

The intuition behind the threshold strategies is the following. A director does
not have incentives to incur communication costs if his signal is close to its
expected value (|xi | < di ) because such disclosure is not valuable for decisions.
In contrast, because the director cares about firm value, he wants others to take
his signal into account if the signal is sufficiently extreme, and hence incurs
the costs of communication. This argument is valid regardless of whether or
not directors have conformity biases.
Importantly, conformity biases give directors additional incentives to share
their signals: di decreases in ri . To see the intuition, let us interpret directors’
actions as votes. Absent pressure for conformity during the vote (e.g., if voting
is by closed ballot), a director who does not share his information with others
will still vote based on this information and hence will affect the outcome

Downloaded from [Link] at Boston College on February 15, 2015

with his vote. However, when the director faces pressure to conform his vote
to others’ (e.g., if voting is by open ballot), he is reluctant to vote based on
his information unless he communicated it to others and convinced them of
his position (see (10) and (11)). Hence, unless the director communicates his
information to others, this information will not be incorporated into the board’s
decision. Because the director wants the correct decision to be made, he now
has stronger incentives to incur the costs of communication.22
Put less formally, when there is no pressure for conformity, the only benefit to
the director of communicating his information is that he makes other directors’
votes, and hence the ultimate decision, more informed. Pressure for conformity
creates an additional benefit of communication. By convincing others of his
position, the director ensures that his vote will be similar to other directors’
votes and hence avoids the costs of dissent.
Using the equilibrium strategies, one can calculate the expected value of the
firm:
Lemma 2. Expected firm value is equal to
N  2 di 
1 1 N ri
E(V ) = V0 −Eε −
2
1− + 2
x fi (x)dx . (13)
i=1
N N 1+N ri −di

Recall that V0 is the maximum possible firm value, 

which would be achieved
if the board made the fully informed decision θ = N i=1 xi +ε. The negative

distribution has several points of symmetry. For example, for a uniform distribution, which is symmetric around
any point, there is a continuum of equilibria characterized by yi ∈ [−ki +di ,ki −di ]. The intuition behind the
multiplicity of equilibria is similar to why there exist multiple self-fulfilling equilibria in rational expectations
models. The director does not have incentives to reveal his signal if it is close to others’ expectations conditional
on no communication (rather than their unconditional expectation, which is zero). Thus, if other directors believe
that conditional on no communication the expected value of xi is yi , these expectations become self-fulfilling.
22 This result does not rely on the assumption that signals are independent. As long as the correlation between
signals is not perfect, conformity induces a director to put less than optimal weight on his signal unless it is
communicated to others. Hence, by the same argument, conformity gives directors more incentives to convince
others of their position. The main difference between models with independent and correlated signals is that
correlation would give rise to a free-rider problem in communication: directors would have weaker incentives
to incur the costs, hoping that other directors with similar information would disclose it.

1501

[18:17 8/4/2014 [Link]] Page: 1501 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

terms in (13) reflect the fact that the board’s decision is not fully informed.
The term Eε 2 captures the fact that part of θ is not known to any director.
Subsequent terms capture the fact that directors’ private signals are not
used efficiently, either due to lack of communication or due to conformity.
Specifically, (13) emphasizes that there are two counteracting effects of the
director’s conformity bias ri on firm value. The negative effect is reflected
N ri 2
in the term 1− N1 + N1 ( 1+N ri
) : if the director did not communicate his signal,
conformity decreases the weight he puts on this signal at the decision-making
d
stage. The positive effect is reflected in the term −di i x 2 fi (x)dx: a stronger
conformity bias encourages communication and shrinks the noncommunication
region [−di ,di ]. The next result shows that in the current setting, the positive

Downloaded from [Link] at Boston College on February 15, 2015

effect dominates for a sufficiently small conformity bias.

Proposition 3. Suppose 0 < ci < ki2 (1− N1 ), where [−ki ,ki ] is the support of
 
xi , ki ∈ (0,+∞]. Then firm value is maximized at r1∗ ,...,rN∗ , where ri∗ is strictly
positive.23

Intuitively, since directors disclose the most important information (|xi | >
di ), the tradeoff of a higher ri is between more efficient use of more important
information by all directors and less efficient use of less important information
by only one director, which explains why the positive effect can dominate.24
The assumption of positive communication costs is crucial for Proposition 3.
As shown in Lemma 1, if ci ≤ 0, director i always reveals his signal, and hence
ri does not affect the communication and decision-making stages. As a result,
conformity biases have no effect on firm value.
The assumption that the decision rule inefficiently aggregates directors’
information is not needed for the result that conformity biases improve
communication (see the online Appendix, which analyzes all linear decision
rules, including efficient ones). It is important only for the result of Proposition 3
that the optimal conformity biases are positive. This is because communication,
which is encouraged by conformity biases, is helpful only when the decision
rule is inefficient.

2.3 Conformity and directional biases

This section studies the joint effect of directors’ conformity and directional
biases.

23 The assumption c < k 2 (1− 1 ) ensures that the communication cost is not too high relative to the support of the
i i N
distribution [−ki ,ki ] and is satisfied for any distribution with infinite support. If this condition is not satisfied,
the director does not communicate any information when ri = 0.
24 The analysis also illustrates the effect of preferences for disagreeing with others, which can be captured by
ri < 0. By (11), if ri < 0, the director overweighs his private signal in his decisions. Moreover, unlike the case of
conformity, there is no counteracting positive effect of contrarian preferences on communication. In fact, (12)
shows that contrarian preferences induce a director to withhold his signal, so that his actions are more likely to
be different from those of the rest of the board.

1502

[18:17 8/4/2014 [Link]] Page: 1502 1486–1532

 Communication and Decision-Making in Corporate Boards

Let ai∗ (b1 ,...,bN ) be the equilibrium action of director i at the decision-
making stage. Lemma A.1 in the Appendix shows that for any distribution of
signals,
ai∗ (b1 ,...,bN ) = gi +ai∗ (0,...,0), (14)
where ai∗ (0,...,0) are the equilibrium strategies in the absence of directional
biases, given by (10) and (11), and the constants gi solve the system of linear
equations (8). Lemma A.2 in the Appendix shows that this system has a unique
solution given by

N
gi = λij bj , (15)
j =1

Downloaded from [Link] at Boston College on February 15, 2015

where λii ,λij ∈ (0,1) if ri > 0, and λii = 1,λij = 0 if ri = 0. For example,
if ri = r for all i, the solution is given by gi = ωbi +(1−ω) b̄−i , where ω =
N −1+N r
N −1+N 2 r
.
Directional biases introduce distortions gi in directors’ actions. According
to (15), gi is a weighted average of the director’s own directional bias bi
and the biases of other directors bj , j  = i. First, gi strictly increases in bi :
the director biases his actions in the direction of his preferred outcome. In
the special case when ri = 0, gi coincides with bi . If, in addition, the director
has a conformity bias (ri > 0), gi also strictly increases in other directors’
biases bj , j  = i. Intuitively, if other directors are known to favor a certain
decision, a director who cares about conformity is more reluctant to oppose
this decision.
This suggests that when directors’ private interests are diverse, there is
an additional benefit from pressure for conformity at the voting stage: it
helps constrain opportunistic behavior. Absent pressure for conformity (ri = 0),
directors vote in a way that maximizes their own interests (gi = bi ). Hence, the
final decision, which is the preferred decision of the strongest group, is far
from the optimal decision from shareholders’ perspective. Conformity induces
directors to be more cautious in pursuing their individual interests, moving the
board’s decision closer to the optimal decision. For example, in the case of two
directors, if r1 = r2 = r and b1 = −b2 , distortions (g1 ,g2 ) are the highest for r = 0
and converge to zero as r increases.
In contrast, pressure for conformity can be detrimental when directors’
private interests are not diverse. In particular, conformity can induce
independent directors, who would otherwise make unbiased decisions, to favor
the policies preferred by their opportunistic colleagues: if ri > 0 and other
directors have positive directional biases, then gi > 0 even if bi = 0.
I next analyze the communication stage. Because the model with directional
biases is less tractable, I assume for the rest of the paper that the distribution of
signals is uniform and discuss the general distribution at the end of the section.
Lemma A.3 in the Appendix analyzes a general distribution.
Assumption 1. Signal xi is uniformly distributed on [−ki ,ki ].

1503

[18:17 8/4/2014 [Link]] Page: 1503 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

The following lemma summarizes the equilibrium at the communication

stage.

Lemma 3. The equilibrium communication strategies are the following:

(i) if bi > ḡ−i , director i reveals his signal if and only if xi > −ki +2δi+ ,
(ii) if bi < ḡ−i , director i reveals his signal if and only if xi < ki +2δi− ,

where ḡ−i = N 1−1 j =i gj and δi− < 0 < δi+ are the roots of the quadratic equation
ci
δ 2 +2δ(bi − ḡ−i )− 1 1
= 0. (16)
1− N 1+N

Downloaded from [Link] at Boston College on February 15, 2015

ri

Thus, in the presence of directional biases, communication strategies are no

longer symmetric around zero. A director who is biased relative to the rest of the
board (bi > ḡ−i ) reveals information that supports his bias (xi > −ki +2δi+ ) and
withholds information that does not support his bias.25 The next result shows
that when communication is costly, directional biases can have a positive effect
on communication:

Proposition 4. If director i is more biased than other directors, bi > ḡ−i ,

then he reveals more information as his bias bi increases further. Similarly,
if the director is less biased than other directors, bi < ḡ−i , then he reveals more
information as his bias bi decreases further.

To illustrate the intuition, I will refer to the example of a target’s board

deciding on the minimum acceptable price. In this example, a director with
bias bi > 0 gets private benefits from keeping the firm independent and hence
prefers to set a higher minimum price than is optimal for shareholders. Suppose
a director is more biased than other board members toward keeping the target
independent (bi > ḡ−i ). Then he discloses positive information about the target’s
value whenever he has it (whenever xi > Ti ≡ −ki +2δi+ ). Thus, if the director
does not speak up, other directors infer that his information is negative,
xi < Ti . When xi = Ti , the director is indifferent between incurring the costs
of communication and convincing others that xi = Ti and letting them make the
more negative inference E[xi |xi < Ti ]. The more biased is the director relative to
other board members, the more he wants them to believe that the target’s value

25 This property would continue to hold if a director received multiple signals and could selectively disclose part
of them. The director would disclose signals that support his bias and withhold signals that do not support his
bias. In addition, he would be more likely to disclose two signals together if these signals reinforce, rather than
contradict, each other. Disclosure in the presence of multiple signals has been examined by Fishman and Hagerty
(1990) and Pae (2005) in the context of disclosure by a single agent to investors.
Note also that if the cost of communication is large relative to ki , then −ki +2δi+ > ki and ki +2δi− < −ki , and
hence the director does not communicate any information in equilibrium. In what follows, I assume that ci is
sufficiently small relative to ki , so that at least some information is communicated by each director.

1504

[18:17 8/4/2014 [Link]] Page: 1504 1486–1532

 Communication and Decision-Making in Corporate Boards

is high. Hence, if bi increases further, the director is no longer indifferent: the

inference E[xi |xi < Ti ] becomes relatively more harmful for [Link] the margin,
this induces the director to disclose more information, decreasing the threshold
Ti . Importantly, an increase in a director’s bias improves communication only
if it further increases the conflict of interest between him and the rest of the
board. This shows the importance of diversity in directors’ private interests,
discussed in more detail below.26
Proposition 4 contrasts the result of the cheap talk literature that conflicts of
interest between the sender and receiver impede communication (e.g., Crawford
and Sobel 1982). While the current setup differs from cheap talk models in
two respects, communication costs and verifiable information, communication
costs are key to the result that conflicts of interest improve communication.

Downloaded from [Link] at Boston College on February 15, 2015

Section 5.3 discusses how introducing cheap talk communication would affect
the equilibrium.
Lemma A.4 in the Appendix shows that expected firm value is given by
 2
1 2 
N N Ti
1 1 N ri
E (V ) = V0 − Eε 2 − gi − 1− + (xi −yi )2 fi (xi )dxi , (17)
N N N 1+N ri ti
i=1 i=1

where [ti ,Ti ] is the interval of signals that are not communicated, and
yi is the expected value of the signal conditional on no communication,
yi = E[xi |xi ∈ [ti ,Ti ]]. By Lemma 3,
  
−ki ,−ki +2δi+ ,for i such that bi > ḡ−i ,
[ti ,Ti ] =   (18)
ki +2δi− ,ki ,for i such that bi < ḡ−i .
Expression (17) illustrates the twofold effect of directional biases on firm
value. At the decision-making stage, directional biases distort directors’actions,
moving the board’s decision away from the optimal decision from2 shareholders’
perspective. This effect is represented by the term − N1 N i=1 gi . In addition,
directional biases
T affect the incentives to reveal information. This is reflected
by the term ti i (xi −yi )2 fi (xi )dxi , which measures the variance of signal
xi over the noncommunication interval [ti ,Ti ]. As Proposition 4 shows, a
stronger directional bias relative to other members encourages the director to
communicate his signal, shrinking [ti ,Ti ] and increasing firm value. The next
result shows that this positive effect can dominate:

Proposition 5. Consider any (r1 ,...,rN ).

26 Although the results were derived for a uniform distribution of signals, the intuition is valid for a general
distribution. Consider any distribution that is symmetric and single-peaked around zero. Lemma A.3 in the
Appendix shows that there exist thresholds ti ,Ti , such that signal xi is disclosed if and only if xi ∈ / [ti ,Ti ]. As
previously, positively biased directors are more likely to disclose positive signals and vice versa. In particular, if
bi > ḡ−i , [ti ,Ti ] is shifted to the left of zero, so that Pr(xi is disclosed | xi > 0) > Pr(xi is disclosed | xi < 0). As bi
increases further, both ti and Ti decrease. However, if the support of the distribution is finite and communication
costs are sufficiently small, the communication strategy is boundary: ti = −ki . Hence, a further increase in the
director’s bias bi improves communication, as for a uniform distribution.

1505

[18:17 8/4/2014 [Link]] Page: 1505 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

(i) Suppose that b2 = ... = bN = 0. Then firm value is maximized at b1 = ±b,

where b > 0.
(ii) Suppose bi = b for i ≤ i0 and bi = −b for i > i0 , where 1 ≤ i0 ≤ N −1. Then
firm value is higher for b > 0 than for b = 0 if b is sufficiently small.

Part (i) shows that even if N −1 directors are unbiased, firm value is higher
if the remaining director is biased than if he is unbiased. Part (ii) shows that
a board that consists of directors whose directional biases are in opposite
directions can be more effective than an unbiased board. For example, consider
the board of a target. Part (ii) suggests that a fully independent board, in
which all directors maximize shareholder value, may be less efficient than

Downloaded from [Link] at Boston College on February 15, 2015

a board that includes both insiders, who are biased against the takeover, and
bidder representatives, who are biased in its favor. The reason is that the two
groups with conflicting interests have strong incentives to reveal information
and convince others of their position, allowing the board as a whole to make
more informed decisions.27
Proposition 5 relies on the assumption of positive communication costs. As
shown in Lemma 1, if ci ≤ 0, director i always reveals his signal, and hence
the only effect of his directional bias is to distort his decision-making behavior.
Thus, if ci ≤ 0, directional biases negatively affect firm value.

3. Extension: Optimal Allocation of Control

In practice, some directors have more control over decisions than others.
Examples include members of board committees, the board chairman, and
the lead director. This section shows that the presence of communication
costs has implications for the allocation of control among directors. The key
factors determining the optimal allocation of control will be the amount of
communication and effective use of directors’ information. Hence, this section
is related to the literature on optimal allocation of control among insiders and
outsiders given the tradeoff between communication and independence (Adams
and Ferreira 2007; Harris and Raviv 2008; Chakraborty and Yilmaz 2011).
However, the tradeoff in my paper is different from the tradeoff in this literature
since it is based on costly communication and no conflicts of interest.
To capture differential decision-making power,  suppose that the board’s
decision equals ai with probability pi , where N i=1 pi = 1, pi ≥ 0. Higher pi
corresponds to greater decision-making power. For example, if pi = 1, director
i is the sole decision-maker. To focus on the allocation of control, I assume

27 It is plausible that biased directors may discourage others from revealing information that does not support their
bias. This possibility would amplify the result that diversity in preferences improves communication. Indeed,
if insiders discourage others from revealing positive information about firm value, this can be counteracted
by including bidder representatives, who encourage communication of positive information and discourage
communication of negative information. In contrast, if all directors except insiders are unbiased, there is no
counteracting effect, and communication is reduced.

1506

[18:17 8/4/2014 [Link]] Page: 1506 1486–1532

 Communication and Decision-Making in Corporate Boards

that directors have only conformity biases. It is likely that when a director is
very influential, other directors are particularly reluctant to deviate from him.
To capture this, suppose that director i’s utility is

Ui (a,a1 ,...,aN ,θ) = −(a −θ )2 −ri p̃ki (ai −ak )2 , (19)
k =i

pk
where p̃ki = 1−p i
reflects the relative weight of director k among the remaining
directors.
Lemma A.1 in the Appendix shows that the equilibrium action of director i
at the decision-making stage is given by (10) for directors who communicated
their signals, and by

Downloaded from [Link] at Boston College on February 15, 2015

   
∗ pi pi
ai = xj + yj + xi + 1− yi (20)
j ∈J j ∈J ,j  =i
pi +ri pi +ri
C NC

for directors who did not communicate their signals. At the communication
stage (see Lemma A.3 in the Appendix), director i reveals his signal if and only
if it satisfies ⎛ ⎞ 1/2
ci
|xi | > di = ⎝ ⎠ , (21)
pi2
1− pi +ri

and hence yi = 0. The greater the director’s control over the decision, the less
effort he makes to communicate his information (di increases with pi ). For
example, if ri = 0, the director who has full control (pi = 1) does not share any
information. To see the intuition, suppose that the decision is binary (whether
to accept or reject a proposal) and the director has a majority of the votes. Such
a director has no need to communicate his information to the rest of the board:
by voting according to his information, he ensures that it is fully incorporated
into the final decision. In contrast, if a director does not participate in the vote
(pi = 0), his information will affect the board’s decision only if he manages to
convince the voting directors of his position.
I next study the optimal allocation of control between directors. In particular,
Iask which vector (p1 ,...,pN ) maximizes firm value among all vectors such that
N
i=1 pi = 1, pi ≥ 0. For simplicity, consider the uniform distribution of signals.
The following result shows that even if directors are identical, allocating full
control to one director can be optimal.

Proposition 6. Suppose that all directors are identical: xi is uniform on

[−k,k], ci = c, and ri = 0 for all i. Then firm value is maximized when one
director has full decision-making power—that is, when pi0 = 1 for some i0 and
pi = 0 for i  = i0 .

Intuitively, when control is allocated to one director, the final decision

aggregates directors’ signals most efficiently. By (21), directors without voting

1507

[18:17 8/4/2014 [Link]] Page: 1507 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

power (with pi = 0) have stronger incentives to communicate their signal to the

director in control than if they had voting power. The director in control then
efficiently aggregates all disclosed information to make the decision. Thus,
since the board’s decision-making rule is inefficient (see Definition 1), signals
that are disclosed to the decision-maker are used more efficiently than if all
directors participated in the vote but did not share their signals with others.
The downside of allocating control to one director is that signals that are not
disclosed to the decision-maker have no effect on the final decision. However,
because directors reveal the most valuable signals (|xi | > di ), the tradeoff is
between more efficient use of more valuable information and less efficient
use of less valuable information, and the positive effect dominates. Numerical
analysis shows that Proposition 6 also holds when ri = r > 0.

Downloaded from [Link] at Boston College on February 15, 2015

Another interesting question is which director should be given control when
directors are heterogeneous. One important factor is directors’ expertise. To
abstract from the effect of expertise, I assume that directors’ signals have the
same distribution and instead focus on heterogeneity in directors’ conformity
biases.
Proposition 7. Suppose that directors are identical in all respects except their
conformity biases: xi is uniform on [−k,k], ci = c. If control is allocated to one
director—that is, pi = 1 for some i—then firm value is maximized when control
is given to the director with the smallest conformity bias: i ∈ argminj {rj }.
Intuitively, a director who cares the least about conformity makes the most
unbiased decisions without worrying about others’ opinions of his actions.

4. Implications
This section presents the implications of the analysis. It starts with policy
implications and then describes empirical predictions and relevant evidence.

4.1 Policy implications

4.1.1 Open versus secret ballot voting and other policies that affect
conformity. Directors’ conformity biases can be strengthened or weakened
by the adoption of certain decision-making policies. One such policy is the
choice between open and secret ballot voting. The open ballot system intensifies
pressure for conformity because the dissenting director’s vote is observed by
other directors and the CEO. In particular, suppose that the firm adopts director
voting by open ballot but does not change any other policies or directors’
compensation contracts. It will then become more costly for a director not to
conform to the vote of the majority for two reasons. First, voting in dissent
is now more likely to worsen the director’s position inside the firm due to
retaliation from the manager and fellow board members and, as a consequence,
smaller perks and lower chances of retaining the board seat. Second, it may
affect the director’s career outside the firm. In particular, having a reputation

1508

[18:17 8/4/2014 [Link]] Page: 1508 1486–1532

 Communication and Decision-Making in Corporate Boards

of a trouble-maker may jeopardize the director’s chances of being invited to

other boards (for example, Marshall 2010 finds that the labor market does not
reward directors for showing dissent). At the same time, a director’s concern
about firm value will not change: concern for firm value is mostly determined
by the director’s contract (e.g., amount of performance-based compensation),
which is unaffected by the change in the voting policy.
Hence, in the context of the model, adoption of open ballot voting is likely to
increase parameter ri without affecting other parameters. Section 2.2 shows that
stronger conformity biases have two counteracting effects on board decisions
and firm value. On the one hand, they encourage pre-vote communication. On
the other hand, if a director does not communicate his position to others, a
conformity bias induces him to vote less honestly and rely on his information

Downloaded from [Link] at Boston College on February 15, 2015

less. The tradeoff between these two effects suggests that the adoption of open
ballot is more likely to have a positive effect on firm value in situations where
communication between directors is more valuable.
One consideration is the nature of directors’ information: whether
information is objective and can be convincingly communicated, or whether
it is subjective and reflects differences of opinion. When information reflects
a director’s opinion rather than objective evidence, the director is less likely
to change other directors’ actions by sharing his information with them. Thus,
communication has less value, and the use of secret ballot could be preferred
because it would allow more efficient aggregation of directors’diverse opinions
in the vote.
Another consideration is the type of decision on the agenda. Schwartz-
Ziv and Weisbach (2013) classify board decisions into “managerial” and
“supervisory.” When the board is making a managerial decision, such as
choosing the scale of investment or setting the acquisition price, it usually needs
to choose the best alternative among many. In contrast, in supervisory decisions,
the board mostly needs to approve or reject a given proposal presented by the
management. To the extent that information sharing between directors is more
important when the board chooses among many alternatives, open (secret)
ballot voting may be preferred for managerial (supervisory) decisions.
Finally, another relevant factor is board structure, and in particular how
diverse directors’ private interests are. As Section 2.3 demonstrates, pressure
for conformity (due to open ballot) is beneficial when directors’ interests are
diverse—for example, when there are two groups of directors with conflicting
interests, such as insiders and bidder representatives on the target’s board. In
contrast, secret ballot would be preferred when interests are not diverse—for
example, when the board consists of biased insiders and independent directors
maximizing shareholder value.
In addition to the choice between open and secret ballot, there are other
policies that affect conformity at the voting stage. One such policy is whether
the CEO is present at the vote. Indeed, an important reason for conformity
is directors’ reluctance to oppose the CEO without support from other board

1509

[18:17 8/4/2014 [Link]] Page: 1509 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

members. Thus, not allowing the CEO to be present at the vote (which is feasible
if the CEO is not on the board or committee) is likely to weaken pressure for
conformity.
Finally, transparency of board decision-making can also affect how strong
conformity biases are. To increase pressure for conformity, a firm could publicly
disclose the votes of its directors: knowing that their dissenting vote will be
disclosed, directors who do not want to be perceived as trouble-makers may be
more reluctant to oppose the majority. In China, for example, the law requires
all listed firms to follow such a disclosure policy (see Section 4.2 for details
about this law).

Downloaded from [Link] at Boston College on February 15, 2015

4.1.2 Executive sessions of outside directors. While open ballot voting
encourages communication prior to the vote, a more direct way to improve
discussions would be to reduce the costs of communication. One regulatory
measure that could have reduced directors’ costs of communicating outside
board meetings is the requirement for regular executive sessions of outside
directors without any management present, imposed on NYSE-, Amex-, and
Nasdaq-listed companies in 2003. To see the effect of this requirement, note
that in addition to the costs of time, discussions outside board meetings
can impose another important cost—retaliation from the manager. Indeed,
directors are likely to initiate such discussions if they have serious concerns
about the manager, and thus, if information about such discussions is leaked,
directors may strain their relationship with him. By making the meetings of
outside directors mandatory, the 2003 requirement has eliminated the negative
signaling role that such meetings had in the past. Consistent with this, NYSE
corporate governance rules state that “regular scheduling of such meetings
is important not only to foster better communication among non-management
directors, but also to prevent any negative inference from attaching to the calling
of executive sessions.”
This argument has implications for the optimal frequency of executive
sessions. Similarly to making these sessions mandatory, increasing their
scheduled frequency is likely to reduce directors’ communication costs. If
executive sessions are scheduled frequently, a director can wait until the next
session to voice his concerns and hence avoid initiating a meeting behind
the manager’s back. However, a higher frequency of executive sessions also
involves a cost because it requires time commitment from outside directors.
The tradeoff between these costs and benefits suggests that frequent executive
sessions would be more valuable in firms where outside directors find it difficult
to communicate outside executive sessions. First, these are firms where outside
directors cannot openly communicate during the full board meeting, which is
more likely when the outside directors are dependent on the CEO—for example,
due to the presence of a board interlock or directors’ business and social
connections with the CEO. Second, these are firms where outside directors do
not have other opportunities to communicate without raising suspicion from

1510

[18:17 8/4/2014 [Link]] Page: 1510 1486–1532

 Communication and Decision-Making in Corporate Boards

the CEO—that is, firms with weak social and professional connections between
their outside directors.

4.1.3 Delegation of decision-making power. Section 3 shows that allocating

decision-making power to a subset of directors can be optimal because it
leads to more efficient use of directors’ information. This result provides a
rationale for the board to delegate decision-making authority over certain
decisions to the manager and retain only an advisory role. When the board
delegates decision-making authority, each director has strong incentives to
communicate his position to the manager. The manager then incorporates
the views of all directors and his own information in his final decision. As
Proposition 6 shows, this can be more effective than if all directors had voting

Downloaded from [Link] at Boston College on February 15, 2015

power.
This result also provides a rationale for the use of board committees. The
standard rationale is that committees allow directors with relevant expertise to
hold more detailed discussions. The above logic suggests that committees can
be beneficial even if all directors have the same expertise.
Finally, Section 3 shows that if directors have similar expertise, it is efficient
to allocate control to directors who care the least about conforming to others
because such directors distort their actions the least. These considerations
may be important for issues where pressure for conformity is particularly
likely. Hence, this result is consistent with the listing requirement that the
compensation, nominating, and audit committees, which are responsible for
many controversial decisions, are composed entirely of independent directors.
It is also consistent with the observed practices, where directors occupying
leading positions, such as the chairman or lead director, are among the most
experienced board members. Due to their stronger reputation, such directors
are likely to have smaller conformity biases.

4.2 Empirical implications

This section discusses empirical predictions of the model and relevant anecdotal
evidence.
As discussed in Section 4.1, a switch from secret to open ballot voting is likely
to increase directors’ conformity biases ri without affecting other parameters
of the model. According to Proposition 1, a higher ri induces a director to
put less than optimal weight on his private information if it was not revealed
and to conform his actions to the expected actions of other directors. According
to Proposition 2, a higher ri also leads to more information being shared at
the communication stage, which can be interpreted as communication prior
to the vote and/or outside board meetings. Hence, the first prediction is as
follows:
Prediction 1. All else equal, the open ballot system (compared to the secret
ballot system) leads to: (i) more communication and dissent prior to the vote
and outside board meetings; (ii) a higher likelihood of unanimous votes.

1511

[18:17 8/4/2014 [Link]] Page: 1511 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

The evidence from board minutes in Schwartz-Ziv and Weisbach (2013) is

consistent with this prediction. In their sample, all votes were conducted by open
ballot. While active disagreement was common in pre-vote discussions, 97.5%
of the votes were unanimous (see Section 6.5 of their paper). The interview
evidence in Sapienza, Korsgaard, and Hoogendam (1997) also shows that split
votes in board meetings are relatively rare. In contrast, consistent with the
prediction, the discussion of most decisions in their sample took place outside
board meetings. Similar evidence is presented in Stevenson and Radin (2009)
and Rock (2004). The case illustrating the opposite dynamics for secret ballot
voting occurred during a judge’s election in Alaska (Vermeule 2010). All seven
members of the election committee spoke in favor of a candidate during the

Downloaded from [Link] at Boston College on February 15, 2015

discussion. However, during the subsequent secret ballot vote, the candidate
was defeated by a vote of 4–3.
While previously the extent of disagreement between directors was hard to
measure since the workings of the board were not observed, recent disclosure
regulations have provided more access to these data. In 2004, the Chinese
Securities Regulatory Commission required increased boardroom transparency
in listed firms. Under the new law, if some independent directors vote in dissent,
the firm must disclose the names of these directors in its annual report (Jiang,
Wan, and Zhao 2012). In the United States, a somewhat similar 2004 SEC
law requires firms to disclose if one of the directors leaves the board due to a
disagreement (Marshall 2010; Agrawal and Chen 2011).
These disclosure laws are likely to have affected boardroom dynamics.
Anecdotal evidence suggests that boards are very reluctant to expose any
internal disagreements to the public. Since the laws required public disclosure
of disagreements, they likely have strengthened pressure for conformity among
directors—that is, in the context of the model, increased conformity biases ri .
Thus, similar to Prediction 1, Propositions 1 and 2 predict the following effects
of disclosure laws:
Prediction 2. If votes of dissenting directors are publicly disclosed, then, all
else equal: (i) there is more communication and dissent prior to the vote and
outside board meetings; (ii) the vote is more likely to be unanimous.
The next prediction concerns the effects of communication costs. The
analysis shows that when communication costs are lower, directors are more
likely to agree with each other at the decision-making stage. Indeed, a lower ci in
the context of the model leads to more information being shared (Proposition 2)
and, as a result, directors’ actions being closer to each other (Proposition 1).
In addition, since lower communication costs improve board discussions, the
board’s decisions incorporate directors’information more effectively and hence
are more likely to increase shareholder value (according to Proposition 3, firm
value decreases in di and hence in ci ).
Since communication outside board meetings has become increasingly
common (Hart 2003), a good measure of communication costs are directors’

1512

[18:17 8/4/2014 [Link]] Page: 1512 1486–1532

 Communication and Decision-Making in Corporate Boards

costs of communicating outside meetings. These costs are likely to be lower

if directors are geographically closer to each other and if their social and
professional ties are stronger. In contrast, busy directors—for example, those
who hold several directorships—are less likely to find the time to communicate
outside meetings. Hence:

Prediction 3. When outside directors are less busy, are geographically closer
to each other, and their social and professional connections are stronger, then,
all else equal: (i) outside directors are more likely to vote in unison with each
other; (ii) the board’s decisions are more informed and are better for firm value.

Downloaded from [Link] at Boston College on February 15, 2015

The evidence in Fich and Shivdasani (2006), which suggests that board
effectiveness is lower when outside directors sit on many boards, is consistent
with prediction (ii).
In addition to communicating outside board meetings, outside directors also
communicate during executive sessions without the presence of any insiders.
There is considerable variation in the frequency of executive sessions across
firms. Some firms set a minimum annual number of executive sessions, ranging
from one to four, while others include an executive session as part of every board
meeting.28 The implications about the optimal frequency of executive sessions
discussed in Section 4.1 can help explain this variation. Assuming that firms set
the frequency of executive sessions optimally, we would expect the following:

Prediction 4. Executive sessions of outside directors are, all else equal, more
frequent in firms where: (i) there are stronger social and business ties between
outside directors and the CEO; (ii) there are weaker social and business ties
between outside directors.

Section 4.1 also discusses the choice between open and secret ballot
voting. Although most boards use open ballot, some variation exists. For
example, Swartz (2007) documents that 47 U.S. firms have passed a charter
amendment that requires voting by secret ballot on certain issues. Assuming
that voting rules are chosen optimally, the discussion in Section 4.1 predicts
the following:

Prediction 5. Secret ballot voting is, all else equal, more likely to be used:
(i) for decisions involving more subjective judgments and less objective
arguments; (ii) for supervisory rather than managerial decisions; (iii) when
directors’ private interests are less diverse.

28 See Shearman & Sterling LLP, “2007 trends in corporate governance of the largest U.S. public companies:
General governance practices,” p. 25.

1513

[18:17 8/4/2014 [Link]] Page: 1513 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

For example, secret ballot voting is sometimes used in the elections of the
chairman and lead director.29 Since information involved in these decisions
is more subjective than objective, the use of secret ballot in these cases is
consistent with prediction (i).
To test the above predictions, it is important to account for all potential
confounding factors. First, one needs to control for relevant elements of the
model that may affect the dependent variable. They include: (i) pressure for
conformity in predictions 1–3 (for example, measured by the influence of
the CEO); (ii) directors’ costs of communicating outside board meetings in
predictions 1–2 (for example, measured by directors’ geographical proximity
and social connections); (iii) directors’ conflicts of interest in prediction 3; (iv)
quality of directors’ information in predictions 3–5. Some of these factors may

Downloaded from [Link] at Boston College on February 15, 2015

be hard to measure. In addition, confounding variables can also include factors
outside the model. For example, in the context of prediction 3, the company’s
industry, strategy, and director ability could affect both the effectiveness
of the board and directors’ busyness, location, and connections with each
other. Ideally, the identification strategy would rely on exogenous variation
in the variable of interest. For example, to establish the causal effect of
outside directors’ connections on board performance, one could use exogenous
variation in directors’ connections due to random assignment of HBS students
to sections (Shue 2013).

5. Robustness and Discussion

5.1 Retaliation from the manager
An important reason for conformity in the boardroom is fear of CEO retaliation.
Directors’ preferences in the basic model do not incorporate this motive
explicitly: a director is reluctant to deviate from others regardless of whether
or not his vote goes against the CEO’s position. In the online Appendix, I
analyze two extensions of the model that capture directors’ fear of the CEO
and show the robustness of the results. In the first extension, a director suffers
a loss if he is less supportive of the CEO relative to other directors, but does
not suffer any loss if he is more supportive of the CEO than others. In other
words, the conformity bias is “asymmetric”: it arises only if the director votes
against the CEO because it is costly to oppose the CEO without support from
other board members. In the second extension, the CEO is one of the directors,
and conformity biases are modeled as a reluctance to deviate from the CEO’s
position.
In both extensions, similarly to Proposition 3, some degree of conformity is
beneficial for firm value due to its positive effect on pre-vote communication.

29 See, e.g., the 2007–2011 DEF 14A filings of The Travelers Companies, Inc., and the 2009 10-K filing of Activision
Blizzard, Inc., Exhibit 99-1. See also “Mixed feelings on secret ballot v. consensus decisions,” Australian
Financial Review, August 8, 2009.

1514

[18:17 8/4/2014 [Link]] Page: 1514 1486–1532

 Communication and Decision-Making in Corporate Boards

The intuition is the same as in the basic model. When pressure for conformity
is strong (e.g., because voting is by open ballot), a director with negative
information about the CEO or his preferred decision will not vote against the
CEO unless he knows that other directors share his concerns. Thus, to be able
to vote against the CEO without suffering retaliation, the director has strong
incentives to convince others of his position by sharing his information prior
to the vote. Because directors share their negative information with each other,
they can be more effective in jointly opposing the CEO than if each of them
acted individually based on his private information (as is more likely under
secret ballot voting).

Downloaded from [Link] at Boston College on February 15, 2015

5.2 Fear of the CEO and conformity at the communication stage
In the current setup, communication costs are fixed and do not depend on
the value of the signal. While such specification captures direct costs of
communication, such as the costs of time and effort, other types of costs
may depend on the information the director has. For example, it may be more
costly to reveal information that conflicts with the CEO’s view. Alternatively,
if pressure for conformity exists at the communication stage as well, it may
be more costly to reveal a signal that deviates from other directors’ expected
signals.30
Importantly, the intuition is robust to this assumption. In particular, for
any given structure of communication costs ci (·), stronger conformity at the
decision-making stage improves communication. To see this, suppose that for
a given signal xi , a director is indifferent between communicating it at a cost
ci (xi ) and not communicating. Suppose now that the costs of voting differently
from other directors increase (e.g., if secret ballot voting is replaced by open
ballot). As in the basic model, the director now has an additional benefit of
revealing his signal because it allows him to avoid the costs of dissent during
the vote. Hence, the director is no longer indifferent and strictly prefers to incur
the costs of communication. Put differently, certain changes in decision-making
procedures, such as those discussed in Section 4.1, allow to increase pressure for
conformity at the voting stage without affecting the costs of communication
and pressure for conformity at the communication stage, and thus allow to
overcome inefficiencies during the discussion.

5.3 Information manipulation

The model assumes that directors’ information is verifiable and cannot be
manipulated. As the next discussion shows, the intuition is robust to this
assumption.

30 Since the expected signal of each director is zero, conformity at the communication stage could be modeled by
a function ci (·) such that ci (x) increases with |x|. Similarly, if the CEO is biased toward higher actions (e.g.,
due to empire-building preferences if a corresponds to the amount of investment), communication costs coming
from the fear of the CEO could be modeled by a decreasing function ci (·).

1515

[18:17 8/4/2014 [Link]] Page: 1515 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

First, allowing information manipulation would not affect the result

that conformity biases encourage communication. To see this, note that
the equilibrium of Section 2.2 would not change even if directors could
misrepresent signals. Indeed, when directors do not have directional biases,
they have incentives to report information truthfully since all board members
share the objective of maximizing firm value.
Second, even if information is nonverifiable, a stronger conflict of interest
with other board members will still increase a director’s incentives to incur
the costs of communication for the same reason as before: the inference upon
no communication becomes more harmful for him as the conflict of interest
increases. In addition, once information manipulation is allowed, this effect will
be augmented by the standard result of the cheap talk literature that conflicts of

Downloaded from [Link] at Boston College on February 15, 2015

interest prevent full communication. Either of these two effects can potentially
dominate.
To see this in the simplest way, suppose that an additional, cheap talk, stage is
added prior to the costly communication stage. At this initial stage, each director
can convey his information at no cost, but not credibly. If such communication
is not fully successful, the director can incur costly effort and credibly convey
his information at the next stage. According to the cheap talk literature, a
stronger directional bias increases the director’s incentives to manipulate his
report and thus inhibits communication during the cheap talk stage. In contrast,
according to the results of this paper, a stronger bias increases the amount of
information conveyed at the costly communication stage. The net effect can
be either positive or negative. To see this, consider three cases. If directors’
interests are perfectly aligned, there is full information revelation at the cheap
talk stage. If conflicts of interest are very strong, no information is revealed
at the cheap talk stage, but there is full information revelation at the costly
communication stage (δi+ in Lemma 3 converges to zero as bi − ḡ−i increases).
Finally, if conflicts of interest are in the intermediate range, only part of the
information is revealed after the two stages.31

5.4 Uncertainty about preferences or information endowment

The results are robust to the assumption that directors’ preferences are known.
Suppose, for example, that directional biases are directors’ private information.
For simplicity, let a director’s bias be −b or b with equal probability, b > 0.
Then, there exists a threshold Tb > 0 such that a director with bias b (−b)
discloses his signal if it is above (below) Tb (−Tb ). Hence, conditional on
the director not revealing his signal, others infer that his signal is on average
zero. A director with bias b and signal Tb is indifferent between letting others

31 I have also analyzed a setup in which at the communication stage, a director chooses whether to send a “cheap
talk” message at no cost or incur a cost ci to convey his signal credibly. Conflicts of interest can encourage
communication in this setup as well: a small increase in conflicts of interest both increases the set of signals that
are conveyed credibly and increases the informativeness of cheap talk messages.

1516

[18:17 8/4/2014 [Link]] Page: 1516 1486–1532

 Communication and Decision-Making in Corporate Boards

make this less favorable inference of zero and incurring the cost and credibly
conveying to them that his signal is Tb > 0. If b increases, the director is no
longer indifferent between these two options and strictly prefers to reveal Tb .
Hence, more information is revealed when the distribution of biases is more
extreme.
The results are also robust to the assumption that directors are always
informed and other directors know about it. Suppose, for example, that
director 1 observes his signal with probability α and remains uninformed with
probability 1−α. Suppose also that other directors do not know whether the
director is informed. Repeating the analysis, it can first be shown that if the
director has no directional bias, his communication strategy does not depend
on α, and hence the effect of conformity remains unchanged. Second, if ri = 0

Downloaded from [Link] at Boston College on February 15, 2015

and the director has a relatively high directional bias (b1 > b̄−1 ), he discloses
2δ + −αk
his signal if and only if x1 > T1 = 12−α 1 , where δ1+ is the positive root of (16).
The threshold T1 decreases in b1 for any α, confirming the result that conflicts
of interest encourage communication. In addition, the fact that other members
do not know whether the director is informed allows the director to withhold
more information (T1 decreases in α).

5.5 Costly information acquisition

The focus of the paper is on costly communication. Another potentially
important friction in board decision-making is costly information acquisition.
Which friction is more important varies across firms. In firms where
management is willing to supply directors with firm-specific information,
directors’ expertise and industry experience allow them to process this
information in a way that gives them relevant private knowledge. In this case,
effective sharing of individual directors’ knowledge is key to efficient decision-
making. In other firms, management withholds or manipulates information it
provides to the board. In this case, it may be more important to elicit information
from management or to encourage directors to obtain information from other
sources.
The effect of conformity and directional biases on directors’ incentives
to collect information is ambiguous. For example, the effect of conformity
is likely to depend on how correlated directors’ signals are. If signals are
independent, as in the current setting, or weakly correlated, conformity can
reduce the incentives to acquire information because information will be used
less efficiently. If signals are highly correlated—for example, are identical—
conformity can encourage information acquisition because knowing the signal
is necessary to act similarly to others. Likewise, the effect of directional biases
on information acquisition depends on whether information is hard or soft. If
information is hard and can be credibly conveyed to others, then a more biased
director is likely to have stronger incentives to acquire information, especially
if he can focus his effort on searching for signals that support his bias. This is
because a more biased director benefits more from changing other directors’

1517

[18:17 8/4/2014 [Link]] Page: 1517 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

beliefs in the direction of his bias. In contrast, if information is soft and can
be manipulated, then stronger directional biases preclude communication and
thus discourage information acquisition.

5.6 Generalized utility specification

In the basic model, a stronger conformity bias does not change a director’s
concern about firm value. In this section, I generalize directors’ preferences
to the case in which a stronger conformity bias can also decrease a director’s
concern about firm value. In particular, extending the specification in Section 3,
suppose that the utility of director i is now given by

N

Downloaded from [Link] at Boston College on February 15, 2015

Ui (a,a1 ,...,aN ,θ) = −w(ri )(a −θ )2 −ri p̃ki (ai −ak )2 , (22)
k=1,k =i

where w(ri ) is the weight put on firm value. The basic model corresponds to
w(ri ) ≡ 1.
The proof of Lemma A.5 in the Appendix shows that director i communicates
his signal if and only if |xi | > di , where di decreases in ri if and only if
 
w(ri )2 pi2 pi ri2
+ 1−p i + w (ri ) > 0. (23)
(w(ri )pi +ri )2 (w(ri )pi +ri )2
This shows that conformity is helpful for communication only if directors care
about firm value. Indeed, if w(ri ) ≡ 0, then ∂d i
∂ri
= 0, and hence conformity biases
have no effect on communication.
In addition, (23) shows that if a stronger conformity bias ri also decreases a
director’s concern about firm value w(ri ), then ri has two different effects on
communication. The first effect, which is emphasized in the paper, is reflected
in the first term of (23): if w(ri ) = 1, as in the basic setup, then w (ri ) = 0 and
hence communication always increases with ri . However, when the director’s
concern about firm value decreases with ri , there is also a second effect: if
w (ri ) < 0, the second term in (23) is negative. Intuitively, the only reason why
a director incurs the costs of communication is because he cares about firm
value. Hence, the lower the weight w(ri ) on the firm value component, the
lower are the director’s incentives to communicate.32
Consider a special case of (22), in which w(ri ) = 1−ri , ri ∈ [0,1]; that is,
a director’s utility is the weighted average of firm value and conformity
terms. The proof of Lemma A.5 in the Appendix shows that di decreases
in ri if and only if ri < r̂i for some r̂i ∈ (0,1); that is, the positive effect on

32 This discussion suggests that in addition to changing the board’s decision-making procedures, the firm can
encourage communication between directors through contracts—for example, through performance-based
compensation. Higher pay-performance sensitivity (w) increases a director’s incentives to share his information,
but not fully: this is because the costs of communication are privately borne by the director, while the benefits
are shared with all shareholders.

1518

[18:17 8/4/2014 [Link]] Page: 1518 1486–1532

 Communication and Decision-Making in Corporate Boards

communication dominates for small ri . Moreover, Lemma A.5 shows that firm
value is maximized at a strictly positive ri ; that is, Proposition 3 continues to
hold.

6. Concluding Remarks
Board communication is impeded by several factors, including time constraints,
the dominance of the CEO, directors’ reputational concerns, and diversity of
directors’ backgrounds. This paper develops a theory of board decision-making
whose key element is that directors need to incur personal costs to communicate
their information to others effectively. I show that directors’ conformity
biases—that is, their reluctance to disagree with other board members—

Downloaded from [Link] at Boston College on February 15, 2015

encourage them to incur the costs and communicate more effectively. This
suggests that the widely used open ballot system is not necessarily inefficient,
even though it strengthens pressure for conformity during the vote and
induces directors to disregard their private information and vote with the
majority. I also show that board communication can be more efficient when
directors’ private interests are more diverse. Hence, fully independent boards
can be less effective than boards where some directors’ interests conflict
with shareholders’ interests. The analysis has novel implications for voting
procedures, boardroom transparency, executive sessions, and the allocation of
control between directors.
While the focus of the paper is on corporate boards, it can also be applied
to other committees where communication costs are important. Notably, there
is considerable diversity in the decision-making rules of different committees.
Whereas corporate boards mostly use the open ballot system, there is variation
in the use of open and secret ballot across university committees and nonprofit
boards. Government agencies are different from most boards in that the votes of
their members are publicly disclosed. It would be interesting to understand why
board decision-making rules are different from those used in other committees.
Is it because boards are inherently different in their structure and functions?
Or is it because board policies are influenced by managers who want to retain
control over decisions?

Appendix
Motivating pressure for conformity
Directors’ conformity biases can be due to a number of reasons. First, they can be caused by
psychological factors such as certain social norms and pressures. In addition, there are several
rational explanations for conformity, coming from directors’ reputational concerns. This section
presents these explanations and discusses why the results of the paper are robust to different reasons
behind conformity.

Reputation for being competent and informed. Directors may want to conform to others if
their actions are used to infer the quality of their private information. In particular, suppose that
the precision of directors’ signals depends on their ability: smarter directors receive more precise

1519

[18:17 8/4/2014 [Link]] Page: 1519 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

signals. If all directors except one take very similar actions, this implies that signals of all but one
director are similar to each other, but the signal of the deviating director is sufficiently different.
Because smart directors tend to receive correlated signals and less competent directors receive pure
noise, this would imply that the deviating director is likely to be less competent. Hence, to signal
his competence, each director has incentives to mimic the behavior of other directors.
Under this rationale, pressure for conformity will arise at the communication stage as well.
Importantly, as discussed in Section 5.2, certain decision-making rules (e.g., the choice between
open and secret ballot voting) affect directors’ conformity biases at the voting stage without
affecting the communication stage. Thus, the result that open ballot voting improves communication
is robust to this rationale for conformity.

Reputation for having opportunistic motives. Conformity can also arise because dissent can
indicate the presence of opportunistic motives. If a director votes against a proposal that is supported

Downloaded from [Link] at Boston College on February 15, 2015

by the rest of the board and does not back up his view with convincing arguments, other directors
may question his impartiality. Thus, to appear unbiased, a director can choose to conform to the
majority.
This argument can also explain managerial retaliation against dissenting directors. By criticizing
the manager without supporting his criticism with convincing evidence, the director indicates a
possible bias against the manager. Because this bias is likely to lead the director to oppose the
manager’s actions in the future, even without objective reasons, it is optimal for the manager (and
the rest of the board) to remove the dissenting director. As discussed in Section 5.1, the results
continue to hold if conformity biases are due to fear of managerial retaliation.

In the examples above, conformity biases come from directors’ concerns about internal
reputation and their position on the board. Conformity can also arise from concerns about external
reputation. Directors may be reluctant to dissent because it could indicate to outsiders their
incompetence or conflict of interest and hinder their reputation in the labor market.
For any type of reputational concerns, a director could partly alleviate the negative inference
from dissent by explaining his reasons for dissent after the vote. However, the ability to
communicate after the vote does not change the result that open ballot voting improves
communication before the vote and hence makes the vote more informative. To see the intuition,
note that the need to explain one’s behavior after a dissenting vote is equivalent to the cost of
nonconformity and arises only if voting is conducted by open ballot. Such explanations would not
be needed if voting were conducted by secret ballot. Put differently, the cost of nonconformity can
now be defined as the cost of voting differently from others plus the cost of having to explain one’s
vote afterward. Since these costs are higher under open ballot, directors have stronger incentives
to communicate prior to the vote under open than under secret ballot.

Desire to avoid confrontation. In addition to reputational concerns, a director may be reluctant

to dissent because of a general desire to avoid confrontation. Even if other directors did not convince
him during the discussion, the director may want to vote with them in order to show that he took
their opinion seriously and is not a trouble-maker. These incentives to conform are stronger when
the director’s vote is observed by others, that is, when voting is by open ballot. Hence, under
open ballot, directors have incentives to voice their opinions during the discussion, realizing that
a desire to avoid confrontation will induce other directors to side with them. This shows that
open ballot voting induces more pre-vote communication than secret ballot voting in this setting
as well.

Proofs
To prove the main results, I first prove several auxiliary results for a more general model with

preferences (b1 ,...,bN ), (r1 ,...,rN ) and weights (p1 ,...,pN ), Ni=1 pi = 1, measuring the influence

1520

[18:17 8/4/2014 [Link]] Page: 1520 1486–1532

 Communication and Decision-Making in Corporate Boards

of individual directors. I derive the equilibrium and expected firm value for this more general
model. The proofs of the main results follow from these auxiliary results.

Auxiliary results: General model. Suppose that firm value is given by


N
V0 − pi (ai −θ )2 , (A1)
i=1

and the utility of director i is


N 
N
Ui (a,θ ) = − pk (ak −(bi +θ ))2 −ri p̃ki (ai −ak )2 , (A2)
k=1 k=1,k=i
N

Downloaded from [Link] at Boston College on February 15, 2015

pk
where i=1 pi = 1 and p̃ki = 1−pi .

Lemma A.1 (equilibrium at the decision-making stage). Suppose that at the communication
stage signals xi ,i ∈ JC were communicated, and that yi is the expected value of xi conditional on
no communication. Denote JN C = {1,...,N}\JC . Then there is a linear equilibrium at the decision-
making stage characterized by the following strategies:

1. If director i communicated his signal, i ∈ JC , his action is given by

 
ai = g i + xj + yj . (A3)
j ∈JC j ∈JN C

2. If director i did not communicate his signal, i ∈ JN C , his action is given by

  pi
ai = g i + xj + yj + (xi −yi ), (A4)
pi +ri
j ∈JC j ∈JN C

where gi solves the system

 
pi pi
gi = bi + 1− p̃ki gk . (A5)
pi +ri pi +ri
k=i

Proof of Lemma A.1. Let us verify that the strategies given by (A3)–(A5) constitute an
equilibrium. Denote the sum of the signals that werecommunicated by X, and the expected sum of

the signals that were not communicated by Y : X = j ∈J xj and Y = j ∈J yj . Also denote by
C NC
Ii the information set of director i after the communication stage. Taking the first-order condition
of (A2), the optimal action of director i is given by
⎡ ⎤
pi ri 
ai = (bi +E[θ |Ii ])+ E ⎣ p̃k ak |Ii ⎦ .
i
pi +ri pi +ri
k=i

First, consider the best response of director i,i ∈ JC . For him, E[θ |Ii ] = X +Y . Also, given the
equilibrium strategies (A3) and (A4) of other players,
⎡ ⎤ ⎛ ⎞
    pk
E ⎣ i ⎦
p̃k ak |Ii = i
p̃k gk + ⎝ i⎠
p̃k (X +Y )+ p̃ki E[(xk −yk )|Ii ].
pk +rk
k=i k=i k=i k∈JN C

Note that k=i p̃ki = 1 and that the last term equals 0 because E[xk |Ii ] = yk . Plugging in E[θ |Ii ] and

E [ k=i p̃ki ak |Ii ] into the first-order condition, we get the conjectured strategy (A3).

1521

[18:17 8/4/2014 [Link]] Page: 1521 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

Next, consider the best response ∈ JN C . For him, E[θ |Ii ] = X +Y −yi +xi and by
 of director i,i
the same argument as above, E [ k=i p̃ki ak |Ii ] = k=i p̃ki gk +X +Y . Plugging in these values into
the first-order condition, we again get the conjectured equilibrium strategy (A4).
The coefficients gi can be found by solving the system of linear equations (A5) for i = 1,...,N.
This system coincides with (8) when pi = N1 for all i.
It is straightforward to prove that the equilibrium (A3)–(A5) is a unique linear equilibrium.
This can be done similar to the analysis of the benchmark case by conjecturing a general linear
equilibrium and plugging in the conjectured strategies into the first-order condition above. The
proof is omitted for space considerations.

Lemma A.2 (properties of gi ).

(i) There is a unique

 solution to the system of linear equations (A5), which takes the form
where λii ,λij ∈ (0,1) if ri > 0, and λii = 1,λij = 0 if ri = 0. The

Downloaded from [Link] at Boston College on February 15, 2015

gi = λii bi + j =i λij bj , 
coefficients λij satisfy j λij = 1 for all i.
N −1+N r
(ii) If pi = N1 and ri = r for all i, then gi = ωbi +(1−ω) b̄−i , where ω = .
N −1+N 2 r

Proof of Lemma A.2. Because (A5) is a system of linear equations on gi with constant terms
pi
equal to p +r bi , the solution to this system takes the form gi = λi1 b1 +...+λiN bN for some λij . To
i i
find (λ1i ,...,λN i ) for a particular i, we differentiate each equation in (A5) with respect to bi and
derive a system of N linear equations on N coefficients λ1i ,...,λN i . The properties of λij in (i) and
the statement of (ii) follow directly from solving this system.

Lemma A.3 (equilibrium strategies at the communication stage).

(i) Suppose that conditional on director i not communicating his signal, other directors believe
that the expected value of xi is yi . Then director i has incentives to communicate xi if and
only if it satisfies Hi (xi −yi ) > 0, where
⎛ ⎞
 ci
Hi (δ) = δ +2δ ⎝bi −
2
p̃ki gk ⎠ − . (A6)
pi2
k=i 1− p
i +ri

For any ci > 0, Hi (δ) has two roots δi− and δi+ , which satisfy δi− < 0 < δi+ .
(ii) In any equilibrium, the strategy of director i at the communication stage is characterized
by an interval [ti ,Ti ] such that xi is communicated if and only if xi ∈
/ [ti ,Ti ]. The necessary
and sufficient conditions for the four possible types of equilibria are:

(a) Equilibrium with −ki < ti < Ti < ki exists if and only if ti −yi = δi− and Ti −yi = δi+ .
(b) Equilibrium with −ki = ti < Ti < ki exists if and only if −ki −yi > δi− and Ti −yi = δi+ .
(c) Equilibrium with −ki < ti < Ti = ki exists if and only if ti −yi = δi− and ki −yi < δi+ .
(d) Equilibrium with −ki = ti < Ti = ki exists if and only if −ki > δi− and ki < δi+ .

Proof of Lemma A.3. (i) Suppose that the equilibrium communication and noncommunication
regions of director i are some sets Ci and N Ci , Ci ∪N Ci = [−ki ,ki ]. That is, the director
communicates his signal xi if and only if xi ∈ Ci . Denote yi = E[xi |xi ∈ N Ci ].
First, we derive the payoff of each director, taking the outcome of the communication stage
as given. Suppose that signals x1 ,...,xN were realized and that during the communication stage
pi
signals xi ,i ∈ JC were communicated and signals xi ,i ∈ JN C were not. Denote Qi = p +r and
i i

1522

[18:17 8/4/2014 [Link]] Page: 1522 1486–1532

 Communication and Decision-Making in Corporate Boards

δi = xi −yi . From (A2) and the equilibrium actions (A3)–(A4) at the decision-making stage, the
utility of director i after the communication stage is
   2
Ui = − pj gj −bi − k∈J δk −ε
j ∈JC
  
NC
  2
− j ∈J pj gj −bi − 1−Qj δj − k∈J ,k=j δk −ε (A7)
 N C  N C
−ri k=i,k∈J p̃ki (gi −gk )2 −ri k=i,k∈J p̃ki (gi −gk −Qk δk )2
C NC

for i ∈ JC , and
   2
Ui = − pj gj −bi − k∈J δk −ε
j ∈JC
  
NC
  2
− j ∈J pj gj −bi − 1−Qj δj − k∈J ,k=j δk −ε (A8)
 NC  NC
−ri k=i,k∈J p̃ki (gi +Qi δi −gk )2 −ri k=i,k∈J p̃ki (gi −gk +Qi δi −Qk δk )2

Downloaded from [Link] at Boston College on February 15, 2015

C NC

for i ∈ JN C .
Consider the decision of director 1 whether to pay c1 to communicate his signal x1 . The director
does not know other directors’ signals and the noise term ε and thus conditions his decision on
all possible values of x2 ,...,xN ,ε. Suppose that among the remaining signals, signals xi ,i ∈ JC
lie in their respective regions Ci and are thus communicated, and signals xi ,i ∈ JN C lie in their
respective regions N Ci and are not communicated. If the director communicates his signal, then
by (A7), his payoff upon communication, U1C , is equal to
   2
U1C = − j ∈J ∪{1} pj gj −b1 − k∈J δk −ε

C
 
NC
  2
− j ∈J pj gj −b1 − 1−Qj δj − k∈J ,k=j δk −ε
 N C  N C
−r1 k=1,k∈J p̃k1 (g1 −gk )2 −r1 k=1,k∈J p̃k1 (g1 −gk −Qk δk )2
C NC

If the director does not communicate his signal, then by (A8), his payoff, U1N C , is equal to
   2
U1N C = − pj gj −b1 −δ1 − k∈J δk −ε
j ∈JC
 
NC
2
−p1 g1 −b1 −(1−Q1 )δ1 − k∈J δk −ε
   
NC
 2
− j ∈J pj gj −b1 − 1−Qj δj −δ1 − k∈J ,k=j δk −ε
 NC  NC
−r1 k=1,k∈J p̃k1 (g1 −gk +Q1 δ1 )2 −r1 k∈J p̃k1 (g1 −gk +Q1 δ1 −Qk δk )2 .
C NC

The director averages these payoffs over all possible values of x2 ,...,xN ,ε and chooses to
communicate his signal if and only if

U1C f2 ...fN fε dx2 ...dxN dε > c1 + U1N C f2 ...fN fε dx2 ...dxN dε. (A9)

If we open the brackets in U1C and U1N C , it is easy to see that the expressions inside the integrals
are some linear combinations of quadratic terms δi2 ,ε2 , interaction terms δi δj ,δi ε, linear terms δi ,ε,
and a constant. Note also that the signal of director k,k  = 1 enters U1C and U1N C with a non-zero
coefficient only if xk ∈ N Ck ; that is, for k ∈ JN C . Also, because δi = xi −E[xi |xi ∈ N Ci ],

δi fi (xi )dxi = 0.
N Ci

Therefore, on both sides of (A9), all linear terms for δi ,i ≥ 2, all interaction terms δi δj ,i ≥ 2,
and all terms including ε integrate to zero. Hence, only quadratic terms ε2 and δi2 , i ∈ JN C ∪{1},
the linear term δ1 , and the constant remain. Note also that the constant terms and the coefficients
for terms ε2 and δi2 ,i ∈ JN C are the same in U1C and U1N C . Besides, the integral over δi2 is taken

1523

[18:17 8/4/2014 [Link]] Page: 1523 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

over the same set N Ci on both sides of (A9). Hence, the integrals over terms ε 2 and δi2 ,i ∈ JN C on
the two sides of (A9) cancel out. Finally, δ12 and δ1 do not enter the expression for U1C and only
enter U1N C . The coefficient for δ12 in the expression for U1N C is equal to −A, where
p12
A = (1−p1 )+p1 (1−Q1 )2 +r1 Q21 = 1− > 0,
p1 +r1
and the coefficient for δ1 is equal to 2B, where
   
B = k=1 pk (gk −b1 )+p1 (1−Q1 )(g1 −b1 )−r1 Q1 g1 − k=1 p̃k1 gk
  
p12 
k=1 p̃k gk −b1 .
= 1− p +r 1
1 1

Hence, (A9) is equivalent to

Aδ12 −2Bδ1 −c1 > 0. (A10)

Downloaded from [Link] at Boston College on February 15, 2015

Because A > 0, (A10) is equivalent to H1 (x1 −y1 ) > 0, where H1 (·) is given by (A6) in the statement
of the lemma. The corresponding quadratic equation, H1 (δ) = 0, always has two different roots
B±(B 2 +Ac )1/2
δ1− < δ1+ , given by A
1 , and δ1− < 0 < δ1+ .
(ii) Because δ1 = x1 −y1 , (A10) implies that the noncommunication region is always some
interval [t1 ,T1 ]. If one of the boundaries (t1 or T1 ) is interior—that is, lies inside (−k1 ,k1 )—then
the director should be indifferent between communicating and not communicating his signal at
this point. This implies that (A10) should be satisfied as an equality and hence t1 −y1 or T1 −y1
should coincide with δ1− or δ1+ , respectively. If the right boundary T1 coincides with k1 , then (A10)
should be violated at k1 , implying that k1 −y1 should be smaller than δ1+ (being positive, it is always
greater than δ1− < 0). Similarly, if the left boundary t1 coincides with −k1 , then (A10) should be
violated at −k1 , implying that −k1 −y1 should be greater than δ1− (being negative, it is always
smaller than δ1+ > 0).
Lemma A.4 (firm value). Suppose that at the communication stage, director i communicates his
signal if and only if xi ∈
/ [ti ,Ti ], and let yi = E [xi |xi ∈ [ti ,Ti ]]. Then expected firm value is given by
 N N  2 Ti
ri
E(V ) = V0 −Eε 2 − pi gi2 − 1−pi +pi (xi −yi )2 fi (xi )dxi ,
pi +ri ti
i=1 i=1
where gi solves (A5).
p
Proof of Lemma A.4. Denote δi = xi −yi and Qi = p +r i . For any given realization of x ,...,x ,
1 N
i i
suppose that signals xi ,i ∈ JC are communicated and signals xi ,i ∈ JN C are not. Using the
derivations in the proof of Lemma A.3, firm value, V (x1 ,...,xN ,ε), is equal to
⎡ ⎤2 ⎡ ⎤2
     
V0 − pi ⎣ gi − δk −ε ⎦ − pj ⎣gj −δj 1−Qj − δk −ε ⎦ ,
i∈JC k∈JN C j ∈JN C k∈JN C ,k=j

and expected firm value is

E(V ) = [ V (x1 ,...,xN ,ε)]f1 (x1 )...fN (xN )fε (ε)dx1 ...dxN dε.

By the same argument as in the proof of Lemma A.3, the integral over all linear terms δi ,ε
and interaction terms δi δj ,δi ε equals 0. Also, because all quadratic terms δi2 ,ε 2 enter additively,
the integral over these terms equals the sum of the corresponding integrals for individual signals.
The coefficient before δi2 for i ∈ JC is 0, and the coefficient before δi2 for i ∈ JN C is −[1−pi +
pi (1−Qi )2 ]. The coefficient before ε 2 is −1. Finally, note that i ∈ JN C if and only if xi ∈ [ti ,Ti ].
Integrating over all possible realizations of x1 ,...,xN , we get

N N 
 
E(V ) = V0 −Eε 2 − pi gi2 − 1−pi +pi (1−Qi )2 δi2 ·1{xi ∈ [ti ,Ti ]}fi (xi )dxi ,
i=1 i=1
which is equivalent to the expression in the statement of the lemma.

1524

[18:17 8/4/2014 [Link]] Page: 1524 1486–1532

 Communication and Decision-Making in Corporate Boards

Proofs of main results

Proof of Lemma 1. Suppose, on the contrary, there exists an equilibrium where director i
conceals his signal when xi ∈ N Ci , where Pr(N Ci ) > 0. Let yi = E[xi |xi ∈ N Ci ]. According to
Lemma A.3, director i finds it optimal to conceal xi if and only if Hi (xi −yi ) < 0, where Hi (δ) is
given by (A6). If Hi (δ) does not have any roots, then Hi (δ) ≥ 0 ∀δ whenever ci ≤ 0, and hence we
get a contradiction. If Hi (δ) has roots δi− ≤ δi+ , then any xi ∈ N Ci must satisfy δi− < xi −yi < δi+ .
Integrating this inequality over N Ci , we get

δi− Pr(N Ci ) < xf (x)dx −yi Pr(N Ci ) < δi+ Pr(N Ci ) ⇔ δi− < 0 < δi+ .
N Ci

However, when ci ≤ 0, the roots of Hi (δ) always lie on the same side of zero: they are both

negative (positive) if bi − k=i p̃ki gk ≥ 0 (≤ 0). This contradicts δi− < 0 < δi+ and proves that there

Downloaded from [Link] at Boston College on February 15, 2015

is no equilibrium with Pr(N Ci ) > 0.

Proof of Proposition 1. The statement of Proposition 1 follows from Lemma A.1 for the case
bi = 0 and pi = N1 for all i.

Proof of Proposition 2. Let yi be the equilibrium expected value of xi conditional on it not

being communicated. According to Lemma A.3 (i) for the case bi = 0 and pi = N1 , director i finds
it optimal to communicate his signal if and only if

ci
(xi −yi )2 > . (A11)
1− N1 1
1+N ri

Hence, there always exists an equilibrium where a director communicates xi if and only if
c
|xi | > di = ( 1 i 1 )1/2 . Indeed, in this equilibrium yi = 0 due to the symmetry of the distribution,
1− N 1+N r
i
and hence by (A11), communicating xi if and only if |xi | > di is optimal.
Moreover, when the distribution is single-peaked at zero, this equilibrium is unique. First, there
is no other equilibrium where the communication interval is interior. According to (A11), any such
equilibrium is characterized by [ti ,Ti ] and yi = E[xi |xi ∈ [ti ,Ti ]], such that Ti −yi = yi −ti = di . It
t +T
follows that yi = i 2 i ; that is, the conditional expectation over [ti ,Ti ] coincides with the middle of
the interval. Because the distribution is symmetric and single-peaked at zero, this is only possible
for yi = 0. Hence, no other interior equilibrium exists. Second, there is no boundary equilibrium.
Suppose, for example, that there is a boundary equilibrium with the noncommunication interval
[ti ,ki ],ti > −ki . According to Lemma A.3 (ii), this is only an equilibrium if ti −yi = −di and
t +k t +k
ki −yi < di . Summing up these two expressions, we get yi > i 2 i , where i 2 i > 0. However, for
a single-peaked symmetric distribution, the conditional expectation over [ti ,ki ] is strictly smaller
t +k t +k t +k
than i 2 i when i 2 i > 0, which contradicts yi > i 2 i . Similarly, there is no boundary equilibrium
in which the noncommunication interval is [−ki ,Ti ],Ti < ki .
If the distribution has more than one peak, there could be multiple equilibria at the
communication stage. For example, for a two-peak distribution that is symmetric around zero,
has peaks at points (−z,z), and is symmetric in the neighborhood of each peak, there are three
equilibria with yi ∈ {−z,0,z} if ci is sufficiently small. For a uniform distribution, the condition
t +T
E[xi |xi ∈ [ti ,Ti ]] = i 2 i is satisfied for any interval [ti ,Ti ], and hence there is a continuum of
equilibria characterized by some noncommunication interval of length 2di .

Proof of Lemma 2. The statement of the lemma follows from Lemma A.4 for the case bi = 0 and
pi = N1 for all i.

1525

[18:17 8/4/2014 [Link]] Page: 1525 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

Proof of Proposition 3. Because contributions of individual directors to firm value enter

additively, we can examine the effect of each individual ri separately. Consider the term reflecting
the contribution of director i:
⎛ ⎞1/2
 2 di
1 1 N ri c
x fi (x)dx, di = ⎝ ⎠ .
i
Vi (ri ) = − 1− + 2
N N 1+N ri −di 1− N1 1+N1
r i
 1/2
c
Because by assumption, i
1 < ki , then di < ki for any ri ≥ 0. It can be shown
1− N
 
3/2  −3/2
that limri →0 Vi (ri ) = fi ci (1− N1 )−1/2 ci 1− N1
1/2
> 0, which implies that firm value is
maximized at some strictly positive ri , potentially infinitely large.
Proof of Lemma 3. The proof is based on Lemma A.3. Suppose that bi > ḡ−i . Then the coefficient
for the linear term inthe quadratic equation Hi (δ) = 0 given by (A6) is 2(bi − ḡ−i ) > 0. It follows

Downloaded from [Link] at Boston College on February 15, 2015

that the roots δi− ,δi+ of this equation satisfy δi− +δi+ < 0.
First, I show that the equilibrium communication strategy of director i is boundary—that is,
either signal ki or signal −ki is not communicated in equilibrium. According to Lemma A.3 (ii),
if the communication strategy were interior, then director i would communicate xi if and only
if xi  ∈ [ti ,Ti ], where ti −yi = δi− ,Ti −yi = δi+ , and yi = E[xi |xi ∈ [ti ,Ti ]]. Because the distribution
is uniform, yi = i 2 i , which would imply that δi− +δi+ = 0, contradicting the fact that δi− +δi+ < 0.
t +T

Hence, the equilibrium must be boundary.

Next, I show that there is no equilibrium where −ki is communicated and ki is not communicated.
Suppose that such an equilibrium exists. By Lemma A.3 (ii), this equilibrium is characterized by
the noncommunication interval [ti ,ki ],ti > −ki , where ti −yi = δi− and ki −yi < δi+ . Summing up
these two expressions, we get ti +ki −2yi < δi− +δi+ . However, yi = i 2 i and hence 0 = ti +ki −2yi <
t +k

δi− +δi+ < 0, which is a contradiction.

Therefore, all equilibria are characterized by a noncommunication interval [−ki ,Ti ], where
Ti ≤ ki .According to LemmaA.3 (ii), in order for an equilibrium with Ti < ki to exist, Ti must satisfy
T −k
Ti − i 2 i = δi+ ⇔ Ti = 2δi+ −ki . Hence, Ti < ki is an equilibrium if and only if δi+ < ki . In order for an
k −k
equilibrium with Ti = ki to exist, ki must satisfy ki − i 2 i < δi+ ⇔ ki < δi+ . Hence, there is a unique
equilibrium characterized by the noncommunication interval [−ki ,Ti ], Ti = min{2δi+ −ki ,ki }.
Following similar arguments, it can be shown that if bi < ḡ−i , there is a unique equilibrium
characterized by the noncommunication interval [ti ,ki ], ti = max{−ki ,ki +2δi− }. Finally, if bi = ḡ−i ,
then the coefficient for the linear term in the equation Hi (δ) = 0 is equal to zero. It follows that
there is a continuum of equilibria characterized by a point yi ∈ [−ki +di ,ki −di ], such that director
 1/2
ci
i communicates his signal if and only if |xi −yi | > di = 1 1 . As shown in the proof
1− N 1+N ri
of Proposition 5, firm value is the same in all these equilibria.
Proof of Proposition 4. Suppose, for example, that b1 > ḡ−1 . Then, according to Lemma 3,
director 1 communicates his signal if and only if x1 > T1 = −k1 +2δ1+ , where
 1/2
c1
δ1+ = −(b1 − ḡ−1 )+ (b1 − ḡ−1 )2 + .
1− N1 1+N1 r
1

According to Lemma A.2, dbd ḡ−1 = N 1−1 N d
j =2 λj 1 , where λj 1 < 1. Thus, db1 ḡ−1 < 1 and
1
d (b1 −ḡ−1 )
db > 0. This implies that b1 − ḡ−1 increases with b1 and hence remains positive as b1
1
increases further. Thus, the communication region continues to take the form [−k1 +2δ1+ ,k1 ].
Moreover, δ1+ decreases in (b1 − ḡ−1 ) and hence decreases as b1 increases. Hence, the director
reveals more information as his directional bias increases. Note also that because b1 − ḡ−1 does
not depend on r1 , δ1+ decreases with r1 and hence the director reveals more information as his
conformity bias increases. The proof for the case b1 < ḡ−1 is similar.

1526

[18:17 8/4/2014 [Link]] Page: 1526 1486–1532

 Communication and Decision-Making in Corporate Boards

Proof of Proposition 5. Using LemmaA.4 for pi = N1 and a uniform distribution of xi on [−ki ,ki ],
expected firm value is given by
 2  
1 2 
N N
1 1 N ri 1 Ti −ti 3
E(V ) = V0 −Eε 2 − gi − 1− + , (A12)
N N N 1+N ri 3ki 2
i=1 i=1

where [ti ,Ti ] is the noncommunication region of director i. Note that firm value only depends on
the length Ti −ti of the noncommunication interval and not on its location.
(i) To prove the statement, I show that limb1 →0+ dbd E(V ) > 0 and limb1 →0− dbd E(V ) < 0.
1 1
(1) First, consider b1 > 0. Our goal is to prove that limb1 →0+ dbd E(V ) > 0.
1
Because b2 = ... = bN = 0, then according to Lemma A.2, gi = λi1 b1 , where λi1 ∈ [0,1) for
1 
i  = 1 and λ11 ∈ (0,1]. For i  = 1, ḡ−i = ( N −1 k=i λk1 )b1 > 0 = bi because λ11 > 0. Also, ḡ−1 =

( N 1−1 k=1 λk1 )b1 < b1 because λk1 < 1 for all k. Since ḡ−i > bi and ḡ−1 < b1 , then, according

Downloaded from [Link] at Boston College on February 15, 2015

to Lemma 3, the equilibrium noncommunication regions are [−k1 ,T1 ], [t2 ,k2 ], ..., [tN ,kN ], where
T1 ,t2 ,...,tN satisfy

T1 = min{−k1 +2δ1+ ,k1 }, (A13)

ti = max{ki +2δi− ,−ki },

We have assumed that ki is sufficiently large, such that the equilibrium is interior: δi+ < ki and
δi− > −ki . Hence, T1 = −k1 +2δ1+ and ti = ki +2δi− . The roots δi− ,δi+ are given by
 1/2
c1
δ1+ = (ḡ−1 −b1 )+ (ḡ−1 −b1 )2 + ,
1− N1 1
1+N r1

⎛ ⎞1/2
ci
δi− = (ḡ−i −bi )− ⎝(ḡ−i −bi )2 + ⎠ .
1− N1 1
1+N ri

N N  
dgi  dg 
i=1 gi = 2 i=1 gi db1 . Because gi = λi1 b1 → 0 when b1 → 0+ and  db i  ≤
d 2
Note that db1
d N
1
max{λi1 }, then limb1 →0+ db 2
i=1 gi = 0. Hence, using (A12) and (A13),
1
 2
d 1 1 1 N r1  + 2 dδ1+
lim E(V ) = − 1− + δ1 lim
b1 →0+ db1 k1 N N 1+N r1 b1 →0+ db1

  2
dδi−
N
1 1 1 N ri  − 2
+ 1− + δi lim .
ki N N 1+N ri b1 →0+ db1
i=2
 
Since ḡ−1 −b1 = ( N 1−1 k=1 λk1 −1)b1 and ḡ−i −bi = ( N 1−1 k=i λk1 )b1 ,i ≥ 2, then

dδ1+ 1 
lim = λk1 −1 < 0,
b1 →0+ db1 N −1
k=1

dδi− 1  λ11
lim = λk1 ≥ > 0.
b1 →0+ db1 N −1 N −1
k=i
 2  2
Since limb1 →0+ δ1+ > 0 and limb1 →0+ δi− > 0, we conclude that limb1 →0+ dbd E(V ) > 0.
1
(2) Consider b1 < 0. Using similar arguments, it can be shown that limb1 →0− dbd E(V ) < 0.
1
(3) Consider b1 = 0. Then, gi = 0 for any i and hence there are multiple equilibria
at the communication stage, characterized by a noncommunication region of length 2di =

1527

[18:17 8/4/2014 [Link]] Page: 1527 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

 1/2
ci
2 1 1 . According to (A12), firm value depends only on the length of the
1− N 1+N ri
noncommunication interval. Thus, firm value is exactly the same in all these equilibria and by
continuity equals limb1 →0+ E(V ) = limb1 →0− E(V ).
Combining cases (1)–(3) together, we conclude that firm value has a local minimum at the point
b1 = 0. Due to the symmetry of the problem, this implies that firm value is maximized at b1 = ±b,
where b is strictly positive, potentially infinitely large.
(ii) It can be shown that for any i0 ∈ {1,...,N −1}, limb→0+ db
d
E(V ) > 0. The proof is similar to
the proof of Part (i) and is therefore omitted.


Proof of Proposition 6. Consider any possible allocation of control (p1 ,...,pN ), pi = 1.
Without loss of generality, suppose that p1 ≥ p2 ≥ ... ≥ pN .
According to Lemma A.4, when the distribution of all signals is uniform on [−k,k],ri = 0 and

Downloaded from [Link] at Boston College on February 15, 2015

ci = c, expected firm value, E(p ,...,p ) (V ), is given by
1 N
 1/2 3  1/2 3
1 
N
1 c c
V0 −Eε 2 − (1−p1 )min ,k − (1−pi )min ,k .
3k 1−p1 3k 1−pi
i=2
c
Note that ( 1−p )1/2 ≥ k ⇔ pi ≥ 1− c
.
i k2
First, suppose that c is sufficiently large: c ≥ k 2 . Then min{( 1−p
c
)1/2 ,k} = k for all i and hence
i
there is no communication regardless of (p1 ,...,pN ). In this case, expected firm value is V0 −
2
Eε 2 − k3 (N −1), which does not depend on (p1 ,...,pN ). Hence, in this case, allocation of control
does not matter. In particular, the allocation (1,0,...,0) is optimal.
Second, suppose that c < k 2 . When control is allocated to one director: p1 = 1,pi = 0 for i > 1,
expected firm value is given by

1  3/2
N
E(1,0,...,0) (V ) = V0 −Eε 2 − c .
3k
i=2

Consider any other possible allocation of control (p1 ,...,pN ),p1 ≥ p2 ≥ ... ≥ pN . Our goal is to
show that E(1,0,...,0) (V )−E(p ,...,p ) (V ) ≥ 0, which is equivalent to
1 N
 ⎡  ⎤
1/2 3

N 1/2 3
c ⎣(1−pi )min c
(1−p1 )min ,k + ,k −c ⎦ ≥ 0. (A14)
3/2
1−p1 1−pi
i=2

There are two possible cases: p1 ≥ 1− c2 and p1 < 1− c2 . Suppose first that p1 ≥ 1− c2 . Let
k k k
M ∈ {1,...,N } be such that pi ≥ 1− c2 for i = 2,...,M and pi < 1− c2 for i = M +1,...,N. Then
k k
(A14) is equivalent to
M    N  3/2
c
(1−p1 )k 3 + (1−pi )k 3 −c3/2 + (1−pi ) −c3/2 ≥ 0. (A15)
1−pi
i=2 i=M+1
 
Note that (1−pi )( 1−pc
)3/2 −c3/2 = c3/2 (1−pi )−1/2 −1 ≥ 0 and hence the last component is
i
non-negative. The sum of the first two components of (A15) is also non-negative:

M  
k M−
3
pi −(M −1)c3/2 ≥ (M −1) k 3 −c3/2 ≥ 0.
i=1

Hence, all components of (A15) are non-negative and thus, indeed, (A14) is satisfied.
Second, suppose that p1 < 1− c2 and hence pi < 1− c2 ∀i. Then (A14) is equivalent to
  
k k
c3/2 (1−p1 )−1/2 +c3/2 Ni=2 (1−pi )
−1/2
−1 ≥ 0, which is satisfied because both components are
non-negative.

1528

[18:17 8/4/2014 [Link]] Page: 1528 1486–1532

 Communication and Decision-Making in Corporate Boards

Proof of Proposition 7. Let f and c be the density of directors’ signals and directors’ cost
of communication, respectively, and let [−k,k] be the support of the distribution. If pi = 1 and
pj = 0 for j  = i, then di = min{( c 1 )1/2 ,k} and dj = d = min{c1/2 ,k} for j  = i. Hence, according
1− 1+r
i
to Lemma A.4, expected firm value is given by

 2 di  min{c1/2 ,k}
ri
V0 −Eε 2 − x 2 f (x)dx − x 2 f (x)dx. (A16)
1+ri −di −min{c1/2 ,k}
j  =i

The first and third component of (A16) do not depend on i, and the second component is a function
of ri . Consider the function

 r 2 d(r)
g(r) = x 2 f (x)dx,

Downloaded from [Link] at Boston College on February 15, 2015

1+r −d(r)

r −1/2 r 2
where d (r) = min{c1/2 ( 1+r ) ,k}. In the region where d (r) = k, g(r) is proportional to ( 1+r ) and
r −1/2
hence is increasing in r. In the region where d (r) = c1/2 ( 1+r ) , it can be shown that g (r) > 0 is
equivalent to

1  r −3/2  r 2 2r d(r)
− c1/2 d 2 (r)f (d (r))+ x 2 f (x)dx > 0. (A17)
2 1+r 1+r 1+r 0

 d(r)
Because f (x) is non-increasing for x > 0, 0 x 2 f (x)dx ≥ d 2 (r)f (d (r))d (r) and substituting
this inequality into (A17), we see that (A17) is always satisfied. Finally, g(r) is continuous at the
r −1/2
point where d (r) switches from c1/2 ( 1+r ) to k. Hence, the function g(r) is increasing in r,
which proves that (A16) is maximized when i ∈ argminj {rj }.

Lemma A.5.  Suppose directors’ utility is given by (22), where w(r) = 1−r. Then firm value is
maximized at r1∗ ,...,rN∗ , where ri∗ is strictly positive.

Proof of Lemma A.5. We start by deriving the equilibrium strategies and expected firm value for
a general function w(r). Repeating the proof of Lemma A.1, it is easy to show that the equilibrium
action of director i at the decision-making stage is given by (10) for directors who communicated
their signals, and by
 
ai∗ = xj + yj +Qi xi +(1−Qi )yi
j ∈JC j ∈JN C ,j =i

w (ri )pi
for directors who did not communicate their signals, where Qi = w . Using these expressions
(ri )pi +ri
and repeating the proof of Lemma A.3, director i reveals xi at the communication stage if and only
c
if |xi | > di = ( Ai )1/2 , where
i

w(ri )pi ri
Ai = w(ri )(1−pi )+ .
w(ri )pi +ri

It follows that di decreases in ri if and only if (23) is satisfied.

Repeating the proof of Lemma A.4, expected firm value is given by


N  2 di
ri
E(V ) = V0 −Eε 2 − 1−pi +pi xi2 fi (xi )dxi . (A18)
w(ri )pi +ri −di
i=1

1529

[18:17 8/4/2014 [Link]] Page: 1529 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

Special case: w(r) = 1−r

Suppose w(r) = 1−r, r ∈ [0,1]. Plugging this into (23), di decreases in ri if and only
   
f (r) ≡ r 2 1−2p +2p2 −p 3 +2r p −p2 +p 3 −p 3 < 0,
 
where 1−2p +2p2 −p 3 = (1−p) p 2 −p +1 > 0. Since f (0) < 0 and f (1) = 1 > 0, then f (r) <
0 ⇔ r < r̂ for some r̂ ∈ (0,1). In addition, using (23) and (A18), it can be shown that

∂E(V )  
= fi ci (1−pi )−1/2 ci (1−pi )−3/2 pi ,
1/2 3/2
lim
ri →0 ∂ri

which is strictly positive. Hence, firm value is maximized at some strictly positive ri , potentially,
infinitely large.

Downloaded from [Link] at Boston College on February 15, 2015

References

Adams, R. B., and D. Ferreira. 2007. A theory of friendly boards. Journal of Finance 62:217–50.

Adams, R. B., B. E. Hermalin, and M. S. Weisbach. 2010. The role of boards of directors in corporate governance:
A conceptual framework and survey. Journal of Economic Literature 48:58–107.

Agrawal, A., and M. A. Chen. 2011. Boardroom brawls: An empirical analysis of disputes involving directors.
Working Paper, University of Alabama.

Almazan, A., and J. Suarez. 2003. Entrenchment and severance pay in optimal governance structures. Journal
of Finance 58:519–47.

Alonso, R., W. Dessein, and N. Matouschek. 2008. When does coordination require centralization? American
Economic Review 98:145–79.

Austen-Smith, D., and T. Feddersen. 2005. Deliberation and voting rules, 269–316. In Social choice and strategic
decisions: Essays in honor of Jeffrey S. Banks. New York: Springer.

Banerjee, A. V. 1992. A simple model of herd behavior. Quarterly Journal of Economics 107:797–817.

Baranchuk, N., and P. H. Dybvig. 2009. Consensus in diverse corporate boards. Review of Financial Studies
22:715–47.

Bikhchandani, S., D. Hirshleifer, and I. Welch. 1992. A theory of fads, fashion, custom, and cultural change as
informational cascades. Journal of Political Economy 100:992–1026.

Burkart, M., and K. Raff. 2012. Performance pay, CEO dismissal, and the dual role of takeovers. Working Paper,
Swedish House of Finance.

Carter, C. B., and J. W. Lorsch. 2004. Back to the drawing board: Designing corporate boards for a complex
world. Boston: Harvard Business School Press.

Chakraborty, A., and B. Yilmaz. 2011. Authority, consensus and governance. Working Paper, University of
Pennsylvania.

Chemmanur, T. J., and V. Fedaseyeu. 2011. Theory of corporate boards and forced CEO turnover. Working Paper,
Boston College.

Cornelli, F., Z. W. Kominek, and A. Ljungqvist. 2013. Monitoring managers: Does it matter? Journal of Finance
68:431–81.

Coughlan, P. J. 2000. In defense of unanimous jury verdicts: Mistrials, communication, and strategic voting.
American Political Science Review 94:375–93.

Crawford, V. P., and J. Sobel. 1982. Strategic information transmission. Econometrica 50:1431–51.

Dessein, W. 2002. Authority and communication in organizations. Review of Economic Studies 69:811–38.

1530

[18:17 8/4/2014 [Link]] Page: 1530 1486–1532

 Communication and Decision-Making in Corporate Boards

Dewatripont, M. 2006. Presidential Address: Costly communication and incentives. Journal of the European
Economic Association 4:253–68.

Dewatripont, M., and J. Tirole. 2005. Modes of communication. Journal of Political Economy 113:
1217–38.

Fich, E. M., and A. Shivdasani. 2006. Are busy boards effective monitors? Journal of Finance 61:689–724.

Fishman, M. J., and K. M. Hagerty. 1990. The optimal amount of discretion to allow in disclosure. Quarterly
Journal of Economics 105:427–44.

Gerling, K., H. P. Grüner, A. Kiel, and E. Schulte. 2005. Information acquisition and decision making in
committees: A survey. European Journal of Political Economy 21:563–97.

Gillette, A. B., T. H. Noe, and M. J. Rebello. 2003. Corporate board composition, protocols, and voting behavior:
Experimental evidence. Journal of Finance 58:1997–2031.

Downloaded from [Link] at Boston College on February 15, 2015

Grossman, S. J. 1981. The informational role of warranties and private disclosure about product quality. Journal
of Law and Economics 24:461–83.

Harris, M., and A. Raviv. 2005. Allocation of decision-making authority. Review of Finance 9:353–83.

Harris, M., and A. Raviv. 2008. A theory of board control and size. Review of Financial Studies
21:1797–1832.

Hart, S. 2003. Directors’ etiquette in a changing boardroom. Directors & Boards 27:61–62.

Hermalin, B. E., and M. S. Weisbach. 1998. Endogenously chosen boards of directors and their monitoring of
the CEO. American Economic Review 88:96–118.

Hirshleifer, D., and A. V. Thakor. 1994. Managerial performance, boards of directors and takeover bidding.
Journal of Corporate Finance 1:63–98.

Jiang, W., H. Wan, and S. Zhao. 2012. Reputation concerns of independent directors: Evidence from individual
director voting. Working Paper, Columbia University.

Kumar, P., and K. Sivaramakrishnan. 2008. Who monitors the monitor? The effect of board independence on
executive compensation and firm value. Review of Financial Studies 21:1371–1401.

Leblanc, R., and J. Gillies. 2005. Inside the boardroom: How boards really work and the coming revolution in
corporate governance. Ontario: John Wiley & Sons.

Levit, D. 2012. Expertise, structure, and reputation of corporate boards. Working Paper, University of
Pennsylvania.

Li, H., S. Rosen, and W. Suen. 2001. Conflicts and common interests in committees. American Economic Review
91:1478–97.

Lipton, M., and J. W. Lorsch. 1992. A modest proposal for improved corporate governance. Business Lawyer
48:59–78.

Lorsch, J. W., and E. MacIver. 1989. Pawns or potentates: The reality of America’s corporate boards. Boston:
Harvard Business School Press.

Mace, M. L. 1986. Directors: Myth and reality. Boston: Harvard Business School Press.

Marshall, C. D. 2010. Are dissenting directors rewarded? Working Paper, University of Richmond.

Milgrom, P. R. 1981. Good news and bad news: Representation theorems and applications. Bell Journal of
Economics 12:380–91.

Morris, S., and H. S. Shin. 2002. Social value of public information. American Economic Review
92:1521–34.

Pae, S. 2005. Selective disclosures in the presence of uncertainty about information endowment. Journal of
Accounting and Economics 39:383–409.

1531

[18:17 8/4/2014 [Link]] Page: 1531 1486–1532

 The Review of Financial Studies / v 27 n 5 2014

Raheja, C. G. 2005. Determinants of board size and composition: A theory of corporate boards. Journal of
Financial and Quantitative Analysis 40:283–306.

Rock, R. H. 2004. Secret ballot. Directors & Boards 28:4.

Sapienza, H. J., M. A. Korsgaard, and J. Hoogendam. 1997. What do new venture boards do? Frontiers of
Entrepreneurship Research, 118–30.

Scharfstein, D. S., and J. C. Stein. 1990. Herd behavior and investment. American Economic Review 80:465–79.

Schwartz-Ziv, M., and M. S. Weisbach. 2013. What do boards really do? Evidence from minutes of board
meetings. Journal of Financial Economics 108:349–66.

Segal, I. 2006. Communication in economic mechanisms. In R. Blundell, W. Newey, and T. Persson (Eds.),
Advances in economics and econometrics: Theory and applications, ninth world congress. Cambridge:
Cambridge University Press.

Downloaded from [Link] at Boston College on February 15, 2015

Shue, K. 2013. Executive networks and firm policies: Evidence from the random assignment of MBA peers.
Review of Financial Studies 26:1401–42.

Song, F., and A. V. Thakor. 2006. Information control, career concerns, and corporate governance. Journal of
Finance 61:1845–96.

Sonnenfeld, J. A. 2002. What makes great boards great. Harvard Business Review 80:106–13.

Stevenson, W. B., and R. F. Radin. 2009. Social capital and social influence on the board of directors. Journal of
Management Studies 46:16–44.

Swartz, M. 2007. Secrecy, collusion and coalition building. Corporate Ownership and Control 4:10–24.

Vermeule, A. 2010. Open-secret voting. Harvard Law School Discussion Paper No. 674.

Verrecchia, R. E. 1983. Discretionary disclosure. Journal of Accounting and Economics 5:179–94.

Warther, V. A. 1998. Board effectiveness and board dissent: A model of the board’s relationship to management
and shareholders. Journal of Corporate Finance 4:53–70.

Zeckhauser, R. 1973. Voting systems, honest preferences and pareto optimality. American Political Science
Review 67:934–46.

Zhang, J., and M. Casari. 2012. How groups reach agreement in risky choices: An experiment. Economic Inquiry
50:502–15.

Zwiebel, J. 1995. Corporate conservatism and relative compensation. Journal of Political Economy 103:1–25.

1532

[18:17 8/4/2014 [Link]] Page: 1532 1486–1532