A capture theory of committees

For excellent surveys, see Gerling et al. (2005) and Li and Suen (2009).

It is well-established in the literature that committees may fail to aggregate diverse information because of strategic (or pivotal) voting (e.g., Austen-Smith and Banks 1996; Feddersen and Pesendorfer 1998).

Stigler presented an influential theory (and empirical evidence) of regulatory capture, which was later refined and expanded by Peltzman (1976) and Becker (1983) and applied to many other settings, including the political economy of trade policy (Grossman and Helpman 1994). For edifying reviews of the regulatory capture literature, see Laffont and Tirole (1993, ch. 11) and Dal Bó (2006).

Like other researchers, we recognize that vote buying is illegal in many societies and organizations, but the inducements offered to committee members need not be explicit. Indeed, there have been many real cases in which committee decisions were doubted or even dismissed owing to the fear of capture. Elliott (2011) reports that, perceived of unduly favoring the industry, the US Atomic Energy Commission was replaced by the independent Nuclear Regulatory Commission (NRC) in 1975. In sports, the international soccer federation’s (FIFA’s) decision to award the 2018 and 2022 World Cups to Russia and Qatar, respectively, were linked to bribery and vote-rigging, resulting in the indictments of several top FIFA officials (Collett et al. 2015). Last, but not least, in the 2016 Rio Olympics, several referees and judges were removed from the boxing competitions after “suspicious results” (Belson 2016).

We later extend the analysis to imperfectly informed experts for whom information aggregation is an issue.

In our baseline model, a larger committee size raises the cost of capture for the agent both because it increases the probability of a sufficiently incorruptible (or socially motivated) member and because it increases the number of bribes to be paid. What ensures a finite committee is a small cost of participation for each member (as in Persico 2004) or a finite pool of available experts. If, unlike our model, experts were purely self-motivated, no committee size would deter capture, as will be clear in the analysis.

Consistent with this finding, a randomized scoring system is used in Olympic boxing competitions, whereby only a subset of judges’ scores are tallied (Belson 2016). See also Amegashie (2006) for a similar rule in Olympic figure skating. Similarly, the verdicts of panels of judges in civil law regimes are announced by the court as a whole; the votes of the individual judges are not disclosed.

As will be seen in the analysis, without the ability to commit to size, the principal has an incentive to scale down the committee under no disclosure. Hence, if partial disclosure is not feasible or credible, the principal may opt for full disclosure of the committee.

The uniform distribution assumption greatly simplifies the analysis, but is not essential for the results.

The fact that the participation cost of a member is small reflects the idea that the committee serves the larger society.

Whether individual votes are secret or public is of no consequence in our setting since, as we will see below, the principal optimally chooses a rule of the unanimity, i.e., \(k=n\), thereby removing the possibility of vote-buying schemes based on casting a pivotal vote, as in Dal Bó (2007).

As is standard in the literature on political influence, we assume that the agent fulfills his promise of bribes even in a one-shot interaction: perhaps because he cares strongly about his “word of honor” or building reputation across a sequence of ad hoc committees; see, e.g., Laffont and Tirole (1993, ch. 11) for a discussion.

Claim A1 in the online appendix establishes that the agent has no incentive to bribe in the original problem if and only if he has no incentive to bribe in the relaxed problem.

Of course, there are other institutional and informational reasons, not modeled here, for adopting a unanimity rule (see, e.g., Yildirim 2007; Bond and Eraslan 2010; Alonso and Câmara 2016; Breitmoser and Valasek 2017).

Since an unbribed committee member would accept only a socially desirable project under the unanimity rule, bribes that target a subset of members would be a pure waste for the agent.

As is common in committee voting problems, a trivial equilibrium exists in which all members reject the project regardless of its social value—i.e., \(\phi _{-i}=0\) for all i. Aside from being uninteresting, such an equilibrium involves weakly dominated strategies for committee members and thus is not considered herein.

Recall that experts incur a negligible but positive participation cost, leading the principal to select the smallest committee that deters bribery.

For instance, the International Olympics Committee (of 98 members), the Tony Awards nominating committee (of 51 members), and university presidential search committees (of 15–21 members) are commonly publicized.

Specifically, Claim A2 in the online appendix shows that under nondisclosure, unless the committee size is conjectured to be one, it would be strictly optimal for the agent to bribe all N experts. Otherwise, as seen in (5), a trivial indifference exists as to the number of bribes, m, offered when the committee comprises one member.

The composition effect identified under full disclosure is internal to the committee and operates regardless of other circumstances.

If the committee prepares a joint report after the vote, then \(c=\frac{C}{n}\) may be considered to be the (decreasing) marginal cost per member. Our conclusion in Proposition 4, however, would not change.

Whether a member decides to become informed before or after receiving a bribe has no qualitative effects, since we focus on the no-bribing equilibrium.

Given that \(s\sim U[-\,S,S]\), the value of information is

In fact, no committee size would deter bribing if \(\eta _{E}>\frac{S}{4}\). The reason is that, with sufficiently high information costs, all members would remain uninformed and vote to accept the project in exchange for a negligible bribe. Anticipating such full capture, the principal would not appoint a committee.

Recall that all informed members observe the same signal, so \(E[s|s_{i},s^{*},I]\) is not conditional on the number of informed members.