Private monitoring, collusion, and the timing of information

• DOI: http://doi.org/10.1111/1756-2171.12114
• Date: 01 October 2015
• Author: Troy J. Scott,Jacques Lawarrée,Fahad Khalil
• Published date: 01 October 2015

RAND Journal of Economics

Vol.46, No. 4, Winter 2015

pp. 872–890

Private monitoring, collusion, and the timing

of information

Fahad Khalil∗

Jacques Lawarr´

ee∗∗

and

Troy J. Scott∗∗∗

When a principal’s monitoring information is private (nonverifiable), the agent should be con-

cerned that the principal could misrepresent the informationto reduce the agent’s wageor collect

a monetary penalty. Restoring credibility may lead to an extreme waste of resources—the so-

called burning of money. A more realistic and efficient outcome is feasible when the private

information arrives in time to rescale the agent’s effort. Rescaling is more effective than pure

monetary penalties because effort has different values to different parties whereas money is

equally valuable to all parties. Furthermore, when rescaling is feasible, private monitoring is

more efficient than public monitoring subject to collusion because nonmonetary penalties are

ineffective to deter collusion.

1. Introduction

Contracting parties can obtain monitoring information at various times during an agency

relationship. In this article, we focus on monitoring information learned privately bythe principal,

that is, private signals, and we show the importance of the timing of information. Private signals

cannot be verified by a third party. For example, the signal could be the principal’s subjective

evaluation of the project’s difficulty or of an agent’s performance, a casual conversation with a

customer or a coworker about the agent’s behavior, or the filtering of the agent’s e-mail. Another

example is the information that internal and external auditors may share with audit committees

in executive sessions with details that go beyond what is documented in publicly released audit

reports. In military procurement, national security concerns may prevent the Department of

Defense from releasing information publicly. In these examples, the private information may

arrive ex post, that is, after the agent has performed his task, or ex ante, that is, before or during

∗University of Washingtonand ERG, Dhaka ; khalil@uw.edu.

∗∗University of Washington and ECARES, Brussels ; lawarree@uw.edu.

∗∗∗RTI International; tjscott@uw.edu.

We thank Yoram Barzel, Mathias Dewatripont, William Fuchs, David Martimort, Roger Myerson, Bruno Parigi, Wing

Suen, Roland Strausz, Xu Tan, Quan Wen, and two referees for helpful comments.

872 C2015, The RAND Corporation.

KHALIL, LAWARR´

EE, AND SCOTT / 873

the performance of tasks. The aim of this article is to investigate to what extent the principal may

use such private signals to improve the agent’s incentives and to highlight the importance of the

timing of information.

When a signal is privately observed by the principal, it provides an opportunity for the

principal to abuse the private information, and the principal’s credibility becomes an issue. For

instance, an employee should be concerned if the employer can reduce his wage or impose a

penalty based on an unverifiable privatesignal. To resolve this concern, we consider two options.

The principal can reassure the agent either by removing her own incentive to abuse private

information, or by relying instead on third-party monitoring.

The main part of our analysis discusses the first option, where a new constraint is needed

to ensure that the principal cannot gain from a penalty based on a private signal. One way to

satisfy this constraint is to force the principal to give away, for instance to a charity, any collected

penalty—the so-called burning money solution.1Although this solution is quite intriguing, it is

also a little disconcerting as it involves wasting resources with respect to the contracting parties.

In this article, we argue that organizations will look for more efficientsolutions than b urning

money to address the credibility issue with private signals. In particular, the principal may have

additional alternatives depending on the timing of private information. If she obtains private

information ex ante, then the agent’s required effort, that is, the scale of the project itself, can

be reduced as an alternative to burning money.2We show that effort can be rescaled so that it

has two effects: it not only penalizes the agent but also ensures that the principal, at that stage,

does not gain from penalizing the agent. We also show that relying on nonmonetary penalties

like rescaling is more effective than pure monetary penalties, which require money burning. The

reason is that effort has different values to different parties, whereas money is equally valuable

to all parties.

To present this intuition, we use an adverse-selection model where a principal observes the

output and can use an imperfect private signal of the agent’s type to induce effort. The signal

allows the principal to penalize the shirking agent but, because it is imperfect, complying agents

are also penalized by mistake. If the penalty has to be burned, such mistakes are costly, and the

signal is less valuable. In the case of an ex post private signal, we show that the signal is not

valuable unless it is accurate enough. In the case of an ex ante private signal, we show that the

signal is valuable even if its accuracy is very low. The reason is that the principal can rely on

rescaling rather than being limited to burning money. If rescaling allows the principal to extract

all rent for low accuracy, then there is no money burning in equilibrium. Otherwise, burning

money can be used to extract the remaining rent.

Because rescaling cannot take place after the project has already been completed, the timing

of information becomes critical: we find that the principal always prefers to receive private

information early, that is, ex ante. In other words, the principal may be willing to incur an extra

cost to receive private signals ex ante rather than ex post.

A second option for the principal to avoid her credibility problem with privatemonitoring is

to rely instead on public monitoring by a third party (e.g., a public accounting firm). If third-party

monitoring was free, this solution wouldalways be preferred. As is well known from the literature,

the first best can be reached irrespective of the timing of the signal.3

However, in a world where information can be manipulated, we consider an endogenous

cost of public monitoring due to the potential for collusion between the monitor (e.g., an auditor)

1See the recent literature on subjective evaluation (Levin, 2003; MacLeod, 2003; Fuchs, 2007). Riordan and

Sappington (1986) propose an analogous remedy: the principal may commit not to accept some portion of the output

ex post, allowing it to go to waste.

2Holmstrom and Milgrom (1987) also challenge the assumption that the agent is required to make a once-for-all

choice of his effort during the relationship without regard to the arrival of performance information. Their focus is on the

agent, instead of the principal, receiving information and adjusting his effort continuously over time.

3See Cr´

emer and McLean (1988) for ex ante signals, and Baron and Besanko (1984) and Riordan and Sappington

(1988) for the case of ex post signals.

C

The RAND Corporation 2015.