The Credibility of Performance Feedback in Tournaments

• DOI: http://doi.org/10.1111/jems.12086
• Published date: 01 March 2015
• Date: 01 March 2015
• Author: Iván Marinovic

The Credibility of Performance Feedback in

Tournaments

IV´

AN MARINOVIC

Graduate School of Business

Stanford University

655 Knight Way

Stanford, CA 94305-7298

imvial@stanford.edu

This paper studies the effect of performance feedback on tournament outcomes, when a possibly

dishonest principal may manipulate the agents’ expectations to stimulate their effort. Under

plausible circumstances, an increase in the principal’s propensity to tell the truth (i.e., integrity)

induces a mean preserving spread in the distribution of effort and leads to a decrease in expected

profits and welfare. More generally, I identify conditions under which a lower integrity can

improve the effectiveness of financial incentives in inducing the agents’ effort, thus leading to

higher expected profits for the principal.

1. Introduction

Performance feedback policies are pervasive within firms. Ederer (2010) cites evidence

showing that between 74% and 89% of business organizations have a formal performance

feedback system. Law firms and consulting firms, where promotion is a crucial compo-

nent of the incentive package, inform workers about their previous performance and

promotion prospects long before the decisions are made. Yet, surveys reveal widespread

discontent with the way performance feedback is conducted in a substantial percentage

of U.S. and U.K. firms (Fletcher, 2001). Two of the most cited reasons for this discon-

tent are feedback manipulation by self-interested appraisers and the lack of meaningful

feedback provided by the process (Barlow, 1989).1

This paper studies the effects of performance feedback on tournament outcomes

when a potentially self-interested principal may manipulate the communication to stim-

ulate the agents’ effort. I build on the model of Ederer and Fehr (2007). At the outset, the

principal chooses the tournament reward. If the agents accept the offer and join the firm,

the principal privately observes a signal of the agents’ talents hence their relative like-

lihoods of promotion. Then the principal communicates this information to the agents

and based on this communication the agents exert effort.

Because the incentives of the principal and the agents are strongly misaligned, the

canonical cheap talk model (Crawford and Sobel, 1982) predicts that no information

transmission can ever take place: any communication should be mere babbling.Soto

I would like to thank Jeremy Bertomeu, Florian Ederer, Ilan Guttman, and Marco Ottaviani for very helpful

comments and Yinqing Xing and Sandro Ambuehl for great researchassistance.

1. For example, according to Jeffrey Pfeffer: “At North Carolina software company SAS Institute, David

Russo, then head of HR, got cheers from employees in the mid-1990s for a bonfire celebration that fed appraisal

forms into the flames. Nothing has changed much since then. Managers don’t like doing appraisals. Employees

don’t like getting them. Perhaps that’s because they all suspect what the evidence shows: such performance

reviews don’t work.”

C2015 Wiley Periodicals, Inc.

Journal of Economics & Management Strategy, Volume24, Number 1, Spring 2015, 165–188

166 Journal of Economics & Management Strategy

enable meaningful communication, I assume the principal has a positive (and known)

propensity to be truthful, which I call integrity. This definition of integrity captures not

only the principal’s moral characteristics, but more broadly the transparency of the firm’s

feedback process, or the amount of discretion available to the principal to manipulate

the communication. In that sense, integrity is viewed here as an observable metric of

transparency arising, for example, from the firm’s prior investments and commitments.

In the model, the tension arises from the agents being uncertain about whether

the principal is truthful, thus having to infer this from the communication itself. When

untruthful, the principal must choose the feedback carefully to avoid being detected, for

otherwise his ability to influence the agents’ effort would vanish.

I use this framework to study how integrity affects social welfare. I focus on the

case in which the agents’ participation constraints bind. Surprisingly, I find that under

plausible and rather general circumstances integrity has an adverse effect over the

principal’s expected profits and the social welfare. The mechanism behind this result is

intuitive. Given the tournament’s reward, the agents’ effort depends on their perception

of relative talent: the greater the perceived gap between the agents’ talents the less

competitive the tournament becomes. Naturally, the perception of a more competitive

environment induces higher effort levels. Hence, when untruthful, the principal would

like to persuade the agents that the tournament is highly competitive. However, the

agents are aware of the potential conflict of interest and they discount the feedback,

particularly if the principal has low integrity and claims high levels of competitiveness

in the tournament. In this context, a principal with higher integrity is more likely to

draw the agents’ beliefs about competitiveness to polar levels. A higher integrity thus

increases the dispersion of the agents’ conditional beliefs about the marginal return to

effort. This is the key to understanding why integrity may have an adverse effect on

social welfare.

For the sake of illustration, suppose the agents’ equilibrium effort strategy is linear

in conditional beliefs. By the law of iterated expectations, a higher integrity cannot affect

expected effort. However, it does increasethe volatility of effort. In turn, a more volatile

effort translates into a higher expected cost of effort: that is, the more volatile the effort

the higher the cost that the agents expect to bear in the tournament. This means that,

given the tournament’s reward, an increase in integrity lowers the agents’ expected

utility from participating in the tournament. In equilibrium, however,this adverse effect

of integrity must be borne by the principal as he must set the tournaments’ rewards so

as to ensure the agents’ participation.

The general analysis is more subtle because the effort strategy may not be linear

in conditional beliefs. Hence, integrity has the potential to affect both the mean and the

volatility of effort. The sign of the mean effect crucially depends on the convexity of the

marginal cost of effort. For difficult tasks, namely those exhibiting a convex marginal

cost of effort, the effort strategy is concave in conditional beliefs. The greater dispersion

of conditional beliefs induced by an increase in integrity results in lower expected

effort, lower expected profits, and lower expected welfare. By contrast, for simple tasks,

namely those exhibiting concave marginal cost of effort, the effort strategy is convex

in conditional beliefs. Hence the greater dispersion of conditional beliefs induced by

higher integrity leads not only to more volatile effort but also to greater expected effort.

A higher integrity is thus detrimental to the agents’ interests but may be beneficial to

those of the principal. In fact, consider the case where the agents’ participation constraint

does not bind, which happens, for instance, when the agents’ outside option is relatively

low but the agents are protected by limited liability. Given the tournament’s reward,