Reputational cheap talk.(private information reporting)

We analyze information reporting by a privately informed expert concerned about being perceived to have accurate information. When the expert's reputation is updated on the basis of the report as well as the realized state, the expert typically does not wish to truthfully reveal the signal observed. The incentives to deviate from truthtelling are characterized and shown to depend on the information structure. In equilibrium, experts can credibly communicate only part of their information. Our results also hold when experts have private information about their own accuracy and care about their reputation relative to others.

1. Introduction

* To foster their careers, experts typically seek favorable evaluations. For example, consider an analyst consulted by an investor about the value of a stock. The investor evaluates the accuracy of the analyst's information by cross-checking the analyst's report with the realized profitability of the investment. Analysts who are believed to have access to more accurate information are rewarded by the market. Similarly, managers and business consultants with better reputations can charge more for their services. Company directors and board members who are believed to be effective are more likely to be reappointed or hired by other companies. Politicians considered to be better informed are more likely to be re-elected and so derive increased private benefits. The model of reputational cheap talk developed in this article can be applied to these and other situations in which individuals are concerned about their reputation for being well informed.

Our model features an expert who observes a private signal (s) about a state of the world (x). The amount of information contained in this signal depends on the expert's ability type (t). After observing the signal (and possibly the ability), the expert reports a message (m) to an evaluator. The evaluator later observes the state and combines it with the message to update the belief regarding the expert's ability. This belief (or reputation) determines the expert's payoff, which the expert aims to maximize. This is a cheap-talk (or costless-signalling) game, in which the expert (sender) does not bear a direct cost from the message sent, but cares about the induced response of the evaluator (receiver). (1,2)

In the context of this model, we aim at addressing the following questions:

(i) When is the expert's concern for ability compatible with truthful reporting?

(ii) In which direction does the expert wish to bias the report, when believed to be truthful?