The Paternalistic Bias of Expert Advice

• Author: John P. Lightle
• Published date: 01 December 2014
• DOI: http://doi.org/10.1111/jems.12070
• Date: 01 December 2014

The Paternalistic Bias of Expert Advice

JOHN P. LIGHTLE

Department of Economics

Florida State University

Tallahassee,FL 32306

jlightle@fsu.edu

Using a theoretical model of noisy expert advice, I show that language inflation can be a rational

response to the vagueness of language. Experts will tend to overstate their positions to a like-

minded decision maker (DM) and this constitutes a Pareto improvement over sending a sincere

message. When the message space is bounded, overstatement may interfere with the DM’s ability

to aggregate the experts’ information, because communication is less precise when the same

message is sent for many states of the world. Despite this, I show that experts are willing to send

either the most extreme message to the DM, or a partially overstated message, because by doing

so the expert can decrease the likelihood the DM makes a suboptimal decision due to his subjective

interpretation of the advice. Because the expert inflates his message toward the policy he believes

the DM would be better off choosing, rather than sincerely revealing his information, I refer to

this behavior as a paternalistic bias.

1. Introduction

Employing the advice of experts is an essential part of being a successful decision maker

(DM) in the public arena. When facing a particularly complicated decision, a decision

maker often faces constraints which make acquiring and/or processing all the relevant

information inefficient or impossible. For example, consider a CEO who must decide

whether or not to buyout a rival company.Analysts will attempt to explain the expected

costs and benefits of acquisition to the CEO, but the ultimate decision is his alone.

Conventional wisdom suggests that, for the good of the company, analysts should be

forthright with their knowledge, giving as accurate a description of the state of the

world as possible. However, this paper challenges that intuition when you include an

often dismissed, yet quite realistic, caveat: communication between analyst and CEO is

inherently imprecise. In other words, the analyst’s intended description of the state of

the world is not equal to the CEO’s interpretation of this message. In this case, using

the model presented in this paper, I identify a novel “paternalistic bias” on the part of

an advisor—that is, an endogenous bias that occurs because an advisor believes that by

sending an insincere message about the state of the world, he can help the DM make a

better choice than had he revealed his information sincerely.

Let us now consider more carefully the decision made by the CEO of Company

X of whether or not to buyout Company Y. Although there may be countless factors

which determine the profitability of such an acquisition, for simplicity we will assume

there are only two relevant pieces of information: the current value of Company Y to

The author thanks seminar participants at the Ohio State University Theory Workshopand the FSU Workshop

on Experimental Game Theory, John Kagel, David Cooper, James Peck, Dan Levin, P.J.Healy, Navin Kartik,

two anonymous referees, and particularly Massimo Morelli, for their expert advice. All errors are my own.

C2014 Wiley Periodicals, Inc.

Journal of Economics & Management Strategy, Volume23, Number 4, Winter 2014, 876–898

The Paternalistic Bias of Expert Advice 877

Company X, and the net increase in future profitability of Company X after purchasing

Company Y. The CEO has one team of industrial experts who will evaluate the quality

and usefulness of Company Y’s physical assets, and another team of economic forecasters

who will evaluate the expected market conditions after the acquisition of Company Y.

We will say that each team of experts conducts a sufficient amount of research to obtain

the true state of the world with arbitrarily high precision, and makes a report of this

information to the CEO. However, because the information obtained is highly technical,

the experts cannot fully communicate their findings to the CEO. In other words, due to

the CEO’s subjective interpretation of the relevant information, the message which the

teams of experts intended for the CEO to hear is not exactly the same message that the

CEO actually hears.

The profitability of the buyout depends on the information obtained by both teams

of experts, and in this scenario each team is a layman when it comes to the other field

of expertise. Neither group of experts is capable of making an informed final decision

and the CEO becomes a necessary part of the decision process if only for his role as a

final information aggregator.The experts and the CEO have perfectly aligned incentives,

both want to proceed with the buyout if and only if it is profitable to do so. Despite this

fact, I show in this paper that the two teams of experts would each rationally choose to

bias their report of the state of the world in favor of whichever outcome they believe is

most likely to be optimal given their information.

To make the example more concrete, let θ1represent the true state of the world

related to Company Y’s current assets, and let θ2be the true state of the world related

to the expected net increase in discounted future profits after acquiring Company Y.

Without loss of generality, we normalize 0 to be smallest possible value of θ1or θ2,

and normalize 1 to be the largest possible value. For simplicity, say that θ1and θ2are

independent and drawn from distributions which are common knowledge. The payoff

to Company X of the buyout is an increasing function of θ1and θ2,sayθ1+θ2,andthe

buyout is optimal only if θ1+θ2>1. The two teams of experts will then make a report

(or a message) to send to the CEO which will be represented as m1and m2, with each

message constrained to the set [0, 1].1The CEO will not know the true messages sent

by the experts, but will instead receive an unbiased noisy signal of each message sent.

Given these two signals, the CEO must decide whether or not to takeover Company Y.

For the purposes of this example, we will say that whether or not the CEO makes

the payoff maximizing decision is an important distinction for the CEO and the two

teams of experts. Rather than being a continuous function of θ1+θ2, we will assume

that utility of both the CEO and experts depend entirely on whether the right decision

was made. One reason for such a utility function could be that the CEO and his teams

of experts will be fired if and only if they make the wrong decision with respect to the

buyout. We can then assume they receive a utility of 1 if the takeover occurred when

θ1+θ2>1 or if the takeover did not occur when θ1+θ2<1, and a utility of 0 otherwise.

Therefore, their expected utility is the probability that the right decision was made.

1. This paper focuses on the case where an advisor has a message space as rich the space from which

her information is drawn. There are interesting examples of insincere behavior when messages are restricted,

for example, voting. Austen-Smith and Banks (1996) demonstrate that even when all voters have common

preferencesover outcomes, sincere voting is not rational, due to the inferences made about others’ information.

(See also, Feddersen and Pesendorfer, 1996; Feddersen and Pesendorfer, 1998.) The addition of a round of

deliberation before a vote takes place can eliminate nonsincere voting behavior (Coughlan, 2000; Gerardi and

Yariv, 2007), which reduces the game to one in which the message space is no longer insufficient.