Biased Recommendations

Abstract

Can experts be trusted to provide useful recommendations? We develop and experimentally test a simplified recommendation game where an expert recommends one of two actions to a decision maker who might instead take no action. Consistent with predictions from the cheap talk literature, we find that recommendations are persuasive in that they induce actions benefiting the expert, but decision makers partially discount recommendations for the action a biased expert favors. Even unbiased experts pander by recommending the action that the decision maker already favors, which decision makers then discount. And if the decision maker is uncertain over whether the expert is biased or not toward an action, unbiased experts follow a political correctness strategy of recommending the opposite action to be more persuasive. However, decision makers do not sufficiently anticipate how uncertainty over the expert's bias gives even unbiased experts an incentive to lie, implying that transparency may be even more important to successful communication in practice than it is in theory.