Lying Aversion and the Size of the Lie

This paper studies lying in a simple framework. An agent first randomly picks a number from a known distribution. She can then claim to have observed any number from the set, receiving a monetary payoff based only on her report. Consistent with previous findings, our participants do not maximize monetary payoff by making the maximal claim dishonestly. The paper posits that this behavior is the result of lying costs and discusses different kinds of lying cost. The paper presents a model of lying costs that is used to generate hypotheses regarding behavior in the experiment. In line with the model, we find that the highest fraction of lies is by reporting the maximal outcome. Reputational concerns matter: More participants lie partially when their outcomes cannot be observed by the experimenter than when the experimenter can later verify the actual outcome, and partial lying increases when the highest outcome is ex ante unlikely. In contrast, the fraction of subjects who lie does not depend on how outcomes are labeled