Lying for votes

To what extent do pivotal nonpartisan voters believe and act upon potentially deceptive messages from partisans who privately observe candidates' ex-ante fitness-for-office attributes? How do nonpartisan voters' responses to messages vary with changing odds of candidates' fitness-for-office attributes? This paper derives contrasting predictions regarding pivotal nonpartisans' reactions to potential lies across two-candidate voting competitions and tests them in the laboratory. We find that the partisan voters lie substantially more when candidates possess unequal than equal ex-ante odds. The inferior (superior) candidate's base transmits more unfavorable (favorable) lies about their opponent (own) candidate. Facing candidates with different odds, the nonpartisan voters largely ignore the message and vote for the ex-ante superior candidate. Yet, lying lowers efficiency compared to when the message is always factual, and the efficiency loss is marginally higher when the candidates are ex-ante unequal than equal. We develop a behavioral framework to explain key features of our data.