Snap Judgments: Voter Inferences Based on Candidate Photos Predict Electoral Success and Politician Quality

Voting is fundamentally a forecasting problem: voters try to predict future performance in office based on incomplete information about candidates. Forecast inputs combine observable professional qualifications with more subjective assessments of confidence and trustworthiness. In developing countries, the amount of information available can be quite limited. This paper explores how well voters do in predicting candidate performance and quality under varying degrees and types of information. It leverages a series of lab-in-thefield experiments in a weak media environment where ballot photos are both the first and last visual impression many voters have of candidates. Inferences based on candidate photos alone predict who later wins actual elections. Further, these inferences are better at identifying trustworthy politicians (i.e. those who divert fewer public resources to personal use) than a suite of professional qualifications. Candidates with more electable faces appear to have stronger persuasion skills, which reflect advantages in both physical appearance and oral communication. Neither snap judgments based on photos nor observable characteristics distinguish politicians along concrete measures of effort. __________________________ * Stanford University Graduate School of Business, 655 Knight Way, Stanford CA 94305, +1 650 725 2167, [email protected]. I thank K. Bidwell and R. Glennerster for collaboration on related projects; and M. Alsan, J. Bendor, E. Blankespoor, D. Broockman, A. Chandrasekhar, J. Fearon, S. Jha, D. Laitin, E. Miguel, M. Morten and S. Seiler for comments. I am grateful for excellent field and research assistance from A. Barnett, N. Eubank, A. Lahai, A. Mansaray, F. Momoh, O. Nabay, K. Parry, C. Sánchez-Martínez and Innovations for Poverty Action. I gratefully acknowledge funding from the Governance Initiative at J-PAL.