Social interactions in voting behavior: evidence from India

We study how voters’ turnout decisions affect the turnout decisions of other voters by exploiting the unique staggered nature of India's elections. Voting takes places in several phases, with constituencies quasi-randomly assigned to phases. At the time that later phases vote, turnout in earlier phases is public knowledge but results are not. Using an instrumental variables strategy, we find that a 1 percentage point (pp)increase in turnout in the previous phase depresses turnout in the current phase by around 0.38 pp. Falsification tests examining the effect on turnout, of either constituencies in the same phase or in future phases in the same election, produce no effect. Our results are broadly consistent with pivotal voter and ethical voter models, and suggest that partial equilibrium evaluations of turnout interventions may overstate general equilibrium effects.