Is Ethnic Violence Self-Perpetuating? Quasi-Experimental Evidence From Hindu-Muslim Riots in India

Abstract

Ethnic riots tend to occur in the same places over time. We study whether this serial correlation exists because ethnic riots tend to be self-perpetuating or because both past and future riots are caused by the same underlying factors that persist through time. To answer this question, we leverage the fact that the timing of major Hindu festivals in India is exogenously determined by the lunar calendar and that when a major Hindu festival happens to fall on a Friday—the principal day Muslims attend mosque—the likelihood of a Hindu-Muslim riot increases significantly. Using this instrument, we find that the well-documented serial correlation in Hindu-Muslim riots disappears entirely (T = 1950–2006). This suggests that the observed recurrence of riots is not driven by the riots themselves, but by underlying conditions that remain unaddressed. Once these confounding factors are accounted for, we find no “additional” effect of past riots on future riots.