This paper explores the relationship between household exposure to riots and social capital in urban India using a panel dataset collected by the authors in the state of Maharashtra. The analysis applies a random–effect model with lagged covariates to estimate the exogenous effect of riots on social capital. Households living in neighborhoods prone to riots are more likely to invest in bridging social capital by joining community organizations but reduce face–to–face contact with neighbors. These effects are driven largely by levels of neighborhood social fragmentation in riot- prone neighborhoods. In these neighborhoods, the salience of social identity is also reduced as individuals attempt to reach out across social divisions. We interpret these results as indicating that households instrumentally use bridging forms of social capital as an insurance against potential future communal violence in socially fragmented contexts where conflicting social groups live alongside each other.