This paper estimates the economic effects of communal riots in India using a new district-year panel for 1994-2013 that combines extended riot data with nightlights, manufacturing outcomes, infrastructure, and district-level credit. To address endogeneity, we use quasi-random variation generated when major Hindu festivals fall on a Friday (Fridays draw large Muslim congregations for weekly prayers) and construct a spatially weighted instrument for local riot risk. Instrumental variable estimates show that higher riot exposure leads to sizable and persistent economic losses. A one-unit increase in distance-weighted exposure reduces nightlights by about 0.105 log points, implying a GDP decline of roughly 0.45\%. We trace several channels, including contractions in firm output and employment, fewer bank branches, less agricultural credit, and a thinner market infrastructure. Using the timing of household surveys relative to riots, we also find that trust in state institutions declines. Together, the results quantify the economic costs of communal riots and identify the mechanisms through which social conflict depresses local economic activity.