This paper explores the macroeconomic impact of social unrest, using a novel index based on news reports. The findings are threefold. First, unrest has an adverse effect on economic activity, with GDP remaining on average 0.2 percentage points below the pre-shock baseline six quarters after a one-standard deviation increase in the unrest index. This is driven by sharp contractions in manufacturing and services (sectoral dimension), and consumption (demand dimension). Second, unrest lowers confidence and raises uncertainty; however, its adverse effect on GDP can be mitigated by strong institutions and by a country’s policy space. Third, an unrest “event”, which is captured by a large change in the unrest index, is associated with a 1 percentage point reduction in GDP six quarters after the event. Impacts differ by type of event: episodes motivated by socio-economic reasons result in sharper GDP contractions compared to those associated with politics/elections, and events triggered by a combination of both factors lead to sharpest contractions. Results are not driven by countries with adverse growth trajectories prior to unrest events or by fiscal consolidations, and are robust to instrumenting via regional unrest.