Gun violence is an important problem across the United States. However, the impact of government policies on gunfire has been difficult to test due to limited and low-quality data. This paper uses new, more accurate data on gunfire (generated by ShotSpotter audio sensors) to measure the effects of juvenile curfews in Washington, DC. Using variation in the hours of the DC curfew, we find that this policy increases gunfire incidents by 150% during marginal hours. In contrast, voluntarily reported crime measures (such as 911 calls) suggest that the curfew decreases gun violence, likely because of confounding effects on reporting rates.