Using a detailed jurisdiction-quarter level dataset, I create a proxy for illegal firearm flows: the number of firearms reported stolen in each police jurisdiction, and map their effect on crime in the U.S. Estimates show a strong, positive impact of increased stolen firearms, in the previous quarters, on firearm aggravated assaults, homicides, and robberies in the current quarter. However, no statistically significant relationship is estimated between firearm flows and non-firearm offenses, providing a crucial falsification test. Various other robustness checks, including an analysis of potential spillovers in illegal firearm flows, find no evidence of a spurious relationship driving the results.