Résumé
Street riots often follow police killings of citizens. Yet, the majority of police killings do not lead to riots. Why do some police homicides lead to riots while others do not? This paper analyzes the intensity of riots in France using monthly data on arsons targeting private property at the department (sub-regional) level from January 1996 to August 2022 (N = 32,080), in relation to an original dataset on the occurrence and characteristics of police homicides. The results suggest that police homicides have a small overall effect on riots: fewer than 0.3% of arsons in a given department can be attributed to police killings in that department. While most police homicides do not have the potential to trigger riots, certain characteristics significantly increase the likelihood that a riot will occur: the victim has a non-European migratory background, is a French citizen, and the perpetrators belong to specific police forces. The impact also increases when there are multiple victims. We provide suggestive evidence that identification with the victim plays a key role among potential rioters, as does the social integration of the victim.