Recent political instability in the Middle East has triggered one of the largest influxes of refugees into Europe. The different departure points along the Turkish coast generate exogenous variation in refugee arrivals across Greek islands. We construct a new dataset on the number and nature of crime incidents and arrested offenders at island level using official police records and newspaper reports. Instrumental variables and difference-in-differences are employed to study the causal relationship between immigration and crime. We find that a 1-percentage-point increase in the share of refugees on destination islands increases crime incidents by 1.7-2.5 percentage points compared with neighboring unexposed islands. This is driven by crime incidents committed by refugees; there is no change in crimes committed by natives on those islands. We find a significant rise in property crime, knife attacks, and rape, but no increase in drug crimes. Results based on reported crimes exhibit a similar pattern. Our findings highlight the need for government provision in terms of infrastructure, social benefits, quicker evaluation for asylum, and social security.