This paper employs provincial data to study the relationship between several crime typologies, namely murder, theft, robbery and fraud, and economic output in Italy. We employ a spatial econometric approach where the spatial proximity is defined by a measure of physical distance between locations, in order to take into account possible spill-over effects. The model used here combines a spatial autoregressive model with autoregressive disturbances. In modelling the outcome for each location depends on a weighted average of the outcomes of other locations. Outcomes are determined simultaneously. The results of the spatial two stage least square estimation suggest that the homicide rate has a negative impact on Italian gross domestic product while theft, robbery and fraud do not affect economic output and that there are beneficial spill-overs from neighbouring provinces.