ÿÈÕ³Ô¹Ï

We use data from the National Incidence-Based Reporting System (NIBRS) to examine how the probability of getting caught when committing a crime, proxied by ambient daylight, impacts criminal activity. We exploit the existence of daylight saving time (DST) to provide within-hour exogenous shock to daylight, using both the discontinuous nature of DST as well as the 2007 extension of DST as sources of variation. Further, we consider both crimes where darkness is likely to play a role in avoiding capture and crimes where darkness would make little difference. Our preferred specification, a regression discontinuity design, shows robbery rates decrease by an average of 51% during the hour of sunset following the shift to DST in the spring. We also find large drops in cases of reported murder (48%) and rape (56%). Effects are largest during the hour of sunset prior to DST (i.e., the hour which was in darkness before but, post-DST, is now light), suggesting changes are due to ambient light rather than other factors such as increased police presence, and we find no changes in crimes where ambient light is unlikely to be a factor. As an additional robustness check, we exploit the variation in the impact of DST by hour and crime to repeat our analysis in a triple-difference framework and show results are largely consistent. Using the social cost of crime, we estimate the 2007 spring extension of DST resulted in $558 million in avoided social costs of crime per year, suggesting investment in lighting such as street lights could have high returns. Finally, we consider if our findings are the result of increased criminal incapacitation or deterrence of criminal behavior, and provide suggestive evidence the majority of the effect is due to deterrence.