To test our theory we use crime data from several different sources. First, we use the Uniform Crime Report (UCR) data, which is a panel data set with violent and property crime rates for each county, split into seven crime categories. Out of these seven, our analysis focuses on homicides, aggravated assaults, and robberies as these crimes are often connected to activities of DTOs and their affiliated gangs (see NGIC, 2011). Given that we focus on identifying supply side effects we abstract from analyzing property crime which might be more likely to be influences by the demand side. Second, we use the Supplementary Homicide Reports (SHR) data, which gives information on the circumstances surrounding homicides committed in the US. This data allows us to see whether homicides are related to drug violence. Both data sets cover the period 1994-2012.
Our main analysis applies a difference-in-difference-in-difference (DDD) methodology where we divide counties in four groups depending on i.) whether the county is located in a Mexican border state or an inland state, and ii.) whether the state introduced M[edical] M[arijuana] L[aws] or not. The DDD methodology allows us to fully control for all shocks to the crime rate that affect all states on the border. Examples of such shocks are increases in border patrols and Mexican law enforcement. In addition, we explicitly control for observable confounding factors that may be correlated to both the introduction of MML and the crime rate, and we include state-linear time trends to control for possible unobservable confounding factors. We augment the analysis, by adding a specification where we interact the treatment dummy for the introduction of MML with the distance to the border. This allows us to verify that within Mexican-border states the effect of MML on crime is strongest for counties located close to the border.
…Turning to our main result, we show that MML lead to a strong reduction in the violent crime rate for counties in Mexican-border states. In these counties the violent crime rate decreases by between 10-20 percent depending on the specification. The decrease is strongest in robberies which decrease by 26 percent, followed by homicides at 11 percent and aggravated assaults with 10 percent. When we consider the distance to the border, we find that the strongest decrease in the violent crime rate occurs in counties in close proximity to the border while the effect weakens with the distance of a given county from the border. We find no robust significant effect of MML on crime in counties that are located more than 350 kilometer from the border.
Our point estimates suggest that crime decreases in all 3 border states that have introduced MML. However, the effect is most robust in California. This may be due to the fact that California has a higher take up rate of medical marijuana, as measured by high density of dispensaries within the state.
Our analysis of the SHR data reveals that MML decrease drug-law, juvenile-gang, and robbery related homicides by 46, 34, and 30 percent, respectively within states on the Mexican border. This result is strongly suggestive of the fact that MML in the Mexican-border region are effective in reducing drug-trade and gang-related crimes (pgs. 3-4).
Drug legalization is doing what economists said it would do. Fancy that.