Does Drug Enforcement Cause Violence? Evidence from Colombia

Drug enforcement may have unintended consequences. This is especially true in source countries where military manoeuvres are utilized to support the activities of the anti-narcotics police. We use Colombian drug war data for 2005 to test whether the anti-narcotics military operations carried out by the Colombian government had any influence on the country’s homicide rate at municipality level. As the government’s military actions are found to be endogenous to the homicide rate, we run IV regressions to determine a measure of military attacks which only represents those actions performed to support coca eradication. In order to accomplish that, geographical characteristics of the Colombian municipalities are used in the first stage regressions as instruments to proxy for the municipalities’ productivity to cultivate coca crops. The underlying idea is that the military attacks that took place in municipalities with a large productivity to produce coca were perpetrated with the aim of reducing that activity. One issue that arises from these IV regressions is that their residuals are spatially correlated. To tackle this problem, we estimated Spatial Autocorrelation (SAR) models. Another issue that arises is that the fitted values that result from the first stage regressions are not necessarily non-negative. We run Tobit models to address this problem. According to our estimations, every additional anti-narcotics military operation in 2005 increased the average homicide rate of the intervened municipalities in nearly 12 homicides per 100,000 inhabitants. In addition, these operations incentivized drug dealers’ counterattacks, which raised the average homicide rate in the affected municipalities in nearly 14 homicides per 100,000 inhabitants per additional attack executed.