What caused the explosion of violence in northern Mexico? Homicides ticked up in 2007 and jumped in ‘08 and ‘09. One common story is that the Calderón administration provoked the violence by its decision to crack down on organized crime right after its accession to office in December ‘06. The causal mechanism is unclear, but it something you hear cited a lot.
Now Melissa Dell, a brilliant student at MIT, has crunched the numbers to test the hypothesis that the national crackdown may have in fact incited the violence. She took daily county-level data on drug-related homicides (generally called “executions” in the Mexican data). She then tested a simple, but not intuitive, hypothesis. If the government crackdown prompted more violence, then we should expect drug-related violence to jump after a county elects a mayor from the Partido Acción Nacional (PAN). The reason is that Calderón’s party is the PAN. Not only did the federal PAN spearhead the assault on organized crime, but PAN candidates generally campaigned on a “tough on drugs” platform, and one would expect PAN mayors to be better able to coordinate law enforcement with the PAN federal government.
So what does she find? Well, here’s a picture:
Violence jumps about 15% after the election of a PAN mayor. The dotted lines are confidence intervals, so you can see that effect is statistically significant. The violence does not occur after the election, but rather the inauguration of a PAN mayor. In other words, the expectation that a Panista will take office does nothing, but the actions the new administrations take after gaining office do something. (Editorial note: I decided to capitalize “panista” when writing in English. I also translated “municipio” as “county” rather than “municipality,” since they are akin to American counties in size. Mexico does not have incorporated municipalities in the American sense.)
Melissa also used data on shootouts between cartel members; they jump after a PAN victory. In addition, reported narcomantas jumped 63% (with a standard error of 32%) in places where PAN just barely won compared to where they just barely lost. Narcomantas are ... well, a photo of one in Torreón is to the left.
The jump in drug-related killings does not happen if a PRI or PRD mayor wins. (The PRI is the Partido Revolucionario Institutional, the old ruling party from Mexico’s bad old days of dictatorship. The PRD is the Partido de la Revolución Democrática, the social democratic opposition party.) It might be that PAN mayors are more corrupt that their Priísta or Perredista competitors, but that seems unlikely. In fact, the 2008 county-level data she analyzed showed no difference in mayoral corruption (as reported by federal law enforcement agencies) across counties governed by different parties. Moreover, in theory one would expect corruption and violence to be substitutes, not complements ... although that is really a hypothesis to be tested rather than an assumption upon which to base a model. She has some evidence that extortion threats against county governments jump after a PAN victory, but it is relatively weak.
Still, her results follow what one would expect. The effect is bigger in border states than elsewhere. It is also bigger in drug-producing states. Better still, it’s bigger in counties along trafficking routes. The cool thing? She derived the routes herself. In her words, slightly paraphrased:
- Central places are those that are relatively costly to go around when transporting drugs from producing counties to the U.S. border.
- For each producing county calculate the least cost path to the nearest U.S. border crossing. This path can pass along any edge in the Mexican road network. Calculate this over all counties for a baseline measure. (Least-cost routes are calculated solely on geographic distance to the border.)
- Repeat this calculation, but do not allow the paths to pass through the county. Calculate this for each county.
- The centrality of a county = (2) – (3). The higher the value, the more central the county, and the more executions should jump after a PAN victory.
And lo, she found it holds up as expected! (She controlled for other factors that one might expect to be correlated with centrality.) Executions jump everywhere if the PAN wins, but they jump more in central counties. Now, executions are a politically-derived variable. (How do you know if a homicide was drug related?) The chance of a homicide occurring in a county jumps about 7% per month after a PAN victory (and overall homicides leap 49%) but there is a lot of variation and the result is not quite statistically significant. So it is possible (although upon first reading of the results unlikely) that PAN mayors are simply more likely than opposition mayors to classify homicides in their county as drug-related.
In short, there is some evidence that the PAN decision to crack down on organized crime triggered the upsurge in violence. Why it would trigger the upsurge is still not clear, and I should point out that her results are preliminary and could change. But it is awesome research, and it is, I think, a great example of how social science techniques can be applied to some of the realest of real world questions.
We would love to hear critiques and suggestions. We would also love to hear what you think the policy implications, if any, might be.