The U.S.-Mexico border fence and homicides in Mexico

On Tuesday, I posted about the effect of coronavirus on the price of illegal goods, particularly in Mexico:

When goods (or people) become more difficult (more costly) to transport, the equilibrium price of those goods and services will rise. That is the case even when those goods and services are illegal.

Stricter policing of border controls make goods more difficult to smuggle. Even something as simple as building a fence will increase the costs of smuggling, even if only fractionally. Building a fence in some parts of a border, but not others, changes the incentives for smugglers, in terms of where the best places to cross the border are located. And if smugglers want to change their location, they may come into conflict with their rivals. That is the context for this recent working paper by Benjamin Laughlin (University of Pennsylvania).

Considering the construction of 649 miles of border fencing over the period from 2007 to 2011 resulting from the Secure Fence Act, Laughlin looks at how the number of homicides changed in Mexican localities within 10km of the border, before and after the construction of the fence. He finds that:

With both dependent variables [logged homicides, and homicides per capita], the localities with access to alternate smuggling routes suffered a significant increase in lethal violence that persisted for over two years. At the same time, localities near the new border fence saw a decline is violence.

Laughlin doesn't make clear in the working paper the size of the effect. Eyeballing his figures, it looks like around 2 additional homicides per 10,000 people per half year. My back-of-the-envelope calculation suggests that the additional border fence increased homicides by nearly 2000, [*] which is an appreciable amount. Laughlin argues that this arises because:

 ...the border fence restricts drug smuggling routes, which increases the value of alternate routes that circumvent the border fence. As a result of this change in the value of territory, competing drug cartels fight for control over territory that provides access to alternate smuggling routes, which leads to a spike in fatal violence. Over time, a new equilibrium is reached as cartels settle on an arrangement over territory sharing, and fatal homicides will subside.

That last sentence is important, because it explains why the increase in violence in areas without a border fence lasted for just two years before returning to the baseline level. His results are robust to the use of different control groups (of localities), and robust to controlling for the intensity of law enforcement in Mexican localities, and the capture of cartel leaders.

When Donald Trump declared in 2016 that the U.S. would build a wall and make the Mexicans pay for it, many people scoffed. However, based on these results it appears that even if the Mexicans don't have to spend a dime on wall construction, they will probably end up paying a high price.

*****

[*] In Laughlin's dataset, the average locality with no new border fence has a population of 2142, and there are 1144 such localities, which means a population-at-risk of 2,450,488. At my estimated two additional homicides per 10,000 population, that is 490 additional homicides. The model is based on half-years, so for two years that is 1960 additional homicides.