The Economics of Assassinations

What matters at the margins

One of the several abandoned projects on my hard drive, this one more than a decade old, took me into a brief literature review on the social science of assassinations. Among the good papers I read for that project was a piece by Benjamin F. Jones and Benjamin A. Olken entitled “Hit or Miss? The Effect of Assassinations on Institutions and War.” Published in the American Economic Journal: Macroeconomics, the piece was extremely clever in seeking to understand the effect of assassinations.

Political violence is reprehensible, of course, but social scientists deal with reprehensible topics all the time (I have read so many descriptions of the details of mass killings and the aftermath of nuclear bombs that a part of my soul has shriveled even as another part of my soul has been enlarged). There’s a reason this newsletter has the title that it has. Politics ain’t the systematic organization of love.

And we might wonder whether and how assassinations have an effect. If you’re a structuralist, and believe that big forces in history and society drive events, then any given individual may be unimportant. If, on the other hand, you view some leaders as particularly important, then the removal of given leaders might have profound influences.

The question is how you might know that this is the case. One strategy social scientists employ to answer that question is to look for times when assignment to treatment is randomized. The idea is that if a given treatment is randomly assigned, then there should be no differences on average between treated and untreated units but for the effects of that treatment.

This is pretty straightforwardly what scientists in the physical sciences do all the time, but they’re aided by the fact that they deal with more or less interchangeable units (one water molecule is the same as others). Similarly, both physical and social scientists often seek to induce the random assignment to treatment, so that they can be sure of the randomness and control other parts of the environment. But when you’re unable to do that, or when doing so would be ahem wildly irresponsible and unethical, you look for similar randomly assigned treatments in nature.

And it turns out that assassinations can be viewed as randomly assigned! The trick is that there’s a lot of assassination attempts but not many assassination successes, and many such attempts can be understood as succeeding as-if random—the bomb mechanism fails, the shot misses, etc. Consequently, as Jones and Olken write, “although attempts on leaders’ lives may be driven by historical circumstances, conditional on trying to kill a leader, the success or failure of the attempt can be treated as plausibly exogenous.” And so the two Benjamins collected data on all publicly reported assassination attempts from 1875 to 2004 (of which there were 298, 59 of which resulted in the leader’s death).

As long as we understand the limits of this approach (for instance, assassinations are likelier to be attempted during times of crisis or war), it’s possible to draw some conclusions. Notably, assassinations of autocrats produce “substantial changes in the country’s institutions, while assassinations of democrats do not”, with transitions to democracy 13 points more likely after a successful assassination than a failed one. It’s unclear whether assassinations have a consistent effect on war—they can intensify small-scale conflicts but lead to the end of larger-scale conflicts already in progress.

On the other hand, “there is some evidence that failed attempts have modest effects in the opposite direction of successful assassinations.” In particular, they write, “failed attempts slightly reduce the likelihood of democratic change”. Because there are more failed attempts than successful attempts, “assassination attempts produce instability in political institutions”.