Is Violence a Disease?
A few years ago an economist colleague of mine debated a professor from our university’s school of public health on gun violence and gun control. My colleague walked through the empirical evidence on the effects of gun control laws on crime, accidental injury, and other social ills, citing well-known studies by economists and legal scholars such as John Lott, John Donahue, Gary Kleck, Ian Ayres, etc. As Circle Bastiat readers probably know, the empirical social science evidence, while not conclusive, suggests that stronger gun control leads to increased rates of gun violence. The public health professor — an MD who also teaches in the medical school — ignored these issues entirely, instead telling emotional stories about ER patients he’d treated for gunshot wounds and how everything must be done to stop this “epidemic” of gun violence.
The two speakers were, of course, talking totally past each other. One approached gun control from a simple, “rational actor,” cost-benefit approach. Gun control laws may make it slightly more difficult for bad people to get guns and for good people to do bad things, accidentally or in the heat of the moment, with a deadly object. But gun control also makes it more difficult for good people to acquire and use guns defensively, to deter crime. The net effect is ambiguous, but the existing empirical evidence suggests that the latter effect probably outweighs the former. (Of course, I’m ignoring for the moment the relevant normative, political, and Constitutional issues.) The other speaker did not see gun violence, and violence more generally, as the result of purposeful human action. Rather, he saw violence as a disease, like the flu or cancer, which — while affected in some ways by people’s preferences, beliefs, and choices — is basically an epidemiological problem.
Indeed, the epidemiological approach to gun control is gaining adherents. “Is It Time to Treat Violence Like a Contagious Disease?” asks Wired in its latest issue. MSNBC breathlessly reports a claim that the National Rifle Association “suppressed” an important study, “rigorously conducted by ten credentialed experts” and “appearing in the prestigious New England Journal of Medicine,” showing that owning a gun makes a person less safe, rather than more safe. But these studies, conducted by public health professionals rather than empirical social scientists, are typically marred by basic methodological flaws such as sampling on the dependent variable, failure to account for endogeneity and unobserved heterogeneity, and other problems that would flunk a first-year econometrics grad student. Jacob Sullum offers a nice summary of some of this literature. Sullum focuses on researcher bias (e.g., selective citation of the existing literature), but also discusses some of the most infamous gaffes, such as a 2009 paper in the American Journal of Public Health showing that gun owners were more likely to be shot than “similar” people not owning guns, but matching only on age, gender, and race, ignoring the fact that people who are more likely to be the victim of gun violence are more likely to own firearms — i.e., that the causation is reversed. (This is a typical problem of what econometricians call “identification,” for which there are several generally accepted remedies, all apparently unknown to the public health research community.)
My sense is that the basic problem is the disease model of violence. Medical professionals are used to running controlled experiments, using clinical trials, and are unaware of how to perform empirical social science research outside a laboratory setting, so they apply the epidemiological model to issues like gun control, not recognizing that the decision to engage in violent acts, like crime more generally, is a form of purposeful human action. It’s time for journalists and politicians to recognize that the disease model is fundamentally flawed, and to turn to the best available research by economists, legal scholars, and other social scientists trained in how to do this kind of work.
(NB: I’m not suggesting that the mainstream law-and-economics research on crime, based on neoclassical economics, is free from problems, like ignoring time preference. See Samuel Cameron’s thoughtful essay.)