As promised earlier, I asked John Donohue for his thoughts on the question of what mechanism might explain his finding that looser restrictions on carrying concealed weapons (“shall-issue” laws) tended to increase crime. (Kieran Healy handily captures the Month’s-Worst-Pun award by suggesting as the answer to this question “the trigger mechanism.”)
It turns out that Donohue will be publishing something on the subject, and so is reluctant to post anything an editor might regard as scooping that published work. He agreed, however, to let me combine some of his thoughts with some of mine. He is not responsible for the what is written below, but he has read it without finding anything to which he objects.
Ayres and Donohue seem to find evidence that liberalizing concealed-carry laws leads to increases in crime, in particular property crime. That creates a puzzle: if this effect is real, rather than a statistical artifact, how does it arise? The same question applies to claims that shall-issue decreases crime: what is the mechanism? In each case, we need a mechanism that is plausible in concept and in extent, and a pattern of crime increases or decreases consistent with that mechanism.
It is easy to tell either a crime-increase or a crime-decrease story for violent crime. The crime-increase story is that more people walk around armed, and some of them give in (perhaps under the influence of alcohol) to momentary impulses to commit acts of violence that either would not occur at all, or would have less devastating consequences, in the unarmed condition. Alternatively, the knowledge that one’s opponent might be armed could be destabilizing, creating an incentive for a decisive pre-emptive strike in a confrontation. But Ayres and Donohue do not actually find statistical decreases in crimes against the person.
On the crime-decrease side, the story is at least equally plausible: knowledge that any random pedestrian might be armed will discourage muggers, street rapists, and auto hijackers. If the John Lott “more guns, less crime” results depended on decreases in those categories of crime, I for one would find them much more convincing. But they do not.
There are two big problems here, for both sides: First, it is not clear that CCW actually changes gun-carrying behavior, or gun-ownership rates, very much. It is hard to see how a policy that has a very small impact on an intermediate variable is going to do much about a final variable. Second, crimes against the person occurring outdoors are only a subset of crimes, and it is not obvious why allowing people to walk around with concealed weapons ought to have any impact on crimes inside the home or inside a business. Yet it is not the case that the outdoor interpersonal crimes show consistently different patterns from other crimes in terms of how rates after a “shall-issue” law passes.
Phil Cook actually came up with an idea about how increased gun ownership might influence burglary rates: a gun is among the things in a home most worth stealing, in terms of resale value per unit weight, so increasing gun ownership in a community might raise the expected rewards to burglary and thus the willingness of some people to engage in burglary. If shall-issue led more people to own guns, the result might then be an increase in burglary. By the same token, a decrease in “indoor” crime might be explained by increased gun ownership due to eased concealed-carry rules, leading to increased fear among criminals of facing an armed homeowner or shopowner. But once again, one would like to see some evidence that (1) liberalized CCW generates significant changes in gun ownership among potential victims and/or (2) that potential perpetrators think it does. To my knowledge, no such evidence is present.
So if Ayres and Donohue had found an increase in property crime specifically for residential burglary, I would take their finding as some evidence to support the Cook hypothesis. But in fact burglary does not stand out in their findings: all the property-crime categories show comparable decreases. I find it hard to tell a story about how shoplifting or auto theft could plausibly rise as a result of liberalized CCW laws. That leads me to suspect that, despite its statistical significance, the positive (i.e., bad) impact on crime found by Ayres and Donohue is actually a null result plus an error term due to omitted variables and noise (e.g., the effects of the crack epidemic and its recession, unequally distributed between liberalizing and non-liberalizing states).
It is possible that the null result reflects offsetting pro-crime and anti-crime effects of liberalized CCW laws. If that is true, we have no assurance that the impact of those laws on crime will remain at zero over time; the balance of pro-crime and anti-crime effects might shift. Nor can we be sure that if Missouri were to pass a shall-issue law it would not have either a measurable increase or a measurable decrease in crime as a result, since the balance of those two effects might be different in Missouri than it was in Texas.
But if, after years of effort, no one has found either a positive or a negative effect on crime rates that is big enough to be convincing, consistent across states and years, robust to the specification problems and coding errors that plague this sort of research, and with a pattern of crimes that fits some plausible mechanism, there is probably not much of an effect to be found.