Which premise is true—that guns increase crime or that the fear of crime causes people to obtain guns? Most of the country’s major newspapers take this problem to have been solved by an article published by Arthur Kellermann and several associates in the October 7, 1993, New England Journal of Medicine. Kellermann is an emergency-room physician who has published a number of influential papers that he believes discredit the thesis that private ownership of firearms is a useful means of self-protection. (An indication of his wide influence is that within two months the study received almost 100 mentions in publications and broadcast transcripts indexed in the Nexis data base.) For this study Kellermann and his associates identified fifteen behavioral and fifteen environmental variables that applied to a 388-member set of homicide victims, found a “matching” control group of 388 non-homicide victims, and then ascertained how the two groups differed in gun ownership. In interviews Kellermann made clear his belief that owning a handgun markedly increases a person’s risk of being murdered.
But the study does not prove that point at all. Indeed, as Kellermann explicitly conceded in the text of the article, the causal arrow may very well point in the other direction: the threat of being killed may make people more likely to arm themselves. Many people at risk of being killed, especially people involved in the drug trade or other illegal ventures, might well rationally buy a gun as a precaution, and be willing to pay a price driven up by gun-control laws. Crime, after all, is a dangerous business. Peter Reuter and Mark Kleiman, drug-policy researchers, calculated in 1987 that the average crack dealer’s risk of being killed was far greater than his risk of being sent to prison. (Their data cannot, however, support the implication that ownership of a firearm causes or exacerbates the risk of being killed.)
Defending the validity of his work, Kellermann has emphasized that the link between lung cancer and smoking was initially established by studies methodologically no different from his. Gary Kleck, a criminology professor at Florida State University, has pointed out the flaw in this comparison. No one ever thought that lung cancer causes smoking, sp when the association between the two was established the direction of the causal arrow was not in doubt. Kleck wrote that it is as though Kellermann, trying to discover how diabetics differ from other people, found that they are much more likely to possess insulin than nondiabetics, and concluded that insulin is a risk factor for diabetes.
The New York Times, the Los Angeles Times, The Washington Post, The Boston Globe, and the Chicago Tribune all gave prominent coverage to Kellermann’s study as soon as it appeared, but none saw fit to discuss the study’s limitations. A few, in order to introduce a hint of balance, mentioned that the NRA, or some member of its staff, disagreed with the study. But readers had no way of knowing that Kellermann himself had registered a disclaimer in his text. “It is possible,” he conceded, “that reverse causation accounted for some of the association we observed between gun ownership and homicide.” Indeed, the point is stronger than that” “reverse causation” may account for most of the association between gun ownership and homicide. Kellermann’s data simply do not allow one to draw any conclusion.
If firearms increased violence and crime, then rates of spousal homicide would have skyrocketed, because the stock of privately owned handguns has increased rapidly since the mid-1960s. But according to an authoritative study on spousal homicide in the American Journal of Public Health, by James Mercy and Linda Saltzman, rates of spousal homicide in the years 1976 to 1985 fell. If firearms increased violence and crime, the crime rate should have increased throughout the 1980s, while the national stock of privately owned handguns increased by more than a million units every year of the decade. It did not. Nor should the rates of violence and crime in Switzerland, New Zealand, and Israel be as low as they are, since the number of firearms per civilian household is comparable to that in the United States. Conversely, gun-controlled Mexico and South America should be islands of peace instead of having murder rates more than twice as high as those here. The determinants of crime and law-abidingness are, of course, complex matters, which are not fully understood and certainly not explicable in terms of a country’s laws. But gun-control enthusiasts, who have made capital out of the low murder rate in England, which is largely disarmed, simply ignore the counterexamples that don’t fit their theory.
If firearms increased violence and crime, Florida’s murder rate should not have been falling since the introduction, several years ago, of a law that makes it easier for ordinary citizens to get permits to carry concealed handguns. Yet the murder rate has remained the same or fallen every year since the law was enacted, and it is now lower than the national murder rate (which haws been rising). As of last November 183,561 permits had been issued, and only seventeen of the permits had been revoked because the holder was involved in a firearms offense. It would be precipitate to claim that the new law has “caused” the murder rate to subside. Yet here is a situation that doesn’t fit the hypothesis that weapons increase violence.
If firearms increased violence and crime, programs of induced scarcity would suppress violence and crime. But—another anomaly—they don’t. Why not? A theorem, which we could call the futility theorem, explains why gun-control laws must either be ineffectual or in the long term actually provoke more violence and crime. Any theorem depends on both observable fact and assumption. An assumption that can be made with confidence is that the higher the number of victims a criminal assumes to be armed, the higher will be the risk—the price—of assaulting them. By definition, gun-control laws should make weapons scarcer and thus more expensive. By our prior reasoning about demands among various types of consumers, after the laws are enacted criminals should be better armed, compared with non-criminals, than they were before. Of course, plenty of non-criminals will remain armed. But even if many noncriminals will pay as high a price as criminals will to obtain firearms, a larger number will not.
Criminals will thus still take the same gamble they already take in assaulting a victim who might or might not be armed. But they may appreciate that the laws have given them a freer field, and that crime still pays—pays even better, in fact, than before. What will happen to the rate of violence? Only a relatively few gun-mediated transactions—currently, five percent of armed robberies committed with firearms—result in someone actually being shot (the statistics are not broken down into encounters between armed assailants and unarmed victims, and encounters in which both parties are armed). It seems reasonable to fear that if the number of transactions were to increase because criminals thought they faced fewer deterrents, there would be a corresponding increase in shootings. Conversely, if gun-mediated transactions declined—if criminals initiated fewer of them because they feared encountering an armed victim or an armed good Samaritan—the number of shootings would go down. The magnitude of these effects is, admittedly, uncertain. Yet it is hard to doubt the general tendency of a change in the law that imposes legal burdens on buying guns. The futility theorem suggests that gun-control laws, if effective at all, would unfavorably affect the rate of violent crime.
The Futility theorem provides a lens through which to see much of the debate. It is undeniable that gun-control laws work—to an extent. Consider, for example, California’s background-check law, which in the past two years has prevented about 12,000 people with a criminal record or a history of mental illness or drug abuse from buying handguns. In the same period Illinois’s background-check law prevented the delivery of firearms to more than 2,000 people. Surely some of these people simply turned to an illegal market, but just as surely not all of them did. The laws of large numbers allow us to say that among the foiled thousands, some potential killers were prevented from getting a gun. We do not know whether the number is large or small, but it is implausible to think it is zero. And, as gun-control proponents are inclined to say, “if only one life is saved…”
The hypothesis that firearms increase violence does predict that if we can slow down the diffusion of guns, there will be less violence; one life, or more, will be saved. But the futility theorem asks that we look not simply at the gross number of bad actors prevented from getting guns but at the effect the law has on all the people who want to buy a gun. Suppose we succeed in piling tax burdens on the acquisition of firearms. We can safely assume that a number of people who might use guns to kill will be sufficiently discouraged not to buy them. But we cannot assume this about people who feel that they must have guns in order to survive financially and physically. A few lives might indeed be saved. But the overall rate of violent crime might not go down at all. And if guns are owned predominantly by people who have good reason to think they will use them, the rate might even go up.
Are there empirical studies that can serve to help us choose between the futility theorem and the hypothesis that guns increase violence? Unfortunately, no: the best studies of the effects of gun-control laws are quite inconclusive. Our statistical tools are too weak to allow us to identify an effect clearly enough to persuade an open-minded skeptic. But it is precisely when we are dealing with undetectable statistical effects that we have to be certain we are using the best models available of human behavior.