Recently I cited the work of criminologist Gary Kleck to rebut claims about defensive gun use and the relative risks of gun ownership; see “Michael Shermer’s Bogus Claims about Defensive Gun Use: Why the ’22 Times’ Statistic Is Nonsense.”
As a follow-up to that article, I contacted Kleck to see if he’d be interested in being interviewed; he graciously agreed. I interviewed him by phone on November 2; the transcript (lightly edited and approved by Kleck) is below. I’ve added some headers for ease of navigation.
One need not agree with all of Kleck’s conclusions to find his remarks enormously insightful. I have four main concerns that might make interesting research topics for later. First, I wonder if factors other than the planning requirements and lethality of suicide methods impact the likeliness of a person attempting suicide. In particular, I wonder if the painfulness (real or perceived) of a method makes a difference. Second, I know quite a few people in Colorado who did not carry a handgun prior to liberalized licensing who do carry now. I find it hard to believe that my experiences are atypical; thus, I’m skeptical of the survey that Kleck cites indicating that carry laws didn’t affect carrying practices. Third, I wonder if Kleck is too skeptical of the possibility of more proactive policing in cases where people make direct threats of violence or articulate plans to commit violence. Fourth, I wonder how much of a difference, if any, expanded records for background checks would make in keeping guns out of the hands of dangerous people. But these concerns are minor relative to the enormous value of Kleck’s expert and deeply researched discussions.
Kleck’s remarks will not settle every debate in these areas, but they are an excellent place to start in thinking through the complexities of gun ownership and crime. —Ari Armstrong
November 21 Update: See also John Lott’s reply and answers to related questions.
Estimating Defensive Gun Use
Ari Armstrong: Over the years various researchers, including you, have attempted to estimate the annual number of defensive gun uses in the United States. Of course, defensive gun uses have probably decreased over the years as violent crime has fallen. In your view, what is the best estimate at this point?
Gary Kleck: I think your premise is correct, that defensive gun uses would go down proportionally as the need for defensive gun use goes down, and thus as the crime rate goes down.
The violent crime rate is about half now of what it was circa 1993, when we did that survey on defensive gun uses, so, best guess, the number of defensive gun uses would be about half. So, if it was 2.5 million then, it would be 1.2 million or so now.
But, I have to tell you, there hasn’t actually been a national survey on the subject that I know of since 2000. It’s as if, once people found out what kind of answers they would be getting if they did a national, probability-based survey, they ceased doing them. They didn’t ask the question anymore.
The defensive gun-use question has been asked in quite a few non-academic surveys—Gallup or whoever would ask the question as a single, isolated question in a survey largely about other topics. But nobody’s done a survey since a Washington Post survey way back in 2000. So we’re fifteen years out of date. So that’s why I have to guestimate what it would be, but I think that’s a pretty reasonable guestimate.
The Cost of Good Survey Results
Armstrong: You did such a study at one point. Would anything tempt you to conduct another such study at some point in the future? What sort of work-hours are we talking about here?
Kleck: Yes, but it’s a matter of somebody has to pay for it. I imagine these days telephone surveys similar to my previous one would probably cost you on the order of $50 for each completed reply.
To a great degree, it’s a matter of calling up, getting no answer, and so on, until you finally have someone to talk to. The interviews didn’t really last all that long when we did it. Maybe the longest ones were on the order of fifteen minutes. The shortest ones were one minute, where all we did is ask if people had a defensive gun use, and, if they said no, for most of those people, we just said “thank you very much” and hung up. Completions aren’t really where the time and labor enters into it.
It’s probably around $50 per completion; all that calling and you finally get a completed interview. We had 5,000 cases, so multiply 5,000 times $50, that’s kind of a ballpark figure of what you’re talking about. It’s an expensive proposition.
Gun Ownership and Violence
Armstrong: Various studies have claimed to show that buying or owning a firearm makes one more prone to being involved with violence. Usually these are in medical journals. What do you think of these studies?
Kleck: The authors didn’t really seem very interested in falsifying their hypothesis. Good researchers make a serious effort to show that their initial hypothesis is wrong, and then, when they fail repeatedly, it’s a strong indication that we ought to tentatively accept the hypothesis or at least not reject it.
In this case, that would mean you would ask a lot of questions about confounding factors, other things that would affect whether or not people got involved in violence besides having a gun in their household, that might also be correlated with gun ownership. You might confuse the effect of these confounding variables with the effect of having a gun in your home.
Since these studies really don’t make any serious effort to control for those factors, you really don’t know much about them.
As far as we can tell, the only reason why people who end up getting murdered were more likely to have a gun in the household is because they live in more dangerous circumstances, and they anticipated the need to have a gun for self-protection. So, if you live in a dangerous neighborhood, or you know dangerous people, or you go into dangerous places, then you are more likely on the one hand to get murdered, but of course you’re also more likely to acquire a gun somewhere along the line for self-protection. So it’s a classic case of a spurious association.
None of these studies has made any serious effort to control for those sorts of factors, things like belonging to a street gang. You’re way more likely to own a gun, and you’re way more likely to get murdered. If you don’t control for whether a person belongs to a street gang, you’re not really going to get a serious estimate of the effect of having a gun.
Probably the best of a bad lot was the famous Arthur Kellermann study from 1993 in the New England Journal of Medicine. All the rest are even worse, but at least he controlled for a few possible confounding factors. But he withheld one crucial piece of information from his readers. He knew that virtually none of the people who had been murdered while having a gun in their home had actually been killed with the gun that belonged to someone in the home. They were almost always killed by someone from outside the home, presumably using their own gun, brought in from outside the home. So whether the victims had a gun of their own in the house had absolutely nothing to do with the event. And Kellermann withheld that information, and a lot of people noticed the problem right away. There were even letters to the editor of the journal asking “what gives,” and he responded with a very evasive answer in his reply to the letters.
The problem became inadvertently evident a few years later when he did another study with overlapping samples, where it became evident that he did have that information, and he knew perfectly well that people are rarely murdered with a gun belonging to someone in their own household.
It’s not usually domestic violence when people are murdered in their home. Instead, it’s more likely to be something like a crack dealer sells drugs out of his own home, and a customer comes in and kills him because he wants to get the drugs and not pay for them. That’s a little more typical of people killed while having a gun in their own home, but, of course, the customer brought in his own gun to murder the dealer.
Substitution in Methods of Suicide
Armstrong: Suicides are horribly tragic however they happen. Regarding methods, obviously there is some substitution effect: If a person cannot easily get a gun to commit suicide, he or she can substitute some other method, such as drugs, hanging, or jumping from a height. What do you think is the best estimate of this substitution effect? In other words, in what fraction of suicides, if a person didn’t use a gun, would the person probably have reached the same result with some other method?
Kleck: I wouldn’t call it an estimate of a substitution fraction, but I can put it this way. There’s really nothing to prevent every last one of those people, who otherwise would have used a gun, from using other methods.
Hanging is the second most common method of committing suicide—a distant second, currently. But, of course, if you got rid of guns, then that would no longer be the case—guns could not be the leading method of committing suicide.
My suspicion is that hanging would be the next most likely substitute, not only because it’s already the next most common way of killing yourself, but also, by reputation as well as in reality, it’s essentially as lethal as shooting yourself. They’re both about eighty percent fatal or less. You can only get an upper limit estimate of what percent result in death, because we always have incomplete counts of nonfatal attempts. Some attempts just don’t come to the attention of the authorities or don’t get medically treated.
Based on what we do know about, based on total number of suicides divided by total number of known suicide attempts, it’s about eighty percent. There’s no statistically significant difference between the two methods, which means presumably you’d have the same percentage dying if they used hanging rather than shooting.
It doesn’t take any special possessions or objects in order to do it—nothing other than a belt or a length or rope would be sufficient to form a noose, and a strong support to hang it from. That’s about all you need. So the materials are even more widely available than guns are. Guns are only in about half of U.S. households; presumably, something you could use as a noose is available in virtually all households. People hang themselves in prisons and jails all the time, and obviously prisoners are very restricted in what materials they have access to.
My judgment is that it would be very rare that someone wouldn’t be able to hang themselves who otherwise would have used a gun.
The only exceptions I might make is if people are really impatient to kill themselves—they had an impulse to commit suicide that didn’t last more than, say, ten or fifteen minutes. Then the preparation time to fashion a noose and find a strong support to hang it from might be sufficient to lead to at least some of those people changing their minds.
Adolescents appear to be the ones who have the most impulsive suicide attempts. They’ll have some transitory crisis in their life—like their significant other dumped them—and if they could just wait it out, or have the experience of living through similar things in the past, they wouldn’t be so quick to attempt a suicide. If literally the impulse to commit suicide didn’t last any longer than fifteen or twenty minutes, you could imagine some fraction of those adolescent suicide attempters not killing themselves quite as quickly as they would have been able to do with a gun. If there’s any saving of lives by change in method, it’s probably most likely to be among adolescent suicide attempters.
Gun Ownership and Crime Rates
Armstrong: Various people have made claims about the relationship between gun ownership and crime in a given region, such as by state. What do you think is the best evidence for what happens to violent crime in a given region (let’s say in the U.S.), other things equal, if gun ownership increases or decreases?
Kleck: There isn’t any direct evidence on that kind of short-term increase or decrease in gun ownership. Usually what you have in the way of pseudo-evidence is what happens when there’s some change in gun control. And the assumption is that if guns were more heavily restricted, then there must have been a reduction in how widely owned they are, or, if there’s a change in gun control in the opposite direction, then there’s a change in the opposite direction in gun prevalence. That’s almost certainly a false assumption.
What we know about the effects of gun laws on gun ownership rates is that they don’t have any effect. They’re not even intended to have an effect. They’re intended to have an effect on a small subset of the population—people with criminal convictions, especially felony convictions, and that’s about it. It’s too tiny a fraction of the population to have much material effect on overall gun ownership.
If you leave aside the research on what happens when somebody adds or deletes a gun control, we really don’t have much in the way of direct evidence.
All we know is that, across areas, there is no effect of gun ownership rates on crime rates, including homicide rates. That’s what we know.
Now, if there was some really enormous short-term increase in gun ownership, who knows? But that’s the point: We don’t have any evidence directly bearing on it. And I don’t think anything like that has been going on anywhere. Gun ownership has probably been pretty stable. It varies enormously from place to place, but it really doesn’t seem to vary much from year to year, when we rely on national surveys.
On rare occasions when surveys seem to suggest a change, it’s usually because there’s been a change in the willingness of gun owners to tell surveyors that they own guns. Something like that apparently occurred after the Brady Act was passed and the assault-weapons ban (on sales) was passed in 1993 and early 1994. There was some implausibly rapid and sharp decline in people reporting gun ownership to surveyors, but I really doubt there was an actual drop in the prevalence of gun ownership, since neither measure was actually intended to reduce that. Again, they were intended to keep bad guys away from guns, not reduce the overall gun ownership rate.
The best estimate we have is that gun ownership rates don’t make any difference.
Shortcomings of Surveys on Gun Ownership
Armstrong: How accurate do you think are surveys regarding gun ownership rates in the U.S.?
Kleck: There are strong indications that surveys underestimate the prevalence of gun ownership. We’ve got all sorts of little bits and pieces of evidence that point in that direction. For example, there have been a couple of studies where researchers get a list of people who are known to be gun owners, because they’ve registered their guns and are licensed gun owners with the authorities. Researchers go and survey them, not telling the respondents that they already know they own guns. They’ll ask them, just as if they were Gallup asking out of nowhere if they own a gun, and something like one in eight legally registered gun owners falsely claim they don’t have one.
But that’s an underestimate of concealing of gun ownership in itself, because remember that’s the most law-abiding subset. If you ask yourself, how many criminal gun owners are likely to deny gun ownership—people who illegally own guns, and are not supposed to be owning any type of firearm because of a criminal conviction—the rate of intentionally concealing gun ownership is almost certainly much larger.
And then we have gun possession by adolescents where their parents don’t know about it. Juvenile gang members, it’s unlikely they’d tell their parents that they’ve hidden away a gun somewhere on the premises. If you add all that up, the prevalence of gun ownership is almost certainly higher than it appears to be when the Gallup pollster or someone like that asks the question.
John Lott and Concealed Carry
Armstrong: What do you think of John Lott’s claims about concealed carry, particularly his thesis as summarized “more guns, less crime?”
Kleck: First of all, making carry laws more lenient did not increase the number of guns out there, or gun ownership, so the whole phrase, “more guns, less crime,” is wrong. Lott did not know that there were more guns, and it’s highly unlikely that there were more guns in the sense that he meant.
He cited one pair of surveys that weren’t comparable to one another, one done before, and one done after. A bunch of states had passed right-to-carry laws, and most of the surveys there didn’t even indicate the increases in gun prevalence that he claimed. There are maybe a couple of states that, if you take the results at face value, seem to indicate an increase, but, since the questions were not phrased the same—and it matters how you phrase the questions—they weren’t comparable.
So you really had zero evidence that gun prevalence increased. Certainly national surveys did not indicate any increase in gun prevalence. The whole premise of “more guns, less crime” was false, because there was no indication that there were more guns as a result of that kind of law being passed in many states.
What increased is the number of people who were legally authorized to carry. But there’s no indication that the total number of people carrying guns for self-protection increased, either. We only know more were legally authorized to do so.
There was a national survey later on that interviewed just people who had carry permits. Surveyors asked people if their frequency of carry had increased or decreased after getting the permit. Of course, if everyone were a law-abiding citizen, they wouldn’t have been carrying before at all. They would have been doing zero carry. So everyone should have said there was an increase. In fact, about the same number of people said it decreased as increased, and a lot of people said it didn’t change at all.
This means the best evidence we have is there was no increase in carrying among people who eventually got carry permits. I don’t know who else was supposed to increase their carrying if not the permit holders. There’s in fact no good reason to think that even the rate of carrying went up. What instead happened is that people who were already carrying guns for self-protection, without benefit of a carry permit, then took advantage of the more liberal rules for getting a permit, and they got a permit. They legitimized the carrying they’d already been doing.
So no increase in the prevalence of gun ownership, no increase in the rate of carrying, and thus no sound reason to expect that criminals were facing more of a risk of running into an armed crime victim.
Having said that, I also want to point out that I’m perfectly sympathetic with the idea that criminals do pay attention to the risk of running into an armed victim. That’s not at all what I’m quibbling with. I’m quibbling with whether or not there was any reason for that perception of risk from armed victims to have gone up as a result of the right to carry laws. I don’t think it did.
I think the best available studies on the subject indicate those laws did not have an effect on the crime rate. Lott was making various mistakes in the research, and, when they’re fixed, then there’s no evidence that the laws actually caused decreases in crime rates.
[November 21 Update: See also John Lott’s reply.]
Gangs and Violence
Armstrong: Let’s turn to questions of where crime is coming from in our country. How much violent crime takes place in the subset of the population we would typically associate with the gang culture?
Kleck: In places like Chicago or Los Angeles, it’s a huge fraction of it. It varies enormously from place to place. It may well be that half or more of the gun homicides in those cities are gang related. But in most places in America, it’s a somewhat more modest fraction.
We don’t have national figures that are of any use. For what it’s worth, in the FBI uniform crime reports data, they do have a category for the circumstance in which the crime was committed. One possible box that local police can check in filling out the homicide reports for the FBI could indeed be for gang-related. But the problem is that the FBI forms require police to check just one circumstance. So if a guy belongs to a gang, and he was selling drugs, and he has a dispute with his customer over the price, and then they get into an argument and one shoots the other, that could go into any of three or four different categories, only one of which is gang-related. So those data are useless.
What we’re stuck with are local estimates, and, as I say, it varies enormously from one locality to another. It’s a huge percentage in a couple of cities. Chicago and Los Angeles have really bad street-gang problems. On the other hand, in Peoria it’s probably a relatively small fraction, certainly well under half.
Armstrong: How much violent crime is associated with domestic violence, lovers’ quarrels, and other cases involving romance or sex?
Kleck: It’s kind of a reverse of what I just said about gang violence. The same places that don’t have much of a gang violence problem are more likely to have a higher fraction of what you could call “ordinary” violence (if you can get your head around that concept).
A place like where I live, Tallahassee, I could well believe that better than half of the homicides are domestic homicides. We don’t have many killings related to other felonies—that is, there aren’t many robbery killings, rape killings, crime killings, or organized crime killings. What dominates instead is what I guess you could call the “ordinary” sorts of violence, and domestic violence is a big share. So in a place like Tallahassee it could easily be half, and hardly any is street gang related.
Violence Resulting from Drug Prohibition
Armstrong: Jeffrey Miron has a 2004 book out, Drug War Crimes, in which he claims that enforcement of drug prohibition is associated with violent crime. In the same way we had some violent crimes increase during alcohol Prohibition. What do you think of his thesis?
Kleck: I don’t know his work in particular, but certainly many criminologists have addressed the same issue. Yes, they definitely are of the opinion that prohibition is responsible for quite a bit of so-called “drug related” violence.
The vast majority of homicides linked with illicit drugs are not due to the killer being under the influence of the drugs, they overwhelming are related to the buying and selling of illicit drugs. That is, they are so-called market related crimes. Obviously that kind of violence would not occur if the drugs were not illegal. I’m not saying that necessarily implies therefore they should be legal, I’m just pointing out the consequences.
You don’t have comparable amounts of homicide among the Anheuser-Busch company versus the Schlitz company or versus Coors. They sell beer, and it’s legal, so they don’t duke it out with violence in the streets and killing one another’s dealers.
But, when the drug is illegal, violence is the only way you have of settling disputes, like disputes over markets. “This is my market share, not your market share; this is my territory, not your territory.” They can’t go to the courts and the police to resolve disputes. If they want to collect a debt, they have only one way to do it: They threaten or use force. And it’s a cash-only business, of course, which is weird in this day and age when everything is electronic transfers through credit cards and the like.
You have prohibition setting up a scenario where there’s bound to be huge amounts of violence with that kind of market. The drugs would be as bad for your health if they were legal—in many ways they would be as bad for your health—but there wouldn’t be that market-related violence, which accounts for most so-called “drug related” violence.
This is an ongoing problem we’ve had for as long as we’ve chosen to make cocaine and heroin and so on illegal. We scarcely notice it, because it’s not anything we’ve changed in decades. It’s been a century since it was legal to buy or sell cocaine or heroin. There hasn’t been any change recently for us to notice what’s the effect of the change. We’ve kind of become accustomed to it. But it’s that background amount of violence that goes on year after year after year that’s linked with the fact that we chose to prohibit those kinds of drugs.
We have an experiment, where we said, let’s see what happens when we do change the legal status. We said, let’s make alcohol illegal, and, for thirteen years, we had huge amounts of violence that were directly related to the organized crime efforts to control the market in beer and liquor. And then we passed the repeal of the Volstead Act in 1933, and all of a sudden the homicide rate starts going back down again. There’s certainly reason to believe that the initial increase in homicide when we passed the Volstead Act, and then the decrease when we repealed it, had a lot to do with those changes.
Problems with Predicting Violence
Armstrong: I have the idea that police should spend more resources investigating serious articulated threats, whether in a mass-murder scenario or in a case of domestic violence. One case I have in mind is the Aurora theater shooting in Colorado, where police were warned, but they just didn’t take adequate action to prevent the crime. Regarding domestic violence, I have the impression that women sometimes go to the police but don’t get much help. Do you think this approach has merit as a potential way to substantially reduce violent crime?
Kleck: I can offer you a little bit of peripherally relevant information, because I don’t think there’s any good evidence on whether that sort of effort actually does reduce violence.
First of all, from the police standpoint, what we all fail to see are the false positives. We see the cases where they fail to take action. That is, they took the negative response of not doing anything, and then it blows up in their face because later on the person involved commits a horrible crime.
We don’t see all the false positives, or potential false positives, where there was some indication the person might do bad stuff and then they never did do bad stuff. And the latter is much more common than the former. That’s the police standpoint. They’re aware of how many complaints they get that don’t turn out to be anything. So the trick is to distinguish one from the other.
The cops are no better than the average criminologist; they’re really not very good at predicting violence of specific individual people. You might get a certain number of reliable positives: That is, you predict someone to be violent, and later on they are violent. That’s the easy part. I mean, right now I could predict that all 320 million Americans are going to be violent in the next year. And everyone who was in fact violent, I would have predicted it. But of course there is the tiny little problem that 99.99 percent of the time I was wrong. So getting correct positives is not the trick; it’s making those predictions correctly but not falsely implicating a lot of other people, the other 99 percent who are in fact not going to commit violence.
So if you’re saying we ought to do something to get a better predictive capability, I’m all for it. That means putting some serious money into research as to what the precursors of serious violence are, so we know what those warning signs are.
Expanded Records for Background Checks
I can suggest one policy change that would be good. It’s not just domestic violence; it’s violence related to mental illness. A certain number of people are not just vaguely predicted to be kind of dangerous. There are some people where, in the past, a court has gone through the process of actually evaluating the evidence. There’s a psychiatric evaluation. It’s not just a neighbor who thinks the guy is nuts and might do some harm. A court has actually been through some evidence with serious legal proceedings, and they declare a person to be dangerously mentally ill. Usually they have some vague phrase like “not capable of controlling their behavior” or whatever, but what it really means is dangerously mentally ill. That’s more than just a casual prediction of future violence.
The problem is, that information does not get widely disseminated. Great, some circuit court has now declared the guy to be dangerous—wouldn’t it be nice if we made sure that kind of guy didn’t get a gun (among other things)?
Well, now you’re a gun dealer, and this guy walks into your store. You have him fill out the usual forms, and you send them in and do the usual background check. Is the fact that he’s dangerously mentally ill going to show up in the computerized records? Ninety-plus percent of the time, no, because we have no reasonably complete, national compilation in computerized form of people who have been declared by a court to be dangerously mentally ill. And that would be nice to have.
For years the federal government has been trying to get the states to expand their computerized records and then make them available to the national database for checking. But the federal government can’t force states to do that; it’s constitutionally forbidden from forcing them to do it. All they can do is give states positive inducements, like giving them money, and they’re certainly willing to do that, but for the most part states have not taken them up on it. States haven’t gone to the effort of making their records more comprehensive and more computer accessible.
I don’t really know the details of why states have not taken that federal money and done that. I suspect there are political reasons, like claims that it might be used for a registration system, and they’ll take our guns away, or something like that.
There’s a policy change that might do some good. It’s boring, though. It’s not an exciting slogan that can fit on a bumper sticker. But it means you can improve the records and the accessibility of computerized records of people who are dangerously mentally ill and shouldn’t have a gun.
If you want to know what’s arguably the most effective thing we can do in the way of gun control, that’s it. But it’s not a matter of passing a new law, it’s making the records better. That would also help with any kind of domestic violence that happens to be attributable to the prospective offender being mentally ill.