Deltoid

Lawrence Kennon writes:

The following documents exactly the kind of “junk science” being foisted off on the public by the medical profession, and in particular the CDC and the NEJM.

It does nothing of sort.

There are dozens of falsehoods, and dozens of claims that are extremely dubious. It would be possible to put these down as honest errors, caused by Suter’s pro-gun bias, except for the following example which can only have resulted from blatant dishonesty on Suter’s part:

Edgar Suter writes:

harmful and unconstitutional nostrums

Crime and homicide rates are highest in jurisdictions, such as Washington, DC, New York City, Chicago, and California, where the most restrictive gun licensing, registration, and prohibition schemes exist. Why are homicide rates lowest in states with loose gun control (North Dakota 1.1, Maine 1.2, South Dakota 1.7, Idaho 1.8, Iowa 2.0, Montana 2.6) and highest in states and the district with draconian gun controls and bans (District of Columbia 80.6, New York 14.2, California 12.7, Illinois 11.3, Maryland 11.7)?(49) (See Graph 18: “Representative State Homicide Rates”)

Precisely where victims are unarmed and defenseless is where predators are most bold.

Got that, folks? Suter implies that gun control is a “harmful nostrum” that caused the homicide rate to be ten times higher in the restrictive places.

What’s wrong here? Well, for one thing Suter has dishonestly chosen to represent “states with loose gun control” by the six such states with the lowest homicide rates, and to represent states with restrictive gun control by a city and the four such states with the highest homicide rates. Why didn’t he choose Louisiana 16.9, Georgia 12.8, New Mexico 10.5, or Mississippi 12.8 to represent states with loose gun control and Rhode Island 3.9, Hawaii 4.0, or Massachusetts 4.2 to represent restrictive gun control states? His graph 18 should be entitled “Misrepresentative State Homicide Rates”.

Could this be an honest error on Suter’s part? Could it be coincidence that he chose to represent the states with loose gun control with the one with the lowest homicide rates? Well, the chance of randomly selecting 6 states from 22 and coming up with the six that have the lowest homicide rate is one in 74613. Coincidence? I think not. Suter must have looked at the complete list of homicide rates and deliberately set out to mislead his readers.

Suter is not to be trusted.

Edgar Suter writes:

Dr. Kellermann’s subsequent research “finding” that a gun in the home increases risk used a method that cannot distinguish between “cause” and “effect.” Kellermann’s illogical conclusion would be like finding more insulin in the homes of diabetics and so concluding that insulin “causes” diabetes. Interestingly Kellermann’s own data show that when a homeowner is killed only rarely is the “gun in the home” the instrument of the homeowner’s death.

Untrue. See table 1 of the paper.

How then can the gun “cause” the death? Does the gun magnetize murderers to the homeowner’s doorstep? Does the gun emit magic rays that cloud the mind of otherwise good people? Of course not.

Amusing speculation, but these bear no relation whatsoever to the actual content of the paper

If we put Kellermann’s research in context with all the other scientific evidence, all we can conclude is that fear of crime causes people in high-risk areas to buy guns for protection — hardly a momentous finding.

The study found that gun ownership was associated with a higher risk of gun homicide and was not associated with a higher risk of homicide by other means. This is inconsistent with your explanation unless you believe that people don’t think that guns will protect them against assailants who don’t have guns.

Even the National Crime Victimization Survey, one of the studies most cited by the anti-self-defense lobby, shows that guns are the safest and most effective means of protection - safer than not resisting or resisting with less powerful means. All this explains why the 28 states that allow law-abiding, mentally-competent adults to protect themselves outside the home with concealed handguns have lower rates of crime for every category of crime indexed by the FBI.

You have apparently found the method for distinguishing between cause and effect that eluded Dr Kellermann. Can you tell us how you were able to eliminate the possibility that injured crime victims may find it more difficult to resist? Or the possibility that high crime rates might cause states to enact controls on handgun carry?

Daniel D. Polsby writes:

Unless I am seriously mistaken, one would find that crack cocaine dealers and other persistent criminals are disproportionately likely to possess firearms and to be murdered by others using firearms. To place firearms at the heart of this story is at best tendentious.

The study controlled for literally dozens of other factors, including criminality and illicit drugs. Furthermore the extra homicide risk associated with firearm ownership was not from shootouts between drug dealers or gangs, but domestic homicides.

Absent a controlled experiment (which is impossible) it is of course always possible for those who don’t like the results to come up with some explanation other than the straightforward one.

Dr. Kellermann’s published work is more diffidently worded than the representations he makes to reporters in subsequent interviews. Moreover, he has in fact made a causal claim for his research at least once in writing, in the July, 1994 Atlantic Monthly correspondence column. He has never made his underlying data public. I respectfully suggest that these are issues that ought to trouble even one who finds his conclusions congenial.

Dr Kellermann is entitled to believe that gun ownership most probably increases the risk of homicide and to offer this study as evidence for this.

I don’t find the conclusions congenial at all. My taxes will go up by $100 next year to pay for this stupid ban on semi-automatic long guns. It would be nice if I could argue that firearms aren’t associated with costs in terms of homicides and suicides.

A limitation on the earlier (43-1) study is not necessarily a limitation on the later case-control study. The authors of the bibliography are quite correct when they state that a case-control study could measure a net protective effect of firearms.

Dr. Paul H. Blackman writes:

The earlier study noted that one couldn’t fully evaluate the protective value of firearms without knowing about their use in non-fatal protective situations. I.e., the authors recognized that a gun could be used for protection without producing a corpse. The later study did not note that one needed such complete knowledge in order to evaluate the protective uses of firearms.

Because the methodology was different. The case-control method can measure net protection from homicides. For example, if all the respondents in Kleck’s DGU survey spoke the truth, then over 500,000 homicides are averted each year using guns. If we make the (perhaps unrealistic) assumption that guns can magically prevent all household members from homicide, we would expect to see zero homicides in the half of households with guns, and over 500,000 each year in the gun-less households. Kellermann would have found no gun owners amongst his case subjects and ended with an odds ratio of 0.00 — indicating a very very strong protective benefit. If, instead of being magic talismans, guns only prevented half the potential homicides in gun-owning households then there would be 500,000 homicides in gun-owning households and 1,000,000 in the gun-less households. A case-control study would find that 33% of the cases owned guns and 50% of the controls owned guns, giving an odds ratio of 0.50 — indicating a very strong protective benefit.

A case-control study which measures only deaths cannot measure a net protective effect of firearms.

Yes it can — see above.

You are simply mistaken. You can only be correct if a gun has no protective value shy of producing a corpse,

Nope, if a gun prevents someone from becoming a corpse, that benefit can be measured.

and if the only crime from which one wishes to be protected is a homicide.

It only measures protection against homicide, not against other crimes. A protective benefit against homicide is ipso facto a protective benefit.

Dr. Paul H. Blackman writes:

But let’s get back to the estimates of gun ownership by the cases and the controls.

OK. Unlike Dr Suter’s straw man argument this is a real threat to the study. If gun ownership of the cases is under-reported more than gun ownership of the controls is under-reported, the correlation between guns and homicide is weakened. If gun ownership of the cases is under-reported LESS than gun ownership of the controls is under-reported, the correlation between guns and homicide is strengthened.

The cases were, of course, proxies for them. But the situation was that a homicide occurred in that home, allowing a certain amount of searching by law enforcement, esp. if guns were involved.

Kellermann did not use reports from law enforcement to measure gun ownership. (If he had, the ownership rates for cases would have been higher.)

In addition, Kellermann et al. waited about a month after the homicide to contact the proxies who were hoped to be persons living in the home, certainly relatives, who almost certainly were going through the remains in the course of the month, thus allowing more truthful statements than might otherwise have been the case, with no temptation to lie.

Fair enough, if case proxies found guns that they were unaware of before the killing, this would weaken the correlation.

Controls: The survey relied upon by Kellermann et al. fails to show that ordinary people admit gun ownership to strangers. Kleck’s appendix on the topic, and the Bordua et al. survey and other studies of Illini indicates that people often don’t tell authorities about guns (in the Illinois case, a FOIA card) but will admit it to survey research,

Kellermann’s study did not rely on something like a FOIA card, but survey research.

and in other instances will lie. Surveys generally show such a dramatically lower reporting of guns by women than of men — when household ownership is the question — as to suggest a substantial understatement when asking “controls” such as these.

Controlling for this factor will strengthen the correlation. Consider:

1. Data for the cases was based solely on proxies while about half the data for controls was obtained from the controls themselves rather than proxies.

2. Cases were mostly male.

3. Controls were matched by sex so were also mostly male.

4. Consequently, proxies were more likely to be female than the controls

5. If women are more likely to under report it follows that this underreporting would be greater for the cases than for the controls, thus strengthening the correlation.

This possibility is, however, undercut by one factor that would weaken the correlation that you did not mention: the proxies for the controls reported a higher rate of gun ownership than the controls did. Kellermann implies that the difference was small, but I would have like to have seen an analysis stratified by proxy/non-proxy.

It would have taken about 11 controls falsely denying guns in their homes for the statistical significance of Kellermann et al.’s crude odds ratio to disappear.

Or, to put it another way, the difference in the crude odds ratio would disappear if 35 more controls falsely denied than cases did.

The odds are good there were 20-40 false denials.

This is a pure guess. We do know that at least 8 cases falsely denied.

The study controlled for literally dozens of other factors, including criminality and illicit drugs. Furthermore the extra homicide risk associated with firearm ownership was not from shootouts between drug dealers or gangs, but domestic homicides.

Dr. Paul Blackman writes:

No. The study measured about 2.5 dozen items, but controlled for about six — with a number of items prevented from measurement by matching the controls (race, age group, sex, etc.)

All right. They controlled for four factors by matching and another six in the multivariate analysis. The other 2 dozen measured items could have been controlled for if they had proved to be statistically significant.

None of the items measured or controlled for involved drug trafficking, as opposed to simple drug use.

Though you would expect this to be strongly correlated with arrested and with drug use.

One would certainly have expected a higher-than-average rate of domestic homicide when the three-fourths-plus of homicides outside the home were excluded — although non-gun and domestic homicides were also minimized a bit by excluding the slayings of pre-adolescents in the home.

Yes, the fraction of homicides that were domestic would be higher than for homicides in general, but that wasn’t my point. The “extra” homicides associated with gun ownership weren’t from drug dealers shooting each other but domestic ones. That is, the alternative explanation: “Drug dealers own lots of guns and are likely to killed by other criminals” does not fit the data.

Edgar Suter writes:

In Kellermann’s most recent study of homicide (Kellermann et al. “Gun Ownership as a Risk Factor for Homicide in the Home.” NEJM Oct 7, 1993; 329(15):1084-1091.) notes “Two hundred nine victims (49.8 percent) died from gunshot wounds.” Prof. Schaffer offers a robust [and successful] effort to show that the “gun in the home” cannot be the actual homicide instrument in every case. According to Table 1, mentioned by Mr. Lambert, this figure includes 1% “unknown firearm” as “method of homicide.” Let’s charitably imagine that these “unknown firearms” actually were the “gun in the home.” Let’s charitably imagine that all of the “handgun,” “rifle,” and “shotgun” methods tabulated were “guns in the home.” Let’s accept the unlikely assumption that not one gun was imported into the home for the homicide.

Still a majority of the homicides were not committed by guns, so could not have been committed by Kellermann’s scary “guns in the home.” How then did the “gun in the home” become a risk factor for homicide.

Because a significant number of the homicides were committed with the gun in the home.

I can imagine no explanation for such a phenomenon other than by invoking silly anthropomorphic magical thinking — the “magnetize murderers to the doorstep” silliness.

You cannot imagine the possibility that someone would use a gun in the home for murder?

Can anyone offer a reasonable hypothesis to explain how a “gun in the home” can increase the risk of being killed by strangulation, stabbing, etc.?

I cannot offer such a hypothesis. Of course, since the the study indicates that a “gun in the home” was not associated with an increased risk of being killed by strangulation, stabbing, such a hypothesis is not needed.

Kellermann asserts that the “gun in the home”is a risk factor in that paper and, as Prof. Polsby noted, was “less diffident” in his post-publication comments. Explaining how “guns in the home” increase non-firearms deaths, since it is clearly part of the “risk” Kellermann invokes, certainly “bear relation to the content of the paper.”

Except that Kellermann only found an increased risk of firearm deaths.

You have apparently found the method for distinguishing between cause

and effect that eluded Dr Kellermann.

Yes, I have. “…in context with all the other scientific evidence…” I simply avoid Kellermann’s self-citations and incestuous citation of only the congenial “public health” literature. In reviewing not only the studies that Kellermann finds congenial, but also the literature that he finds uncongenial or ignores, the correct order of causality can be deduced — and is the only order of causality that does not require invoking any magical thinking. The studies that show an inverse relationship between gun density and crime disallow the “guns cause crime” hypothesis.

The relevant hypothesis is “guns make crime more deadly”. The scientific evidence on this is never going to be conclusive, but it certainly tends to support it. For example, Kleck found that “gun present” was associated with a 1.4% increase in the risk of death in violent incidents (Point Blank table 5.11). Knives were associated with a very much smaller increased risk (0.3%).

Can you tell us how you were able to eliminate the possibility that injured crime victims may find it more difficult to resist?

Certainly a possibility, but of what relevance? Looking at endpoints, gun defenders have fewer injuries than non-resistance or resistance with less powerful means.

Or, to put it another way, those that are injured are less likely to use guns. Does gun use prevent injury? Or does injury prevent gun use? Which is the cause and which is the effect? Correlation alone cannot decide this.

Or the possibility that high crime rates might cause states to enact controls on handgun carry?

Please note Clayton’s earlier post about the admixture of high, average, and low crime states among the reformed states.

This just means that the correlation is not that strong. It does not help us disentangle cause and effect.

Clayton does properly challenge my use of “static” data. In a “more diffident” context I offer the “every category of crime… is lower” comparison only to undercut the anti-self-defense lobby’s claim that progressive carry reform “will” or “might” lead to “blood running in the streets.” While the static observation does not allow me to make a claim of a causal relationship (that carry causes lower crime), the observation does undercut the anti-self-defense lobby’s claim that carry increases crime (particularly when coupled with Cramer & Kopel’s data). I should have made the less expansive claim, “All this mayexplain_ [rather than my original “explains”] why the 28 states that allow law-abiding, mentally-competent adults to protect themselves outside the home with concealed handguns have lower rates of crime for every category of crime indexed by the FBI.” Mea culpa.

Based on the data you presented, I find that the average homicide rate in the permissive carry states is 7.9 and in the restrictive states it is 10.2.

States with low populations tend to have low homicide rates and permissive laws. This is the reason why the average is lower in the permissive states. If we control for the size of the state’s population by just considering states with populations between 1 and 10 million we find that the average in the permissive states (8.3) is very slightly higher than in the restrictive states (8.2). (There are too few restrictive low population states and too few permissive low population states for meaningful comparison amongst high and low population states.)

The authors of the bibliography are quite correct when they state that a case-control study could measure a net protective effect of firearms.

EdgarSuter writes:

“could” if death were the only legitimate measurement of the protective benefits of guns.

Wrong. “could” if protection from death is a legitimate use of a firearm.

Mr. Lambert’s quibbling about the definition of “only rarely” bypasses the main point of the letter.

Since an indisputable majority of Kellermann’s homicides were not committed using those scary “guns in the home,” one must invoke magical thinking to explain how these unused “guns in the home” were risk factors for the majority [non-gun] of homicides.

For the fourth time: THEY WERE NOT.

I would ask Mr. Lambert to address this fundamental problem, rather than a diversion indulging in hair-splitting over how small a minority of cases involved the “gun in the home.” [shivers in fright at the mere mention of “gun in the home”] Please address the fundamental problem.

Addressed. Yet again.