pageok
pageok
pageok
"Guns in the Home and Risk of a Violent Death in the Home":

Yet another study — which I learned about because it was cited to me by a fellow academic — that shows a correlation between home gun ownership and homicide and suicide risk to occupants without controlling for some obvious confounding factors.

Consider, for instance, criminal record (both its presence and its magnitude). Criminality is highly associated with extra risk of being killed (your business associates and rivals are criminals, too) and extra risk of suicide. At least for certain types of criminality, it's also likely to be associated with an extra likelihood of gun ownership, for instance because the gun is a crime tool or because the criminal needs the gun to defend himself against his criminal associates and rivals. Does the study control for this? Nope.

There are other problems with the study as well; Gary Kleck, for instance, has pointed to evidence that in many studies, quite a few respondents (perhaps 5-10% or more) conceal their gun ownership if they can. That's especially likely to happen when there hasn't been a gun death in the house, so that concealing the gun ownership is easy; so as a result, the gun ownership among the control group (where there has been no gun death) is underestimated, and the correlation between gun ownership and gun death is overestimated.

But the failure to control for the obvious criminal history variable strikes me as especially glaring. And yet such studies are read, reported on, and believed.

Alex Denmark (mail):
The inability (and, as in this case, the conscious choice not to) control for every meaningful variable is why I pretty much ignore these sorts of correlative studies.
7.1.2008 5:39pm
Mark Buehner (mail):
Were suicides included? Thats half the gun deaths in the U.S. right there. Accidents account for something like 3%.

This brings up an interesting question- Justice Stevens mentioned suicide quite a few times in his Heller dissent. How does that jive with his (concurring) opinion in Vacco v. Quill where he opens the door to assisted suicide under certain circumstances, which would certainly indicated an underlying right to suicide in general?

In other words how do you argue that suicide may be a fundamental right but that it is reasonable to suspend another fundamental right in order to combat it?
7.1.2008 6:03pm
Eugene Volokh (www):
Suicides were included; I've revised the post to make that clear.
7.1.2008 6:06pm
Mark Buehner (mail):
Justice Breyer, not Stevens, my mistake.
7.1.2008 6:12pm
bobby b (mail):
"In other words how do you argue that suicide may be a fundamental right but that it is reasonable to suspend another fundamental right in order to combat it?"

It's easier than it used to be. Increasing levels of rabid partisanship, combined with an increasingly polarized society, have reduced the need for an actual logical progression in any argument. Thus, "the sky is blue, but not at night, and so BUSH LIED!!!!", or "three times two equals six, and two times three also equals six, and so OBAMA IS A MUSLIM!!!" won't hardly get you an askance look anymore. Your conclusion is obviously correct in the minds of your self-selected cohorts, and so, why would anyone dig deeper?
7.1.2008 6:23pm
Curt Fischer:
From the abstract of the study:


Those persons with guns in the home were at greater risk than those without guns
in the home of dying from a homicide in the home (adjusted odds ratio = 1.9, 95% confidence interval: 1.1, 3.4).



If you read the study, you find that the increased risk faced by gun owners of dying from a firearm-related homicide is also significant (odds ratio 3.5, 95% confidence interval 2.0, 6.1 -- see Table 4).

Comparing these two odds ratios, 1.9 and 3.5, is highly interesting. The ratio of these two numbers is a measure of how much of the increased risk of homicide that falls on gun owners is attributable to guns, controlling for all general factors which lead to increased homicide risk.

3.5 / 1.9 = 1.84. That is easy to calculate. Of course, the confidence intervals on this ratio are critically important. By making several undoubtedly inaccurate assumptions [i) that the odds ratios are normally distributed, ii) that the standard deviations of the odds ratios are equal to half the difference between the CI extrema and the mean, and iii) that the number of observations of the odds ratio is 1) and this page, I calculated the 95% confidence interval as being from 0.584 to 5.663.

If the CI spans unity, it's tough to say much about whether gun ownership increases risk of dying from firearm-related deaths per se, or if a hidden, unmeasured variable is responsible for most of the variation. Of course, my method of calculating the CI is highly suspect. I just hope my analysis inspires a real statistician to do a good job of it.

Secondly, Prof. Volokh's criticism seems overwrought, because the study itself says in its conclusions the following:


The
gun in the home may not have been the gun used in the death.
This possibility seems less likely with suicide, but, with
homicide, it is certainly plausible that someone brought a
gun into the home. [...] It is possible that the association between a gun in
the home and risk of a violent death may be related to other
factors that we were unable to control for in our analysis. For
instance, with homicide, the association may be related to
certain neighborhood characteristics or the decedent's
previous involvement in other violent or illegal behaviors.
Persons living in high-crime neighborhoods or involved in
illegal behaviors may acquire a gun for protection. The risk
comes not necessarily from the presence of the gun in the
house but from these types of environmental factors and
exposures.



It seems like Prof. Volokh's criticism is squarely addressed by these remarks. Of course, I'm not sure why the study authors were "unable to control" for some of those factors in the manner I suggested above.
7.1.2008 6:26pm
30yearProf:
Whoops!

QUOTE:
Second, the gun in the home may not have been the gun used in the death.


Incredible! They interviewed the family members! They could have easily found this out by asking the very people they interviewed (who would know).

After all, the implication sought from ALL these pseudo-scientific "studies" is that YOUR gun will hurt YOU. Here, they had the chance to ask whether or not it was YOUR gun that "caused" YOUR death and they failed to ask.

Logically, that raises a strong suspicion that they did not ask because the researchers either knew or strongly suspected that the answer would not support their desired result.
7.1.2008 6:41pm
30yearProf:
BTW, knowing that your gun will hurt you sometimes is meaningless without a comparison to how often your gun will protect you.

Chainsaws will hurt you sometimes (ask any firefighter or logger) but they also allow increased productivity that seems to overwhelm the danger. Chainsaws aren't "necessary," the lumber to rebuild Chicago after the Great Fire was cut by hand tools but they are desirable.

Also, off the top of my head, it seems that the "bad" odds get smaller with every new study. Wasn't it 43 to 1 once?
7.1.2008 6:53pm
Don Miller (mail) (www):
For the past two years, I have been a volunteer EMT in my community.

I am well aware that anecdotal evidence isn't statistically proof, but my experiance makes me wish there was a researcher to study the situation I have seen, talking to other EMTs and Paramedics, I think my observations are not unique.

Firearms are highly effective for suicide. I think that is beyond dispute. Of the 20 suicide attempts I have responded to in the past two years, only the ones involving guns resulted in death. In every case though, the family was warned that the person was suicidal and asked to remove all firearms from the home. Only one time did a family take those steps. Actually turned over all the guns to the Police department. That individual purchased the gun he used from a friend that day.

My suggestion for researchers, what percentage of firearm suicides happen after the family was warned to get the person help and to take precautions to remove their access to firearms, and then failed to takes those precautions.

My supposition, it is fairly high.
7.1.2008 6:56pm
Thomas_Holsinger:
"...that raises a strong suspicion that they did not ask because the researchers either knew or strongly suspected that the answer would not support their desired result."

That is often, if not generally, the case for news stories as well as so-called studies, and both sides do it. Very, very, few news stories about the recent shootings of two burglars in Pasadena, Texas, by a neighbor named Horn have ever mentioned that they were shot on Mr. Horn's property, not the property which was burglarized. The stories, including those on pro-gun blogs, have emphasized that Horn left his home armed with a shotgun for the purpose of shooting the perps for the crime of burglarizing his neighbors' house, and that this was apparently authorized by Texas "new castle" law.

But it was terribly inconvenient to mention facts showing that, despite Horn's intent, his shooting the perps dead was justified at the precise moment he did it. His intent didn't matter if his act was justified under common American legal principles, but both pro and anti-gun publicists want to play up his intent.

Note how the stories about this incident go into great detail about Horn's conversation with the 911 dispatcher but say nothing about the actual shooting.

We see the same pattern in so-called studies about gun use, gun fatalities, global warming, etc.
7.1.2008 7:05pm
hawkins:
While it may be subject to the same criticism (whether the homeowners have criminal acquaintances), the below statistic seems to make a fairly persuasive argument that having a gun in your home creates more danger than it avoids.

Entirely irrelevant to the issue of legality though, unless Congress wishes to consider it next time it amends the Constitution.


In Atlanta, a city where approximately a third of households contain guns, a study of 197 home-invasion crimes revealed only three instances (1.5 percent) in which the inhabitants resisted with a gun. Intruders got to the homeowner's gun twice as often as the homeowner did.


7.1.2008 7:06pm
hawkins:
7.1.2008 7:17pm
Curt Fischer:

Incredible! They interviewed the family members!


The authors only interviewed the kin (or kith) of 83% of the decedents in the study. For the other 17% they relied only on the death certificate, which presumably does not indicate ownership of any firearm(s) used in the homicides.

The level of outrage I've seen so far on the thread is out of proportion to the severity of the study's flaws, at least as those flaws have been discussed here thus far.
7.1.2008 7:17pm
cboldt (mail):
I wonder if this is the same study.
.
Between June 1 and August 31, 1994; Atlanta Police Department reports were screened to identify every case of unwanted entry into an occupied, single-family dwelling. Cases of sexual assault and incidents that involved cohabitants were excluded.
.
RESULTS--A total of 198 cases were identified during the study interval. Half (99 cases) involved forced entry into the home. The victim and offender were acquainted in one third of cases. A firearm was carried by one or more offenders in 32 cases (17%). Seven offenders (3.5%) carried knives. In 42% of cases, the offender fled without confronting the victim. Victims who avoided confrontation were more likely to lose property but much less likely to be injured than those who were confronted by the offender. Resistance was attempted in 62 cases (31%), but the odds of injury were not significantly affected by the method of resistance. Forty cases (20%) resulted in one or more victims' being injured, including six (3%) who were shot. No one died. Three victims (1.5%) employed a firearm in self-protection. All three escaped injury, but one lost property.


Weapon involvement in home invasion crimes. [JAMA. 1995]
7.1.2008 7:17pm
cboldt (mail):
-- Intruders got to the homeowner's gun twice as often as the homeowner did. --
.
That doesn't pass the smell test, which is why I poked around to see if I could find more details pertaining to the study.
.
"And the crowd booed"
7.1.2008 7:22pm
Mike S.:
There are often constraints in doing studies that limit how well you can control for everything. For instance, it is probably very hard to get an honest answer to questions like: "Are you a criminal?" "Are your friends and business associates criminals?" You do the best you can. Which makes it important to recognize the limits of your study (which the authors seem to have done, though not the reporters)
7.1.2008 7:27pm
cboldt (mail):
Heh. Even funnier, it's Arthur Kellermann in the June 29, 2008 OpEd, and Arthur Kellermann the author of the Atlanta "1994 hot break-in" study.
.
So, A firearm was carried by one or more offenders in 32 cases (17%), and Forty cases (20%) resulted in one or more victims' being injured, including six (3%) who were shot. I guess that means 100% of the six victims who were shot, were shot with their own guns.
.
Who believes this guy?
7.1.2008 7:36pm
Dr. T (mail) (www):
Curt Fisher is not performing correct statistics. Probabilities and odds ratios have binomial distributions, not normal (Gaussian) distributions. The method he used for calculating the confidence intervals for ratios of odds ratios also is incorrect. An F test is needed. It is possible to do this with the data given, but I'm not interested enough in the study to do the work.

I have read numerous articles in the psychiatric literature about depression and suicide. I believe that easy access to a handgun does increase the odds of suicide. When suffering from severe depression (especially anger-depression as opposed to lethargy-depression), the ability to end it all with one fast, dramatic gesture makes suicide seem more appealing. This partly explains why police officers have high suicide rates.
7.1.2008 7:46pm
Gerg:
Wait, why would controlling for "criminality" be a useful thing to do -- even if it were a meaningful concept to begin with?

"criminality" is not some external factor, it's not some immutable property people are born with. And it's very much not unrelated to owning the gun, as you point out. Readers of the study may be deluding themselves "oh but I'm a law-abiding citizen and shouldn't be compared with those 'criminals'" but in fact those "criminals" were once law-abiding citizens too and probably still considered themselves law-abiding citizens.

The basic premise of your argument is that an individuals decision on whether to own a gun or not will have no affect on whether they one day find themselves being considered a criminal. But of course that's not true at all, as you point out. If they own a gun they're much freer to engage in the kinds of activities that someone might need a gun to engage in or to protect themselves afterwards. They're also more likely to find themselves "accidentally" a criminal because of a gun accident or gun law violation.

This is all a bit like saying statistics on drunk driving should only include people who don't own cars.
7.1.2008 7:48pm
Someone who reads (mail):
Arthur Kellerman being the author of the Washington Post editorial linked to above, the author of the study referenced in Atlanta, and the author of the Washington State study that began the "why didn't they control for obvious variables" questioning in the first place, I'm inclined to take with a grain of salt any of his conclusions.
7.1.2008 7:53pm
Paul Hsieh (mail) (www):
A coworker sent me the same story, and here is part of my reply:

The Spring 2007 issue of the Harvard Journal of Law and Public Policy includes the article by Kates and Mauser which asks the question:

"Would Banning Firearms Reduce Murder and Suicide? A Review of International and Some Domestic Evidence"

In short, their answer is "no". From the abstract:

"This article examines a broad range of international data that bear on two distinct but interrelated questions: first, whether widespread firearm access is an important contributing factor in murder and/or suicide, and second, whether the introduction of laws that restrict general access to firearms has been successful in reducing violent crime, homicide or suicide. Our conclusion from the available data is that suicide, murder and violent crime rates are determined by basic social, economic and/or cultural factors with the availability of any particular one of the world's myriad deadly instrument being irrelevant."

In the section on suicide rates, some points the authors discussed include:

* In the absence of firearms, suicidal people simply substitute other means.

* In the 1980s, suicide among teenagers and young adults spiked in the U.S., and many blamed firearm availability for the increase. What they failed to mention was that suicide among young adults was rising throughout the developed world--regardless of gun availability--and in many places was rising far faster than in the U.S.

* Among English youth, for example, suicide increased 10 times as fast as among American youth, yet the preferred method of suicide there was car exhaust asphyxiation.
7.1.2008 7:54pm
Modus Ponens:
Aren't you the same clever people who spent much of the past 5 years telling me that anthropomorphic climate change was a nullity?
7.1.2008 8:01pm
great unknown (mail):
And in Japan, the current rage is suicide via self-produced poison gases, which can and has caused considerable collateral damage. Japan, of course, is gun-free.
7.1.2008 8:04pm
Curt Fischer:

Curt Fisher is not performing correct statistics. Probabilities and odds ratios have binomial distributions, not normal (Gaussian) distributions. The method he used for calculating the confidence intervals for ratios of odds ratios also is incorrect. An F test is needed. It is possible to do this with the data given, but I'm not interested enough in the study to do the work.


I much appreciate the reply Dr. T. I was sure my analysis was full of errors, but I figured it would be better to put it out there and get corrected than to not report it at all.

The use of the F test sounds interesting to me, but how can the odds ratios follow a discrete distribution like the binomial? Aren't odds and probabilities both continuous variables?
7.1.2008 8:14pm
AnonLawStudent:
An interesting tidbit for those who would use this study to support "reasonable" gun control regulations. From p935:

There were no significant differences between those with
only handguns in the home and those with only long guns or
both handguns and long guns, those with two or more guns,
and those having one gun in the household; and between
those who stored one or more guns unlocked and those who
stored all guns locked.
7.1.2008 8:14pm
style, not substance:
It's refreshing to see the use a proper object after "cited to." Thanks for not writing, "which I learned about because a friend cited to it." Like nails on a chalkboard.
7.1.2008 8:33pm
frankcross (mail):
EV, these criticisms make sense for homicides but not for suicides. I'm not sure criminals are more likely to commit suicide (though there may be other omitted variables) and the odds ratio is so extraordinarily high that it is hard to explain away.

The Harvard article is a very biased analysis of suicides that does not appear reliable. I haven't reread, but my recollection is that it omitted a great deal of the social scientific literature.

Ultimately, though, I'm intrigued by these analyses of the data. People always seem to take the position on a matter of fact that suits their position as a matter of political preference (both liberals and conservatives). That doesn't seem right.
7.1.2008 8:42pm
Dennis Nicholls (mail):
I've never been sure what these studies are purported to mean in respect to constitutional rights....

If we came up with a study that "proved" women voting caused lots of social problems, could we simply ignore the 19th Amd.? If we came up with a study that "proved" that black people committed a disproportionate share of all violent crime, could we just send them off to a concentration camp and ignore the 14th Amd.? Scalia nailed it....a constitutional right removes certain policy decisions from the table.
7.1.2008 8:55pm
Toby:

Aren't you the same clever people who spent much of the past 5 years telling me that anthropomorphic climate change was a nullity?

Usually I hate the language police (perhaps because my own postings are frequently their fitting subject), but I must admit I am intrigued by what anthropomorphic climate change is and what it must look like...
7.1.2008 9:16pm
Eugene Volokh (www):
Frank Cross: I'm sorry my link under "extra risk of suicide" wasn't quite right; I've corrected it, and it now links to a study that reports -- quite as an aside, so there seems to be no axe to grind there -- that 40% of the suicide victims in the study (all South Carolina suicides in 2004, if I read this right) had criminal arrest records. I've seen similar statistics elsewhere; criminal record is highly correlated with suicide risk.
7.1.2008 9:53pm
H Bowman, MD:
The involvement (including demise) of 'criminals' with illegally possessed firearms is irrelevant to non-criminals with lawfully possesed means to defend themselves.

Frankly, I want my family to be able to defend itself against a criminal.
7.1.2008 9:53pm
TDPerkins (mail):
Toby--

An increased incidence of people shaped clouds.

TDP, ml, msl, &pfpp
7.1.2008 9:56pm
Nunzio:
Given the number of police officers in this country, is there a study about how many accidental firearm deaths there are in police-officer homes?
7.1.2008 9:57pm
JK:
AnonLawStudent, that could very likely be due simply to a small n for each of those sub-categories. It could also very likely be the case that there is no correlation, but you're implication is right out of "how to lie with statistics."
7.1.2008 10:11pm
DeezRightWingNutz:

The basic premise of your argument is that an individuals decision on whether to own a gun or not will have no affect on whether they one day find themselves being considered a criminal. But of course that's not true at all, as you point out.


No. He's saying gun ownership may be correlated with criminality, not that it causes it. So if being a criminal causes one to get shot, and being a criminal is correlated with having a gun, you need to control for criminality.

Let's say I want to determine if driving in snow increases traffic accidents. Well, I'd better control for the driving records of people who drive in blizzards. Maybe the risk takers who get into accidents in good weather are the only ones willing drive during blizzards. Maybe not. Either way, you need to control for driving record, just like you need to control for criminality.
7.1.2008 10:37pm
Doc W (mail):
Curt Fischer says (at 5:26):

"Secondly, Prof. Volokh's criticism seems overwrought, because the study itself says in its conclusions the following:

The
gun in the home may not have been the gun used in the death.
This possibility seems less likely with suicide, but, with
homicide, it is certainly plausible that someone brought a
gun into the home. [...] It is possible that the association between a gun in
the home and risk of a violent death may be related to other
factors that we were unable to control for in our analysis. For
instance, with homicide, the association may be related to
certain neighborhood characteristics or the decedent's
previous involvement in other violent or illegal behaviors.
Persons living in high-crime neighborhoods or involved in
illegal behaviors may acquire a gun for protection. The risk
comes not necessarily from the presence of the gun in the
house but from these types of environmental factors and
exposures.

It seems like Prof. Volokh's criticism is squarely addressed by these remarks. Of course, I'm not sure why the study authors were "unable to control" for some of those factors in the manner I suggested above."

THE PROBLEM WITH THIS DEFENSE IS THAT IT LEAVES DAHLBERG ET AT WITH NO BASIS FOR THE FOLLOWING CLAIM, WHICH THEY MAKE ANYWAY:

"Simply having a gun in the home increased the risk of a
firearm homicide or firearm suicide in the home."

AND RE GERG AT 6:48--THE OLD "GUN SICKNESS THEORY" IS SURELY THE LAST REFUGE OF ANTI-GUN PROPAGANDISTS. WHEN CONFRONTED WITH THE LIKELIHOOD THAT VIOLENT CRIMINALS ARE MORE LIKELY TO GET GUNS AND MORE LIKELY TO GET SHOT (AND THAT'S WHAT EXPLAINS ANY CORRELATION BETWEEN THE TWO), FLOAT THE NOTION THAT JUST HAVING A GUN TURNS PEOPLE INTO CRIMINALS. PATHETIC.
7.1.2008 10:52pm
Curt Fischer:

THE PROBLEM WITH THIS DEFENSE IS THAT IT LEAVES DAHLBERG ET AT WITH NO BASIS FOR THE FOLLOWING CLAIM, WHICH THEY MAKE ANYWAY:

"Simply having a gun in the home increased the risk of a
firearm homicide or firearm suicide in the home."



I take it your problem with this statement is its implication that having a gun in the home causes increased risk of firearm homicide. Causation is unsupported by the data -- if that's your beef, then I agree with you.

This kind of language appears all too frequently in correlative studies such as this one. But in my estimation it is not specific to studies on gun deaths, nor is it indicative of bias against guns on the authors' part. I just think its a trap that many statistical researchers fall into, to make their conclusions sound more important.
7.1.2008 11:05pm
AnonLawStudent:

AnonLawStudent, that could very likely be due simply to a small n for each of those sub-categories. It could also very likely be the case that there is no correlation, but you're implication is right out of "how to lie with statistics."

The "no significant difference" certainly does have implications for the maximum value of such a correlation. The exact figure depends on how the study was powered. Unfortunately, Dahlberg et al never discuss statistical power - or conscious design to achieve a specified power - a common and long-critiqued defect in epidemiological studies. Cf. Bailar &Mosteller, Guidelines for Statistical Reporting in Articles for Medical Journals, 108 ANN. INTERN. MED. 266, 266-67 (1988). If the defect is based on the numerosity of subcategories combined with failure to adjust for multiple hypotheses testing, the study displays even greater statistical sin. See, e.g., Newman v. Motorola, Inc., 218 F.Supp.2d 769, 779 (D. Md. 2002) ("[W]hen there is a high number of subgroup comparisons, at least some will show a statistical significance by chance alone.").
7.1.2008 11:57pm
Doc W (mail):
Curt:

Yes that's my beef. Not indicative of bias on the authors' part? Come on, they didn't fall off the proverbial turnip cart. Everybody knows that the news media and pundits pick up on statements like this and trumpet them as proof that guns cause crime.

And actually, if the authors don't have a good case for causation, if they don't even have a methodology that could possibly establish causation, then why are they (as conscientious scientists) publishing a paper on this at all? Why would a journal accept it? Anybody can plot regressions of anything against anything else. So what? Papers like this have been getting published on a regular basis for over 20 years (remember Seattle vs Vancouver?) and they don't do anything except stir the pot for gun control propagandists.
7.1.2008 11:59pm
Curt Fischer:

Not indicative of bias on the authors' part? Come on, they didn't fall off the proverbial turnip cart. Everybody knows that the news media and pundits pick up on statements like this


I never said the authors had no bias. I said the authors misleading wording was not necessarily indicative of bias against guns.

I think that in general, authors' bias is likely to be towards writing articles that pique the interest of their readers. Researchers like to get famous and writing articles that are widely discussed is one way to do it.

This is a trend/problem of overhyping one's work transcends the issue of gun policies--it's something you'll find in all fields of research.
7.2.2008 12:37am
theobromophile (www):
And actually, if the authors don't have a good case for causation, if they don't even have a methodology that could possibly establish causation, then why are they (as conscientious scientists) publishing a paper on this at all? Why would a journal accept it?

Semi-related thought: that's probably why EV takes the time to go through the methodological flaws, even if those flaws are stated in the study. As posters above have noted, people will read some of the study and either accept the conclusions of the authors, which are unsupported by the evidence, or take it at face value but not think about how the design flaws will change the result.

It is tough, when you hear about a study with a design flaw, to explain to people exactly what it is. EV took about 250 words, a few links, and a few random statistics (that he may or may not have needed to look up) to discuss only one of the design flaws (albeit the major one).

In the ideal situation, studies with such severe design flaws wouldn't be performed anyway, or would be reported only if the results were so overwhelming that it would appear as if no variable could account for those. How best to bring about that result? Perhaps by critiquing studies with obvious design flaws, even if those flaws are mentioned in a part of the paper?
7.2.2008 12:42am
Tony Tutins (mail):
I thought the CDC was denied funding for their studies of the microbe theory of homicide. To me, the highwater mark of the Kellermann et al. studies was this 1993 one, which showed that renters were three times more likely than gun owners to be homicide victims. Yet Kellermann et al. did not trumpet this conclusion, nor was their work cited by the Tenancy Policy Center or the Realtors of America. In contrast, this 2004 study, (which used 1993 data -- what's up with that? Certainly memories fade after a decade), cherry picked their variables of interest, not analyzing any control variables, such as education, religious practice, household income, political affiliation, renting vs. buying, dog ownership, etc.

The case-control methodology is unique to mortality studies. Basically, you take a dead guy, and compare him to a not-dead guy who lives nearby, but not next-door. If you are sincerely trying to track the cause of death, you look for all differences -- lead paint, empty bottles of vodka, molds, McDonald's wrappers vs. Whole Foods reusable bags, bottles of steak sauce vs. bottles of fat-free vinaigrette, worn-out running shoes vs. video games, etc. etc. You do not make a priori assumptions about what could cause the deaths, or at least you don't zero in on your pet hates. From the newspapers, drug dealing is highly correlated with homicide -- but these anti-gun studies are not motivated to analyze such data.
7.2.2008 12:46am
Gerg:

Chainsaws will hurt you sometimes (ask any firefighter or logger) but they also allow increased productivity that seems to overwhelm the danger. Chainsaws aren't "necessary," the lumber to rebuild Chicago after the Great Fire was cut by hand tools but they are desirable.


Sure but this is precisely where youre intuition is letting down. Home invasions are scary so the mind inflates the actual risk of them which is virtually none. Deaths or injuries from your own gun are quite common.

In the analogy above it would be kind of like suggesting that because chain saws are useful to firefighters that everyone should be issued a chainsaw at the age of 10. The likelihood of a random 11-year-old being a firefighter is so small to the likelihood of him hurting himself or someone else that the net will not be positive
7.2.2008 1:51am
cboldt (mail):
-- Home invasions are scary so the mind inflates the actual risk of them which is virtually none. Deaths or injuries from your own gun are quite common. --

.

"Quite common?" Are you Arthur Kellerman?

.

Guns are scary and some minds inflate the actual risk of them.
7.2.2008 7:00am
Ryan Waxx (mail):

The level of outrage I've seen so far on the thread is out of proportion to the severity of the study's flaws, at least as those flaws have been discussed here thus far.


The level of outrage is mostly directed at the fact that the authors seemingly didn't even attempt to get it right.

The outrage is also directed at the multiple groups of people who decide that such sloppy studies nonetheless tell the story they want to hear, and therefore welcome them into the mainstream, much like the outrage that would happen if someone was dumping partly-treated sludge into the collective water supply.
7.2.2008 8:39am
Happyshooter:
The Academy accepted fraud that was obvious in the case of "Arming America" and even gave the fraudster professor many awards. It was because he gave them the anti-gun position they wanted.

The fraud in that case was so great that untrained readers spotted the lies. It didn't matter. It was only when the average NRA member started openly laughing at the profs that they took action.

Of course the Academy is going to accept and embrace a study with an anti-gun result where there is just a method to argue over and not a glaring lie.
7.2.2008 8:44am
Ryan Waxx (mail):
Sure but this is precisely where youre intuition is letting down. Home invasions are scary so the mind inflates the actual risk of them which is virtually none. Deaths or injuries from your own gun are quite common.


Perhaps you aren't aware that home invasions (while the occupant is home) are unusually rare in the U.S. as compared to similar countries, and that the most likely reason is precisely because the homeowner may be armed.

It seems rather perverse for you to be citing the results of the deterrent effect of widespread firearm ownership as a point against their ownership.
7.2.2008 8:45am
Happyshooter:
but I must admit I am intrigued by what anthropomorphic climate change is and what it must look like...

Mother Gaia is a living being. The use of coal, oil, nuclear, and hydro power interferes wih her love force towards all living things on the planet.

In reponse, she is punishing humans by raising the temps, killing the polar bears, and melting the ice caps to drown the evil coastal cities and force humans to move into flyover county and live with a simple roof and wood stove.

Once most of the humans so move and start living with only a wood stove and manual farming tools, she will be happy again and our lords and masters can live in peace in Boston, New York, LA, and Washington, making decisions that Mother Gaia approves of and that we masses carry out.
7.2.2008 8:51am
corneille1640 (mail):

And in Japan, the current rage is suicide via self-produced poison gases, which can and has caused considerable collateral damage. Japan, of course, is gun-free.

Assuming that Japan is entirely gun-free and that the suicide rate is not zero, then something other than guns is likely to be a common means of suicide. That fact, by itself, does not necessarily prove that owning a gun or having ready access to one does not increase the chances of a successful suicide.
7.2.2008 9:38am
Mark Buehner (mail):

Home invasions are scary so the mind inflates the actual risk of them which is virtually none. Deaths or injuries from your own gun are quite common.
There are about 300 million guns in America. There are maybe 1200 people killed a year in gun accidents.

There are about 140 million cars in America. About 42,000 people are killed in auto accidents a year.

Heck, around 4000 people die in motorcycle accidents every year.

The math is not hard. You might want to keep thinking about inflated vs actual risk with this in mind.
7.2.2008 9:53am
byomtov (mail):
Incredible! They interviewed the family members! They could have easily found this out by asking the very people they interviewed (who would know).

The authors only interviewed the kin (or kith) of 83% of the decedents in the study. For the other 17% they relied only on the death certificate, which presumably does not indicate ownership of any firearm(s) used in the homicides.

Read the paper. The authors didn't interview anybody.

They used existing data, the National Mortality Followback Survey of 1993, compiled by the CDC. Since the CDC didn't ask about criminal records the authors couldn't include them in their analysis. So drop the outrage.
7.2.2008 10:22am
byomtov (mail):
Incredible! They interviewed the family members! They could have easily found this out by asking the very people they interviewed (who would know).

The authors only interviewed the kin (or kith) of 83% of the decedents in the study. For the other 17% they relied only on the death certificate, which presumably does not indicate ownership of any firearm(s) used in the homicides.

Read the paper. The authors didn't interview anybody.

They used existing data, the National Mortality Followback Survey of 1993, compiled by the CDC. Since the CDC didn't ask about criminal records the authors couldn't include them in their analysis. So drop the outrage.
7.2.2008 10:23am
Curt Fischer:

The authors didn't interview anybody.


Good catch byomtov.

I'd like to note that the quoting in your post at 9:23am was a very confusing. The second paragraph [...kin (or kith)...] was me, replying to the first paragraph [Incredible!...], which was written by 30yearprof. The goal of my post was in fact to temper the outrage -- so I have no outrage to drop. Sorry.
7.2.2008 10:29am
FAR (mail):
As a scientist I am frequently outraged that a refereed journal will accept for publication a paper that is missing information that may have changed the authors conclusion. This is sloppy work and only proves the bias of the editors. This is a frequent occurrence in the medical literature.
7.2.2008 10:42am
byomtov (mail):
As a scientist I am frequently outraged that a refereed journal will accept for publication a paper that is missing information that may have changed the authors conclusion. This is sloppy work and only proves the bias of the editors.

Few papers have all the possibly relevant data, especially when they are not reporting laboratory studies. When dealing with questions of social science you often have to rely on what's available.

That's what they did. And they carefully noted a number of things, including criminal history, that might have changed the results but that they had no information about.

So it's fair to discuss how those things might have affected the results, but the accusations of sloppiness, bad faith, and dishonesty are simply unjustified.
7.2.2008 11:44am
James Gibson (mail):
They used existing data, the National Mortality Followback Survey of 1993, compiled by the CDC.

God it took forever for someone to notice. This new study was actual published in 2004, using data generated in 1993 at the start of the Clinton, I'll pay for anti-gun studies. Like Justice Breyer, who was used the crime statistics for the early 1990s (at the end of the crack wars), want the right conclusion for your study- select your data to fit the result.

Regarding suicides in other countries, our gun owning nation has essentially the same total suicide rate as England, Canada, and Australia (we actually are between England and Canada). Japan has a rate 50% higher then ours, yet there are no gun related.

As for accidental deaths of Police, thank the Clintons for the answer. In the 1990s, to show the harm guns do to Police, the FBI was required to document officers deaths by all means and circumstances. The Brady people hoped to show that guns were the primary cause of deaths. They ended up showing that vehicle accidents (either being struck by or operating) tend to be the primary cause of death among officers. http://www.fbi.gov/ucr/ucr.htm
7.2.2008 12:10pm
Charlie (Colorado) (mail):

The use of the F test sounds interesting to me, but how can the odds ratios follow a discrete distribution like the binomial? Aren't odds and probabilities both continuous variables?


Curt, I don't see an answer to this one, so let me just say "no, not necessarily." Consider the simplest example: Let's say that you toss a fair coin ten times: the odds are 50/50 that you'll have more than five heads; the probability that you'll have five and a half heads is zero. There are 1024 discrete possibilities, and 5.5/4.5 is not one of them.
7.2.2008 12:19pm
josh:
Don't criticisms of these studies actually go to "weight" rather than "admissibility"? I find it hard to believe Prof. V is arguing that their findings are not newsworthy, and, like most news, should be afforded any (or no) credibility as the reader sees fit. I suppose a fair criticism of the reporting is that they don't include the flaws, but I find most good reporters usually tend to quote critics of such studies, who convey those issues for the reader.
7.2.2008 12:26pm
The Ace (mail):
Home invasions are scary so the mind inflates the actual risk of them which is virtually none.

Really? None?
Is this because you say so? Tell that to this guy:

Nearly a year after a deadly home invasion in Cheshire outraged the state and made national headlines, the lone survivor talked with Eyewitness News about how he is keeping his family's memory alive.

Dr. William Petit gave a heartfelt "thank you" to more than 200 people that spent Monday afternoon and evening at the second annual Petit Family Foundation Golf Tournament.
The foundation was formed soon after Petit's wife, Jennifer Hawk-Petit and two daughters, Hayley and Michaela, were killed in a home invasion.



Deaths or injuries from your own gun are quite common.

Is this because you say so? Because you don't have a shred of evidence supporting such a silly statement.
7.2.2008 12:58pm
Curt Fischer:

Consider the simplest example: Let's say that you toss a fair coin ten times: the odds are 50/50 that you'll have more than five heads; the probability that you'll have five and a half heads is zero. There are 1024 discrete possibilities, and 5.5/4.5 is not one of them.


Thanks for the reply, and I agree that in your example the possibilities are discrete. (Although the chances of more than five heads should be less than 50/50 for a fair coin).

But I think that perhaps we mean separate things by "probability" and "odds", at least in this case. The difference is that you are assuming a certain type of coin (a fair coin with 50-50 odds per flip, say), and looking forward ("prospectively") to calculate the likelihood of different discrete outcomes given your knowledge of the coin.

This study is "retrospective". To paraphrase it into a coin example, the researchers knew how many heads and tails they got, and from this (without assuming anything about the coin in advance) tried to calculate the probability that the coin was fair.

Since no fundamental law of nature precludes the construction of coins that have 50.1-to-49.9 odds, or 50.0000001 to 49.99999999 odds, it only makes sense to treat the probability that the coin was fair as a continuous variable.
7.2.2008 3:06pm
Tony Tutins (mail):

When dealing with questions of social science you often have to rely on what's available.

This article was published in the American Journal of Epidemiology. Thus it purported to be a medical study, not a social science one. The peer reviewers included medical doctors and masters of public health.


That's what they did. And they carefully noted a number of things, including criminal history, that might have changed the results but that they had no information about.

So it's fair to discuss how those things might have affected the results, but the accusations of sloppiness, bad faith, and dishonesty are simply unjustified.

Publishing this article adds nothing to mankind's stock of knowledge. This publication follows a lot of similar question-begging studies of criminal victimization published in medical journals by three researchers: Kellermann and Wintemute, MDs, and the economist Hemenway, PhD. Johns Hopkins, the publisher of the American Journal of Epidemiology, has a "Gun Violence" center. One hopes that criminologist Gary Kleck will similarly operate in someone else's sandbox, and someday publish a study of the spread of AIDS.
7.2.2008 3:25pm
byomtov (mail):
it purported to be a medical study, not a social science one.

Call it what you want. The point is that it was clearly and explicitly based on an existing dataset. There was no deception here. No "purporting."

Publishing this article adds nothing to mankind's stock of knowledge.

Of which you have a complete understanding? Of course it added a bit. No one's claiming it's Nobel-worthy. It investigated some data and pointed out some possible flaws in its results. It's interesting how inflamed commenters, and EV, are over this article. Could it be that they just don't like the results, however qualified?

One more thing: It's truly rich for people who publish in law reviews to criticize the standards of other academic journals.
7.2.2008 5:05pm
FAR (mail):
"So it's fair to discuss how those things might have affected the results, but the accusations of sloppiness, bad faith, and dishonesty are simply unjustified."

A someone who teaches at a school of public health, I think my comments are accurate, especially on gun control and lead in the environment...when it can be used to shut down ranges.

They are supposed to be real scientists not social "scientists."
7.2.2008 5:17pm
Tony Tutins (mail):

Could it be that they just don't like the results, however qualified?

Lawyers are trained to scrutinize every link in the chain of causation. Causation is missing entirely here. The basis for all such studies is the thought that "What if the 'epidemic of gun violence' was more than a metaphor?" Bacteria lack volition, but the guy pulling a trigger doesn't. Dietary fats do not decide to plug up your arteries. Further, inanimate objects like guns cannot control the minds of human beings. Belief in sympathetic magic has its place, I suppose, but I expect people with doctorates to think more sophisticatedly than Cargo Cultists.
7.2.2008 11:51pm
Snowdog99 (mail):
Epidemiological studies are 99% horsesh*t for many of the very reasons Eugene cites in his post - that is (in essence), that the chain of causation is difficult if not impossible to accurately characterize empirically.

It reminds me of a "study" done a few years ago, whereupon the researchers found that young teenage women sporting tattoos were more likely to give birth out-of-wedlock. The media dutifully interpreted these results to mean: "tattoos cause unplanned pregnancy."
7.3.2008 3:30pm