December 27, 2012

"[W]e estimate the elasticity of homicide with respect to gun prevalence as between +0.1 and + 0.3."

Some numbers.

32 comments:

Paul said...

But I see NOTHING in the study on legal use of guns to defend the same families in their homes nor on the street.

So the study stinks. Oh, and even one gun, with ammo, cost more than $180! Were they taking about the cost of the guns themselves as 'social cost' (yes a joke but considering the cost of a AR-15 now days and a cheap Bryco .25, well the range of social cost just about matches the cost of the guns.

kentuckyliz said...

I work with a guy who successfully defended his home, wife, and young children from an armed home invasion. This wouldn't show up in their statistics.

Kevin said...

I posted this on Tyler's blog:

This is a 2006 paper by Philip J. Cook and Jens Ludwig, well known anti-gun "researchers". Their research was funded by the Joyce Foundation, which has long funded gun control "studies" and other efforts:
http://en.wikipedia.org/wiki/Joyce_Foundation#Gun_violence_prevention_and_gun_control

A 1999 paper by the National Center for Policy Analysis found a considerable benefit to firearms ownership:
http://www.ncpa.org/pub/st223?pg=6

Quote:
The total social benefit from defensive gun uses is equal to the annual savings from non-deadly gun uses plus the annual savings from defensive gun uses in which the criminal is killed. And any number selected is greater than the cost of firearm violence.72 For instance, even under a best-case scenario for lawsuit proponents, using the NCVS firearms crimes number (915,000), and the Tarrance Group's defensive gun use statistic (764,000), the net benefit from defensive gun use ( the total benefits less the total costs) ranges from just over $90 million annually to $3.5 billion per year. Using Kleck's more credible estimate for annual defensive gun use, the net benefit ranges from $1 billion to $38.9 billion annually. [See Table IV.] The annual net benefit of civilian gun use in the prevention of crime is even greater if the NCVS overestimates firearms crimes and the Bureau of Justice Statistics figures (483,000) are more accurate.

dbp said...

In the paper, it looks like they simply correlate gun ownership with gun crime. It doesn't seem to occure to the authors that people who live in high crime areas might respond to that by purchasing fire arms.

A better approach would be to look at changes which happen after a major change in the law. So, compare states which tightened their gun laws with those who loosened them and use states which made no changes as a control.

My perusal of the literature has shown that there has been declining violence nation wide, but it has declined even faster in states which relaxed their gun laws.

Brent said...

As one of the commenters noted, the data presented is based only on the 200 largest counties in the US. I'm guessing that excluding rural counties (high gun ownership, low crime rate) may bias the results a bit, no?

kentuckyliz said...

I have a theory about population density and opinion on this issue. The police are 20-30 minutes from me, assuming they can respond immediately. I am well raped, robbed, kidnapped, or dead by then.

I might feel differently if I lived in a high population density area with a cop on every corner.

Why should urbanites impose their views on the flyover zone people?

Levi Starks said...

There is a problem with with gun statistics That can never be corrected.
We will never know what our nation would look like had their been no 2nd amendment.
It's only possible to say what wouldn't have happened if guns had been effectively removed from the hands of citizens. There is another side that we can never know.
We can look at oppressive governments where citizens were denied the right of self determination via firearms, But we can't know that any of those things would have happened here. But it would be a mistake to deny that they could have.

SteveR said...

Let me know when the CliffNotes come out.

Lauderdale Vet said...

Guns are used defensively, too. As a tool, this is their intended purpose, whether sctually fired or not.

"Middle estimates have estimated approximately 1 million DGU incidents in the United States" each year. (Defensive Gun Use)

Paul said...

Say.. what is the 'cost' of being dead? Is there a pricetag on that?

Home invaiders kill family? What is the cost? Freeked out father kills family, what is the cost?

And what if the killers used OTHER weapons? Baseball bats? Knives? Firebombs? Hammers?

The whole study is a stupid idea by intelectuals who know nothing about real life. It's just a political hack job.

Richard Dolan said...

Lots of talk about the correlation between suicides and guns, as well as homicides in connection with criminal activity (drug gangs, typically) and guns. As for the suicides, you would need data about the rate at which suicides would not occur but for the availability of a gun (hard to ask someone who has already committed suicide, but economists are clever sorts who might get past that) to figure out net social cost. As for felon-on-felon murders with a gun, measuring net social costs is just an exercise in deciding what counts as a 'positive' or a 'negative' cost. Not so much economics as it is an exercise in moral philosophy.

Economic studies like this can be useful in getting past the same-old talking points, by showing the correlations between social pheonmena (not the same as causation but still worth knowing). They are less useful when they simply dress up those talking-points in fancy statistical dress.

As with the AGW disputes, the stats and models used to generate the bottom-line numbers in this study are open to lots of criticisms. Given the unlikelihood that Congress will do anything about federal gun regulation this session, what with all the fiscal, tax and other problems they will have to deal with, I suspect this study will just end up on the shelf with lots of others. Think of it as part of the 'fairness to dust' campaign since even those little motes need a place to land.

J.P. said...
This comment has been removed by the author.
Moose said...

Lies, damned lies and statistics.

Kylos said...

dbp, as I read the study, the authors lagged their "gun prevalence" statistic behind the homicide rate to mitigate chicken and egg effects. I don't know if the lag was sufficient, however, since prevalence was also measured by a proxy (gun suicides) that may lag true prevalence.

Brent and Kentuckyliz are onto something by noting that rural regions are excluded from the study. I assume this was done simply because the measurements used in the study would have produced stats with too large errors for less densely populated regions. However, it seems quite likely that these regions will have much lower homicide rates relative to gun prevalence.

Another case where data conditioning to avoid anomalies in the data may have caused a noticeable change in the results was their decision to remove Oklahoma from the input data during the year of the OKC bombing. Such spikes would adversely affect a linear regression, but it seems like it should still count toward total homicides. I'm not a statistician but perhaps there is some way to average it over several years, so that the bombing is still taken into account, but without screwing up the analysis.

William said...

It doesn't take much research to prove that a gun in the hands of a madman or criminal is a bad thing. But how do you quantify the value of a gun to a law abiding citizen? Are divorce rates lower among gun owners? Are gun owners less inclined to drink big gulp sodas? If I could prove that gun owners live on the average 2.3 years longer than gun control advocates would that change anyone's opinion?

Unknown said...

The Althouse lemmings struggle with basic numeracy skills.

mikee said...

When Kellerman, et al., published their 1990's study on gun ownership being more dangerous to the gun owner than to anyone else, their study also ranked renting your home, prior domestic violence,living alone, drug use in the home, and living with a criminal as higher risk factors associated with homicide than gun ownership.

In other words, gun ownership is less likely to be associated with murder than renting.

I urge a ban on rental residences immediately.

mikee said...

Joan Peterson, Brady board member and blogger at www.commongunsense.com is famous among gun bloggers for her inability or unwillingness to recognize that when gun violence decreases and overall violent crime increases, a bad thing has happened.

I would hope that the commenters here do not fall into similar error of thought.

Emil Blatz said...

Is he safe? Is David Gregory safe?

MadisonMan said...

Gotta love a scientific paper that uses the word varmints.

Sayyid said...

Significant problems with that study.

(1) They are using suicides by firearms as a proxy for gun ownership levels. I'm not entirely sure why that's an accurate measure, let alone a more accurate measure than others. They mention that they had regressed that proxy against registered firearms and found a correlation, but if they had registered firearms already why use the proxy? Why would the suicides-by-gun be a more accurate measure of firearms prevalence then ... registered guns?

(2) Also, why isn't there any discussion in the study about what happens if they just go directly and regress registered firearms against deaths? If you're going to use a proxy variable in your model, you must tell me why it's more valid than alternatives, and one way to prove that it's more valid is to show what other related variables come up with and why that's not right. They haven't made any real effort to do that.

(3) They time-lag the suicide rate before regression. I don't see why that's necessary. They say it's because murders in the prior year could cause more gun ownership in the next year. But they're using state-level data! Does the murder rate actually change that much year-to-year on a state level?

(4) They've put %urban and %black as controls. Why not %nonwhite instead of %black? Or why not add a third control for %Hispanic? Why not add 5poor?

Remember -- what they actually found was a link between suicide by gun and homicide by gun. If we're going to have two gun-use variables as the backbone of our study, we better be damned sure to control for culture -- but they didn't!

I could go on, but suffice it to say this study is just plain shoddy.

Sam L. said...

My cost has been zero my entire life; well, other than purchase of guns/rifles/shotguns/ammo/cleaning supplies and associated shipping costs.

Never shot nobody. Never had the opportunity to. Psychic upper, that.

Had to carry at work, sometimes. Had to wear one my first married Thanksgiving. I accessorized with a Jr. Deputy Sheriff's badge. And a bandana.

kentuckyliz, I don't think there are urban areas with a cop on every corner. Pretty sure there are some that ought to. And why would urbanites want to impose their views on others? Because they can. Because they can't imagine their views not being applicable to others, and that's a kind of diversity they just hate!

Paul Brinkley said...

"estimates"
"prevalence"
"utilize"
"proxy"
"validate"
"elasticity"

I estimate the elasticity of $5 words to graduate degrees as +100 to a million.

chickelit said...

Adamant Jake, the only Althouse commenter left with both Irritable Vowel and Inflamed Bush Syndrome (based on the avatar), contributed: The Althouse lemmings struggle with basic numeracy skills

Adamant Jake is apparently impressed with elasticity -- especially sphincter elasticity.

Unknown said...

I'm honored that Dumb Cluck is fantasizing about my anus.

chickelit said...

@Jake: You seemed afflicted with epistolic closure -- fixated on Saul and not Paul.

Peristaltic closure is a secondary diagnosis because you're mostly full of it.

Michael said...

Jake Diamond. As a lefty math guy why not explain the variances alluded to in the headline. Do you view them as trivial and if not why not.

mark said...

Jake Diamond said...
The Althouse lemmings struggle with basic numeracy skills.

Curious. What level of Mathematics do you teach? And if not teach, what is the highest course in Statistics you took?

Michael K said...

"We will never know what our nation would look like had their been no 2nd amendment.
It's only possible to say what wouldn't have happened if guns had been effectively removed from the hands of citizens. There is another side that we can never know."

That is an unknowable story since this continent was settled by Europeans in modern times opposed by a hostile indigenous population. The Indians, except the Caribs, were not initially hostile but resisted the culture of the more advanced settlers. The Plains Indians were far more hostile than the eastern tribes.

The only other places where this occurred were South Africa and Australia. The Polynesian islands are another example but they were not settled by the Europeans. The Australian aborigines were not hostile but did not adopt the culture of the newcomers.

Similar events in Europe and Asia occurred so long ago that no historical record survives but they were no more peaceful, as the archeology of the Indus Valley suggests.

Tim said...

Jake Diamond said...

"The Althouse lemmings struggle with basic numeracy skills."

Ah, yes, indeed so.

Because "the average annual marginal social cost of household gun ownership is in the range $100 to $1800" is so specifically narrow as to completely validate the underlying study.

Trolls gonna troll.

Chip S. said...

A couple of brief points about the study:

1. A proxy is used b/c it's based on data that are available for more units of observation than the direct household survey measure. It's a perfectly sensible approach.

2. The author of the original study demonstrating the usefulness of the suicide-by-gun proxy for gun ownership explicitly advises against its use in "panel" studies like the one linked here. It's a good proxy for cross-sectional variation in gun prevalence, but not very good as a proxy for changes over time. Yet, panel estimation specifically exploits differences in changes over time across different units of observation for purposes of estimation.

3. Even if the suicide-by-gun proxy were valid in this application, the question of causation is not particularly well addressed in the study. Merely using a lagged value of the proxy is inferior to other widely-used techniques unless the lagged value is a highly precise instrument for the current-period variable.

4. the study does address many of the criticisms raised in this thread. In particular, it does not neglect the beneficial aspects of gun use, b/c it estimates the difference b/w the negative and positive effects. It also takes pains to sort out the degree to which guns merely substitute for other weapons vs. the extent to which they increase overall homicides.

5. If, for the sake of discussion, we gave credence to the study's findings, it appears (if I've done the back-of-the-envelope calculations right) that the bulk of the increased homicides due to more guns is concentrated among 15-19 year-olds--which is to say, likely gang members and their victims. This probably explains the wide gap in perceptions of the utility of guns among people in very different locations.

Henry said...

I wonder what the elasticity of homicide with respect to car prevalence is? Someone run the numbers.