[Part II; Part III]

Recently there have been some new analyses of gun control and its effects on violence: one from the Center for American Progress (CAP), one from a two criminologists and a judge (free version here), and one from the economist John R. Lott in his book The War on Guns. Over the next three days I’ll be explaining and nitpicking them in a trio of blog posts.

Up today is CAP, whose new report received a fawning New York Times mention. The piece paraphrases an anti-gun public-health researcher, Daniel Webster of Johns Hopkins, as saying the research is “scientifically sound”; for a response, the reporter went to the NRA instead of a qualified researcher with pro-gun sympathies. Even a halfway skeptical anti-gun researcher might have raised some concerns: Garen Wintemute of UC-Davis has criticized studies similar to CAP’s in the past.

Anyhow, this chart represents the thrust of the report:

capgraph

Basically, states with stricter gun-control laws have less gun violence. This is an interesting exercise, but it fails to address any of the myriad problems that plague gun research.

First of all, the technical-sounding “Gun Violence Index” is really just a measure of gun deaths—the overall gun-death rate is actually one of the index’s ten components, and all but one of the others are just specific types of gun deaths (gun suicides, gun homicides, gun accidents, police shootings, etc.). The overall gun-death rate correlates so strongly with the Gun Violence Index that the two are nearly identical. My chart, their numbers:

gvi

I’m pointing this out because there are two major problems with studying “gun deaths.” The biggest is that we shouldn’t care about reducing gun deaths; we should care about reducing total deaths. If we reduce gun deaths, but people simply commit violence with other weapons instead, we’re just spinning our wheels—and yet studies of gun deaths will say we’re making progress.

The other major problem is that gun deaths comprise two very different phenomena. About two-thirds are suicides, with most of the rest being homicides (plus some accidents). Suicide and homicide are completely different things, often targeting the opposite demographic groups (old vs. young, rural vs. urban, white vs. black). They also raise different political questions, as many of us hesitate to restrict people’s freedom for their own good. Suicide and homicide need to be studied separately, even though they both involve guns.

We can see the importance of this second point in CAP’s own data. Most strikingly, the report’s central claim—that states with stricter gun laws have less gun violence—does not hold for homicide, which is what most people think of when they hear “gun violence.” Not that most readers could tell that from the report: the finding is relegated to a footnote, and even there the authors fail to note that the correlation is far too weak to be considered statistically significant.

(Technically speaking, it’s 0.13, with 50 observations, one for each state. That translates to a “p-value” of 0.37, far above the traditional 0.05 cutoff. If you generated the data at random, there would be more than a one-in-three chance of getting a correlation this strong.)

None of this is necessarily to deny that weak gun laws or high gun ownership—CAP’s report makes no attempt to distinguish between the two*—can have bad consequences. They do seem to correlate with suicide rates, as well as some specific types of homicide, including police and intimate-partner violence. Even when it comes to overall homicide rates, some more advanced studies (though far from all) argue there’s a correlation there once the data have been adjusted to account for differences in demographics and culture.

Of course, such adjustments are crucial if we are trying to discover causation, and not mere correlation. But CAP’s analysis makes no attempt to perform them. And even if we could definitively say that strict gun laws reduce violence, the analysis would say nothing about which gun laws in particular help, because they’re all compressed into a single scale.

But don’t despair. Over the next two days, I’ll look at two other new studies that are a bit more intricate.

Robert VerBruggen is managing editor of The American Conservative. 


*I suspect that to a large extent, strong gun laws simply reflect the fact that the people of a state don’t own many guns, and thus don’t oppose such laws. I can’t report an actual correlation between CAP’s gun-law measure and gun ownership, though, because the full gun-law data are not provided in the report and I have been unable to get them. They come from the Law Center to Prevent Gun Violence; the lead author of the CAP report told me she doesn’t have permission to share them, and several emails and a voicemail to the Law Center have gone unanswered.

For whatever it’s worth, though, I did try to digitally extract the data from CAP’s scatterplot and match them to the correct states. It’s not perfect, but my correlations between gun-law weakness and CAP’s gun-violence indicators are very close to the ones in CAP’s footnote, and my gun-law measure correlates strongly (0.8) with a 2004 gun-ownership survey.