[Part I; Part III]

Yesterday I explored a new report on gun control from the Center for American Progress. Today I’ll be looking at an academic paper (ungated draft here) by Gary Kleck, Tomislav Kovandzic, and Jon Bellows that uses very different methods and reaches different conclusions. Kleck, the lead author and a criminologist, has been well-known in the gun debate for decades; he’s often considered pro-gun, but he has a respectable willingness to anger “his” side, for example by supporting universal background checks.

Of the three studies I’ll be looking at in this series, this is arguably the most complicated. Instead of looking at the 50 states, it looks at every city in America with a population of 25,000 or more—1,078 in all. Instead of collapsing all gun-control laws onto a single scale, it separately measures the effects of 19 specific laws. And instead of just comparing raw crime rates, the authors build elaborate models that account for demographic and cultural differences between the cities, and that also include gun ownership so we can be sure the results stem from the laws themselves, rather than just reflecting the fact that states with strong gun cultures tend to have less gun control.

The key findings:

  • As Kleck and co-authors have found in other studies as well, gun ownership by itself does not cause higher crime rates. This finding relies on some arcane statistics, unfortunately—the authors measure gun ownership through a “proxy” (the percentage of suicides committed with guns), and in addition employ a technique called “instrumental variables” to separate correlation from causation.
  • Most gun-control laws don’t work. There’s even some (weak) evidence that a few of them increase crime.
  • A couple laws do seem to reduce crime, however—most notably, requiring a license to possess a gun, a measure strongly opposed by gun-rights groups. The authors note that such laws entail background checks on prospective gun owners.

All of that squares with my own prior views, so I’m tempted to just agree with the study and leave it at that. But in my first post I promised nitpicking, and I shall deliver. Fortunately, nitpicking this study is easy to do, because the authors themselves are very thorough in explaining its limitations.

If this is the most complicated study we’ll be looking at this week, it’s also the ugly duckling in some ways. Perhaps the oddest thing about it is that it relies on data more than a quarter-century old, from 1990. The authors do this because good suicide data (which they used to estimate gun ownership) are not available at the city level in more recent years.

A lot has changed since 1990, though; crime has plummeted, and we’ve seen developments in federal law (such as the requirement for all licensed gun dealers to perform background checks and the creation of a system for performing those checks). It’s hard to say for sure how well the findings translate to the 2016 context. And by using a single year of data, the authors miss an opportunity to compare crime rates before and after the laws at issue went into effect, though in one of their models they do control for the 1980 crime rates of these cities, a somewhat similar idea.

Also, many of the state and local laws studied here, such as those keeping guns from criminals, minors, and the mentally ill, were similar to federal law even in 1990. To be fair, the authors note that local controls are often stricter than what the feds require—they might cover misdemeanants in addition to felons, or set the cutoff for “minor” higher—but when the study finds these laws to be ineffective, that means they’re ineffective in relation to the federal baseline, not to, say, the complete absence of any law against criminals buying guns.

My final note is that some of the variables are somewhat imprecisely measured. Despite going back to 1990 in search of usable suicide numbers with which to estimate gun ownership, for example, they report that “it was not possible to compute city-level measures of [their gun-ownership proxy] for most of our cities.” Instead, they used county-level data when necessary. And for cities missing crime data, they “estimated missing values by computing the average crime rate for that year in cities in the same census region and same population group.”

It’s not the authors’ fault that the data are less than ideal, but readers should bear these issues in mind. Nonetheless, this is an important contribution to the gun-control debate—suggesting that background checks, rather than other forms of gun control, should be a focus for those who want to reduce violence. But tomorrow, I’ll take a look at some research saying even background checks aren’t effective.

Robert VerBruggen is managing editor of The American Conservative.