The Brady Campaign to prevent gun violence has compiled a scorecard for each state, according to how they (the Bradyites) view each state’s gun laws. These statistics are through 2009.
More restrictive states received higher scores. Utah received the best—or worst, depending on your perspective—score (0), and California received the worst (79). The score had several components, such as whether laws were in place to forbid bulk purchases, and whether there were “assault” weapons bans. Complete details are available at their site.
The Guardian, doing the job American papers have forgotten to do, and using the Fed’s data, compiled each state’s total murders, and the number of those that were perpetrated with firearms or by other means. Firearms murders were separated by handguns, shotguns, rifles, and unknown types; the other means included knives, hands and feet, or unknown. These statistics were from 2008.
I then went to the Gallup organization and retrieved the results from two polls: the Democrat minus Republican voting percent intentions (numbers greater than 0 indicated a Democrat advantage, etc.), and the percent population self-identifying as “conservative” (anywhere from 29-49%; other possibilities were “liberal”, “moderate”, and “none of your business”). These were from 2009.
The caveats first! The Fed’s murder and population data is pretty solid, enough so that we can assume measurement error is not especially important. The Brady scorecard is obviously arbitrary. And unidimensional: two different states can have equal scores but they might differ in the composition of its gun laws. Comparing unidimensional scores is always dangerous!
The Gallup polls are not as arbitrary, but they have all the usual flaws of surveys. Stated rates of plus/minus error are in the four-point range, but all experience suggests that it’s far safer to double these.
Worst of all, statistics at state levels ignore the difference between the urban and rural. New York is the starkest example: New York City is night and Upstate is day, yet they are lumped together as one. But, given all these distinct—and very real—possibilities for misinterpretation, here are some pictures.
Here’s a shot of the percent identifying as conservative versus the Brady score. It shows what we’d guess: except for Vermont, where the people have some catching up to do, legislatively speaking. The blue “liberal” dots represent those states that have fewer than the median 39% self-identified conservatives; the red dots have 39% or more. (The picture is similar for Dem-Rep voting intentions.)
The previous picture informed this one, the real meat. (Download PDF.)
It shows how the Brady State Gun Law Scorecard is related to the handgun murder rate (per 100,000). I specifically excluded shotguns and other types of firearms from this picture. Handguns account for anywhere from 11% to 99% of firearm murders, with a median of 64%. They account for 0% to 78% of all murders, with a median of 39%. (Did you think it would be more?)
In absence of the dot-coloring, the data show two chunks. Those with Brady scores from 0 to about 20, and those with scores greater than 40. The low-Brady-score chunk shows very little relation to the handgun murder rate. That is, the whole range of murder rates are found in the low-Brady score chunk.
But the high-Brady-score chunk shows a rough correlation (I use that word in its plain English sense): higher Brady scores are associated with higher murder rates. Why? The data do not say.
The coloring of “conservative” and “liberal” might be helpful, since this division was so predictive of the Brady score. The blue/liberal dots show the same positive association: in general, the higher the Brady score—that is, the harder it is for citizens to own guns legally—the higher the handgun murder rate.
But the red/conservative dots show no such association. This does not prove a gun law/murder rate association doesn’t exist, though. It’s easy to be fooled in cases when a score has a cutoff, like the Brady score does at 0. Don’t forget: the Brady score’s intention is to measure restriction, not freedom.
There might be a lot of difference between states with a score of 0 and those with a score of 2 (the next lowest), but we can’t see it. What we’d need is an extension to the Brady score which awards “negative points” for states which have even fewer restrictions than are noted by the Brady organization.
Overall: there’s some indication that more restrictive gun laws (as measured by the Brady score) are associated with higher handgun murder rates. We cannot say, with this data, why this is so.
Stop right there! Any kind of formal statistical analysis of this data would be extremely unwise. Especially one which assumed a linear relationship in the data. Any formal analysis is practically guaranteed to instill more confidence than is warranted. Remember those caveats? There’s no way to incorporate them reasonably into any formal model. We’ll have to settle for vagueness. That’s life.