Over at 538 today, Tom Schaller mentioned that, outside of DC, "the ten states with the lowest per-capital gun death rates voted for Obama." He referenced this chart of firearm deaths per 100,000 residents by state. If we cross-reference these data with the percent that voted for Obama in each state (here), we get a rather striking picture.
DC is a notable outlier -- it's no surprise that there are more gun deaths there than anywhere else. But if we look at only the 50 states, the correlation between gun deaths and the percent who voted for Obama is a strikingly high -.63, (p<.001). If we graph only the 50 states, it's readily apparent that the relationship is quite strong.
The data are almost too perfect -- there don't even appear to be any real outliers. Indeed, if we look at the original chart in a new way -- color coded by the political leanings of the state -- it again appears almost too perfect. In the chart below, the state is blue if it voted for both Kerry and Obama, purple if it voted for Bush and Obama, and red if it voted for Bush and McCain in '04 and '08.
red: 13.83
purple: 12.1
blue: 7.47
In other words, blue states have almost half the rate of gun deaths as do red states. If we regress gun deaths on the percentage voting for Obama, we find out that a 10 percentage point gain in Obama voting is associated with a decrease of 2.9 gun deaths per 100,000 people, or .67 standard deviations (r squared = .402).
So, what does this mean? Well, I think it's fair to assume that the percentage of people in a state who voted for Obama is fairly reasonable proxy for how liberal a state is -- which should also be a fairly reasonable proxy for how strict a state's gun control laws are. So it seems almost certain that, controlling for nothing else, states with stricter gun control laws tend to have fewer gun deaths per capita. Which means that gun control works . . . right? Not really. Like I said, this controls for nothing else. One potentially mediating variable is gun ownership. I have to believe the rates of gun ownership are fairly strongly correlated with how conservative a state is -- hence states with more guns were less likely to vote for Kerry or Obama. Though even if that's true, I see two problems with arguing that it invalidates these statistics:
1.) One of the arguments against gun control is that people can -- and do -- safely own guns. And that guns don't kill people -- people kill people. But if states with higher rates of gun ownership also have higher rates of gun deaths that calls that assumption into question.
2.) One of the goals of gun control is to reduce both the number of people who own guns and the number of guns each person owns. The narrative I've heard in the media is that it's not very successful at doing this -- which I find quite plausible -- but it's still possible that gun control reduces gun ownership a little bit, which would mean that gun ownership would be associated both with how much people feel they need guns and rules regarding gun ownership.
One of the commenters on 538 argues that gun deaths is a misleading statistic, and that we should use homicides involving guns instead. But I disagree. Gun control doesn't aim only to prevent murders with guns, it also aims to prevent gun accidents (e.g. by requiring child-proof safety locks). It's certainly worth looking at both, but I'd put the latter above the former in terms of importance.
I find it most interesting that despite having more urban areas (which usually have higher rates of crime), more liberal states have fewer gun deaths. Indeed, if we were going to conduct a rigorous investigation of whether gun control works or not, we'd certainly have to factor in urbanicity, poverty, non-gun crime rates, gun ownership rates, and a number of other variables. We'd also want to know if there were any other effects of the gun control policy beside that on gun deaths (e.g. were more people killed with knives, or were there more home invasions b/c robbers knew homeowners were less likely to have guns?)
Anyway, here are the possible conclusions we can draw from these data (and given the limited scope we can't be sure that any one of these conclusions is better than the rest):
1.) The data are faulty -- real life isn't usually that neat and clean, and the list could have been doctored or simply made up
2.) Gun control works -- states that have stricter gun control laws also tend to have fewer gun deaths despite being more urban
3.) Gun control is irrelevant, demographic factors matter a lot more
4.) Gun ownership is dangerous, regardless of whether one is in an urban or rural location
5.) Nothing. Far more research would be needed to draw any conclusions.
I'll stick with number 5 until I see more evidence.
