On Gun Violence, the Data are There, if Anyone Wants to Look
Updated: Jun 3, 2022
In the wake of horrific mass killings in Buffalo, Uvalde, Tulsa, and elsewhere, politicians, journalists, pundits, and interest group leaders have wasted no time in flooding the airwaves with their theories about just why these things happen in the United States and almost exclusively in the United States. Unsurprisingly, these explanations haven't involved a lot of research or reflection -- the killings mostly just prompted people to double down on what they already believed.
But for a moment, I'd like to take some of these ideas seriously and attempt to measure and test them. (I did this with my undergrad methods class and the discussion was pretty interesting.) Let's examine some of the leading recent ideas about why mass shootings happen in the U.S.:
Violent video games. At the recent NRA convention, Sen. Ted Cruz (R-TX) said, "Tragedies like the events of this week are a mirror forcing us to ask hard questions, demanding that we see where our culture is failing.... Social media bullying, violent online content, desensitizing the act of murder in video games, chronic isolation" may be responsible for the recent violence.
Porn. Ohio Senate candidate J.D. Vance recently suggested that the use of pornography was to blame for the violence, and called for a total ban on porn.
Mental illness. Texas Governor Greg Abbott attributed the Uvalde shooting to mental illness: "We as a state - we as a society - need to do a better job with mental health. Anybody who shoots somebody else has a mental health challenge. Period."
Religiosity. Sen. Ron Johnson (R-WI) said that Uvalde was the result of "the secularization of society" and "loss of faith," adding, "I think the solution is renewed faith." Fox & Friends' Rachel Campos-Duffy said, "We took God out of schools and we wonder how this evil comes in," referring to a 1962 Supreme Court case.
Guns. The gun control perspective tends to argue that the presence of shootings is, quite simply, a function of the presence of guns. As The New Republic's Jason Linkins said, "Unlike everywhere else in the world, there are too many readily available guns in America."
Obviously there are more theories out there, but these strike me as the most common or at least the loudest assertions over the past week. So let's try to operationalize these. The variables below are far from perfect, but I've attempted to find state-level measures that allow us to test the above hypotheses.
Gun deaths: For this, I'm using the CDC's statistics on firearms deaths per 100,000 residents in 2020.
Video games: This is imprecise at best, but CenturyLink put together an index of states' video gaming communities, measured by the number of video gaming conventions and Meetup gaming groups, as well as download speeds and other factors. They then rank the states, with 1 having the most gaming, and 50 the least.
Porn: In 2014, Pornhub published the per capita page views for their site by each state. This measure is far from perfect, in that it's somewhat old and Pornhub isn't the only porn source out there by a long shot. Also, there's a methodological problem in that Kansas appears to have far more porn views than it should. (These data are geo-coded from IP addresses, and when the system can't identify the address, it defaults to the geographic center of the country.) But it's still a reasonable measure.
Mental illness: Mental Health America publishes a ranking of all the states based on the incidence of mental illness and the range of mental health care provided. States with lower numbers have the best resources and lowest incidence of mental illness.
Religiosity: Pew has an index of religiosity for each state based on several questions, including importance of worship, frequency of prayer, and belief in God. This is the percent who are considered “highly religious.”
Gun ownership: RAND put together a database of the estimated proportion of state residents with a gun in the home from 1980 to 2016.
You can access the whole dataset here. I've thrown all the independent variables together to try to predict rates of gun violence. I'm just going to post some Stata regression output here:
The above regression analysis shows us a number of important lessons:
Gaming and porn appear to be unrelated to gun deaths. The coefficients are small and not statistically significant.
The mental health rank is positive and statistically significant, suggesting that states with higher incidence of mental health problems and poorer mental health care options do see more gun deaths. This supports what Gov. Abbott was saying, although doesn't necessarily reflect his policy agenda. This doesn't mean that the mentally ill are disproportionately violent -- a great deal of evidence argues otherwise. But it is possible that those struggling with mental illness might have an easier time accessing care and a harder time accessing weapons in some states.
Religiosity is highly statistically significant and is positively related to gun violence. The more religious a state, the more gun deaths it has. This doesn't prove a causal relationship, but it also really doesn't support the idea that increased faith will reduce shootings. (Also, the idea that shootings came into the classroom because we drove God out is a particularly monstrous claim, however the mechanism is supposed to work, portraying God as either callous or vicious or homicidal.)
Gun ownership is highly statistically significant and positively related to gun violence. It also has the largest beta coefficient of all the variables, suggesting it has the greatest impact on gun violence. Basically, this supports a main contention of the advocates of gun control: the most shootings happen in the places where the most guns are. A more armed society is a less safe one. Below is what that looks like in a simple scatterplot:
Now, as I suggested above, these measures carry a few problems, and I certainly wouldn't advocate making serious policy changes based on a few state-level measures I compiled in a few minutes of Googling. Policy experts would want to examine these in finer detail, looking at county-level or individual-level data, and also cross-nationally. Many have.
The point, rather, is that some useful evidence is both readily available and easy to understand. And it points in some helpful directions. For one thing, it debunks a lot (though not all) of the arguments gun rights advocates make in the name of maintaining the status quo. Politicians do not have to make these arguments, and reporters do not have to accept them uncritically. For another, it suggests that poor mental health care and widespread gun ownership are serious problems in America, and it would be helpful to look into solutions to address these.
But this also reminds us that many in politics are simply not motivated by empirics. They work very hard to justify the opinions they already have -- even when clearly clutching at straws -- rather than to approach anything like empirical truth. No, we don't necessarily expect elected officials to act like social scientists, but nor should they be acting like the opposite.
UPDATE: Tom Pepinsky ran these same numbers and included regional controls, finding that the effect of mental health care washed away but religiosity and gun ownership remained strong predictors of gun violence. And the R-squared goes up to .82.