Guns Kill People

In this post, I’ll review the evidence for whether higher levels of gun ownership cause more people to die. I’ll save any discussion of what to do about it for another day.

For context, there are nearly twice as many gun suicides as gun homicides. Nearly all gun homicides are not part of a mass shooting. Accidents account for about 2% of gun deaths, which, even in recent years, is much larger than the number of deaths in mass shootings. I’ll focus on total gun suicides and homicides without limiting the discussion to mass shootings.* Because of the huge number of people killed each year by guns, nearly all outside of mass shootings, we have much better data for what causes total gun deaths rather than mass shootings deaths in particular.

I’ll focus on the evidence about whether a high availability of guns makes death more likely, not just by a gun, but overall. It’s important to distinguish studies of gun homicides and suicides from all homicides and suicides. Opponents of gun control argue that deaths by gun would simply turn into deaths by another method if we had aggressive gun control. We have to look at the evidence to see if they’re right.

Household Comparisons

This meta-analysis of studies with individual-level data on gun ownership found 16 retrospective, observational studies. Fourteen studied suicide and six measured homicide victimization (that is, did a person who lived in a house with a gun become the victim of homicide, not did they commit homicide). After controlling for confounders like a history of mental illness, having access to a gun was correlated with a more than three fold increase in risk of suicide. Owning a gun was also correlated with a doubling in the risk of becoming a victim of homicide, though this effect was much larger for women than for men.

The limitation here is that case-control studies are retrospective. They ask people (or people who can speak on their behalf) who did or didn’t commit suicide whether they had access to a gun. People who didn’t commit suicide may not remember that they actually do have access to a gun. There may also be larger bias in the selection of the control group in retrospective studies than in prospective studies.

International Comparisons – Homicide

International comparisons get a lot of attention, but honestly these data sets aren’t very useful. It’s very difficult to gather data on gun deaths in the United States, let alone across many countries. I didn’t find any papers that make international comparisons while controlling for things like poverty and non-violent crime rate that could plausibly explain gun deaths independent of the gun ownership rate. A paper I’ll discuss more below argues these studies “are severely limited because of inadequate adjustment for confounding factors.”

This weak study finds that developed countries with more guns may or may not have more homicide, depending on how you measure gun ownership. Another weak study of female homicide victims finds that the correlation mostly goes away without the U.S. The U.S. is a huge outlier in both the gun ownership rate and the gun death rate. If you exclude the US from your analysis, there isn’t much correlation left. Some people would argue that self-evidently proves that gun ownership causes gun deaths. That’s wrong. There are plenty of ways the US is a major outlier compared to other developed countries. If you plotted gun deaths versus number of corn farms, you could similarly “prove” that corn farming causes gun deaths. We need to look for better evidence one way or the other.

Inter-State Comparisons – Homicide

Comparisons between the 50 states give us much better data. Three scientists in particular – Matthew Miller, David Hemenway, and Deborah Azrael – have led the way. In a 2002 paper they studied an indirect measure of gun ownership and a direct measure of homicide rates from 1988 to 1997. They found that states with higher levels of gun ownership have higher overall homicides, whether or not you adjust for a range of other factors.

Their even stronger 2007 study (full-text) with data from 2001 to 2003 uses a more direct measure of household gun ownership. Again, they find that states where more households own guns have more homicide, controlling for aggravated assault, robbery, unemployment, urbanization, alcohol consumption, and poverty. The results hold separately for the killings of men, women, and children and for everyone pooled together.

Screen Shot 2019-08-08 at 1.27.54 AM.png

They found that the firearm homicide rate increases proportionally with the fraction of homes that own a gun, with no offset in non-firearm homicide. There’s nearly a 1-1 relationship between relative increases in firearm ownership and overall homicide. This dataset is superior to the international comparisons because it has a better measure of gun ownership and controls for confounding variables.

Their results are more robust than the international comparisons as well. Dropping the states with the highest or lowest gun ownership rates doesn’t change the results, unlike the case where dropping the US from country-level data changes the overall result dramatically. Also, using a simpler model that only adjusts for poverty and urbanization doesn’t change the results.

Admittedly, this doesn’t prove causality. However, the gun ownership rate is uncorrelated with non-firearm homicide, robbery, or assault, so we can probably rule out the argument that gun ownership is a response to a high level of violence instead of the cause of gun violence.

How about more recent research? This 2013 study uses state-level data from 1981 through 2010 and again finds that higher gun ownership correlates with higher gun homicide rates. Oddly, they use non-firearm homicide as a confounder for firearm homicide, which means their study is really about the proportion of homicides that are committed with firearms and not about the total number of homicides.

Monuteaux et al. 2015 uses state-level data to show that the states in the highest quintile of gun ownership have over twice the overall homicide rate of states in the lowest quintile of gun ownership, whether or not they adjust for income, sex, race, ethnicity, age, education, poverty, population density, region, and bordering state ownership. They add that assault and robbery with a gun is also five to seven times higher in the states with the highest gun ownership.

Inter-State Comparisons – Suicide

Remember that nearly twice as many gun deaths happen from suicide as from homicide. Most people would probably guess otherwise because homicides make the news and suicides rarely do. This study of state-level data from 2008-2009 controls for the rate of suicide attempts in each state and finds that the number of completed suicides is positively correlated with the rate of gun ownership. The suicide attempt rate is actually uncorrelated with the gun ownership level. The most straightforward interpretation is that people decide to commit suicide whether or not they have a gun, then use whatever means is most available. If they have a gun, they use it, and they’re much more likely to die.

This study of state-level data from 1981 to 2013 finds that in both men and women higher gun ownership is correlated with higher suicide by gun but lower suicide by other methods. For women, they found the overall suicide rate was uncorrelated with the prevalence of gun ownership, but for men a higher prevalence of guns was correlated with higher overall suicide.

The length of their dataset allowed them to use both the variation in gun ownership and suicide between states and within states over time. They controlled for region, urbanization, religion, income, alcohol use, crime rates, and incarceration rates.

Using a simple model that only used gun ownership rate or that clustered by state left the results unchanged. Changing their measure of gun ownership also didn’t change the result. They estimated that if Wyoming’s gun ownership rate declined from 73% to the average of 41%, the overall male suicide rate there would decline 16%. The supposed reverse causality that a high homicide rate causes gun ownership doesn’t make any sense as an explanation for higher suicide rates in areas with lots of guns.

Guns Cause Suicide

None of the above evidence can directly demonstrate causality. Two studies can. First, changes in policies of the Swiss Army abruptly reduced the access that men had to guns in their own home. This wasn’t randomized, but the sharp boundary in time offers something of a natural experiment. This paper documented a large drop in firearm suicide among young Swiss men, with no matching change in non-firearm suicide or suicide by women.

A second, similar natural experiment occurred in Israel. A sudden policy change required soldiers to leave their guns on base when they returned home for the weekend. The suicide rate among men in the age range for military service plummented 40%, including a 70% reduction in suicide on the weekends.

Neither one of these is a perfect experiment. If we wanted to prove that taking people’s guns away reduces suicide, we’d have to randomly take away a subset of people’s guns. That isn’t feasible, though, so this is the closest we can get.

Overall, the evidence is quite strong. Being in an area where many people have access to guns makes it more likely that someone will kill you. Having access to a gun also makes it much more likely that you’ll commit suicide. Suicide attempts are often impulsive and having access to a gun makes them much deadlier. None of this tells us the best way to reduce access to guns. It does tell us that if we could snap our fingers and get rid of guns in America, we’d save tens of thousands of lives per year.

*There is research on the connection between gun control and mass shootings, but the small number of mass shootings makes it difficult to say much about what causes them with confidence. Mass shootings are so rare that a study comparing the rates of mass shootings between different US states – which would be 0 in nearly all cases – wouldn’t have enough data to work with. Instead, these studies typically make international comparisons of mass shooting rates. But there are not reliable data sources that compile data on mass shootings all over the world, so this requires the researchers – who start with a hypothesis in mind – to gather data manually. This process is more subjective than you might think. The research above, for example, which got a lot of news coverage for linking gun ownership rates to mass shootings, makes the unusual choice of excluding any mass shootings involving more than one shooter. The result has intuitive appeal, but the data set is quite poor.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s