Do firearm norms explain the variation in interstate teen suicide deaths?
They clearly matter, but we need a multifactorial explanation centred on capability.
Different US states have very different teen suicide rates. Why is this?
Jeremy Horpedahl: Teen suicide rates vary dramatically across US states, from a high of 30 per 100K in Montana to a low of 4 per 100K in New York. What are the best explanations for the variation? (Data from CDC WONDER for years 2020-22, ages 13-19)
My initial thought was firearms. That is - states in which firearms are more prevalent and accessible will have higher suicide rates. I suspect this is the first thought for a lot of people. It was on my mind in part because of a footnote from Waldron & Currie (2023) where the authors suggested regionally specific factors may be driving much of the overall increase in teen suicides; they specifically noted the variance in rates of gun suicides and gun suicide attempts across states and congressional districts.
On Twitter/X, others also suggested that variation in teen suicide rates might closely track firearm norms across states. But Jeremy Horpedahl was sceptical that this was a sufficient explanation.
MarketUrbanism: It probably tracks teen gun ownership very closely
Jeremy Horpedahl: Probably a factor, but teen girls generally don't use guns to commit suicide
I don’t usually associate teenage girls with guns. But is firearm use actually a less prevalent suicide method among young women?
There’s work on this. Catherine Glenn, Evan Kleinman, and colleagues published a 2019 meta-analytic review of global adolescent suicide rates, which included the following figures on variation in suicide methods.
Take a look at Figure 1c. In the US, half of the young women who died by suicide used a gun. This proportion is equal to or slightly higher than that of young men.
My guess is those firearm suicides are unevenly distributed across states in a way that reflects differences in firearm norms. The map of states that have adopted policies requiring secure storage or child access prevention for firearms looks eerily similar (though not identical) to the map of suicide rates across different states.
Here’s the map of suicide deaths by state, again, for comparison:
The inverse pattern is pretty consistent.
There is also these older data (1997-2006) from Ohio that suggests many women who make suicide attempts do use firearms. There were gender differences in using firearms for suicide; men use firearms for suicide more than women. However, firearms were still the method used by nearly 40% of women who died by suicide. The mean age in the sample was 45, so maybe teenagers and young adults behave differently. But the newer work above suggests that’s not the case.
Based on the findings I’ve presented so far, my guess would be that interstate differences in firearm norms could be the primary factor driving the interstate variance in teenage suicides.
These data are in line with a strong theoretical argument for what leads to more suicide attempts and deaths since increased access to firearms would likely lead to higher practical capability. Capability is a concept that’s less discussed outside of suicide research directly, so let me expand on that point a bit next before coming back to consider other factors that seem to relate to interstate variations in suicide deaths.
The impact on capability for suicide
Most people who think about suicide don’t go on to make a suicide attempt. And most attempts don’t lead to death. So, what leads to more suicide attempts and deaths among people who are thinking about suicide?
Findings from suicide research in recent years have clarified that the biggest difference between people who want to end their lives and people who move toward a potentially lethal suicide attempt is whether the suicidal person is capable of taking lethal action. Even for people who feel a strong desire to end their life, acting on that desire is difficult and terrifying. This is where the concept of ‘capability’ comes in. Some people are more capable of acting on their desire for suicide.
What leads to someone having high capability?
Researchers have proposed three different aspects of capability as potentially relevant to differentiating between people with a high desire to end their lives and people who go on to make a suicide attempt. Some theorists suggested ‘acquired capability’ (reduction in fear of death through habituation to pain) or ‘dispositional capability’ (lack of pain sensitivity and harm avoidance based on temperament) might be important differentiating factors. But it now seems likely that practical capability (i.e., access to means, knowledge about methods, privacy or lack thereof) is the key factor that explains the difference between the set of people who want to die by suicide and the subset of those individuals who go on to make a potentially lethal suicide attempt.1
If it’s easier to access firearms in a given state/district, people who live there and are experiencing strong suicidal desire will have a higher practical capability and, therefore, be more likely to move from suicidal desire to making a suicide attempt.
A large effect of firearm regulation across states
To get a sense of the impact of firearm norms on the rate of teen suicide across different states, I ran some analyses using the CDC Wonder data. Based on the Everytown research map above, I roughly operationalized ‘gun control norms’ as a binary categorical variable. Operationalizing gun control norms as a binary categorical variable will necessarily shrink the effect size, so do keep that caveat in mind.
Still, even with that variance-shrinking potential, I found a significant difference and a large effect of gun control norms on deaths by suicide comparing states with vs without requirements for secure storage or child access prevention to firearms, t(45) = -3.60, p <. 001, d = -0.91.2
Considerably fewer teenagers died by suicide in states that required secure storage or child access prevention to firearms (8.16 per 100,000 vs 12.86 per 100,000).
Other contributing factors: What’s with the elevation-suicide relationship?
In addition to firearm norms, I imagine poverty would also influence suicide rates by making life more painful and that there might be an urban vs rural effect that could make social isolation more likely3 and impact connectedness. However, both of these would be more likely to impact suicidal desire rather than necessarily leading to more suicide attempts across states. Still, they could contribute to more people in a state reaching the point where they might seriously consider suicide.
Elevation is another factor with a moderately strong relationship with suicide. Some studies have found a r = 0.5 (p <.001) correlation between elevation of US counties and suicide rates. It’s a striking finding, as this correlation is relatively large, and it’s not very intuitive why it would relate to suicide.
Some researchers have begun to put forward potential explanations for the relationship between elevation/altitude and suicide. The explanations are definitely in conjecture territory, but they are interesting. Reno and colleagues (2018) note that three studies on this topic have suggested increased hypoxia (insufficient oxygen) at higher elevations as a potential biological explanation. They say hypoxia might exacerbate depression or negative mood. However, recent work has clarified that although depression relates to suicidal ideation, it does not explain the variance between people who have suicidal desire and the subset of those individuals who go on to make a suicide attempt. Therefore, based on what we understand about how suicide works, it seems unlikely that hypoxia leading to worse depression would lead to many more deaths by suicide.
Betz and colleagues put forward an alternative explanation rooted in person-level differences among people who live at high altitudes. They found several factors that differed between people living at high and low altitudes. People who died by suicide at a high elevation were twice as likely to have had a crisis in the two weeks before dying. They also found this population had a potentially dangerous combination of a) 2x odds of depressed mood while b) having no increase in mental health diagnosis or mental health treatment.
Betz and colleagues also argue there is no substantial evidence to support the hypoxia explanation and that the various studies that have suggested this explanation all used aggregate data without controlling for person-level factors in each death.
Both these potential mechanisms suggest ways in which people at high altitudes might be more likely to develop strong suicidal desire (through impacting pain, hopelessness, and, in the latter case, also connectedness). However, neither explanation points to why people at higher elevations might be more likely to act on a desire for suicide.
One possible additional explanation that occurs to me based on what we know about the impact of capability is that perhaps people who live at a higher altitude may be more likely to have privacy and, therefore, have a higher practical capability to attempt suicide.
Do firearm norms impact suicide beyond covariates?
After talking with my PhD supervisor, David Klonsky, about this piece, he reminded me that Michael Anestis has consistently published on this topic for over a decade. I dug into his work, and it provides some really great additional detail on the strength and uniqueness of the relationship between suicide and several factors I’ve discussed above.
In many papers on this topic, Michael Anestis and his colleagues have done much more robust adult-population versions of the analysis I did above.
This table of correlations from their 2018 paper on the association between gun ownership and state suicide rates stuck out as especially relevant:
The three factors most related to suicide were being a veteran, gun ownership, and elevation. Education, poverty, and population density all also have a moderately large relationship with suicide.
That said, both in this study and previous work, when we include elevation, poverty, and population density as covariates, the unique impact of firearm norms on suicide rates gets much smaller.
Firearms still matter, but the additional variance explained is smaller when all the covariates are included. This is a huge list of covariates, but it does lend credence to the idea that this is a multifactorial problem.
There are two important caveats to note here, though.
When controlling for the first set of factors, the impact of firearms on deaths by suicide is much larger than the impact of mental illness and suicidal desire on suicide. So, the impact of firearms is only “small” if we’re ready to entirely discount the impact of mental illness and the desire to end one’s life, and I don’t think anyone is prepared to dismiss those factors.
This analysis demonstrates the limited unique marginal explanatory power of firearms over and above elevation combined with all the other listed factors. As far as I can tell, a similar examination of the unique variance explained by elevation over and above firearms has not been published. My guess is that we would see a similarly (small) amount of unique variance explained by elevation.
Longitudinal impacts of changing firearm norms
I would be remiss not to point out that although the additional cross-sectional variance explained by firearm norms over and above all the other potentially relevant factors is not huge, changing firearm norms through policy seems to impact suicide at a clinically meaningful level.
A particularly vivid example comes from a 2010 study by Lubin and colleagues, which found that stopping soldiers from taking guns home with them on the weekend led to a 40% reduction in suicide deaths among members of the IDF.
Gun control laws implemented across different states between 2009 and 2012 also approximate a natural experiment and show increases and decreases in suicide in the anticipated directions.
For reference, 3996 people died by suicide in California in 2011. So, the 3.5% reduction in suicide the year after the open carry restriction was the equivalent of 103 more people who stayed alive in 2012.
Anestis and colleagues have also examined the longitudinal trajectories of states with and without universal background checks and mandatory waiting period laws for firearms more broadly. They found the increase in suicide was less steep in states that had these laws in place.
So, do firearms explain this map?
Firearms are not the whole story here. But differences in firearm norms across states seem like an important and relatively malleable factor that impacts the practical capability of teenagers and the rest of the population and contributes to this unfortunate map. Elevation, poverty, population density, and other factors all clearly matter as well.
I hope more researchers work to clarify why elevation, in particular, is related to suicide since this factor sticks out as less intuitive. If future work supports the possibility that people at high elevations are more likely to make a suicide attempt based on increased privacy/isolation impacting practical capability, possible interventions could involve trying to reduce that isolation.
My read on these data is that interstate differences in firearm norms meaningfully impact the number of suicides in each state. But it’s not the only factor that matters.
Special thanks to David Klonsky and Sam Mitchell for helpful feedback on this post!
See section 2.1 of Klonsky et al. (2021) for more on these findings regarding practical capability.
Here’s the GitHub repository with the code for reference.
If you fall out with one group of friends in a place where fewer people live, you have fewer options to start over again with a new set of people without uprooting your life.
There is a theory about lithium accumulation in drinking water along the water shed that tracks with the suicide–altitude correlation (https://slimemoldtimemold.com/2021/08/02/a-chemical-hunger-part-vii-lithium/).
Woodard's American Nations "Far West" cultural grouping roughly corresponds to this map. Woodard divides states into smaller units (I think counties) so the eastern chunk of California, Oregon and Washington are different from the western portions. The "Far West" also generally correlates with the elevation factor. You would expect his "Appalachian" culture to be somewhat similar, but that does not play out here. Since Woodard is going specifically to the organizing cultural assumptions, it would probably relate to the gun culture adopted in those regions.