Guns, education, religion, and suicide

My earlier post indicated that gun laws in the US had little impact on the homicide rate, when demographic factors were taken into account. This makes sense – if I want to kill somebody, the lack of a gun will merely prompt me to choose another weapon. But what about suicide? The impulse to suicide is often brief, and easy access to a gun during a suicidal episode may increase the chance of dying.

To test this, I extended my previous dataset with data on educational attainment, data on religiosity, registered gun ownership data from the ATF, age-adjusted suicide rates from the CDC, poverty rates, unemployment rates, and other demographic data. I ran all that through a regression tree analysis, using R.

Suicide rates in the chart (click to zoom) are indicated by colour, ranging from 8 per 100,000 for New Jersey and New York (yellow) to 23.7 for Montana (black). Having a college degree seems to have a protective effect – states on the right of the chart, with more college degrees, had lower suicide rates. This may relate to the higher employability of college graduates. However, states at the top of the chart, with higher high school graduation rates, had higher suicide rates. I am not sure why this is the case.

Among the states with fewer college graduates, religion had a protective effect (this is consistent with other studies). States where 77% or more of the population said that religion was “somewhat important” or “very important” to them are indicated on the chart by triangles. For the states with fewer college graduates, the suicide rate was 13.6 per 100,000 for religious states, and 17.5 for less religious ones.

Finally, the highest-risk states (fewer college graduates and less religious) split according to gun ownership. States with more than 0.008 registered guns per capita are marked on the chart with an inner dot. Among the highest-risk states, the suicide rate increased from 13.9 per 100,000 to 18.6 when more guns were present. This group included Alaska (23.2 per 100,000), Arizona (17.5), Idaho (19.2), Maine (17), Montana (23.7), Nevada (18.6), North Dakota (17.3), Oregon (16.8), and Utah (21.4). Among the more religious states, registered gun ownership did not seem to have an effect (although, of course, registered gun ownership is a poor indicator of true gun ownership).

Thus the data does seem to suggest a link between gun ownership and suicide risk, but only when other risk factors are present (low religiosity and no college degree). This is exactly what we expected, and it means that suicidal (or potentially suicidal) people need to be kept away from guns.


Advertisements

Do gun laws save lives?

Somebody pointed me at this interesting data the other day. The chart above (click to zoom) combines the “Gun Law Score Card” from the Law Center to Prevent Gun Violence in the US with homicide rate data from Wikipedia and voting results from the last US election. Do gun laws reduce the chance of being murdered?

Obviously, “Blue” states tend to have stricter gun laws than “Red” states (an average of B− vs D−). “Blue” states also have lower homicide rates than “Red” states (4.5 vs 5.9), and this is statistically significant (p = 0.012). There is a weak (R2 = 6%) correlation between gun laws and homicide rates, but this relationship is not statistically significant.

Whatever it is that makes you less likely to be murdered in some states than others, it does not primarily seem to be the gun laws. Poverty may be one of the relevant factors, however – median household income explains 22% of the variance in homicide rates, and when this is taken into account, any effects due to gun laws or election results disappear. “Red” states are, on the whole, simply poorer (and, conversely, poor states are more likely to vote Republican and have weak gun laws). Other demographic factors, such as the number of people with college degrees, also seem to have explanatory value as far as the murder rate is concerned. However, the phenomenon of murder does not seem to be understood as well as it could be.