Firearms and Suicides in the United States: County-level Statistical Analysis

Chart: The chart displays the relationship between “gun shops per 100,000 people” (x-axis) and “suicides per 100,000 people” (y-axis) for counties in the United States. The area of each bubble is proportional to county population. Each bubble’s color is proportional to the 2016 Presidential election result for each county: dark-to-light red indicates solid-to-marginal Trump majority, dark-to-light blue indicates solid-to-marginal Clinton majority.
Analysis: Are high rates of firearm ownership linked to higher suicide rates? Although we do not have county-level gun ownership rates, we do know the number of gun shops in each county (courtesy of the Bureau of Alcohol, Tobacco, Firearms and Explosives; ATF) and can therefore calculate gun-shops-per-100,000-people. This is likely a good proxy for firearm demand, which is likely a good proxy for gun ownership rates. And suicide rates are reported by the Centers for Disease Control and Prevention (CDC).
There is a reasonably strong visual relationship between gun shop rates and suicides. There is also a strong statistical relationship. Simple linear, power law, and exponential fits return r-squared values of 0.21, 0.22, and 0.26 respectively (i.e., between 46% and 51% of the variation in suicide rates can be explained by gun shop rates). When we weight the data by population, the r-squared values rise to 0.36, 0.59, and 0.51 (59% to 72% of the variation). Over-weighting the data by population further improves r-squared values to 0.79, 0.83, and 0.65 (i.e., 81% to 91% of the variation).
In addition to the strong relationship between firearms and suicide, there is also a clear differentiation in how each county voted in the 2016 Presidential election: solid Democrat counties typically have both low rates of gun shops and suicides, while increasing Republican vote is associated with higher rates of both. This trend also follows population or urban density; large urban counties tended to vote Democrat with lower values of gun shops and suicides, and rural counties tended to vote Republican with higher values of gun ownership and suicide. The urban density trend also explains the significant success of weighting and over-weighting the data by population.
Methodology: Gun shop data were sourced from the ATF¹ and suicide data (1999–2006) from the CDC Wonder Tool². Election data were taken from Secretary of State websites for each state. County population data were taken from ESRI³. The data were compiled and visualized using custom VBA code in Microsoft Excel⁴.

1. https://www.atf.gov/resource-center/data-statistics
2. https://wonder.cdc.gov/
3. http://www.esri.com/data/esri_data
4. https://products.office.com/en-us/excel