About this time last year, I asked ChatGPT to provide a graph of firearm availability versus homicide rate at the national level. It is here.
Just for fun, I thought I’d see if updated information would make a change. Here’s the 2026 result. The data only goes to 2024 because the 2025 ATF data is not finalized.
I spent an afternoon with ChatGPT playing with the statistics. The raw negative correlation is obvious even visually. However, the question becomes whether there’s some real meaningful correlation or we are just overlaying two independent data sets, one of which is generally rising and one of which is generally decreasing. So, you can do correlations of first differences per year or you can do regression and fit to varying time frames. It is also possible to look at net additions to gun availability rather than crude cumulative availability. In other words, if you change the number of guns during a particular time period, is there an associated change in homicide. The results for this is that lookng at first derivative data or regressions to shorter time frames, there’s not much there.
The big problem is the COVID spike. Starting with 1993 data, if you include the COVID years, there’s a slight positive correlation (note that starting in 1993 removes the huge historical decline that I believe is due to atmospheric lead — see my previous post on this). If you exclude the COVID years, there’s a slight negative correlation. Again, based on net addition analysis rather than crude numbers. A distributed lag model also shows heavy influence of 2024, 1996, and 2016, all of which are negative influencers. 2024 is overweighted because it’s an end point.
In essence, then, there is no evidence for a causal relationship between gun availability and homicide using this data. Homicide rates are, and always have been, driven by other things. In the 1990s, it was the generational effect of mass poisoning. In the 2020, it was/is the generational effect of the COVID virus and the dismal governmental response and associated societal degradation. IMHO, of course.