I'm seeing two arguments that keep getting repeated in interpreting this phrase: "where there are more guns there is more homicide."
One is that country A has more guns (in total, per capita, whatever) and more homicides (in total, higher rate, whatever) than country B, then that means that having "more guns" means "more homicides". The flaw in this comparison is that there are many other variables regarding the differences between country A and country B that aren't included in the analysis.
If you read the methods sections of most of those studies you will see that many of the various other factors are accounted for and included in the models.
But this is actually a great topic. In any given study there are factors you know and factors you don't know about. Given that you cannot model what you do not know the best way to determine any MISSING variables is to look at the NOISE in the data. The
residuals.
Let's look at the earlier data set based on your preferred data inputs (WISQARS and ATF). Using that data the model generated could only account for about 7% of the total variance in the data. That indicates there are a LOT more things to consider.
There's also a metric called "lack of fit" in many data analyses.
If a model fits the data better you have what you need:
sufficient factors to explain a large majority of the data's behavior. That's all you can ever really know.
The second version of the argument is that in the US, when we get "more guns" we have "more homicides". Sometimes someone will cite a study substituting gun ownership rate, but "more guns" isn't "higher gun ownership rate", and the latter can't even be accurately measured.
In the real world of science this is not an uncommon thing. That is why proxies are developed. Proxies are used ACROSS all the sciences and used rather effectively. Proxies are not perfect (nothing is, even direct measurement!)
Again, the best available data and the best available models show that more guns correlates to more gun homicides.
Just nitpicking at what you THINK might be a problem is insufficient to establish that the model is unsound.
For this argument, the data is unequivocal - the number of guns, based on ATF manufacturing, expert and import reporting, shows that the number of guns in the US increased by about 270 million from 1986 (the date of the earliest ATF industry report). Using that as our starting date, CDC WISQARS data clearly shows that the homicide rate between 1986 and 2019 (the latest date of CDC data) declined significantly.
Wrong. You have shown no such data. You have failed to show that it is a statistically significant regression and you most egregiously fail to show that YOUR preferred
sui generis metric of gun prevalence is superior to other proxies. In fact your proxy is seriously flawed if you think every gun manufactured in a year (or imported) is sold that year. In fact I've worked in consumer facing industry for many years now and I know that isn't even CLOSE to reality.
With the Second Amendment in place, any such studies are worthless.
With the Second Amendment removed, any such studies are needless.
It is a true irony when a data analyst admits that data is useless when debating a topic. One could almost think that data isn't even important to you.
