Wednesday, August 26, 2015

Forget Murder, Gun Ownership More Related to Suicides/Accidents

Another day, and another shooting is the top news story.  The shootings in the news constantly inspired me to download gun and shooting data, and what I found was highly surprising, yet depressing.

As a quick piece of personal background, I know two people who have been shot in the last 18 months.  And it's not like I am living a hardcore life to know people who have been shot. I'm an analytics manager at a financial company. I am also a gun owner. 

BACKGROUND

I've looked at mass shooting data before, but mainly just the twitter conversations in the weeks after the shootings.  I never had much of an interest in the actual data, largely because I was sure it would be depressing, and I like to have a good time with data.

This morning, after the shooting, I was confronted by data from Ezra Klein at Vox regarding the relationships between gun ownership rates by state and gun deaths.  It was a fairly tight positive correlation.  It was compelling data, but I have seen Vox misleading with data in the past, so I tried to verify.  Data wasn't hard to find, here's my initial analysis.



Crap.  Effectively no correlation.  That doesn't look ANYTHING like Klein's graph, did I scramble States?  Is Klein lying? Did I get bad data?  I dug in.  The answer is no.  In my haste to verify Klein's numbers I pulled a different metric:  GUN MURDERS.  Not the same thing.

LEARNINGS

Down the rabbit hole I went.  A couple of surprising things I found:
  • Gun suicides outnumber gun murders at a ratio of about 1.75:1. It's possible that Klein is still right, and gun deaths and murders work in much different ways statistically. 
  • Suicide by gun has a higher "success" rate, at about 85%.  Everything else is below 70% (depressing, this is why I don't like looking at these statistics).  So access to a gun is key in successful suicides.
My prior chart showed very little correlation between gun murders and gun ownership, but could I replicate Klein's chart on total gun deaths and gun ownership rates?  Yes, here you go.



Per my prior graph, though I know that there must be some noise in this data, due to an uncorrelated subset problem (non-murder deaths).  So I backed them out, to come up with non-murder deaths, thus achieving our highest correlation of the day, and a conclusion: gun ownership rates are highly associated with accidents and suicides.


THEORIZING

The data is somewhat surprising, not necessarily that suicide and accidents are highly correlated to gun ownership, but that murder is not.  I may dig in in the future, if I have a desire to get depressed again, but here are generalized thoughts on this:
  • Conservatives would likely argue that any increase in access to guns in order to commit murders would be offset by the deterrent effect of more people having guns to defend themselves with.  
  • Guns create an additional risk for accidents and suicide simply by availability, and the more families with guns has a somewhat linear impact on risk, especially if suicides occur at the heat of the moment.   
  • Liberals may argue that it's the easy access to guns for people who want to kill that creates high gun-murder rates in the United States.  Essentially, if someone wants to get a gun, that process is so easy that background ownership rates don't matter. 
  • I may think underlying demographic correlations are to blame.  Urban districts with low gun ownership rates have high murder rates for other reasons such as poverty.  Rural areas with a lot of guns may be depressing, thus creating suicides.  White people have much higher suicide rates.






No comments:

Post a Comment