Thursday, March 4, 2010

Brady Scores and Crime

Over at Snowflakes in Hell, Sebastian takes a stab at showing there is no correlation between Brady Score and Violent Crime Rate. I saw it and instantly said, "Why use one graph when four will say the same thing?" So I went to work.
First to understand the Brady Score, the Brady Campaign has developed a list of gun control laws they think should be implemented and grade each state from 0 to 100 (0 being "bad"). Supposedly these laws help decrease gun violence. Based on the fact that 80% of the states score less than 25, most of the governments in the US do not agree. The best score is from California, with a 79. The only other state with a passing grade (which I define as >60 - or a D-) is New Jersey. So, the Brady Bunch is not too happy with the way gun laws in the US are going.

Let's put them to the test by examining a couple of things. First, Sebastion graphed out violent crime rates. I decided I wanted to expand on that. If gun control laws are going to have any affect on crime, they are going to decrease gun crime primarily. So for my crime data I went to the FBI Uniform Crime Reports. This data is reported by law enforcement organizations to the FBI. They have tables that break down crime by the type of weapon used and the state. I looked at murders, robbery, and aggravated assault. Rape is normally included in the violent crime statistics but I elected not to include it since based on other FBI data less than 10% of rapes involve someone using a firearm. And compared to robbery and assaults there are much fewer rapes so the numbers (and rates) won't be affected too much. Using this same logic, I could exclude murders as well (there aren't that many compared to robberies or assaults), but roughly two thirds of murders are with a firearm so including it might help the Brady Bunch out (but not really since the number of robberies and assaults committed with a firearm is 35 times the number of murders).

Now, to simplify things I have to delete a few data points. For one, the Brady Campaign doesn't rank District of Columbia. By throwing this out it is definitely to their advantage as DC has stricter gun laws than California and have a very high violent crime rate - which is contrary to what the Brady Bunch would have you believe about the effect of gun control. Illinois also has to be thrown out since they have spotty reporting to the FBI. Their Brady Score isn't anything to write home about (28), but without consistent data compared to the other states there is no sense in including them. Finally, Florida gets thrown out as well since they didn't have a murder rate listed in the FBI tables. I have no idea why but I didn't want to go hunting for another data set just to include them. They score a 6 by the Brady's reckoning so in effect we have thrown out a low, medium, and high Brady score, but we still have 48 data points.

So I begin by re-creating Sebastian's graph with my data. As you can see, it is mostly similar to his, although my linear trendline is showing a slight decrease whereas his showed a slight increase. The R-squared value is the same 0.0005. In a nutshell, this says that 0.05% of the variability of the violent crime rate from the trendline can be explained by the Brady Score. In other words: there is no correlation! (You would expect something at least above 0.5 before you started claiming correlation.)

But, gun control laws would primarily affect gun crime, not necessarily all crime. So, let me run the numbers again. This time, the R-squared value has increased to 0.0176. That is light-years ahead of all violent crime, but still very dismal as far as correlation goes: there is no correlation!
Now, before we just leave this issue lets take into account population. Not with the rates, but with the Brady Scores. Let me create a population weighted Brady Score. By dividing the Brady Score by the population of the state and then multiplying by 5,000,000 (the average population of the states) I get my Brady Population Weighted Score. This helps to minimize the effects of those small states like North Dakota (by giving them a higher score). Unfortunately for the Brady Campaign, it does nothing for a state like Utah that has a score of 0. Violent crime has an R-squared value of 0.0525.

And firearm crime has an R-squared value of 0.0902. This is the absolute best that I could get the Brady Scorecard to show some sort of correlation to gun crime. Once again, there is no correlation.
To sum up the data: Gun control laws do not have any observable affect on violent crime or crimes committed with firearms. Remember, correlation may not be causation, but if there is no correlation, there cannot be any causation.

1 comment:

  1. Maybe that last graph goes like e^-B where B is the Brady score?