Monday, August 10, 2009

What caused the 90s reduction in gun deaths? - Part 4

Part 1, 2, and 3 went over some of the statistics related to gun deaths and deaths in general. Basically showing in multiple ways that to pin the decrease in gun deaths on the Brady Bill or the Assault Weapons Ban would require one to believe that the Tooth Fairy is going to help the US get our of debt. Now I would like to take a slightly more serious approach and try to make some inroads into the problem.

First, to try to say that crime has a single cause or even that it has a major cause that if we eliminated, then crime would be dramatically reduced or eliminated is nonsensical. As with most social issues, there are a host of factors that influence it to varying degrees and at varying times. If you are hoping that I will prove conclusively what is causing crime in this post, you are going to be disappointed.

With the hard sciences (math, physics, chemistry, biology) one can form a hypothesis and then test the hypothesis through experimentation. Once correlations are determined, further testing can be done where other factors are controlled such that causes can be found. Unfortunately, in most cases it wouldn't be ethical to do such experiments in the soft sciences (psychology, sociology, economics), because there isn't a laboratory that you can control all of the factors in. So the soft sciences are left with computer modeling and tracking surveys. Invariably, they have their work cut out for them. Criminologists fall into this category.

To begin this journey, let's forget that we started talking about gun deaths and look at crime in general. Why do this? Because, besides accidents (which are already steadily declining), crime is the easier to understand and take action on. Using the FBI uniform crime statistics, I graphed violent crime and property crime figures for the period that I have the other graphs from 1981-2006 (dashed lines use right scale, solids use left scale):
I would try to not include any sarcasm here, but I must. THE LINES ARE THE SAME FREAKING ONES AS GUN DEATHS. Don't believe me, check here. Now back to being serious. Violent crime and property crime on their surface would tend to be different. This information would show they are not as different as you think. The decline in property crime started in 1991, violent crime in 1992, and gun deaths in 1993 (although it was basically flat between 1991 and 1993 for overall gun deaths and homicides). Going back to the original question, the Brady Bill and the AWB were passed in 1994. The Brady Bill had nothing to do with crime in general, the AWB was bundled with a larger crime bill, but the decline in crime had already begun 2 to 4 years before it came into effect. So my question is, what other things in our society might drive people to crime? Since we know that gun availability has been steadily rising and crime in general fell in the 1990s and leveled off in the end, then gun availability cannot be correlated to gun crime.

When I saw these charts, I knew that I had seen the shape before so let me go through my thought process. First, crime has been linked to economic prosperity before so I ran some numbers to see. Here is the average yearly unemployment rate from the Bureau of Labor Statistics during the time in question:
OK, it is not quite like the crime graph, although there much better correlative properties than gun availability. We see a peak in 1991 and then a steady decline until 2000, which is nearly what crime in general did. However, there is the other peak in 1982 that we don't have a corresponding peak for crime.

Next I decided to look at another economic indicator. The S&P 500. In order to compare it to rates, first I determined the yearly average based on the daily closing averager. Then, I needed to take the value and divide it by some population coefficient to account for the increasing population (more workers should equal more value in the S&P 500). Next, the S&P 500 is a positive indicator of the economic health (directly correlated) as opposed to unemployment rate which is a negative indicator of economic health (inversely correlated). Therefore, I could just take the inverse of the manipulated S&P value, but inverses tend to add distortion, so I also wanted to see what the curve would look like undisturbed. So I also graphed a manipulated value with a negative. The scales are not so important (because you would need my coefficients to be able to back track to the S&P value). The result is this graph:
The negative curve is showing the steep decline starting about 1995, while the inverse curve is showing a leveling off after about 2000. Not perfect, but the indicators of correlation are there. I would say that because of the way that the S&P 500 is calculated (a weighted average), one would be hard pressed to correlate it to anything straight across. Some normalizing has to be done. However, I believe it is valid to look at trends over short periods of time.

To end, I can't make a firm conclusion of causation about the economy and crime. However the factors I looked at are better correlated than gun availability was (which showed no correlation at all). So from that I would conclude that if one was really serious about decreasing both crime in general, as well as gun deaths, then perhaps you should focus more on policies that would improve the economic situation of people rather than trying to control who has the guns. If you want to believe the article of faith that in order to control crime you have to control who has the guns, then keep on believing. The evidence isn't there to support you.

Too many people don't want to hear that crime follows the economic cycles. To some that is admitting defeat in that as long as we have a free market, we will continue to have these economic cycles and crime will ebb and flow. The easy way is to say that guns (or drugs, or the boogeyman) is the primary driver, and if we just control or eliminate that item, then all will be well. Tinman, pay no attention to the man behind the curtain...

As with any of my analysis, I am more than happy to take large sums of money to do this full time from my couch while I watch reruns of MacGyver. If you're interested, please contact me. If you have complaints about how I did the analysis, email me and I'll send you the spreadsheet and you can do your own. If the facts I have presented are wrong, correct me and I will update these posts.


  1. Great work. My only complaint is your need to use different scales on the same graph. I totally understand why you did it, but maybe instead of two scales just start bothgroups at zero and run the y axis on a Delta value. Then you'll just be looking at the change in value rather than the total, as that's all the antis are talking about. I mean it's not like crime has ever STOPPED because of ANY laws, pro or anti-gun, but the change in crime around laws or events is relevant to discussion.

  2. 2nd to last paragraph, boogy not boody man. Although...the boodyman has a certain ring to it.

  3. Great work, I notice that absence of comments by the pro-ignorance crowd. Guess they don't want to let facts get in the way of their agenda.