Thursday, August 6, 2009

What caused the 90s reduction in gun deaths? - Part 1

There has been a lively debate going on in some blogs I read about the information presented by Linoge here. There are a few points among the commenters that I wanted to expand on. To begin, the graphic that Linoge created is a compilation of several datasets. It is an excellent display of the wealth of information that can be shown by a graphic. I am a table and formula man myself, but I'll try to constrain myself in this post to only graphics.

The first comment I would like to address is that the Assault Weapons Ban and the Brady Bill were part of reason for the gun death decline in the 90s. Even though some commenters criticize the links (note they don't criticize the evidence, just the links) as being to pro-gun cites here are links to the reports (not someone's interpretation) that show the two bills could not be correlated to the decrease in crime. Here and here. I will attempt to show it in a different way.

First, the data Linoge displays is total gun deaths, which includes homocides (45%), suicides (50%), legal intervention (less than 5%) and accidents (less than 5%). First let us understand what the Brady Bill and the Assault Weapons Ban were. The text of the bills are here and here. I'll summarize, but if you don't believe me, feel free to read the whole thing.

The Brady Bill initially made a 5 day waiting period for the purchase of a handgun to do a background check (until the NICS came into effect). This only applies to licensed dealers selling to private individuals.

The Assault Weapons Ban was part of a larger crime bill. The "ban" part only banned the manufacture or importation of "assault weapons." "Assault weapons" was defined by a firearm having a minimum number of parts. None of the parts had anything to do with the actual firing capabilities of the firearm, they were all cosmetic. In other words, the "assault weapon" sold before the AWB looked slightly different from the other one, but used the same ammunition, had the same accuracy, and the same operating mechanism (in some cases, target shooters found some of the features superfluous and were happy to be able to get a rifle without them). The "ban" did not remove a single "assault weapon" from the US. It only banned the import and manufacture.

So, let us look at the hypothesis: That the Brady Bill and the AWB were a cause of decreased gun deaths. First, let me think of how this could be. The arguments put forth at the time and still put forth by supporters is that criminals could no longer buy their guns from legal sources (even though FBI statistics show they weren't buying them from legal sources) and hence wouldn't have a gun to commit a gun crime.

The Brady Bill only dealt with guns, and handguns at that. But the vast majority of gun crimes are committed with hand guns so we won't worry about rifles and shotguns. So this one is easy. A quick way to see if the Brady Bill is responsible (fully or in part) for the decrease in firearm deaths is to compare it to other deaths (falling, stabbing, suffocation). Except that the vast majority of firearm deaths are homicides and suicides, so we would really have to compare it to other common causes for homicide and suicide (stabbing, suffocation).

The AWB is a little more tricky. Since it was part of a larger crime bill (which included putting more cops on the street), we need to isolate it. We'll start by ignoring that assault weapons are used in less than 3% of all gun related crimes (which means at most the change we see should be 3%) and rarely, if ever in suicides. Next, because of the sunset provision of the AWB, we should see an uptick in the amount of gun deaths when the "ban" expired because all of the other provisions remained. All other forms of death typical of homicide (stabbing, bludgeoning, suffocation) should remain the same since they were unaffected by the AWB.

For the dataset, I'll be using the CDC WISQARS database (same as what Linoge used) from 1981 to 2006. The great thing about WISQARS is that besides just deaths, I can look at injuries as well (since not everyone dies from a gunshot wound). Their injury reports only go back to 2001 so there is limited amount of data to analyze.

So let's get started looking at some charts. All of the charts have a rate per 100,000 population so that you can get an accurate year by year comparison. This first one includes all deaths by the cause listed (homicide, suicide, legal intervention, accidental).
You can click on it for a larger image. The bold line is number of firearms per 100 people. Notice that there is a steady increase (indicating that the number of firearms in the country is growing faster than the number of people). The dashed line for all deaths uses the right hand numbers, all other lines use the left hand numbers. I know all of the lines make it busy, and since we are looking for correlations, the actual value of the datapoints is not that important, what is important is the shape of the line.

Now the Brady Bill was passed at the beginning of 94 was the first full year of its effect. The major decline of firearm deaths began in 93, however there is a slight increase in the slope of decline so if this was our only dataset, then I can see how people may make the argument that the Brady Bill helped. That still doesn't explain the previous decline. The AWB was passed at the end of 94. Interestingly enough, there is no increase in slope from 94 and 95. To me this indicates that the AWB had no effect (which as I said before was bundled with a crime bill that did a lot of other things).

Now let's look at the backend. The NICS came into effect at the end of '98. This ended the 5 day waiting period, a background check was still required. At this point the gun deaths level off. One explanation that could be made is that without the "cooling off" period, people were committing more crimes and suicides. Except that I said the deaths leveled off, there has been a very gentle fluctuation since then but essentially the rate has remained the same. In other words, gun deaths are happening at the same rate as immediately before the NICS came into effect. For the AWB, it ended in 2004. There is a very slight increase but the flucuation is no more than the increase after 1999 or the decrease after 2002. Again during this entire time, the number of firearms per person in the country has been increasing (and presumably the number of firearms per criminals as well? maybe...).

Another interesting factoid, more people are now die of poisoning than firearms. The last thing to point out with this graph is a couple of the lines that are looking similar to the firearm: cut/pierce and struckby/against. Let's blow those up to look at them better.
Pretty cool huh! Firearms use the right scale, the others are on the left. It is easy to see the correlation between cut/pierce and firearm deaths. Almost like they are related. Struck by/pierce is not as easy to see since the rate is so low, but if we blew the scale up more, the general trend is there but the shape is note quite the same. What is important here is that the Brady Bill and the AWB had absolutely nothing to do with knifes or blunt objects. So if these are showing the same trends as firearms, then one would be hard pressed to explain the mechanism for the Brady Bill and AWB causing knife and bludgeoning deaths to decrease.

So in summary, when we only look at firearm deaths, it appears that there may be evidence to support the Brady Bill affecting the rate of firearm deaths, however once deaths by other means are looked at it is found that they have similar trends and were not affected by the Brady Bill. So more than likely, the change in firearm, knife, and bludgeoning deaths have a common cause. But let's not jump to conclusions just yet!

On to Part 2 ...

1 comment:

  1. Reputo,

    Great analysis. I fing it interesting that people who support rights tend to find, analyze and study the data available, yet those who try to restrict rights downplay or don't use facts.

    Keep up the great work Sir