Can medical advances explain the falling murder rate?

7/08/2013 06:27:00 PM
In a previous post, I discussed the fact that the 2010 murder rate was the lowest since 1964, and that falling murder rates appears to be a long-run trend since the 1990s that is likely to continue. Understanding the cause of that decline is important, because different causes have different policy implications.

One possible cause I suggested was that medical advances in treatment of trauma patients contributed to the falling murder rate--better treatment means fewer deaths and, by definition, fewer murders. In that post, I linked to a Wall Street Journal article that made precisely that argument, claiming in particular that murder rates have fallen even while gun violence has risen. I was happy to receive a comment from none other than Rick Nevin (famous for his research on the effects of lead levels on crime), who pointed to this post attempting to debunk the Washington Post's claim that medical advances explain the falling murder rate.

Let me start by saying that, by any standard, the Wall Street Journal article was shoddy reporting. Though the thesis remains plausible, the fact is that the author did not cite a single scientific study substantiating the claim. I did my own brief search of the medical journals and found no studies comparing the effectiveness of medical treatment for gunshot wounds over time. Hence, the Wall Street Journal article was pure speculation--what my macroeconomics professor would call a vigorous assertion--with absolutely no scientific basis for the grandiose title that proclaims "In Medical Triumph, Homicides Fall Despite Soaring Gun Violence." This may not be a medical triumph at all.

That said, I find the rebuttal to the article inconclusive as well. Here is the incidence of non-fatal gunshot victims from violent assault for 2001 to 2011:

The data comes from the CDC, estimated from data on visits to emergency rooms in hospitals across the country. You can see the general upward slope to which the WSJ alluded (though this is clearly way less than the 50% increase they claimed ); however, I want to make three points:
  1. The trend may not be statistically significant. ER data on wounds is not as accurate as data we have on murders, and the standard errors for each of the datapoints in the graph above are huge. For example, the upper bound of the 95% confidence interval for the lowest year 2002 is 19.27, higher than the estimate for the highest year 2008. That is, we can't even say that the difference between the lowest and highest estimates are statistically significant. That doesn't necessarily mean that the trend is statistically insignificant, but I can't really assess that without access to the full dataset. 
  2. Finding that gun-crime rates have risen while the murder rate has fallen is not, by itself, evidence of improved medical effectiveness. A variety of outside factors could easily explain this apparent contradiction, including a change in the types of guns people use to commit crimes, or a change in the type of victims typically targeted for crimes.
  3. A rise in non-fatal gunshot wounds could be consistent with a fall in gun crimes if the decline in the murder rate exceeds the improvement in effectiveness of medical treatments (which the data does seem to show).
At the same time, I'm also somewhat skeptical of Rick Nevin's claim that lead levels are the main driver of violence levels. This is a popular idea in the wonkosphere, but not so popular among the medical researchers I work with on a daily basis, for one reason in particular: lead fails to explain the differences in violent crime across countries. What really matters, developmentally and behaviorally, is the blood levels in children. A quick second of googling brought up this chart from the World Health Organization on blood levels of children in various European countries:

Some of these are relatively recent measurements, but most enough are long enough ago that those children would now be at prime crime-committing age. Here's a graph of the murder rates, over time, in those same countries, from the United Nations:

This certainly isn't a scientific statistical analysis, but you can plainly see that lead levels is uncorrelated with crime across countries. The US actually has generally had lower blood levels than most of Europe (my guess is that WWII had a big impact on this), yet has some of the highest murder rates. Moreover, for the few countries for which we have two measurements of blood lead levels taken at different times, there is no within-country response of the murder rate to the change in lead levels.

Of course, this isn't meant as a refutation of Rick Nevin's thesis per se. But I do think that the data suggests that lead levels are not the biggest determinant of crime rates. I tend to think that more old-fashioned explanations of the crime rates, like race relations, fit the data better. Yet, we should recognize that lots of things, probably including lead, all affect the crime rate in different ways and to varying degrees.