Statistical Homicide

This post appeared on the Nashville Scene‘s Pith in the Wind blog.

In a Tennessean op-ed last week on the proposed budget for Nashville and the tax increase it contains, local businessman Nathan Massey made a statistical claim that more money for policy officers reduces crime:

Let’s look at crime statistics. There were 51 homicides in Nashville in 2011. That is the lowest number of homicides since 1966. That suggests a direct relationship between investment and return.

I won’t comment on the merits of the budget or the tax hike,* but I can put on my social scientist hat (or striped shirt, to upgrade the metaphor) and referee the merits of a statistical argument. And on the matter of a lower number of homicides supporting a relationship between investment and return I’m throwing yellow flags: extremely illegal procedure.

Massey’s factoid may be accurate — the number of homicides has receded to the terrority it occupied back in the 1960s — but the meaning of those numbers is anything but clear. If you want to understand trends in violent crime, murder is perhaps the least reliable thing to look at. It is well understood that drops in raw homicide numbers can be explained in significant measure by advances in medicine and technology that make victims of attempted murder likely to survive. A landmark 2002 study, “Murder and Medicine,” published in the academic journal Homicide Studies (and who doesn’t want a subscription to that next Christmas?) showed that it’s not violent assaults that dropped over the last four decades of the twentieth century — nationally the rate of aggravated assault quintupled! —  but rather the lethality of assaults. The study (summarized by the NIH here) found that but for the dramatic drop in trauma mortality, murder rates by the turn of the century would have been much higher than they were.

A glance at long-term trends in violent crime in Nashville (using the FBI’s Uniform Crime Reporting statistics, synthesized for Metro Nashville here) bears out this skepticism about the meaning of a drop in homicides. Looking at rates of violent crime in Davidson County, rather than numbers, we find that while, yes, the murder rate (per 100,000 population) has dropped by more than a third since the mid-1960s, rates of aggravated assault, rape, robbery, and violent crime overall have more than doubled. On an upbeat note, though, these rates have noticeably backed off their peaks during the 1990s.

As a New York Times writer put it at the time the “Murder and Medicine” study came out, we “aren’t any less murderous — it’s just getting harder for us to kill one another.” It is, by extension, statistical malpractice to use long-term trends in the raw number of homicides as evidence either for or against the efficacy of law enforcement spending.

*_{No can do, owing to spousal unit on Metro Council who will vote on the thing.}