Beware of those Crime Statistics
Matthew Martin 7/01/2013 08:53:00 AM
Update: In response to Rick Nevin's comment below, I have written a bit more on the possible causes of the falling murder rate here.In a previous post, I critiqued the limitations of causal interpretations in the social sciences. In that context, you can consider this post a counter-point, warning about the importance of having an empirically grounded causal interpretation for your statistics.
The media has made much ado lately of reports that the murder rate is on track to fall to its lowest level in a century. Now, that preliminary estimate might be over-optimistic, but it certainly is true that the murder rate in 2010 was the lowest it has been since 1964:
All of these critiques are false. The data comes from medical examiners' and coroners' reports, which are extremely accurate and, more importantly, consistent over time in how they report deaths from murders. Moreover, the data sets are publicly available, and scientists would be able to tell if the published figures are inaccurate or inconsistent. However, the point remains that people who lived in the 1960s don't feel nearly as safe today as they did back then. But should they?
Actually, their sense that the world continues to get more dangerous is actually consistent with the empirical observation of a falling murder rate, for several plausible reasons, and that's where causality comes in. Turns out, unless you are a politician currently congratulating yourself for a job-well-done fighting crime, the actual cause of that falling crime rate has important real-world implications about how you should behave. Here is a non-exhaustive list of possible causes of the falling crime rate:
- a decrease in lead levels, causing fewer people to want to commit crimes
- an improvement in medical prognosis for gunshot and other wound victims
- an increase in law-enforcement capacity
- an increase in law-enforcement efficiency
- an improvement in citizens' crime-avoidance strategies.
Notice that each of these has different implications. If you are a city council member voting on whether to cut the police force, then if the cause is (1) or (4) you might want to support the measure, whereas if the cause is (2), (3), or (5), you probably should oppose the measure.
I don't want to discount any of these possibilities, but I do think that (5) has been the dominant effect. Think of all the strategies people have adopted since the 1960s to minimize their chances of being murdered: we lock all our doors now, we arm ourselves for self-defense, we've moved out of the inner-cities, we no longer talk to strangers, we don't stop to help people on the side of the road, we don't pick up hitch-hikers, we maintain distance from suspicious-looking gangs or people, we don't let people see our money or wallets in public, we take self-defense courses, etc. All of these crime-avoidance strategies are things we have learned to do every day, but that people generally didn't do back in the 1960s. And the thing is, they work.
All of this is to say that if crime-avoidance strategies like locking your doors are what caused the drop in murder rates, then you cannot look at the falling murder rate and conclude that you actually are safer. The data does not, in fact, say it is ok for everyone to start leaving their doors unlocked again.