Right-thinking lefties everywhere have had their knickers in a twist over the House Republican proposal to curtail National Science Foundation funding for social science (especially political science), while maintaining funding for the “hard” sciences like physics, chemistry, etc. Well, perhaps papers like the one from Matthew Ranson, posted the other day on the Social Science Research Network, suggests why someone might look askance at social science, whether funded by the government or not. To paraphrase that great line from Randall Jarrell, you have to read it not to believe it. Herewith the abstract to “Crime, Weather, and Climate Change”:
This paper estimates the impact of climate change on the prevalence of criminal activity in the United States. The analysis is based on a 50-year panel of monthly crime and weather data for 2,972 U.S. counties. I identify the effect of weather on monthly crime by using a semi-parametric bin estimator and controlling for county-by-month and county-by-year fixed effects. The results show that temperature has a strong positive effect on criminal behavior, with little evidence of lagged impacts. Between 2010 and 2099, climate change will cause an additional 30,000 murders, 200,000 cases of rape, 1.4 million aggravated assaults, 2.2 million simple assaults, 400,000 robberies, 3.2 million burglaries, 3.0 million cases of larceny, and 1.3 million cases of vehicle theft in the United States.
Now, the idea that climate change would lead to more crime isn’t exactly brand new—it’s been on the infamous WarmList of all the things caused or made worse by climate change for some time now (though the Warmlist link to previous crime predictions has gone dead). And it makes some intuitive sense: of course assaults go up in hot weather. (But did we need social science to tell us? This seems another instance of social science “proving” “with rigor” what most people know by common sense.) On the other hand, this suggests those Freakonomics guys may have missed a possible cause of the declining crime rate of the last 20 years—the increasing use of air conditioning! Which also suggests a partial remedy in the event of significant climate change. In fact, buried in the text of the full study is this “why-am-I-bothering-to-read-this?” caveat:
It is also worth emphasizing that the estimates presented here do not take into account longer-term adaptation mechanisms. If climate change does cause a permanent increase in the frequency of crime, people in affected areas will have the opportunity to modify their behavior to avoid being victimized. Furthermore, it is likely that law enforcement agencies will respond with increased policing activity. The potential for such actions suggests that the estimates in this paper should be viewed as an upper bound on the potential impacts of climate change on crime.
Above all, can anyone really give credence to the specificity of this 90-year projection, given that there isn’t a social scientist anywhere who can convincingly explain the rise of the crime rate in the 1960s-70s-80s, let alone the subsequent unexpected fall in the crime rate of the last 20 years?
Climate economist Richard Tol has been having great fun with this, tweeting “How long before a rapist argues ‘I’m innocent, climate change made me do it’?” (Tol, who is cited in the Ranson paper—as he was in the infamous Stern review that he also shredded—has serious criticisms to make, too: “Ranson paper confuses weather and climate, ignores seasonal integration, and suffers from omitted variable bias.”) I’ll look forward to Roger Pielke Jr. weighing in on this, too.
And what, pray tell, is a “semi-parametric bin estimator”? My occasional writing partner Ken Green emails to say that “It sounds like something Marvin the Martian would use to attack Duck Dodgers in the 24th and a halfth thentury! ‘Stand where you are, Earthling, or I will have to shoot you with my semi-parametric bin estimator!’”
I’m sure Mr. Ranson’s semi-parametric bin estimator represents the acme of contemporary statistical rigor (if you want to get a quick headache some time, try to have someone explain the fine points of “unit root” statistical analysis), but that’s exactly the problem—specialists in the social sciences tend to write for each other rather than an intelligent reading public, and can never be bothered to decode their methodology for non-specialists, which makes it hard for non-specialists to tell genuine rigor from postmodern jargon designed deliberately to obfuscate.
By the way, when we finally get around to having a trophy company work up the actual Green Weenie statuettes that we’ll send all the winners, I think we’ll make it in the shape of a semi-parametric bin estimator.