LAPD gets...well...Freaky
By Ian
The popularity of Freakonomics has done two bad things, in my estimation. First, it cast the work of economics that looks at things other than say income, GDP, or currency markets as aberrant. And second, the focus on "economics" downplays what I think is the real power behind Levitt's work: statistics. Sure, sure, econ is concerned with incentives and all that; so are a vast number of other social sciences. But almost without fail, the tidbits in Freakonomics are the result of assiduous and often inventive efforts at working with data. (Even, in my estimation, the Sumo examples -- I know of no place in statistics classes where students are told not think about why certain patterns or correlations are occuring; there is nothing unique to the economics profession that resulted in thinking about the win/loss record for on-the-bubble wrestlers.)
I quote the inestimable Will Wilkinson's "Caesar's Bath" contribution:
Levittolotry. Yes. He's smart and interesting, but his work isn't that unusual, and he doesn't walk on water. He's a super-clever, McGyver-esque technician, able to conjure up a useful empirical study out of a paper clip, a length of string, and a stick of gum. It's sweet, but not filling. I want theory.
Far from the dry work of people "who are good with numbers but lack the personality to be accountants" (as the old joke goes), the work of a well trained statistician could be every bit as "freaky" as that of an economist.
To wit:
LOS ANGELES -- Dogged by scandal, the Los Angeles Police Department is looking beyond human judgment to technology to identify bad cops.
And what is this magical machine doing, one might wonder, to ferret out bad cops?
Statistics.
The system, developed by Sierra Systems Group and Bearing Point, mines databases of complaints, pursuits, lawsuits, uses of force and other records to detect patterns that human eyes might miss or choose to ignore.
In other words, there are correlations between certain kinds of "flags" and the tendency for a cop to be a problem. The resorting to some mechinized process seems to be more a political decision -- "Hey, man. I didn't say you were bad. The computer just spit your name out. No one working for the Rat Squad here, boyo." -- than anything else. Statisticians could very well be saving lives and making the police force a better group all around.
Please note that the tetchy sound of this is not meant in any way to castigate the economist himself--the credit and rewards are his due, and the heavy-weather is not of his own making (aside from going along with the terrible title). Reading his actual papers is gripping, even when some of the technical details fly over my head. But of course, as even he is wont to say, I must consider the economics of the situation. If this is what people think all economists do -- and is a skill or field of work available only to those trained in economics -- then there could well be a surge in the number of people appling to econ programs. And the last thing I need is more competition.
For continually interesting talk about modeling and stats, I refer you to the Statistical Modeling, Causal Inference, and Social Science weblog.
Comments