Uneasy Rhetoric

Freakonomics by Steven Levitt and Stephen Dubner purports to use the tools provided to economists to reexamine some interesting life questions.

For example, Levitt ponders why the crime rate fell in the 1990s when most criminologists expected an increase. His answer? The children most likely to commit crimes had never been born. Fetuses aborted as a result of the Roe v. Wade decision would have been reaching their prime crime-committing years in the early 1990s. (Throughout, I refer to “Levitt,” because, as I understand it, the work was his. Dubner was just a journalist who helped him get it all down.)

In another chapter, Levitt plays with the numbers to prove (or at least to strongly suggest) that there is cheating going on in sumo wrestling. And, the authors demonstrate how real estate agents cheat their customers by hastening them into a selling position. The authors show that real estate agents tend to hold out longer for better prices when selling their own homes.

Often, academics think within their own disciplinary boxes. Levitt, with no particular political, sociological, anthropological, or cultural ax to grind, can look at the data without preconceived assumptions.

For this reason alone, Freakonomics is worth reading. Data analysts should chew it over carefully. Casual data observers should heed the authors’ warnings about correlation and causation.

If the book has a fault, it lies in its authors’ elevation of economics over other social sciences. Levitt concludes his book by restating something he said at the outset: “if morality represents an ideal world, then economics represents the actual world.” Despite his cleverness, despite his ability to use data to “think outside the box,” Levitt has fallen victim to a conceit shared by many economists and even by the public at large. Economics, with its basis in probability, statistics, and calculus, can only ever approximate the real world, ceteris paribus.

The kind of data analysis Levitt does - nothing overly mathematical or theoretical (except for a bit of regression) - and that much basic economic theory is based on, requires rationality, and the “actual” world just isn’t rational. Anyone who has ever done regression analysis knows its biggest weakness: there is always one more variable for which to control. Even if you’ve controlled for every variable in your dataset, there might be one more that simply wasn’t observed. Life is messy that way.

Levitt’s observations are compelling if not downright convincing. But shouldn’t we take his observations with the same grain of salt he insists we use on the observations of criminologists, policy analysts, and others who use data analysis to reach their own, possibly different, conclusions?