[Books]

Freakonomics.

I finally got around to reading Freakonomics, the most-hyped book in recent memory. You may detect in my characterization high expectations, and indeed I had them. Everybody on the planet--readers, reviewers, my officemates--seemed to regard this book, about Steven D. Levitt's economic theories, as some kind of oracle. Even the book itself, in which chapters begin with little hagiographic bits about Levitt--did its best to hype the theories as just short of devinely insightful.

It was ... all right. Actually, it was better than that: it was a good, interesting guide for the uninitiated reader into the world of statistical analysis. What people who work with piles of quantitative data know--and what few people who don't work with data recognize--is that research isn't about good math. It's about good questions. The best (catty readers might say only) example is a hypothesis Levitt has that the drop in crime is not related to any of the direct efforts to address it--more cops, tougher laws, more prisons. It's the unintentional result of abortion, which removed millions of unwanted babies (who are disproportionately likely to become muggers) from the streets.

I have no idea if this is true, because Levitt doesn't show his work. He seems pretty successfully to show the effects of other efforts to dull crime had weak or no correlations with the dropoff--though again, we're getting the made-for-TV numbers. The exercise is good at showing how you can use piles of numbers to test hypotheses, and how the results are little correlated with conventional wisdom.

What the book doesn't do is offer anything near the kind of revolution readers, reviewers, and the book itself promises. Levitt may be an interesting guy, and his mind is lively and unexpected, but don't pick up the book expecting to find answers to life's great riddles.