~~Moved from GR~~
I have seriously mixed feelings about Freakonomics, so be prepared for a very opinionated review. Like The Tipping Point, it was written by a journalist, and is an extremely engaging and entertaining read. It certainly has more scientific merit and empirical backing than Gladwell's book. It also presents several important statistical points that it is critical to understand: for example, at several points, the book reiterates that correlation does not imply causation. There is also an intuitive, if vague, description of Bayes' Law. What I loved most about the book was its quirky and off-the-wall thinking, the entertaining narration, and some of the very clever insights into the world of economics. And no matter how much I rail against his abuse of statistics, I picked up the second book from the library, which goes quite far in showing how well-written and entertaining I found it.
Unfortunately, the impact of the book was seriously lessened by what I saw as a mixture of dogmatism and hypocrisy that undermined all of the authors' conclusions. Take one example: in the start of the book, and in a later chapter, Levitt discusses the sudden fall in crime rates. He belittles all the other theories (more policemen, better policing, etc), claiming that they use correlation to imply causality. He presents his own: abortion, which he proclaims is "the" correct theory and "the" reason. Wait. So he knocks down and belittles all other scientists for using correlation to imply causality...and then uses correlation to imply causality? And dogmatically claims to have found the "right" explanation? If anyone tells you they have the "right" solution without having proved that their solution is both necessary and sufficient, everything they say is bull. I'm not really sure if this dogmatism really stems from Levitt; apparently the journalist author wrote most of the book, and he seems to practically deify Levitt. Every chapter begins with an excerpt from an NYT article that spews hyperbolic praise of the young economist, and throughout, Dubner speaks of Levitt's insight with a kind of adoring awe I found rather irritating.
I admit that I enjoyed at least thinking about some of the "shock factor" ideas that Levitt presents; for example, the life of a crack cocaine dealer. But again, his dogmatism, his insistence on being right, spoiled a lot of this for me. In my opinion, when your experiments and theories involve sensitive issues, you need to be very careful and realize that there is no way for a human being to live on the earth and not have biases and preconceptions. There is no such thing as being unbiased. Period. Levitt, for example, seems to have a real hangup about race. Yes, he goes on to show that (duh) there is no difference in intelligence between races as soon as socioeconomic factors are equalized. However, in almost every statistic in the book, he splits the numbers along racial lines. He almost never looks at age, socioeconomics, or any of a dozen factors; he usually splits statistics by race. And if that doesn't tell you about his preconceptions, I'm not sure what will.
I enjoyed the book (even though parts clearly made me boilin' mad) because he presented enough of the facts that I could at least determine when he had correctly used or horribly misused statistics. Not only did he keep treating strong correlations as causal truths; he also continually failed to apply Bayes' Law correctly. Take the following example, where he compares swimming pool safety to gun safety. He says that there is 1 drowning per 11,000 residential pools and 1 child killed by gun for every 1M+ guns. He therefore concludes that it's safer for your child to come into contact with a gun or be in a household with a gun than in one with a pool. Right?
Well, only if you have a ridiculously simplistic viewpoint. Because what he concludes is NOT what he calculates. What probability is there that a person who has a pool also has children? I bet it's higher than the probability that a person who has a gun having children. But Levitt ignores this. He just takes the total number of guns and the total number of pools in the US, and the total number of deaths from each type. Exaggerate this spin technique a bit and you can also argue that a child is more likely to get Alzheimer's in Florida than in NY. Why? Because the number of people with Alzheimer's is higher in FL than in NY (the population is older). There are probably at least as many children in NY as FL, so the probability of a child having Alzheimer's' is much higher in FL, right? Uh, no.
In conclusion, although I enjoyed reading it, I found the book dangerous. Levitt presents this adorably naive idea that math never lies, and that his conclusions are therefore totally unbiased. As he says, "freakonomics style thinking doesn't traffic in morality." What he appears to miss totally is that it is ridiculously easy to misuse statistics even without intending to, because it does traffic in your preconceptions.