In many ways, a brilliant book. If the 2nd half was as good as the first half, I'd have given it a 5 star rating.
Steven Levitt knows how to look at data, and better than that, he knows how to find interesting results in the data that anyone can access. He asks the right questions, he finds the best way to get the answers that he wants, and fortunately for the rest of us, the facts he finds interesting are interesting for the rest of us.
His big revelation was the effect abortion was having on the crime rate, but there are lots of other things that he looked at that were immensely cool and interesting. How do you know if sumo wrestlers are cheating? We've got all this data to look at, what would a sumo wrestler cheating look like? How about teachers cheating on standardized exams? How about real estate agents selling their own houses? How are crack organizations like McDonald's franchises?
Levitt knows exactly how to look at his data sideways, and how to eliminate irrelevant factors that might confuse anyone else. It really is a remarkable display of setting up an experiment, even if it is with events that have already happened.
Unfortunately, by the middle of the book, it felt like Levitt had hit a limit and didn't have anything quite so interesting to say. Sure it was pretty cool - but not nearly as mind blowing as the first half. After the 1/2 way point, he simply took lots of data and did correlation analysis of the results and how one variable was related to another (do the names people get affect their success in life for example). Still - the ultimate causes to the effects remained unknown, and we're left with Levitt's speculation. As insightful as his speculation is, it's still just speculation.
Still - well worth your time, especially for the first half. A truly original book.