Levitt and Dubner apply economic reasoning to questions outside conventional economics, using incentives and data analysis to test claims that people take for granted. The chapters cover cheating among schoolteachers and sumo wrestlers, the information asymmetry used by estate agents and by the Ku Klux Klan, the economics of crack dealing drawn from Sudhir Venkatesh's gang financial records, the causes of the 1990s crime decline, the effect of parenting on child outcomes, and the significance of names.
The method is the point rather than the topics. Levitt looks for natural experiments and for data that reveals behaviour people do not report. The teacher cheating analysis identified suspicious answer patterns across Chicago classrooms. The sumo analysis found win rates in matches where one wrestler needed a single victory that reversed in later meetings between the same pair. The estate agent analysis compared how long agents kept their own houses on the market against their clients' houses.
The most contested chapter argues that the legalisation of abortion in the United States after 1973 explains a large share of the crime decline in the 1990s, through the reduced birth of unwanted children who would have been at high risk of offending. This produced substantial academic dispute, including a coding error identified by Foote and Goetz in 2005 that weakened the original result, though Levitt and Donohue have continued to defend the finding with updated data.
The book is a set of case studies with no unifying theory, which the authors state. Some chapters have aged poorly and the parenting analysis rests on correlational data. Its lasting value is demonstrating how incentives and data can test claims that everyone assumes.
Key takeaways 8
Incentives explain behaviour across contexts that appear unrelated, and identifying who benefits explains outcomes that moral explanations cannot.
Experts exploit information asymmetry against their own clients, which estate agents demonstrate by keeping their own houses on the market longer for higher prices.
Cheating can be detected in data by finding patterns that honest behaviour would not produce, including answer sequences on tests and win rates in specific matches.
Most street-level drug dealers earn below minimum wage, and the gang financial structure resembles a franchise with a tournament payoff at the top.
Conventional wisdom about causes is frequently wrong, and the crime decline of the 1990s is not explained by the policing and economic factors usually cited.
What parents do matters less to measured child outcomes than what parents are, since most measurable effects track parental characteristics rather than parenting practices.
Names correlate with outcomes because they signal the parents' characteristics rather than because they cause anything.
The abortion and crime finding is the book's most contested claim and has been challenged on coding and specification grounds since publication.
Who should read it
Readers who want a demonstration of how incentives and natural experiments test received explanations, useful as method rather than for its specific conclusions.