"Over the past two decades, economists have tried to structure their work, as much as possible, to resemble scientific experiments. This “credibility revolution” is an attempt to explicitly link causes to effects, and sweep aside the old criticism that correlation is not the same as causation. One of the advantages of the large tax database Chetty and his colleagues constructed is that it allows “quasi-experiments”—clever statistical methods that approach the power of a true experiment without requiring a researcher to, say, randomly assign children to live in different cities"