Freakonomics has really popularized the idea of ‘natural experiments’ that economists can use to examine social questions. By using fancy econometric methods, they can give simple answers to complicated questions. I’ve always been dubious of this; it’s really a lot of correlational work, and as we always say, correlation is not causation! Case in point, one of the underlying assumptions of some of this work – that your month of birth is independent of income – is in fact wrong. Of course, that doesn’t mean the results are wrong (though the p value certainly is). It just means that one needs to be very careful when evaluating these things. But can I say what a cool finding this is?