Dani Rodrik uses his own book's Amazon ranking to run an experiment that seems to show taking out book advertisements in the Sunday New York Times is meaningless. He's surprised by that outcome, and I'm surprised by his surprise. Who makes book buying decisions off a tiny ad? That said, no one has yet run any kind of experiment showing that book ads in The American Prospect don't work, and such, I suggest that all new authors buy many of them, preferably at inflated prices.
