Of course there's something special about the sample Levitt chose: it's all real estate agents. If the statistical results are sound then we can infer some cha

Of course there's something special about the sample Levitt chose: it's all real estate agents. If the statistical results are sound then we can infer some characteristics about the population. But in this case, the sample came from a population of real estate agents. There is a fixed effect found across the whole sample: they're in the housing business, not the headhunting business. No recruiters in that population, so we can't infer anything about recruiters in general. Sorry to argue semantics, but this is a basic sampling concept that people forget about.