considers that possibility in his recent column. About one in four bartenders has some kind of degree. Orszag draws heavily on this paper by Beaudry and Green and Sand, which postulates falling returns to skill. It’s one of the more interesting pieces written in the last year, but note their model relies heavily on a stock/flow distinction. They consider a world where most of the IT infrastructure already has been built, and so skilled labor has not so much more to do at the margin. This stands in noted contrast to the common belief — which I share — that “IT-souped up smart machines” still have a long way to go and are not a mature technology. You can’t hold that view and also buy into the Beaudry and Green and Sand story, unless you think we have suddenly jumped to a new margin where machines build machines, with little help from humans.
Rather than accepting “falling returns to skill,” I would sooner say that education doesn’t measure true skill as well as it used to.