In his recent New York Times article on scientific studies of baseball, Alan Schwarz discusses a study showing that sacrifice bunts benefited the 2008 Mets, who bunted a lot, but would have hurt the 2005 Red Sox, who sacrifice bunted very rarely.
This seems to me to be a mundane illustration of what we call the Roy Model: given two choices, some people will choose one, some the other — and they will select themselves into different treatments.
In many cases, no single treatment works best for everybody. Perhaps more people should get college degrees; but many of those whom public policy might encourage to attend college are better off stopping at high school. Oftentimes the individuals know best what treatments will benefit them, so twisting incentives by offering subsidies to alter outcomes in such cases will not be socially beneficial.
Too often in thinking about policy we treat people as if they are identical along dimensions that we cannot observe.
They aren’t, and I imagine baseball is just one of the many non-economic illustrations of how the Roy Model applies.