My friend and co-author Murat Iyigun points me to an interesting theoretical paper by Jess Benhabib and Boyan Jovanovic on optimal migration from a global perspective. Contrary to what comes out of textbook models, freedom of labor mobility does not maximize global output in their model. The reason is that low-skill workers are taken to exert a negative externality on average productivity in their host economy. From an output-maximizing perspective, it would then be optimal to geographically segregate low-skill labor completely--i.e. close off all international labor mobility. Only if we care about the global poor sufficiently, will the distributional considerations compel us to open our national borders.
Another reminder that economics is rarely as simple as its popularizers often make it seem.