I think we should embrace the anti-federalist mood swing. Allow states such as Pennsylvania to embrace talent immigration as each sees fit. Better yet, let cities decide. H-1B visas effectively tie foreign born labor to the employer through sponsorship. Municipalities could act in the same capacity, making the visa contingent on urban residence and site of work. Various schemes could be concocted to enhance geographic mobility. Green cards would issued after a few years, well before the end of the federal queue.
This does make a certain amount of sense. In the realm of actually existing subnational immigration policies, I'm most familiar with that of Québec, where--as I wrote in June of last year--concerns over the growth of English led to the adoption of a policy explicitly favouring Francophone and French-leaning immigrants over others (French and Senegalese and Congolese versus Britons and Indians and Guyanese, say). This shift has arguably made immigration more popular in Québec, removing the fears of language shift from French, and has the potential to provide Québec with the workers--including skilled workers--that it will need as the provincial population ages. The rest of Canada obviously doesn't share the priorities of Québec and wouldn't be as responsive to local concerns. So, inasmuch as the subnational jurisdiction of Québec's control over immigration policies go, it seems a relative success.
The problems with sub-national immigration policies? Immigrants don't necessarily fit the slots allotted to them. As I noted above, the gap between immigrant and native-born wages in Québec is even worse than in the rest of Canada, a product of many things including the non-recognition of skilled workers and difficulties with social integration. In many cases, it's not especially clear that local control over immigration would be an improvement.
The second problem is that of mobility. For municipalities to have control over immigration, as the above blog goes on to sugget, strikes me as a very bad policy move. Immigrants have to be mobile, geographically as well as socially, and a municipality doesn't necessarily offer sufficient scope. In Québec the overwhelming majority of immigrants may be concentrated in Montréal, but this isn't because they're forced to live there. Rather, immigrants are concentrated in Montréal because that's where immigrant communities have formed neighbourhoods, dense social networks, and the like. Restrictions on mobility are especially problematic if--as some propose--immigrants are assigned residences in hinterlands in an effort to try to boost stagnant or declining populations. The waste of potential in those cases is arguably as much a moral problem as an economic one.
Finally, there's the question of whether immigrants will stay in their localities once probationary periods are up. My native Prince Edward Island has an immigrant retention rate of 25%; for cited reasons of wages (the poorest province in Canada) and social integration, most immigrants do not stay. Many sub-national jurisdictions may not keep as many immigrants as planned--in Canada, the differences between the have and have-not provinces on this metric is notable.
In conclusion? Sub-national immigration may be a useful idea, but it's one that definitely has its serious issues. It may produce short-term gains, but those gains can quickly be dissipated with bad planning and bad underlying conditions. Beware.