http://blog.robpitingolo.org/2010/05/where-smart-people-live...
the specific graphs being:
http://1.bp.blogspot.com/_UnYG5zOkgEY/S_bRIxhp8CI/AAAAAAAABn...
http://3.bp.blogspot.com/_UnYG5zOkgEY/S_bRVVyNVmI/AAAAAAAABn...
They look like you'd expect.
(smart/area) / (population/area) = (smart/population)
I don't really care if the population is more educated if they're really spread out. Assuming I have some way of selecting people who are smart—which I definitely do in NYC—I'm going to get to see a lot more of them in a place with a higher smart density.
Does that compete with NYC and SF? No. However, density undervalues western cities and overvalues dense ones. Unfortunately, there's no statistics for "people with whom you can have coffee within 20 minutes of travel."
I'd say it comes down to whether you're interested in a bounded product (infrastructure) or an unbounded one (innovation). A good argument could be made for spreading degrees around geographically, focusing on area instead of population, but it couldn't be stated with any accuracy without a good model of the likely nonlinear effects of degree densities on both types of goals.