It is commonly noted that there is an urban wage premium, perhaps because of agglomeration effects on productivity, or to compensate (in utility terms) for higher traffic and land costs. S. Lee presents evidence that, at least in some fields, there is a negative urban wage premium for high skill workers. The story is thus: high skill workers have greater taste for variety in consumption, and cities provide such variety; if the extra utility from consumption variety grows faster than rents, and utility is equal in rural and urban areas, then urban areas could show lower wages for high skill workers. Since high skill workers are relatively cheap in cities, they will be used in firms in higher proportion than low skill workers. Data on the medical profession bears this out: hospitals use a higher ratio of doctors to nurses in cities, the doctors are paid less in real terms in cities, and the nurses are paid more in real terms in cities than in rural areas. Further, these urban doctors tend to work in higher skill specialties, and be graduates of higher ranked medical schools, than higher paid rural doctors. This is yet more evidence of the consumer-driven rationale for cities: production is not everything.
I should note, however, that the data in the paper violates my major urban econ pet peeve by using MSA level data. Look at a map of MSAs in the US: they bear very little resemblance to what we might call “cities”, and it’s tough to argue that a worker 50 miles from the city center is somehow taking lower wages in exchange for urban amenities. I understand that a lot of data is given at the MSA level, but we really should be focusing on tighter definitions; my preference is the Census “urbanized area”.
http://strategy.sauder.ubc.ca/lee/papers/ability_sorting_and_consumer_city.pdf (Final WP – published version in JUrbE 68 (2010))
A Fine Theorem 
But, presumably, the specialization/division of labor we observe in cities, as opposed to the more generic occupations in rural area are about some kind of increasing return with fixed costs and size-of-market effects rather than preferences/tastes-for-variety which can be the same across locations.
Agreed, but check the data in the paper. The ratio of doctors to nurses (just standard doctors, not surgeons or anything) is higher in urban hospitals. It’s tough to justify this on division of labor grounds. Further, ooking at wage data, the doctors are relatively cheaper, not more expensive, vis-a-vis the nurses in urban areas (perhaps for consumption reasons) and therefore urban hospitals higher proportionally more doctors (only relative prices matter, of course).