Cities are not built on flat, developable circles, but are rather constrained by geography: steep slopes, oceans and wetlands do not permit new housing. Saiz develops a new database of geographically-constrained land within 50 miles of US metro area centers, and shows that increased geographical constraints both increase housing prices and increase the housing supply price elasticity. This effect is not simply a coastal effect: it is evident even within the subset of coastal cities. An immediate question, then, is why anybody lives in constrained locations. In a clever model, Saiz shows that geographically constrained cities which we witness in the data are seen to exist ex-post because of favorable productivity and amenity shocks in the past. The model suggests that, in order to still exist as a city, geographically constrained locations must show higher productivity and higher amenities, and further that there should be no correlation between population and geographical constraint. All three of these are evident in the data. Further, Saiz confirms the “homevoter hypothesis”, whereby stricter land use regulations are endogenous to higher home prices – in this way, restrictive geography increases home prices directly by constraining supply, and indirectly by increasing land use regulation.
http://real.wharton.upenn.edu/~saiz/GEOGRAPHIC%20DETERMINANTS.pdf
A Fine Theorem 
I certainly would like to see the data.
My priors would be closer to a demand model, or at least attempt to incorporate the effect of demand on price (can’t Saiz give a little room, literally, figuratively?)
I don’t get why there should be poor pop/constraint correlation?
Maybe it’ll be clearer when I read the paper.