I smell a research project

Transit outcomes don't really depend on how much of this low density stuff there is in a city. They depend on how much high density stuff there is, and the average density of the whole city will tell you nothing about that.

 Average density is a useless statistic when it comes to transit planning. While the WFRC model has a field for % of pop within a half mile of the transit station, that may not be the best metric for transit planning.

Suggested:

..residential density is ideally the density of a single parcel: one house or apartment building. What percentage of the population lives on detached homes on quarter-acre blocks? What percentage lives in buildings taller than five stories? 

 WFRC has pretty good data for the UrbanSim model, data I've recently been told exists not at the Traffic Analysis Zone level, but on the parcel level. This changes everything, because it permits a housing units/households comparison on a per-parcel basis, giving a pretty good idea of the location of transit suitable density.

 I've never seen a map published containing the data, so I presume there must be some sort of privacy issue, although I can't see how the number of housing units per parcel could be considered invasive.

I wonder what kind of data exists as Public Use Microdata from the American Community Survey? The Census collects data on a per-household basis, but the information is geo-coded to a specific address. If the geo-code for census data can be matched to parcel data, it would be possible to connect the two. I assume it would be possible to do some sort of database operation to cluster all the numbered units into a single parcel ID....