Over 60% of transit trips are for work (See Commuting in America III). This is compared to just under 20% of overall trips. This means that the focus on where people work is important in monocentric cities such as Nashville. I'm not saying that residential density is ultimately unimportant. But I believe its less important for starting a transit system and more important for growing it. There are lessons on this in previous research works that we tend to ignore.
To be honest I hadn't really read Zupan in full until more recently on account of there is just too much to read in general. But when I caught up on it, the findings are quite interesting and get you wondering if we've been looking at this whole transit and development thing all wonky. In their seminal work Public Transportation and Land Use Policy (1977), Jeff Zupan and Boris Pushkarev made the following observations based on the existing data at the time:
Pushkarev & Zupan Pg 174-175:
1. Clustering or dispersing nonresidential space. Suppose 10 million square feet are to be added to a growing urban area. One option is to put the floorspace into two highway oriented non residential clusters, each 5 million square feet in size. Another is to create a new downtown of 10 million sq ft. In the second case, per capita trips by transit within a 3 to 5 mile radius will be 50 to 70 percent higher than in the first case, keeping residential density the same.All of this says that increasing your downtown size, and putting dense housing near downtown is likely to increase transit ridership. Now this goes so far when we're talking about polycentric regions and employment clusters that we see today such as Tyson's corner etc. But ultimately I think this work has lessons for those places as well. It will be interesting to see where the next decade of TOD research heads because ultimately I think this is a part of the research that needs to be explored in greater detail.
2. Enlarging downtown size or raising nearby residential density. Suppose the options are to double the size of a downtown from 10 to 20 million square feet, or to double the residential density within a few miles of it from 15 to 30 du/acre. The former will increase per capita trips by transit three to four times more than the latter.
3. Increasing residential density near downtown or farther away. Suppose the options are to double non-residential density from 5 to 10 dwelling units per acre either within one mile of a downtown of 10 million square feet or at a distance of 10 miles from it. In the first case, public transit trips per capacity in the affected area will increase seventeen times as much as the second case.
4. Scattering apartments or concentrating them near transit. Suppose a rapid transit station is located 5 miles from a downtown of 50 million square feet of nonresidential floorspace (the 1976 size of Newark). At a density of 15 du/acre, the square mile surrounding the station will send about 620 trips a day downtown by transit. Suppose speculative development scatters apartments throughout the square mile, raising its density by 20%. This will increase transit ridership at the station by about 24 percent. Yet if the apartments are clustered within 2000 feet of the station, preserving the rest of the neighborhood intact, transit ridership will increase by 34% or more; at least a car load of 62 people a day will be added not from any increase in average density within the square mile, but only from a new arrangement of the new development within it.
"Thus land use policies which will do most for public transportation are those which will help cluster nonresidential floorspace in downtowns and other compact development patterns. Downtowns of 10 million square feet of gross non residential floorspace, if confined within less than one square mile, begin to make moderately frequent bus service possible and to attract an appreciable proportion of trips by transit. By contrast, downtowns of 5 million square feet can support only meager bus service. Spread suburban clusters of nonresidential use can only occasionally support meager bus service, if they contain shopping centers, or if they are surrounded by residential densities in excess of about 7 du/acre.
Residential density is less important for transit use than residential location, ie proximity to a downtown of substantial size or proximity to a rail transit line. If greater transit use is the goal, it is more important to put housing close to a downtown than make it high density. In fact, moderate residential densities in the range of 7-15 du/acre can support moderately convenient transit service by any of the transit modes reviewed in this book. Of course, densities higher than this will support better service, as well as more trips on foot. Thus, a strongly transit-oriented city such as Montreal has an average density of 35 dwellings per acre; attached two-family houses form an important part of its newly developed neighborhoods. Evidence from New York suggests that the shift from auto to transit diminishes, and reductions in total travel per capita cease, at densities above 100 du/acre. This density can be represented by 13 story apartment houses covering 20% of their site; on transportation grounds, there appears to be no need to exceed this density. It is important to emphasize, though, that a 13 story building located amid open fields will make no contribution to transit; it will only make a contribution if embedded in existing urban fabric, close to downtown or a rail station.”