Density and Public Transport

Arguments about density and public transport usage are notorious. Paul Mees dedicated an unfortunately large amount of his most recent book, Transport in Suburbia, to a rather implausible argument that there is little relationship between density and public transport usage – although in the process making a pretty helpful argument that service quality is also extremely important. A recent paper by Eric Guerra and Robert Cervero tackles this forever vexed issue, looking at the relationship between density and the cost-effectiveness of public transport provision. With Auckland’s future density somewhat up for debate, and the nature of where we invest our future transport budget very much a matter of great debate, this is an issue worthy of further analysis.

Guerra and Cervero set the scene:

Comparing the costs and the number of passenger-miles traveled for 54 American rail transit investments since 1970, we found wide variation in cost-effectiveness. The worst-performing system costs nearly 50 times more per passenger-mile than the best-performing. What factors distinguish the most successful transit investments?

Dense concentrations of people and jobs around transit stations are particularly important. Outside of Manhattan, Chicago’s Loop, and a few other urban pockets, however, most Americans dislike density. Many loathe it. For them, the “D” word means traffic congestion, crowded sidewalks, packed schools, long lines at the grocery store, and high crime rates. Without density, however, high-capacity transit tends to attract too few trips to offset the high price tag. As a result, there is a great interest in the minimum densities needed to support transit.

While Perth has shown that you can make rail work well in low density areas, this has required massive park and rides, comprehensive feeder services plus very widely spaced stations to ensure trains have a high average speed. All this clearly comes at a cost, plus Perth’s rail system hasn’t really ‘shaped development’ in the way that a system more focused on its walkable catchment might.

The issue of ‘minimum densities’ to support transit is particularly interesting when we look at Auckland’s future development areas – especially in the south where the railway line runs through the biggest area identified for greenfield development in the Council’s spatial plan. There has been research into this issue for quite some time:

In 1965, John Meyer, John Kain, and Martin Wohl wrote, “nothing is so conducive to the relative economy of rail transit as high volumes and population density. High population density increases the costs of all urban transportation, but substantially less for rail than for other modes.” They and other scholars found that rail transit, with its high up-front investment and high capacity, costs less than buses or cars only in corridors with high travel demand. Thus they found that rail was more cost-effective than buses or cars in high-density cities, while cars were more cost effective in low-density cities. The majority of job and population growth, however, was occurring in newer, low-density cities and in distant suburbs.

A decade later, Boris Pushkarev and Jeffrey Zupan estimated minimum density thresholds for different types of public transit. According to their calculations, net residential densities of 12 households per acre surrounding a 50-million square-foot central business district (CBD)—roughly the size of Los Angeles’ or Newark’s downtown in 1970—could support a cost-effective heavy-rail investment. Nine households per acre surrounding a 20-to-50 million square-foot CBD could, at that time, support a minimal light-rail investment.

12 households per acre is a bit higher than the general development standard we see in much of Auckland (around 400 square metre lots or 10 households per acre) – but not by too much.

To measure ‘cost-effectiveness’, the study looks at using a measurement of ‘net subsidy per passenger mile’ (including capital cost). As alluded to earlier, this results in an incredibly wide spread of cost-effectiveness results:

As there is such a massive range in cost-effectiveness, we should be able to get some interesting results when comparing this with the densities of the different areas the projects serve. But first, putting all the stations and stops of the various projects completed since 1972 together and sorting them by density shows a rather surprising result – the absolute vast majority of new stations/stops have been constructed in very very low density areas:

The authors then undertook a regression analysis between density and cost-effectiveness, as well as developing a ‘threshold of cost-effectiveness’ – which is a useful tool as it describes the level at which the new project was worth it, rather than just investing more in the existing system. This is described in a bit more detail below:

Using our recent investment and system data, we calculated minimum land use density thresholds for otherwise-average transit systems to be highly cost-effective. We defined cost-effective investments as investments that increased passenger-miles for a smaller estimated subsidy than either fare reductions or increased train frequencies on existing systems. Fare reductions, less expensive than increased frequencies, required an estimated $0.58 subsidy for each new passenger-mile. Roughly a quarter of the investments met this cutoff. They carried 57 percent of passenger-miles on the 54 investments in 2008. 

The relationship between density and cost-effectiveness is shown in the graph below:

The paper goes on to develop some pretty high density thresholds, but notes that these should be treated with a bit of caution as there are some big exceptions to the trends:

Transit-supportive density thresholds need to be viewed with caution. There is no one hard and fast rule that can be applied across all projects. Regression-based models mask considerable variation. For example, despite low surrounding densities, the Franconia-Springfield extension of the Blue Line in Washington, DC, is one of the best performing investments. Low capital costs, a plentiful supply of parking at stations, frequent train service, and good access to downtown jobs contribute to low costs per rider. By contrast, the Buffalo light-rail system is one of the least cost-effective, despite above-average job and population densities.

I think the most important thing to take out of the study is that some pretty widespread analysis of US cities has shown a pretty clear connection between density and the cost-effectiveness of public transport projects. This may end up being most useful in driving land-use policies to support public transport – particularly in relation to the level of upzoning around existing stations or target densities for new development in places that are within walking distance of train stations.

So while density may not be everything (after all you can have dense sprawl), it seems that if you’re trying to ensure investment in rail is going to be cost-effective, density certainly helps. (And this isn’t suggesting Auckland has insufficient density – as Matt pointed out recently, Auckland’s density is actually surprisingly high.)