Facebook Google Menu Linkedin lock Pinterest Search Twitter


Public transport

Nov 7, 2012

How cost-effective are new rail transit projects?

Many newer rail-based transit projects aren't cost-effective. Too often it's assumed one transit project is as good as any other. More effort needs to go into building the right projects in the right places.


Net cost per passenger mile of building and operating transit by density around stations (Source: Guerra and Cervero)

Erick Guerra and Robert Cervero from UC Berkeley have published a new paper, Transit and the “D” word, examining the relationship between urban density and the cost of constructing and operating rail-based transit.

They looked at 54 light rail and heavy rail projects constructed in 20 US cities between 1970 and 2006. The paper is a summary of a larger one published here (the latter one’s gated though).

Transit projects with low capital costs look attractive but some also have low ridership. Some that cost a lot have correspondingly high patronage. For example:

The first section of the Red Line in Los Angeles cost more to build per route-mile than any other investment but had below average costs per passenger-mile. Because of its low ridership, San Jose light rail had among the highest costs per passenger-mile despite low investment costs per route-mile.

The authors accordingly look at both capital and operating costs and compare them against ridership to arrive at a reasonable measure of cost-effectiveness.

After netting off fare revenue, they find the lowest combined capital and operating subsidy is $0.14/passenger km for the Denver Central Corridor light rail project. So a commuter who travels 20 km each way five days a week relies on an average weekly subsidy of $27.

But the highest subsidy is almost fifty times greater – $6.48/passenger km for the Newark light rail extension to Broad Street. A commuter who travelled the same distance on this system requires an average subsidy of $1,296 per week!

The average subsidy across all 54 projects is $0.84 per passenger kilometre (around $170 per week for my notional commuter). Half the projects require a subsidy of more than $0.50/passenger km and 20% exceed $1.00/passenger km.

The authors emphasise that “a system with few passengers and a high price tag is, by most accounting, a poor investment economically, environmentally, and socially.” In their view, the key to cost-effective rail-based transit is high population density and, especially, high job density (hence their reference to the “D” word).

However they find the density of neighbourhoods around the transit stations they studied is very low. Only 26% of 526 heavy rail stations and 19% of 261 light rail stations meet the influential minimum population density thresholds for successful transit proposed by Boris Pushkarev and Jeffrey Zupan.

As the exhibit shows, Guerra and Cervero modelled jobs and population density against subsidy cost per passenger mile and arrived at their own thresholds.

The results…..suggest that, on average, light rail is more cost-effective than heavy rail in areas of up to approximately 28 residents and jobs per gross acre. With system-area densities near or below 20 residents and jobs per acre, Atlanta, Miami, and Baltimore appear better suited for light than heavy rail, while heavy rail is the appropriate choice in the San Francisco Bay Area and Washington, DC.

They argue many recent transit investments have failed to recognise the importance of urban development patterns. Many of them lack the concentration of activity required for cost-effective transit.

All too often, rail transit investments in the US have been followed by highway-oriented, rather than transit-oriented, growth…..Despite the unease many citizens, planners and politicians have with density, if costly rail and BRT investments are to pay off, larger shares of growth—particularly jobs—must be concentrated around transit stops.

Nevertheless they acknowledge that density is only one factor, albeit an important one. Other aspects of transit design and management matter too. 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.

Unfortunately, this paper doesn’t provide a lot of supporting technical information so it’s hard to evaluate the author’s claims. For example, I can’t tell if their recommended density thresholds make sense; I’m not sure adding job and population density into a single figure works; and I suspect they’ve under-estimated construction costs for older projects.

With that caveat, here’re the key things I take away from this paper. Density really does matter for rail-based transit, most especially employment density.

It’s quite possible to provide an acceptable level of rail transit without any high density stations, but it might not be cost-effective. But density isn’t the whole story – there’re other important dimensions to providing and operating cost-effective public transport too e.g. connectivity, frequency, parking.

The size and nature of the “market” being served should determine the type and scale of the public transport “solution” that’s suitable. Guerra and Cervero’s findings on per passenger kilometre costs suggest that’s not always – or perhaps even often – the case.

Cost-effectiveness matters. I don’t know what’s a reasonable level of subsidy per passenger kilometre for the benefits of public transport. But the available financial and political capital is frustratingly finite and great care and attention should accordingly be given to maximising the cost-effectiveness of projects.

I think the key message (the authors are strong transit advocates) is precious financial and political capital shouldn’t be squandered on rail-based projects that don’t deliver, no matter how glamorous they might be. If a project doesn’t stack up, other options need to be considered.

(If someone can e-mail me the more extensive gated version published in the Journal of the American Planning Association I’d appreciate it. Address is in About This Blog. Update: have a copy – thanks Matt).


We recommend

From around the web

Powered by Taboola


Leave a comment

9 thoughts on “How cost-effective are new rail transit projects?

  1. IkaInk

    Krammer – We could easily double the usability of the bus network by simple rationalising the route structure and putting in some better priority systems. Melbourne’s bus network is currently a complete mess. We should be following the example of Auckland recently and looking at the current “dropped spaghetti tangle” and trying to turn it into something that makes more sense.

    Ideally routes should:
    * go in straight lines, along arterial roads (avoiding unnecessary zig-zagging)
    * be frequent, every 15 minutes at a minimum most of the day
    * anchor at either end at a major trip generator or connecting transit hub
    * have dedicated bus lanes and traffic light priority systems

  2. Krammer56

    Interstingly, most of the conversations about rail imply it is a “good thing”, so then we have to try and make it work. This then, infers based on the paper and conventional planning, that we need lots of concentrated jobs and housing. Of course, if we have lots of concetrated jobs and housing then we need rail as roads will be too congested.
    This is a nice circular argument that reinforces the prejudices of the inner-city lifestylists and the big transport project apologists.
    Unfortunately, in Australia more than 90% of our existing housing and 85%of our existing employment investment doesn’t fit this model. THe vast majority of trips are actually quite short and local. How about we try and develop a public transport system that serves what we already have?
    That to me means buses. In Melbourne, just redirecting the annual interest costs and operating payments for the mooted $10B metro (or even better, the EW road tunnel) would allow us to more than double the bus services across Melbourne, making it a much more usable system!

  3. Dudley Horscroft

    Density is often suggested as an important parameter as to the viability of a transport system. But this is not really the case. What really matters is the population (jobs/dwellings) within the catchment area. A general rule of thumb is that for buses you consider the area within a quarter mile, and for trams a half mile radius of each stop. For trains, a bit further, but this is for ‘walk on’ passengers only. Add a bike (and proper bike storage at the stations or permit carry on bikes), the size of the catchment area substantially increases. Ditto for a feeder bus service to tram or train. And as the Blue Line example shows, parking spaces can be very important. Think the very well used and very cost-effective new rail lines in Perth, which have both parking spaces and bus feeder services.

    Perhaps the best arguments that density is not really relevant have been given by Paul Mees, see “Transport for Suburbia” in which, IIRC, he relates the stories of isolated villages in Switzerland which have good public transport services because they are part of an integrated system where feeder services connect to main feeders and so on up the chain to the main line rail system. As a result, some mini-villages, pop 250 IIRC, have as good or better services than some suburbs of Melbourne. (Might be pushing it there – but read the book and see.)

    Re the extension to the Newark City Subway, this was extremely costly because much was in tunnel – which should not be used if at all practical for light rail, and preferably not for heavy rail.

    Probably the worst least cost effective project in the past would be the Boston “Big Dig”, some $24 billion and counting (intended to cost about $2.5 billion) and in the future the “Waste Connex” road subway proposed by Infrastructure NSW for Sydney.

  4. Tom the first and best

    It is not only the density of a station catchment that has an effect but also the closeness of the station to the density.

  5. Alan Davies

    melburnite #4:

    Re Newark, I’d love to see what sort of future density would claw $6.48 per passenger kilometre back to a reasonable level (say to the median $0.50/Pkm)! Actually, would be an interesting exercise to look at all the projects in these terms.

    I’ve got the full version of the paper now so I’ll read it when I get a chance – maybe it will have some info on this question. Obviously there are problems with an evaluation in how you deal with prospective claims – will they eventuate? And when?

    They do say they looked at projects back to 1972.

  6. melburnite

    Yes this paper raises a few more questions than it answers (albeit just going by your summary alan). Eg. How do they measure jobs+pop / acre ? Does it extend one acre around every station on the line, or many, or is it the average of all development along the line ? Interesting that the Washington Blue line is most cost effective, at least partly because of lots of parking – so the acreage catchment must be large (always wondered how a relatively low density city like washington afforded an expensive metro, but they did and its heavily used).

    And to echo IkaInks point, if you build in the expectation of greater densities in the future, then this calculation doesnt apply – or rather gives you a figure to aim for in future development. So the Newark Line may become more ‘cost effective’ over time, since its meant to revitalise downtown districts. That’s something thats not measured in Melb – eg. a Doncaster line should be developed along with policies for higher density development at each station.

  7. hk

    “…political capital shouldn’t be squandered on rail-based projects that don’t deliver, no matter how glamorous they might be. If a project doesn’t stack up, other options need to be considered.”? One challenge for Australian academics would be to better define the political capital, if any, that exists to support particular rail development scenarios.

  8. Alan Davies

    IkaInk #1:

    I don’t think they exclude that possibility, in fact I take it for granted they’d embrace it. Robert Cervero is the co-author of the 1996 book, Transit villages in the 21st century.

  9. IkaInk

    I’m yet to have read the paper, but what if your aim with building a transit service is to encourage a different type of development? It seems their logic would only hold true if you’re following a this is the land use that exists therefore we should build this style of service logic. If instead you’re following a T.O.D style plan then at least in the shorter term virtually all projects will fail any of the cost-benefit analyses suggested.