Human Transit has an excellently detailed analysis of public transport operating costs, which comes from Jarrett Walker’s upcoming book that’s bound to end up being a bible for public transport planners around the world. Each year we spend close to $150 million subsidising public transport in Auckland, so it’s utterly essential for us to ensure we’re running the most efficient system possible and making best use of that money. We need to have a good idea about how PT operating costs work and also look at where we can save money without reducing service quality/quantity. This is an issue I touched upon in this recent post.

Jarrett notes that there are a number of components to operating costs:

  • Time-based costs vary based on how many transit vehicles are operating and for how long. The dominant time-based cost is the wages and benefits of the driver and any other on-board employees, which we pay for by the hour.
  • Distance-based costs vary with the odometer reading of the transit vehicle. As in cars, most of transit’s maintenance and fuel costs are distance-based.
  • Fleet-based costs vary with the number of transit vehicles owned. Fleet size is based on the number of vehicles needed to run the most intensive part of the service day, typically the commute period which transit planners call the peak. Fleet size drives some maintenance cost, but it main impact is the cost of the vehicles themselves, and of the facilities needed to store and maintain them.
  • Finally, there may be some administrative costs unrelated to any of these, though in fact most administration costs are roughly proportional to the other measures of size.

One thing to keep in mind, when thinking about operating costs, is that getting an additional vehicle on the road/track at peak times costs a lot more than getting that vehicle in service outside the peak – because of ‘fleet-based costs’, and to a lesser extent ‘ time based costs’ (need more staff on expensive/messy split shifts).

Generally the biggest section of operating costs is spent on labour, especially for bus based systems where you have fewer passengers per employee. Labour is obviously a time-based cost, and if we can reduce time-based costs (like shifting to a ticketing system for our trains that doesn’t require a huge number of on-board staff) then we can save a huge chunk of our operating costs. I’m looking forward to seeing our per-trip rail subsidy plunging over the next few years as we shift to the new ticketing system. It is time-based labour costs which I suspect will end up providing the ‘tipping point’ for North Shore rail becoming financially viable: it will become hugely expensive to run hundreds and hundreds of buses from the North Shore to the city at peak times in the future. A driverless metro, like what Nick suggested in this post, would have massively lower operating costs than continuing to add and add buses.

Another key consideration in operating costs is what Jarrett calls “lumpiness” – where trip length is just above or just below allowing a logical number of vehicles you need to operate the service:

Lumpiness has important consequences when designing lower-frequency networks, such as local bus routes in low-density suburbs. In these cases, good planning designs routes to be of a certain length, so that they will run an efficient cycle. If our network of local routes is meant to all run every 30 minutes, for example, we try to design routes that cycle in 29 or 59 minutes, but not 31 or 61.

A small deterioration in speed can cause sudden big changes in operating cost. If we’re running 30-minute frequencies on a route that cycles in 29 minutes, that will require one vehicle. But if for some reason the line slows down just a little, so that it now cycles in 31 minutes, we have to add a whole additional vehicle and driver, doubling the cost of running the line. A mere 7% increase in the cycle time has become a 100% increase in operating cost. In that case, a planner may try to redesign the route to make it shorter.

The Western Line is a classic example of this, with a 53 minute running time between Britomart and Swanson allowing 15 minute frequencies with 9 trains and three and a half minute layovers at each end – but, the running time to Waitakere station of 58 minutes being just a bit too tight. This means an extra train is needed for Saturday services compared to what would otherwise be required, rather a waste of money and probably partly explaining why Saturday train frequencies on the Western Line are still a pathetic hourly service.

Of course one way to increase frequencies at no cost is to increase the speed of a particular service. If we think about Northern Express buses, because the busway allows them to travel so quickly we can get pretty high frequencies out of many fewer buses than would be required if each service took a lot longer to complete its route. This is the magic of bus lanes and other bus priority measures: not only do they make the trip faster and therefore more attractive for users (probably increasing farebox recovery rates and requiring a lower subsidy), but also the faster trip means that it takes fewer buses and fewer drivers to maintain a certain level of frequency.

To finish, Jarrett highlights perhaps the three most important aspects of thinking about operating costs when designing networks:

  • Every increase in frequency is an increase in service hours, and thus in operating cost. If you want to increase service on a line from every 30 minutes to every 15 minutes, that will double the cost of running the line. This is why most transit agencies would like their service to be more frequent, but have trouble affording that frequency.
  • Every increase in average speed is a savings in service hours, and thus in operating cost. If we can cut the cycle time of a line by 25%, that cuts its operating cost by 25%. This is why transit agencies are always trying to control delay.
  • At low frequencies, operating cost is lumpy. Because you can’t run a fraction of a driver, small differences in speed or frequency can create large differences in operating cost, if the overall frequency is low.

As I noted at the start of the post, it’s critical for us to think deeply about these issues if we want to improve our PT network at relatively low cost.

Share this


  1. the driver cost is why small buses cost almost as much to operate as a full size bus, running costs are not significantly lower, but their earning potential makes them unattractive in the peak

    the old ARA practice of deleting routes and then trying to cover the area with an adjacent route only made running times longer, deterred passengers and led to other inefficiencies, there is still some legacy of this practice on the Shore

    I remember riding the old 945 route some years ago, every time we came to an intersection I thought “the bus should go that way” and it didn’t

  2. One thing I would love to see is AT release on a regular basis (i.e. monthly) figures for things like PT distance travelled, fares collected and total costs. The reason for this is that I am really keen to see how costs are being impacted by patronage growth etc.

    One other thing I think AT should actually consider doing might sound weird at first but would be really innovative would be to offer a decent prize (say $1m) to anyone or team that can design a series of routes and and timetables that do things like provide better coverage, better frequency, x% more passengers and at x%less cost than what we have now. It might sound like a lot of money but if done right could have massive implications for PT provision in the city and a prize of $1m or more would attract talent from around the world to work on the problem. Something similar to the netflix prize

  3. Here are some of the costs I found for the Vancouver skytrain when researching the driverless metro post which really illustrate the impact of staffing costs on time-based costs, these are for 2008 and in Canadian dollars:

    • Operations & Maintenance budget = $82.7M (60% labour, 11% power, remainder inventory and overhead)
    • 73.5 M passengers = $1.12 / cost per passenger
    • 35.4 M revenue car-kms = $2.34 / cost per rev car-km

    You can see how they manage to run an operational profit at the cost of only $1.12 per passenger / $2.34 per revenue kilometre, a flat fare of $1.20 per trip would cover things nicely!

    If we had the same costs for a hypothetical 16km long line from Albany to Britomart, then each run would only cost $37.44 and the train would only need to pick up around ten people along the way to break even (while the remaining 190 potential passengers would be just profit)! I wonder what a single run of the NEX costs, probably more than thirty-seven bucks.

    1. Sorry, that should read the remaining *90* passengers for a single car train. A two car set would cost $74.88 per run and require about twenty passengers to break even.

    2. From what I can find in the business case for the CRL it suggests that each 3 car EMU would do roughly 80,000 km per year and that annual operating costs would be about (staff, energy, cleaning etc.) would be about $385,000 which puts operations at about $4.8 per km.

      For a 6 car EMU:
      Swanson to Britomart at 27km from Britomart would then cost about $260 per run
      Papakura to Britomart at 31km is about $300 per run

    3. Given the high-costs of vehicle drivers and other staff, you would think there would be more uptake on driverless transit systems, whether its light metro as described by Nick for high-capacity transit or PRT for low-to-potentially-high-at-some-point-in-the-future operations.

  4. the operating costs of road based transit pale into insignificance when compared to the capital costs of LRT, driverless or not, I think some of you have the wrong end of the stick

    there seems to be alot of either/or in these discussions, the answer is both/and, the answer is to use each PT mode appropriately to its strengths

    1. Of course Steve, horses for courses… I wouldn’t frame it as an either/or discussion, I think I speak for most people here with the assumption that road based transit will form the bread and butter of Auckland’s transit system now and in the future (in the form of buses on regular streets and on bus lanes, plus perhaps some trams in a few niche corridors). That’s where the strength of road based transit lies, at street level, with regular stops, finely grained coverage and good accessibility. An absolutely essential component.

      But what we tend to talk about more is the trunk rapid transit routes, where road based transit cannot supply the necessary capacity, speed and efficiency of the intensive core of the system. Not much point identifying the fact that road based transit has lower operating costs when road based transit isn’t a realistic option for what we are talking about. Quite frankly if you tried to take a bus based solution and operate it to the level of rapid transit you’d find that it has huge operating costs (mostly due to the staff-passenger ratio) plus you’d also find the capital costs of providing a top-tier route to be very high as well.

      At the rapid transit level the capital costs of busway, LRT, metro, heavy rail are all an order of magnitude higher than street based solutions, although there can be significant differences between them depending upon the specifics. For entirely new rapid transit routes in Auckland (as opposed to upgraded or extended ones)the best solution is likely to have a combination of ‘light’ corridor design requirements (i.e. busway, light rail or light metro) to minimise capital expenditure and low staffing requirements (either driverless or vehicles that have high passenger to staff ratio) to minimise operating expenditure. To me it seems that the existing sort of heavy rail fails on capex, while busways fail on opex.

      1. Or to put your last sentence another way: Both rail and busways are equally expensive so it comes down to which system we prefer…..

        1. Well my conclusion was more like neither is particularly affordable regardless of which we prefer, so sure we can chose either if we’re happy to shell out but otherwise we might need to look at something else.

          However we need to look at what is more feasible given current or future funding arrangements, a cheap capital project that slugs us with ongoing high costs (the usual story) or a big one-off grant from a friendly government that is cheaper to run in the long run (where things might head).

        2. yes and I guess my suppressed conclusion was that the one with the higher capex but lower opex is also way more desirable on other measures; environmental [electric] and especially urban form and congestion in the city.

          By the way. I never cease to marvel that driving fiends so often end up lobbying for buses; the most irritating things on the road. Why not?: Trains man! Get those PT losers out the way!

  5. I would have thought that the “desirable other measures” you mention like less congestion, pollution, accidents and other externalities are just as much a cost of operating a transport system as are the driver’s wages. Shouldn’t they be quantified and have equal prominence in any cost-benefit analysis?

  6. A 3-car electric train should cost a similar amount to operate as a bus, based on energy cost*. Differences are

    1. The train has a capacity of about 300, whereas the bus only 50
    2. The train has a much greater capacity to attract “choice” riders if frequencies are attractive
    3. More choice riders improve cost recovery
    4. Choice riders attract other choice riders, and reduce security and graffiti problems

    By designing the off-peak transport system around the rail network, the overall subsidies for off-peak operations should decrease. Train frequencies of 15-20 minutes have resulted in high weekend patronage in Melbourne and Perth. These people may be attending events in the city or at stadiums.

    Arguments put this way are more likely to be listened to by conservative politicians.

    * Fuel efficiency data from Wikipedia article “Fuel efficiency in transportation”, diesel bus 39L/100km, electric train 20.6 MJ/ Other information: Diesel price $1.47/km, electricity price $30/MWh. Bus costs $0.57/km, 3-car train costs $0.513/km. Mechanical maintenance costs for trains are often covered under the purchase cost for the first few years of the contract.

Leave a Reply

Your email address will not be published. Required fields are marked *