UN 173: How much does public transport cost to run?

We know how much transport infrastructure costs to build but transport operating costs get less publicity. Especially how much (or how little) it costs to add service and what factors cause this to vary.

Back in March I went through 20 years of state budget papers and population data to see how annual service kilometres per capita had changed over time. I found we were in a per capita service recession for our busiest metropolitan transport modes. Melbourne's grown fast in population, built lots of infrastructure but has hardly added service. This contrasts with Sydney that backed infrastructure with service such that waits for its trains, trams and some bus routes are now about half ours, especially at night. 

Annual cost

Last week I returned to past budget papers to examine how much we pay our transport operators (which, except for V/Line, are private companies operating under a franchise or contract agreement).  

In round figures this totals around $3.8 billion per year. Fare revenue would be well under $1 billion, meaning an average farebox recovery ratio of 20-something percent across the network. 

When assessing the value of this one needs to consider the broader transport system context including the public and private capital costs and negative externalities of roads and driving. While both are very expensive mega-projects, the North-East Link will have far less people throughput than the Suburban Rail Loop with the latter also offering accessibility and agglomeration gains. 

Active transport is the cheapest for all concerned yet gets the slimmest slice of the transport spending pie. It's also politically difficult with cycling and walking improvements frequently blocked by trader-dominated local councils scared at losing parking spots for transport's least space-efficient  and dearest to cater for mode.

And at the state level the most cost-effective public transport improvements tend to be overlooked due to a nearly decade old bias of prioritising peak-oriented infrastructure builds over good all-day service. That  might have made sense given pre-pandemic peak growth and premier Andrews' determination not to repeat the cautious 'too little too late' thinking that arguably cost premier Brumby office in 2010.

However changing travel patterns and the rise of off-peak and weekend travel is making most of our train and bus timetables increasingly look like museum pieces with 30, 40 and even 60 minute gaps between services widespread at some important times people are still travelling. All this is fixable. And as we'll discuss later, adding off-peak frequency to a peak-biased public transport network makes financial sense as existing assets is worked harder with fixed costs spread over more riders. 

The divide between regional and metropolitan transport roughly reflects the state's population split. In this case regional transport services cost roughly $1b to run with metropolitan close to $3b per year. The farebox recovery ratio is likely much lower on regional services, especially since the statewide fare cap was introduced last year. 

Today's main topic though is overall dollars per service kilometre for the various modes. This is calculated by dividing how much we pay operators per year (in millions) by the annual service kilometres (also in millions).  

Below is a table I did, comparing operator payments and service kilometres scheduled. Click to improve readability. Or scroll down for bigger versions, split by mode. 

Cost and value per mode

Metropolitan train is the dearest per service km, at around $55. Buses are the cheapest, with both regional and metropolitan around $6 - 7 per service kilometre. Tram and regional train/coach are in the middle at about $20 - 30 per km. 

Being cheapest is however not the same as the best value, especially when you factor in capacity and speed. One bus driver might carry 60 people in a packed standard bus. Whereas a train driver can carry 1000. Trams are in between. Which makes rail modes cheaper per person carried despite the higher cost and having to maintain their own lines and overheads. Then there's speed and comfort, with the train's superiority being enough to encourage people to travel to it from a wider area than a slower bus in mixed traffic. 

Mean versus marginal

Even more important is that the costs per kilometres is a mean figure. If we're interested in adding service the relevant figure is the marginal cost. Factors that cause the marginal cost to vary greatly from the average cost include:

(a) whether the network has high fixed costs and capacity constraints (more rail than bus issue), and 

(b) the time of day, with off-peak below average cost and peak service being much dearer, especially if it needs capital as the line is at capacity and all rolling stock is used.  

Especially on rail systems with substantial station officer, signalling and security staffing, expanding operating hours is cheaper than improving peaks but much dearer than boosting off-peak frequencies within existing operating hours. 

Having said that there are certain cases where longer operating hours are essential for network usefulness. Buses are the most glaring example given the late starts and early finishes on more than 90% of routes. For trains the main operating hours issues are with Geelong and Melton V/Line services, where a 60 to 90 minute earlier weekend start would make a big difference. A similar problem also afflicts Good Friday and (mostly) Christmas Day Metro services as these have inherited late-starting pre Night Network timetables. 

Speaking of which, the value for money we got from longer operating hours was perhaps less in the Home Safe/Night Network political promise, and its subsequent honouring, to operate Metro Trains 24 hours on weekends. The service we got was hourly, meaning that only a few train drivers were needed. However vast numbers of station staff and PSOs were needed, with the quantity being the same regardless of whether the service ran frequently or not. The policy was generally regarded as a success and no one batted an eyelid at its cost. 

While there are definite legibility benefits of operating trains all night, the opportunity cost was not  resolving our world worst train frequencies at other times of night and even Sunday mornings. That would have been excellent value given the numbers of passengers who would benefit.

Luckily this option remains open to us at modest marginal cost, with potential for us to have the best of both worlds. In the meantime we can give Night Network credit for finally fixing the Metro rail network's notoriously late Sunday morning starts (leaving just frequency as another, as yet unresolved, matter!) and for demonstrating that where there's the political will then something can happen, regardless of opportunity cost.   

Metropolitan train ($55/km)

A cut-down version of the above table, for metropolitan trains only, is below. 

The numbers start too late to show operating costs under Rail Franchising Mark 1 (effective for a few years up to 2003). That spectacularly collapsed because of a shared delusion about the extent to which costs could be cut while growing patronage and thus revenue under incentive-based franchises. Aggressive low-bidding foreign operators (after market share) and  bureaucrats (eager to please their political masters by getting a lot for very little) both contributed to this mess. 

Survivor Connex, whose operations were a little less financially parlous than National Express, got the keys to the whole network and a more generous contract, funded by tolling Eastlink under Rail Franchising Mark 2. This restored organisational stability but there were many network transition issues and reliability plunged as patronage surged. 

Connex became possibly Melbourne's most hated brand. The government dumped them in 2009,  refranchising with MTR from Hong Kong. MTM's first couple of years was rocky but reliability revived and train problems rarely made the front pages. The new franchise was also much more generous as can be seen in the payment leap between 2008 and 2009 that outstripped the rise in service kilometres. In 2016 the Auditor-General found that service reliability improved but saw scope for better value. 

There's another jump between 2017 and 2018. The current franchise commenced in this period, with MTM retaining the business. However the 2017 budget has a footnote saying that the payment excluded the revenue portion for train and tram. The current franchise was to come up for renewal at the end of this year but Metro Trains has been given an extension so it can bed down Metro Tunnel operations. 

As noted before, service kilometres for Metro Trains has remained flat, trailing population growth. It remains to be seen what gains the Metro Tunnel timetable will bring. But we can say that rail service costs around $55 per kilometre with the marginal cost for shoulder and off-peak service additions likely significantly lower given rail's high proportion of fixed costs. There may also be rostering efficiencies possible, especially for shoulder peak services, that would further lower costs if realised. 

Metropolitan Tram ($22/km)

Trams operating costs are below. 

Even more so than metropolitan trains, annual service kilometres for trams has hardly grown in many years. The main factor affecting per kilometre costs is the outcome of the tram franchising. Currently  there's three bidders in contention, comprising Keolis Downer (the incombent), Go-Ahead (including well-known bus operator Kinetic) and Transdev/John Holland (who have a poor reputation to overcome from Transdev's previous time here with buses). We're expected to know the successful bidder any time soon.   

Metropolitan Bus ($7/km)

Now we come to buses, which are the cheapest metropolitan transport mode per kilometre. Payments per kilometre are a gentle rise, as can be seen below. 

Buses have lower fixed costs than other modes (as there's no tracks and overheads), however on a cost per passenger basis they can still be dearer. For example we pay more to run buses than trams each year even though tram patronage is higher (admittedly including many very short trips).

However bus routes and timetables aren't as productive as they should be, with inefficient overlaps and  poor weekend service relative to usage contributing to patronage that is attainable.   

A bus upgrade example - what can we do with $1 (or $10) million?

A typical suburban bus route might be 15km long. One trip on it costs $100 each way, or $200 return to operate. This being based on 15km x $6.80 per km x 2. 

Sat it runs 15 hours per day. Service is every two per hour weekdays and every hour on weekends. That translates to 30 return trips per weekday and 15 per weekend day, meaning 180 return trips per week.

Multiply that by $200 to get a weekly cost of $36 000. Or by 52 to get an annual figure of $1.82 million for a route like that.

To prove that we're not way out, if we multiply that by 350 (the number of bus routes in Melbourne) we get $655 million per year. Which is not that far off the $897 million total given (a) some bus routes are much longer than 15 km and (b) some bus routes, including the very long orbitals, are about double the frequency of our example so would cost more. 

Boosting the weekend service on our example route from every 60 to every 30 minutes needs 15 extra return trips for each weekend day, or 30 more per week. In other words a rise from 180 to 210 return trips per week. That lifts weekly cost from $36 000 to $42 000, or $312 000 per year to halve maximum waits across the week. 

$1 million per year could upgrade 3 average length routes, while $10 million per year could upgrade 30, with gains all across Melbourne. And these are millions, not billions. The sums associated with big construction projects have stopped us thinking what we can do for a few million (or less) but this is an essential skill to retain, especially in tough budgetary times. 

Costs would be even lower if routes that already have good Saturday service just need a Sunday upgrade. They'd also be less if the existing base is higher, eg you just looking to upgrade a bus that's every 40 minutes to every 30 minutes.

Conversely, if you've got really strongly performing routes you might boost them from hourly to every 20 minutes on weekends, instead of just being satisfied with 30 minutes. Going from 60 min (1 bus per hour) to 20 min (2 more buses per hour) needs double the resourcing (or can be done on half the routes) compared to going to 30 min (1 more bus per hour). Still the cut in maximum wait from 30 to 20 minutes makes that well worth it for strongly performing routes from major weekend destinations. 

Bear in mind that weekend upgrades are marginal costs. Whereas I'm using the average per km cost. So the cost of service uplift may well be less since you're working the existing bus fleet harder. However weekend trips do have some extra costs with driver penalty rates, making them somewhat dearer than adding (say) extra midday weekday trips. However patronage productivity numbers often show stronger patronage numbers on weekends, making adding to weekends more worthwhile, especially on lesser served but high patronage potential routes.

Maximising cost-effectiveness with simpler routes

Opportunities to get results just from a simple upgrade to an existing route are highest where the route is already an above average patronage performer, it goes to useful destinations, it's already reasonably straight and it doesn't inefficiently overlap other routes.

If it's the first three and not the fourth then network simplification can help deliver even better value frequent service across a wider area. In some cases it may even be possible to do it with the same or less live kilometres per week, especially if there is scope to redistribute service hours from weekdays to weekends.  

Six strong opportunities to provide cost-effective 7 day frequent service along main roads to Melbourne's biggest suburban destinations include:

* Footscray-Sunshine along Ballarat Rd (220/410 simplification)
* Sunshine - Highpoint via Braybrook (408/465/468/903 simplification)
* Footscray - Highpoint (223/406 simplification)
* Footscray - Laverton via Millers Rd (232, 411, 903 simplification)
* Coburg - Preston - Northland - Heidelberg along Bell St and Murray Rd (527 and 903 simplification) 
* Chadstone - Monash - Dandenong (802/804/862 simplification)

Because of substantial network overlaps big 7 day frequency upgrades are possible by consolidating and simplifying routes without significant coverage loss. This type of thinking would allow many bus corridors to run (say) every 10 minutes weekdays and 15 minutes weekends without substantial additional cost.

The resultant network gains and usage uplifts mean there's no excuse even for an apparently cash-strapped state government to check its 'hollow logs', with only a pittance required for setup costs like a few moved bus stops and signage changes plus likely some small extra operating costs for (highly desirable) operating hours improvements. 

Extended operating hours for buses are vastly cheaper than the same for trains. That makes getting bus operating hours to match trains and trams a no-brainer, particularly on the main routes. And, looking further ahead, if Melbourne was to get 24 hour 7 day transport you would at least start with buses in the wee hours (like Sydney already does on some bus routes). 

Regional Rail + Coach ($31/km)

A mix of V/Line trains and coaches. Costs have risen by a slower rate than metropolitan trains, as can be seen here:

Regional Bus ($6/km)

Your regional city buses, such as run in cities like Geelong, Ballarat, Bendigo and many towns down to around 10 000 or even less population. Slightly cheaper than metropolitan buses per kilometre to run. 

Conclusion

This has been our look at annual running costs for public transport in Melbourne and Victoria.

The key takeaway messages is the scope for much improved cost-effective all-day better service to make the network vastly more useful for more people and for more diverse trips. 

In arranging this the need to understand the difference between average and marginal costs, and to exploit the latter where low is also essential. And, particularly for buses, network simplification can be a powerful tool in maximising the extent of the frequent network and thus peoples' travel freedom. 

I'm not the world's tidiest spreadsheetist but those who want my data for further analysis can find it here. 

Index to other Useful Network items here