Does Automating the Metro Save Lots of Money?

July 27th, 2017 by ant6n

As a continuation of my recent post on the cost of operating the Montreal Metro, I wanted to look at the merits of automation. The recent discussion of the planned REM light metro has brought that subject up repeatedly, with the implication that drivers are super expensive and that you can save a lot of money by running completely unattended trains. The PR keeps touting that it would be one of the largest automated metro systems, often along the claim that transit will be profitable from now on.

But can automation actually turn a heavily subsidized transit system into a suddenly profitable one?

A while back I spoke to the technical director of CDPQInfra (the private company which wants to build the REM), at one of their poster sessions. I was doubtful automation would reduce operating cost very much. How expensive can the drivers possibly be, compared to the rest of the system? The technical director emphasized that back in Lyon, where he gained a lot of his experience, they had three automated lines, and one non-automated one — and that line always gave them problems, and was much more expensive to operate.

I looked it up later, and it turns out that this one manually operated line of the Lyon Metro, the Line C, is a tiny, 2.4km-long cog-railway, with steep inclines of 17%, operated with a tiny fleet of 2-car trains that were built specially for this operation. Of course this line is going to be expensive to operate!

Lyon Metro Line C

Lyon Metro Line C (source)

I find it rather strange that this example was being brought up at all in the discussion of automation. It means a line that is inherently very expensive is being used as an example to show the expensiveness of manual train operations.

Back to the Montreal Metro. My previous article on the cost of the metro already showed that it is likely cheaper to operate per passenger-km than the REM will be, despite one having drivers and the other not. This is probably thanks to its concentration in urban areas and resulting high ridership. But how much could we save if we automated the metro and get rid of the drivers?

One way to estimate this is to calculate the cost of metro drivers, and compare that to the cost of converting to automatic operation.

The Cost of the Metro Drivers

In 2011,the STM employeed 308 metro drivers – and more than 9000 employees overall. Already there’s a hint that 3% of the employees will probably not have a very large financial impact overall.

Assuming 100,000$ cost per metro driver in 2015 [1], and assuming that the number of metro employees increased by the same amount as the amount of service between 2011 and 2015 [2], this means the drivers cost about 31 million $ per year.

The table below shows the cost of drivers compared to the operating cost and total cost of the metro:

STM Metro Drivers 2015     cost Cost per
% of
operating cost
% of
total cost
Total pay of Metro drivers 31M$ 1.4¢ 7.6% 3.5%
Metro operating cost 406M$ 18.3¢ 100.0% 46.2%
Metro operating + capital cost 878M$ 39.7¢ n/a 100.0%

(see spreadsheet with complete calculations)

$31 million may sound like a lot, but it only represents about 7.6% of the metro operating costs, and only about 3.5% of the total operating + annualized capital costs of the metro system.

The number of metro drivers is relatively small — there are many more people tasked with maintaining the trains and tracks. There are also a lot more ticket booth attendants than driers: 447. It would probably be easier to save money by removing those.

Still, let’s say saving a couple of percent on the operating cost of the metro would be great. A penny saved is a penny earned!

In order to know how much it will save, we need to know how much it will cost.

The Cost of Automation

The problem with automation is that in order to create a completely unattended system, the industry best practice is to install platform screen doors to ensure that nobody can ever enter the tracks.

Platform screen doors at Sacoma station in Sao Paolo (source)

The STM estimates the cost to install platform screen doors to be about 10M$ per station, or 5M$ per platform. (In Toronto a detailed study came up with the same amount.)

Let’s assume this is still a realistic cost estimate, that this is the only cost needed to get to full automation and that the Montreal signal system is already ‘ready’ for automation. This would give us the following cost:

Platform screen doors cost per platform 5M$
Number of metro platforms 146
Total capital cost 730M$
Annual cost of capital 4%
Annual cost of platform screen doors 29.2M$

So assuming that we can get rid of all metro drivers by automating, and assuming that we can automate by simply installing platform screen doors, we find that the annual cost of drivers matches the annual cost of paying off the upgrades (~30M$).

In our completely optimistic calculation, the already small savings are being completely undone by the costs.

But the reality is more complicated than that. Firstly, you can’t get rid of all employees in the metro system. In Vancouver’s automated Skytrain network, a large number of attendants are roaming the system to ensure safety and deal with emergencies.

Further, salaries paid to employees living in Montreal have a way of circulating back to the budget (via taxes), but also circulate in the local economy. I don’t think it’s the role of the state to employ people, but I think this consideration does skew the equation towards paying people rather than buying things from international vendors where most of the money will get exported away.

So if we’d take these points into account, we will actually lose money every year by converting the metro to automatic operation, based on the cost of platform screen doors alone.

There’s another problem: Platform screen doors have to be placed exactly to match the location of the doors of the trains. But the new Azur trains of the Montreal Metro have 3 doors per car instead of 4 like the existing trains. So any automation effort can really only be done after the line is converted to all new trains.

Are there Arguments for Automation besides Cost?

This analysis so far has only looked at the cost of drivers. But are there other arguments in favor of automation?

When talking about automation, we generally separate “automatic train operation” (ATO), where a computer drives the trains to the next station, from “unattended train operation” (UTO), where a computer also opens and closes the doors.

So under ATO, there are still “drivers”, who open and close the doors, ensure the safety of passengers, and deal with emergency situations. But the driving itself is done by a computer. It ensures that the trains respect safe distances and speed limits, accelerates and decelerates the trains, and makes it stop at the next station.

The ability to increase the frequency of trains and thus capacity is the main argument in favor of automation, not necessarily driver cost.

The Montreal Metro is actually already operated this way, the drivers mostly ensure safety of the passengers, but the trains drive themselves to the next station.

Being able to increase the frequency of the metro would be incredibly useful, especially on the overcrowded Orange line. But the bottleneck is not the computer or signalling system that drives the trains. The main problem is the safety in the tunnels.

The metro uses only a single tunnel for both directions, so there’s no separate one that can be used for egress in emergency situations. Therefore the system has adopted a rule that a train may only leave a station when the path to the next station is clear.

If we want to increase frequency and thus capacity, we would have to upgrade the safety systems and plans to allow multiple trains in the tunnels between stations.

So in the end, the main benefit of automatic train operation isn’t the removal of drivers, but the increase in capacity. But for the Montreal Metro, further automation would not help at this point, because the bottleneck today is due to the safety rules in the tunnels — which we would have to fix first.

[1] Every of the 9374 employees cost the STM about 90,774$ on average, including all benefits (see page 50 of the 2015 budget).

[2] total vehicle-km increased by 1.7% between 2011 and 2015, from 77 million to 78.3 million vehicle-km.

10 Responses to “Does Automating the Metro Save Lots of Money?”

  1. Fraser Harris Says:

    Vancouver is the obvious counter-example for requiring platform screen doors. The biggest benefit of ATO is it enables all-day frequency, which equates to freedom for users to live their lives without working to a timetable. Skytrain’s frequency at say 1a on a Saturday night is every 2 minutes. No system with drivers comes close to matching that.

  2. Julien Gatien Says:

    Great article as usual.
    I believe one of the main reason for service slowdown is due to track obstruction at the stations from customers which platform screen doors would prevent.
    So the service increased reliability should be factored in when analysing their cost.

  3. Jason Says:

    Given the capriciousness, if not outright sadism, of STM metro drivers (they seem to delight in making the doors slam shut just as people are entering the trains), it’s a shame that the STM doesn’t save on operating costs and let the trains operate themselves. It doesn’t cause any issues in Vancouver, which has a significantly higher on-time performance than Montreal’s metro. I’ve also never experienced door issues there, whereas in Montreal I’ve witnessed many occasions where the doors slam shut with passengers still in the threshold. There have also been incidences in Montreal where the trains go from one station to another with the doors still open!

    Passenger Screen Doors are not a pre-requisite for automation, and result in many benefits. They prevent debris from accumulating on the tracks, which is a major cause of delays and drives track maintenance costs. They prevent people from committing suicide (something that occurs fairly often but is not publicized), and prevent other person-on-tracks incidences (e.g. someone going down to retrieve something). They allow more efficient climatisation of the stations, and prevent the accumulation of brake dust in the air in the stations (a recent report showed high levels of particulate matter in the air in subway stations, comparable to China’s air quality.) They also provide safety in the event of crowding. Overall they improve the reliability of the system, which is a significant benefit for their cost.

  4. Yogi Says:

    The metro is already automated. In normal operation, the only thing the drivers do is close the doors after which the metro will automatically depart and follow a programmed schedule (acceleration, top speed, deceleration) to the next station where the doors will open automatically again.

    So why the drivers? Some say it’s due to big contracts signed under pressure from the union that covers the STM employees. But while they are there, they do have their hand on the emergency stop in case someone throws him-/herself on the rails. And in case of a disruption, they’ll override the auto pilot to catch up with the schedule (driving faster than the program). But if the STM manages to install the platform doors, there will hardly be any disruptions any more with people blocking doors (harder to hold two sets of doors) and people jumping on the tracks.

  5. Fraser Pollock Says:

    Vancouver’s Expo Line has very high frequency of service during the day and evening peak periods but much lower service frequency than both the Montreal Metro and Toronto Subway Systems during non peak times, especially in the late evening and weekend time periods.

  6. Jason Says:


    The expo line branches, dividing frequency. So yes, if you happen to be on one of the 6 stations on a branch then you could be waiting 8-10 minutes for a train. But the core of the expo line is the combination of both branches, so you never have to wait more than 5 minutes for a train.

    The Montreal metro has an absolutely abysmal off-peak schedule (night time, weekend) with a more than 10 minute wait between trains. The difference is that you could be downtown at McGill station waiting 12 minutes for a train, whereas in Vancouver you always get good service if you are in a central location. The branches also mean that there is no transfer needed, so that 12 minute wait in Montreal could be times 2 (I work on the green line and live on the orange line, but usually it’s faster to take a bus at night than deal with the wait and transfer.)

  7. Roman Says:

    If you are interested in comparing true costs, maybe reach out to Taipei authorities. The city has both automated, Vancouver-like, bombardier line, and a couple of classic underground lines with hoomans.

  8. gerard Says:

    I am living in Lyon (France) and the claims are totally false.
    There are 4 lines: A (the main one by far), B, C (the tiny steep one indeed), D (the only one automated). There is talk about automating the B line (it exists since at least 20 years so I am not holding my breath).

    The D line is better than the other by far because of the higher frequency and the reliability (something too often forgotten in discussion of transport modes)
    You get a a far better travel performance with the D line, but it may be more expensive as well, because what you save by not having drivers to pay, you lose in energy and maintenance by running it more.
    That’s why the whole argument is inane; it’s not a matter of cost per se , it’s making metro more attractive versus individual modes: being faster and more reliable wins it for metro.
    If you have drivers you need to rest them, there is the question of them doing an en masse stop if there any fracas between a passenger and an employee (‘droit de retrait’), and the strike question (both problems are somehow similar). Lyon’s drivers have played the strike dance very often, and usually at the most annoying moments as you can guess; the favourite one was always before the main touristic event (the 8 december at Lyon). The town has always kowtowed before the drivers, I guess because of a french law that forces it to re-hire the same drivers if it contracts with another company.
    I wish that all the lines and in particular the A line could be automated, but it is not going to happen. The drivers are just too powerful.

  9. Caciara Says:

    C’mon, I have never heard Metro company executives saying “oh, we would wholeheartedly upgrade the system to bleeding edge, but those pesky unions don’t want us to do it!”

  10. Angela Says:

    As gerard said, the main argument for automating the metro in France is that strikes have less impact on the system.

Leave a Reply