Creative Accounting With Subway Operating Costs

The Toronto Star’s Royson James writes today about automation of the TTC’s subway service and the elimination of train crews. His article includes a figure taken from a paper published by the Neptis Foundation which claims:

Converting the TTC subway to UTO [unmanned train operation] could save about $200 million per year, or $2 billion NPV [net present value].

Installation of PSDs [platform screen doors] might cost another $300 million to $500 million.

[Page 51.]

I wrote briefly about the Neptis paper last year, and keep meaning to return to it if only to debunk some of its more outrageous claims. However, the emergence of fantastical statements about the potential benefits of total automation force me to address this separately.

First off, the cost of PSDs is considerably higher than stated in the report. When this was still part of the TTC’s “above the line” budget, the cost stood at roughly $1-billion (about $15m per station).

The Neptis report says that automation could save “about $200 million per year”. This is wildly inaccurate as can be proven in various ways.

The total TTC operating budget for 2014 is $1.6-billion. Of this, at most 80% of the costs are for labour, and only half of that will be for operators who make up roughly 50% of the workforce. This means we are starting with a total cost of everyone who drives a bus, streetcar or subway train of $640-million. Saving almost one third of this by eliminating crews on the subway simply is not credible.

How much would a subway crew have to be worth to generate $200m in savings per year? Here are a few “back of the envelope” calculations.  Note that throughout this I have tried to err on the conservative side to avoid accusations of bias and cooking the books to suit my argument. Without question, the numbers can be refined with more detailed background information, but they are certainly accurate to the level needed for this discussion.

Peak      8 hours x 95 trains =     760 train hours
Off Peak 13 hours x 66 trains =     858
Total per day                     1,618
Weekdays per year                   250
Total per year                  404,500

         21 hours x 66 trains =   1,386
Days per year                       115
Total per year                  159,390

Grand total                     563,890 train hours

For the sake of simplicity, round this to 564,000 train hours.

If we are going to save $200m annually, then the crew for one train hour must be worth $355. This is quite substantially more than they are actually paid including benefits. In this calculation the number of train hours is overestimated because all off-peak service is based on weekday mid-day levels. Other adjustments such as the shorter Sunday operations and higher rate of pay for Sunday work would also be needed. However, the number is certainly “in the ballpark” and, if anything, overestimates the amount of service on which savings can be achieved.

Another way of looking at this is to build up from operator wages and benefits.  For this we must calculate how many operators are required to drive the trains by looking at their annual working time.

An operator’s shift includes paid breaks and other time that is not spent actually driving a train with passengers such as dead-head trips to/from carhouses, and the time allowed for reporting for work and travelling to the point where a train is picked up.  The numbers here are for purposes of this calculation and are almost certainly conservative (i.e. will make calculated costs higher than actual).

Driving hours per day            6.5
              per week          32.5
Weeks worked per year           49
Driving hours per year       1,592.5

Next there is the question of what an operator costs per year.

Hourly wage                $    35.00
Benefits @ 30%                  10.50
Total                           45.50

Weekly cost @ 40h/wk         1,820.00
Yearly cost @ 52wk/yr      $94,640.00

Now we convert train hours to operators and hence to the cost of staffing based on 1 operator per train.

Train hours/year              564,000
Operator hours/year             1,592.50
Operator years                    354
Annual cost               $33,502,560

In other words, elimination of the guards on existing train crews would save about $33m annually. This is something the TTC plans to do anyhow. A further $33m is all that is left to save for elimination of the train driver.

This saving would not pay off the substantial cost of installing platform screen doors.

The Star is irresponsible to publish a claimed “saving” which is so vastly out of line with reasonable estimates and so transparently in error.  This does not serve discussions of transit policy well at Council or at the TTC.

22 thoughts on “Creative Accounting With Subway Operating Costs

  1. There are most probably other factors and/or costs as well. Although unrelated – I have asked a bus driver a few years ago to let me out on “my” side of the street in a situation where (a) traffic signal was on RED; (b) his bus was tightly at the curb; (c) the stop was (is) on the opposite side of the street from my house. He said, that he would not with a reason: that if I had broken my leg or had fallen on my “gentle behind”, the TTC insurance costs would go up and/or he could be disciplined. I didn’t ask again.

    My questions therefore would be: (a) what are potential implications for insurance costs? (b) what are or would be potential maintenance costs of these new machines especially given that some subway stations are supposed to work efficiently and frequently in a wide span of temperatures (Warden, VicPk)? I do not want to be nasty to Mr. James and/or TTC, but TTC was not able to keep the “movator” at the Spadina station going for eternity while that “beast” was fully underground.


  2. Hi Steve,

    Thanks for taking the time to debunk the egregious subway ATO cost savings claims in the Neptis Foundation paper.

    I never gave their paper a lot of credibility in the first instance; their stated metrics were clearly those of an organization/author with no knowledge of, let alone an informed understanding of TTC operations & service costs.

    Rather than “creative accounting” I’d call Neptis’ claims “baseless projections” as they clearly aren’t based on any actual TTC service plans & their detailed cost accounting metrics, as are your estimates.

    What surprised me more was TTC CEO Andy Byford’s (normally a straight shooter) statement that platform edge doors (PED) were a prerequisite for ATO (automatic train operation), a new twist on the old TTC canard from the last Commission that PED were needed to reduce TTC suicides (Priority Ones in TTC-speak).

    When I returned to my Vancouver hometown for Expo86, I was amazed to ride the Skytrain with ATO—without PED—in 1986, 28 years ago! So, it’s a spurious claim that PED are needed for ATO.

    The real reason for spending $1B on PED isn’t to reduce suicides, or now ATO, but, to minimize operational delays from paper fires, that require a self-imposed (2000s) “smell-of-smoke” (SoS) delay protocol for TTC & Toronto Fire to attend & investigate the “smell” of smoke — not a fire — before resuming service — virtually ensuring a 15-minute service interruption — or 5–6 held trains resulting in crush loadings that slow down all subway station stops and take over an hour+ recovery to get back to 2:21 m:s scheduled subway peak service.

    In the 60 year history of the TTC subway there has never been serious passenger injury or death due to smoke or fire — before or after the 2000s implementation of the SoS protocol — yet the TTC allows the remote possibility to disrupt daily subway service and claims the only solution is a $1B subway relief line—10+ years in the future!

    Even worse is the origin of all the SoS incidents — an overreaction by the TTC to remove its platform concrete/metal/tile garbage containers in the early 2000s (to minimize shrapnel injuries/fatalities from a terrorist bombing) without a replacement, that resulted in an explosion — of platform litter — overwhelming track-level paper catchers, despite the TTC’s Pollyannaish ad campaign exhorting TTC riders to carry their litter off the train, up stairs & escalators to remote waste/recycle disposal bins.

    The simple solution would be for the TTC to add more clear plastic garbage/recycle bins to stations & platforms; redesign the paper catchers & clean both of them more frequently until SoS incidents rarely happen.

    This is an example of the TTC Commissioners not challenging TTC Staff assertions, as they have short tenures, no subway expertise or business Governance experience & fail to ask probing, challenging questions.

    No wonder a subway (downtown) relief line has so much public support: riders experience the crowded, crush peak conditions daily, without knowing how much is due to line overcrowding and how much is due to the TTC shooting itself in the foot with not only 15 minute SoS delays, but, also similar 10-15 minute delays for the TTC, EMS/Police to investigate PAA (passenger assistance alarms); also the subject of a bound-to-be-ineffective TTC public education ad campaigns to not press the PAA, unless really, really necessary!

    A Commissioner with extensive subway operations experience & business expertise, such as David Gunn (ex-CGM: TTC/NYC/DC/Boston/Philly public transit, and Harvard MBA) would eviscerate Staff’s rationale for wasting $1B in scare capital on PED that could be achieved with greater attention to train, station & platform cleanliness and common-sense subway safety protocols.

    As for the savings claims, TTC Commissioners rely on TTC Finance Staff to give them financial projections, a inherent conflict-of-interest, they don’t have their own independent financial staff (CoT has one TTC financial analyst) with public transit expertise & experience to vette the TTC financial submissions — to separate out the absolutely necessary from the rainbow pot-of-gold wish list. This is common in private sector Board governance, as I experienced with PepsiCo, Inc.

    As for the Toronto Star professional standards & responsibilities, Royson James is a columnist, writing an opinion column, he is not a financial analyst, such as those that cover publicly listed companies on stock markets like Bell, Apple, etc. Ditto for the transportation reporters I’ve known covering the TTC: Joe Hall Jr., Kevin McGran, David Bruser (briefly), Tess Kalinowski are all reporters, reporting verbatim on the TTC Commission reports & press releases, without the wherewithal to critique the cost, revenue & metrics claimed by TTC Staff, Commissioners, politicians or public transit commentators. Ditto Christopher Hume, its Urban Affairs columnist.

    With TTC operating & capital budgets and fares increasing at several times the CPI & increase in available property tax revenue — with no consensus on increased user fees & taxes — it’s time for TTC governance to have built-in financial reality checks with tenured Commissioners having the business, financial & public transit expertise to set service priorities that ensure maximum rides & revenue are achieved for the capital & operating budgets set by government: whether municipal & provincial–only, or with (unlikely so far) federal contributions too.

    Steve: A few points here.

    First off, I believe that any journalist (and I include myself in that description) has a duty to at least think about the reasonableness of what they write. The simple fact that a claimed annual saving is equal to 1/8 of the TTC’s entire operating budget (a well-known number, and easy to find especially for someone who covers City Hall and knows where to look) should raise a big red flag. However, the Star has been treating the Neptis paper as some sort of holy grail since they got to write an exclusive about it. Imagine if Robyn Doolittle applied the same journalistic rigour to the Star’s investigation of the Fords.

    Second, we already have Commissioners who have business backgrounds. On rare occasions, they ask questions with a business spin, but they don’t do it often enough. Part of the problem, no doubt, is that they may know business, but they don’t know transit. Any good businessperson has a sixth sense about proposals that just don’t look right, and enough understanding of their business to do a preliminary analysis and ask the necessary questions. At the TTC, the brains trust has a long history of dealing with Commissioners who defer to their expertise.

    All of those delays are part of a problem that was mentioned in an international peer review of the TTC to which the Commission regularly points with pride. What they want you to notice is how well ranked they are for “efficiency” (e.g. cost per customer), a factor that comes directly from the relatively high riding relative to the extent of the system, and the fact that a lot of costs are sunk via old capital projects. However, the same review also looked at the TTC’s ability to execute a reduction in headways and increase in capacity, and said, in brief, “get your act together on service reliability, or don’t even bother trying”. The TTC mentions that much less often.


  3. Steve: Your calculated savings would be even less based on the actual 2013-14 operator hourly rate of $31.23 (not $35.00) and still make the Neptis numbers look like they were pulled out of very thin air. When I looked at the biographies of Neptis staff and researchers, I found not one person with a public transit background.

    Steve: The paper was done by a consultant based in London, UK. I cannot help wondering why Neptis had to go so far afield to get this paper done, and why it contains so many transparent errors (although there are some good points too). I would love to know who actually paid for it.

    I used $35/hour deliberately because it is higher than the current rate, even if I add in one more year’s worth of an arbitrated increase. Throughout the calculation I was trying to err in such a way that the most favourable possible comparison would exist between my numbers and the $200m/year claim.

    All that said, anyone who writes a paper implying that there is a 1.0-1.5 year payback for platform doors ($200-300m capital, vs $200m in annual savings), and uses blatantly wrong numbers for both factors, is either seriously deluded, deliberately deceiving his readers, or monumentally incompetent.


  4. In response to Jiri and the Spadina station Movator; It was installed in 1978 and removed in 2004, 26 years old. Maintenance costs were over $100,000. per year and replacement parts were becoming harder to find. Varying temperatures did not affect it, but it was exposed to a harsh winter environment of salt, sand and water from the foot traffic causing excessive wear and tear. Only a few thousand people used it daily and today it would cost approx. $6,000,000. to replace, money better spent elsewhere.


  5. There have got to be significant benefits to making Eglinton grade separated and fully automated as the Neptis report advocates. It would allow more frequent trains (every 90 seconds) doubling the line’s capacity. It would get rid of the problem of accidents, which are a problem on every light rail/streetcar or bus system, and must cost the TTC or other systems like Calgary Transit a significant amount of money in compensation costs and repairing rolling stock. Even if the cost of a human driver is not enormous, automation allows service to be increased rapidly in response to surges in demand (like in Vancouver), and it eliminates disruptions due to strikes. I can’t imagine automation costs very much if a system already has ATO, and drivers are being retained on many ATO systems solely because of opposition by the union to automation.

    Obviously if you want to build the outer sections of Eglinton grade separated you would have to make them elevated, which some residents might dislike, and it would cost more than light rail but less than tunnelling, but somehow people in Vancouver are willing to accept this. It might be better to just get rid of the eastern section due to the lack of development around it, particularly if extending the Sheppard subway would cannibalize it. It seems really unusual what Toronto is doing; lots of cities are building light rail systems that typically are low capacity and typically entirely on the surface, and lots of cities are building subway systems, but a partially underground light rail line costing over $5 billion is unheard of outside Toronto, and a light rail line on a route like Sheppard where there is already a subway is rare.

    A lot of cities have installed half-height platform screen doors on existing lines which are less expensive than full height doors. For instance, Paris Metro line 1 was recently converted to automated operation and has half-height PSDs. Full height PSDs typically require structural modifications to the subway and are a lot more costly.

    Steve: It’s important to remember that the Eglinton line was proposed in a very different context. First off, it was to run from the Airport to Kennedy Station with a leg turning north to replace the SRT, and another LRT line continuing on the surface east and north on Morningside. The central tunnel is unavoidable because of the nature of the part of Toronto it would run through, but LRT surface construction elsewhere would save a ton of money over a full subway.

    Elevated structures bring with them the extra cost of stations and the likelihood that they will be further apart than on a surface route leading to access issues and demands for parallel service. Also, many of the Vancouver locations with an elevated are in areas where it has little effect, or has been incorporated into nearby development.

    I agree with your point about capacity and the ease of adding more service, but that capability comes at a cost of higher capital and a larger fleet. However, how much reserve do we want to build into the line, and is that the best use of available capital?

    Depending on how a project is funded, capital may simply disappear off of the city or transit authority’s books with some other government paying the tab. Cities tend to be more tight-fisted when they’re paying with their own money.


  6. I’d even say that considering there to be tradeoff between the door guard crewman and platform doors is disingenuous, as they serve different functions. The most common argument in favour of PSDs is that it stops suicides and/or accidents before or as the train arrives in the station, which crew on board the train cannot do anything about.


  7. I like your calculations. More people should do them, especially before publishing in a newspaper. We will have the public believing the $200 million value, and politicians may try to play on this belief.

    Let us think of some other consequences of having no TTC staff in the train. Imagine a power failure. No lights, no computers, no communications. Currently if that happens, there are two staff members on a train, to deal with the safety of 1,000 passengers. It does not happen frequently, so riders don’t worry about it. I think the TTC would worry, and would keep at least one employee on the train. So saving “millions” by going to ATO may be overrated.

    Keep up the good math.


  8. I’ll just chime in as a professional Accountant, “The world’s only profession where creativity is officially discouraged.”

    I took out my calculator to see if I could get $2 billion in NPV, even assuming the absurd $200 million in savings.

    Steve has already shown how reality-impaired the $200 million is, and I will only say that I agree with his analysis.

    On page 20 of the Neptis paper, the authors assert that they are using a discount rate of 5% to do the NPV calculations. UTO and its NPV is discussed on page 50, but this discussion omits some vital details.

    In order to calculate the NPV of this proposed project, it is necessary to know the cost of installing the UTO system, its annual maintenance costs and its lifespan.

    Even assuming that the phony $200 million is actually true, the only way that I am able to get anywhere near the alleged NPV value is to assume that the cost of installing the UTO system is zero, its annual maintenance costs are also zero and its lifespan is infinite.

    I would suggest that there is a faint possibility that these assumptions may not be realistic.

    Steve: Accountants are also known to have a bone dry sense of humour.


  9. Vancouver does not need platform screen doors for its driver-less trains. So why would they be needed for automatic train control on Toronto subway trains?

    Andy Byford wants to phase out the guard on subway trains starting on the Sheppard line. I notice that the guard keeps his/her head out the window for several seconds as the train leaves the station. Presumably this is to verify that no passenger is being dragged along the platform with clothing caught in the doors. London subway trains have some automated sensors to detect such a problem. Do the older T1 cars have such a sensor?

    Steve: TTC is concerned that crowded stations need doors, but they also want to cut down on debris blowing onto the tracks. As for the guards, you may have noticed the green, yellow and red markers on the platform walls that indicate that the train is correctly positioned relative to the guard to open the doors (green), that the point has been reached during departure where the guard can stop watching the platform (yellow), and the point where the guard had damn well pull his/her head in *now* (red). These were installed many years ago in response to an accident at Spadina Station eastbound where someone alighting was caught by doors closing on the straps of their backpack (pack inside, person outside), dragged along the platform and killed. The guard didn’t see this because he had already started to switch to the other side of the train for St. George’s centre platform.

    I cannot count the number of times I have been on trains where guards have closed doors on passengers, not because they are dawdling, but because the crew wants to get in and out of the station as fast as possible, and the guard could not possibly be paying attention to the passengers. What I have observed, however, is that this behaviour has declined quite a bit in the past year or so, at least in my travels.

    When the drivers take over responsibility for the doors, there will be a set of video monitors to show them the status along the train and, one hopes, they will not attempt to close doors on passengers who are simply making a reasonable effort to board.

    One important issue here, too, is that the subway doors do *not* have “sensitive edges” that cause the doors to automatically reopen if blocked.


  10. Byford is correct in maintaining a driver on the train who would act as eyes and ears for Transit Control in case of an incident or passenger illness that caused the PAA to be activated. The driver can assess the situation in seconds and advise Transit Control accordingly. How many minutes would it take for emergency services or a TTC supervisor in his van to attend especially in heavy traffic? Another issue that people may not know and one of many reasons that the driver on the “automated” Scarborough RT was kept is how many times the old computer hardware and software fails bringing the train to a halt between stations. Within seconds the driver puts the train into “manual” operation and continues on. I don’t want to think of a driverless train getting stuck between stations and passengers suffering from the heat or cold or worse, risking injury by jumping down to track level and walking along the tracks. How long would it take for TTC staff to attend such an incident and attempt to restart the train? Always better safe than sorry.


  11. Here is another back-of-envelope calculation to look at the question from the opposite perspective.

    The Service Summary shows the number of vehicles in service at any given time of the day, broken into six broad time periods per day. This enables you to calculate a rough total number of operator hours in service over the course of a weekday, Saturday and Sunday, and then over a full normal week. (One has to make broad assumptions on how many hours are covered in each time period, based on the boundaries stated at the start of the Service Summary document, and it doesn’t include factors such as deadheading, overtime, etc., but it is good enough for this calculation.)

    For example: during the weekday AM peak period, there are 1,546 bus operators, 202 streetcar operators, 190 subway operators (2 per train) and 6 RT operators (1 per train?), operating for approximately 3 hours, 5 days per week, leading to about 29,000 operator-hours.

    Based on these calculations, there are a total of about 162,500 operator-hours over the course of a typical week, of which roughly 18,000 are subway and RT operators, or about 11% of the total.

    You estimate that operators cost about $640-million per year ($1.6 billion x 80% x 50%). If subway and RT operators make up 11% of total operating hours, this works out to roughly $70 million. This works out pretty close to what you calculated ($67 million).

    Steve: This is a good example of the sort of calculations I do now and then to verify numbers, and it’s good to find a completely different path (the back of another envelope) that leads to a similar result. If our two figures had disagreed, that would be a good reason to dive into both calculations to understand why, and if nothing else, this would lead to a stronger understanding of the factors that drive costs.

    Your calculated hours per week value is quite close to the actual budgeted hours/week values.

    Neither of our methods is perfect, but they show one does not need an elaborate computer model nor access to detailed internal information to do basic calculations as long as the basic assumptions and methodology are sound.


  12. Steve said:

    “One important issue here, too, is that the subway doors do *not* have “sensitive edges” that cause the doors to automatically reopen if blocked.”

    If I am not mistaken the TR trains have features built into them relating to the doors which are both a hindrance and a benefit to crewless operation. The doors on the TR from what I understand after talking to Andy and Brad (and having seen it with my own eyes) stop when they hit something. If someone or something gets caught in the door they stop at that point and cycle through. They can do this up to 3 times before the door goes O/S for safety reasons and a 99 is required.

    It does not automatically cycle through I think but no doubt if the train was crewless and fully automated it would.

    Steve: Yes that is true, although it is still possible for very thin objects (typically carrying straps or bits of clothing) not to be detected.

    I remember many a time running for an H4 where the doors had the mind of a guillotine. The doors themselves always shut hard (from what I recall the M series through to H Series all had doors that did) it was less of a gentle closing and more of a abdomen severing slam. This of course was compounded by the fact that operators wanted to get to the end points rather than sit around and take their time. This was primarily an issue prior to Russell Hill when ops took things a little fast and loose however now whether they like it or not ops cannot leave a station before a red has fully cleared.

    That being said I agree with Brian regarding dealing with emergencies on a crewless train. I was once on a powerless RT train in the middle of winter between Midland and Ellesmere Stations and would not have enjoyed waiting for a crew to come rescue us. We had to walk back to Midland Station in the same deep snow that caused the RT to konk out. That would not have been entirely safe or feasible if there was not a person aboard who knew what to do and how to to it safely. It would not have been enjoyable either to wait for someone to find us and then come assist (that would have taken awhile).

    Now that my rant is done I will sum things up by saying this … it will be a cold day in a pig flying hell when the TTC runs without crews onboard. I doubt the Ministry of Transportation would allow it on the basis of emergency response. I can see the Province legislating mandatory manned trains at all times long before this is ever put into practice.

    There are some things you can get away with and others you cannot in the name of saving a few dollars. Safety is not one of them. If you cut manned operation from the system you are playing a dangerous game with peoples lives.

    I shudder to think what would happen if trains were crewless. If someone has a heart attack onboard a train who would help them? If there is a PLAN A who would evacuate the train and get people to safety? If all hell breaks loose and the system needs to be keyed who will operate the trains … will we shut entire lines down due to signal issues?

    Steve: Just as the Station Collectors are to be redeployed as Customer Service agents, the headcount for operators won’t disappear, it will just change into another function such as a roving train supervisor. This would keep someone onboard for emergencies, but would remove their primary function as a driver/guard. Skytrain has roving staff, although not one per train. Their equipment tends to be more reliable partly because some of it is newer, the signal system is certainly in newer/better condition, and snowdrifts are uncommon.


  13. It’s true that Vancouver’s SkyTrain has fully-automatic trains, with no crew at all aboard, and doesn’t have platform-edge doors.

    But I’m pretty sure that SkyTrain does have extensive intrusion-detecting sensors, not just in stations but all along the right-of-way, so that if there is a body or a tree limb or a discarded swan boat or whatnot on the tracks trains will stop before striking it.

    I doubt the TTC or any other North American transit property could get away with running crewless trains without either platform-edge doors or intrusion detection. In fact I bet the TTC would have to have intrusion detection even with platform-edge doors in any place that anything could possibly fall on the tracks, which means any place that isn’t in a completely-enclosed tunnel, which is a substantial part of the existing subway network.

    Has anyone any idea how much the Vancouver sensor system cost to install or costs to maintain, or how much a similar system might cost for Toronto?

    Steve: I don’t know about costs, but do know that from day 1 there was intrusion detection technology on Skytrain both along the fences, and with sensors at track level in stations. It was the same technology used for border monitoring, and as I understand it, the Skytrain folks couldn’t get it from the Americans who, at the time, considered it too sensitive to release. They sourced it from Israel.


  14. If the TTC wants to save money then they should get the fare system modernized, at least to 20th century technology if not 21st. I have seen stations like Queen with 4 collectors in the booth and at least 1 chrash gate collector plus whatever is at the Albert St. Entrance. With more modern equipment they could get down to 3 agents, Albert Queen east side and Queen west side. There are lots of other stations where they can reduce collectors.

    Modern equipment would allow them to be out of the booth to help passengers more. The TTC fare control system is really antiquated.


  15. Steve:

    First off, I believe that any journalist (and I include myself in that description) has a duty to at least think about the reasonableness of what they write. The simple fact that a claimed annual saving is equal to 1/8 of the TTC’s entire operating budget (a well-known number, and easy to find especially for someone who covers City Hall and knows where to look) should raise a big red flag. However, the Star has been treating the Neptis paper as some sort of holy grail since they got to write an exclusive about it.

    When I started writing about public transit issues in Malaysia in 2006 it became very clear that education was going to a very important piece in the foundation of building awareness among the general public, and educating the media would be a cornerstone of those efforts.

    At the time whoever would report on public transit would simply quote press releases without investigating or asking questions. What made it more challenging for us was that articles related to public transit were rarely found in the ‘news’ section and would most often be found in the ‘Metro’, ‘Business’ and (surprisingly) the Property (Housing and commercial development) section.

    It became necessary to vet reporters before answering any questions from reporters … ensuring that they could get comments or quotes from us but we would have to get a face-to-face meeting to discuss public transit within a month.

    I would have loved to hold a Transit Camp like the one held in Toronto (and indeed, I wish there were more of those here in Toronto) to share information and ideas about public transit with reporters, columnists and the general public, but never had the opportunity before leaving Malaysia.

    Cheers, Moaz


  16. Hi Steve,

    I read your blog post with much interest. You have accumulated a wealth of information about TTC and I welcome your interest in my paper. I have reviewed my calculations and you have indeed found an error. The NPV figure of $2 billion, which is the one used in arriving at the conclusions, is correct, but the annual figure should be $100 million. This is 50 per cent more than your figure of $66 million, but half my figure of $200 million. In any case, the potential savings are huge.

    Turning to your estimate of $66 million per year in savings from eliminating drivers and guards, I think there are other things that need to be considered:

    • The sunshine list suggests that pay rates are higher. You estimate that TTC driver hourly wages are on average about $35 and add a 30% uplift for benefits, bringing the rate to about $45. From the 2012 “sunshine list” we know that there were 217 TTC operators with taxable earnings of more than $100,000. At $35 per hour this suggests some staff are working 2,857 hours per year, or 57 hours per week, 50 weeks per year. I don’t think this is credible. And some apparently earn considerably more than $100,000. The average wage rate (including overtime, premiums for Sundays, etc.) must be closer to $45 per hour. The $45 per hour rate will include taxable benefits but would not include pension contributions (which are not taxable), from the employer and employee, which add another 20%.

    • Absenteeism is currently 7%, implying that each staff member is sick, on average, about one day every three weeks. Not only does TTC have to pay drivers who stay home sick, it also needs to roster extra staff to cover for staff who do not turn up. Sickness is unpredictable, so the net impact is to increase TTC costs another 15%. In the UK and Germany, where I have experience as a Director operating bus and rail companies, we were able to get absenteeism below 4%, and I expect this is Andy Byford’s ambition.

    • Driver supervision (inspectors and the control room) and driver training will add another 10%.

    Altogether these raise your $66 million estimate to $100m.This translates into a Net Present Value (at a 5% real discount rate) of $2bn which is used in arriving at the conclusions. Given that TTC is likely to have to pay drivers even more in future, the net present value figure could be even higher.

    As for the PSDs (platform screen doors), you refer to a TTC estimate of about $1bn, which is two to three times my estimate of $300m to $500m. My figure was based on published costs of other PSD retro-fittings. One would expect costs to install PSDs in Toronto to be fairly low, given that the system is relatively modern, with straight and wide platforms, compared with say Paris, where they were installed on narrow curving platforms on a line built over a century ago. In any case, I never said PSDs are required for driverless operation, just that they might be worth having to reduce suicides. Vancouver has driverless trains without PSDs.

    Kevin Love has correctly pointed out that we need to take into account the costs of operating and maintaining the UTO (Unattended Train Operation) system. In practice, it is typically cheaper to install and maintain a train control system that operates entirely in UTO mode, than one that allows for use with drivers. I was involved in both the Vancouver and London Docklands installations, and the insistence of London Transport that it install a system that allowed for manual driving added substantially to cost and complexity, for a feature that is rarely used. TTC is now replacing the signalling on the entire subway system; now is the time to convert to driverless operation. Very likely there will be further savings in operation and maintenance of the signalling system that I have not included in my analysis.

    I do hope that you will point out to me any other figures you think may be incorrect.

    The report I produced for the Neptis Foundation was meant to inject some new thinking into the region’s transit debate, with analysis illustrated through examples of international best practice.

    It was meant to begin a discussion on how to ensure that the Big Move delivers on its promises.

    As far as “back of the envelope” calculations go, if transit agencies were required to publish transparent accounts, we would not need to speculate about these figures. Transport for London publishes separate accounts for each modes (bus, underground (subway), commuter rail, and even water and bicycle services), and GO and TTC should be required to do the same.



    Steve: Thank you for clarification of your calculations. However, I still disagree on some points.

    First, and most importantly, the TTC had already planned to eliminate one of two crew members as part of ATO implementation before your study was published. The saving from that change is not available to be counted as an “efficiency” in the sense of a saving your recommendations could achieve beyond what the TTC has or is planning to do anyhow. In the public’s mind, as this has been presented in the press, the entire saving is achievable, and this is not true. Only the saving from going from one to zero crew members is applicable as an offset to PSDs. Moreover, you make no allowance for the replacement of train crews with roving customer service agents or security staff, an expense the TTC does not now bear, but which would almost certainly be demanded for unattended trains.

    Second, you rely on the presence on the Sunshine list of operators with very high wages. You divide $100k by $35/hour to achieve an hours worked value of 2,857/year. However, the extra time will all be at overtime rates of at least 1.5 if not 2.0 times the base rate, and therefore the actual hours worked will be considerably lower than what you calculate. The base pay at $35/hour for 52 40-hour weeks is $72,800 leaving $27,200 to be made up by overtime. At $52.50/hour (1.5 times base), this requires 518 hours of overtime or about 10 hours/week, something that is quite within reason. (As an aside, it is worth noting that benefits do not scale 1:1 because many of them are fixed costs per employee, not per hour paid. As a result, the all-in cost of overtime is lower than the multiple would suggest.)

    The fundamental reasonableness check on any claim you make must be that there is only about $640-million available as operator costs, including all overtime, benefits, etc., based on the total budget and headcounts, and the vast majority of these operators drive surface vehicles, not subway trains.

    It is basic errors in your analysis that undermine the credibility of your statements. These were factors I can find with only a few minutes’ work and a simple counter-example.

    (There is a strong parallel here to other “efficiency” studies that miscount “savings” that have either already been achieved, or which depend on false assumptions. Your goals may be noble, but you come into a political context where such claims are at least suspect if not downright dishonest.)

    Your statement that “pay rates are higher” does not accord with the actual amounts in the ATU contract, only that some operators work overtime, a not unheard of situation on a large transit system. There are roughly 5,000 operators, and if only a few hundred have made it above the $100k mark, then there is certainly no justification for assuming that the value of an operator is on the order of 50% higher than what I used.

    Pensions are included in the 30% allowance I have made. Your statement about what is included in that 30% arises from your erroneous use of a higher hourly rate. The “Sunshine list” numbers are for direct pay and taxable allowances (not applicable to operators) only, not for benefits.

    I agree that absences such as sickness and training will reduce the number of productive days/year per average staff member. However, to get a solid figure, we would need to know the value specific to subway crews, not to hourly-rated TTC employees as a whole. It is a matter of record that absenteeism is higher among some groups than others. 6.82% is the current system average as shown in the January 2014 CEO’s report. I do not agree with your arbitrary inflation of this number to 14% because of the way the TTC uses spare operators to cover otherwise unmanned pieces of work. If there are more spares available than needed to cover for absences, there is other work for them to do.

    With respect to the PSD cost, my number is taken from the TTC’s capital budget estimate. Indeed, that $1-billion total sat erroneously in (and artificially inflated) the TTC’s list of “state of good repair backlog” for a few years before it was finally reclassified. (There are problems with other “SOGR” projects that don’t really belong under that heading, but the PSDs were the most egregious example.)

    The question of who said “PSDs are required for ATO” gets us into how the story about your report was written plus statements made by the TTC, notably by Andy Byford. When they were first proposed by TTC management some years ago, the justification was not related to suicides, but to track fires caused by debris blown from platforms to track level. Suicide prevention came along somewhat later as a rationale, but the number of “smoke at track level” incidents is substantially higher than the number of “injury at track level” situations. A related point here is that an international peer review of the TTC noted the high number of equipment failures, fires and on-board delay-causing incidents as a significant problem for a TTC attempting to increase the level of service (or reduce headways) to bump capacity.

    To put things another way, there are many issues that the TTC needs to address to improve overall service reliability, and hence regularly delivered capacity. PSDs would make a nice project to keep the engineering and construction folks busy for years, and they would make a handy excuse for the absence of progress on other fronts while the project ambled along. The TTC has a very long history of this sort of behaviour — finding some external “cause” or “solution” that has little to do with upgrading its own day-to-day practices.

    You may also be interested to know that Andy Byford is not 100% sold on PSDs because his experience at crowded stations (the very place they are intended to limit safety issues with trains entering at speed) is that waiting passengers crowd around the doors as these are known points of access. With the more random stopping locations of manual operation, this crowding does not occur and platform space is more evenly used.

    There is the question of whether we would install half-height doors simply as access controls, or full-height doors. If the latter, this introduces very substantial ventillation issues in stations that now depend on the piston effect of trains for circulation.

    Finally, there is the point that we will not be in a position to go with 100% ATO until sometime late in the next decade when resignalling of the Bloor-Danforth subway is completed. At this point, it is unclear whether this will even include ATO capability. That’s a debate in its own right, but almost half of the system’s savings for UTO would be due to that line’s conversion. Therefore only about 60% of whatever number might be on the table is available as a saving in the near term.


  17. On the topic of SkyTrain intrusion sensors, I personally witnessed a basketball taller than the reaction rail lodged in the track between stations. Nothing stopped the line from operating and nobody was aware of the problem. While a train likely would have simply burst the ball if it did hit it, there was a chance that train and rail damage could have occurred, particularly to the sensitive and closely-aligned LIMs. Put a different way, if a solid object the size of a basketball could make it onto the gideway without any warning then the detection system is wholly inadequate and clearly shows why on-board live spotters are critical to safety. This raises another question – If an automated train derails one axle how does it know immediately to stop?

    Royson James’ intentions were made clear in a follow-up article today where he wondered why the existing SRT technology is off the table for the Eglinton line and beyond. I was suspicious his earlier comments may have been leading to this and now it is clear he is off the mark on every account. Is all of this the last gasp of the behind-the-scenes movement conspiring to force a switch to sole-source ART technology, or is Mr. James just a loon is in own right?

    Steve: Bombardier and friends in the government have been flogging an Eglinton ICTS as long ago as the early days of Transit City. I would really like to know who paid for the Neptis report, and more generally where the talk of ICTS in the Star has support.


  18. Royson James also does not understand property tax, but he has sufficient influence at The Star so that his misunderstandings make it into the news columns as well.

    I think that his misunderstandings are instructive though, in fighting for causes – including transit. Mr James is definitely better read and more knowledgeable than the average voter. The fact that he is so far off the mark (or more accurately you and many readers believe he is off the mark) shows the amount of struggle and importance of perseverance if our views are to prevail. If Mr. James can be, apparently, so off the mark, it is no wonder that a large number of voters may believe that subways can be buit with fairy dust gifts from the “private sector”.


  19. Did anybody notice the factual error in James’ column?

    “From the east, you bus it to McCowan and change mode to an LRT heading west to Don Mills; then change to the Sheppard Subway before switching trains to go downtown; or hop a bus to go farther west, to hook into the subway at Downsview. Madness.”

    Wrong. From the East, one “buses it” to Conlins Rd., assuming one is coming from that far East; then LRT to Don Mills, then Sheppard Subway. The far more likely case is simply LRT to Don Mills. How can anyone who holds themselves out as a knowledgeable person in good faith not know that the LRT plan goes further than any proposed subway plan?

    Also, failing to mention that the transfer at Don Mills is extremely easy (same platform) is, while not strictly speaking a factual error, something that in my mind raises questions about the objectivity of the writer.


  20. Michael Schabas says:

    “The sunshine list suggests that pay rates are higher. You estimate that TTC driver hourly wages are on average about $35 and add a 30% uplift for benefits, bringing the rate to about $45. From the 2012 “sunshine list” we know that there were 217 TTC operators with taxable earnings of more than $100,000. At $35 per hour this suggests some staff are working 2,857 hours per year, or 57 hours per week, 50 weeks per year. I don’t think this is credible. And some apparently earn considerably more than $100,000. The average wage rate (including overtime, premiums for Sundays, etc.) must be closer to $45 per hour. The $45 per hour rate will include taxable benefits but would not include pension contributions (which are not taxable), from the employer and employee, which add another 20%.”

    I cannot believe that someone who served “…as a Director operating bus and rail companies…” would not realize that overtime gets paid at a higher rate than base salary. I guess he is too accustomed to all those incentives and perks management gets for making regular workers do more for less.

    I would hate to think of the length of a delay if a train went disabled between Eglinton and Lawrence or between St. Clair West and Eglinton West and a supervisor had to drive to the station and walk down the tracks to the train. It does not happen often but it does not have to happen often for the benefit of one crew member on a train to become apparent.

    The idea of driverless trains is a red herring. The area where real savings can be made is in automating and improving the sale of fare media in stations. In most other jurisdictions all fares are purchased from machines, not operators. The Chicago system has personnel in its station but they will not sell you anything or make change. This is true in Cleveland, Washington and many other cities as well.

    Ten years ago in London I first came across the concept of station agents as I was arriving in Earl’s Court Station after a long flight. On seeing my wife and me with our suitcases he came over to inquire where we were going and told us the best way to get to our hotel. When we came back later in the day he explained the various fare and pass options and showed us how to purchase the passes. While he was doing this regular riders could buy fares from the machines with out having to stand in line behind us. It is a lot cheaper to automate fare sales than trains and the fallout if one fails is not quite as severe as when the ATO conks out.

    Steve: As you no doubt saw from my reply to Schabas, I think his credibility is in tatters when he makes fundamental arithmetic mistakes like this. He certainly has his biases, and a love for unattended train operation is clear. I must assume that his inability to do basic analysis of budgetary information is through inadvertence, not from a desire to deliberately misrepresent the argument for his position. He is a professional consultant, after all, and we are only amateurs.


  21. Steve – I’ve wanted to respond to your response to my response but been unable to do for a few days. We seem to be in what the British call “violent agreement”. I don’t think PSDs are required for driverless operation either. So what are we arguing about? I personally argued against their installation when (back in 1989) we were planning London’s Jubilee Line Extension, but lost out to the advocates from Hong Kong. I allowed for them in my “subway modernisation package” to show that even if they were required for driverless operation, there was still a good business case. PSDs have some other benefits, which I did not include in my analysis, although they have dis-benefits too. Personally I am not a great fan, because they seem to increase dwell times.

    I also don’t see why we are arguing about average pay rates. Overtime is normal in transit industries; regular 8-hour shifts don’t match the passenger traffic very well, and it is often cheaper to pay overtime than to hire additional staff. You can’t just assume TTC pays all its staff the minimum applicable hourly rate; it never works that way. I simply use the “sunshine list” to estimate the actual average hourly cost, including overtime premiums. This seems to be about $45 per hour on the TTC, not the $35 rate you use.

    As for the actual savings from driverless operation, even your lower figure will still far exceed the (fairly negligible) net costs (especially if we don’t need PSDs). Features like platform CCTV, intrusion detection, and roving attendants that usually accompany driverless operation have some additional costs, but also give other benefits (and probably should be implemented anyway). Note that in Vancouver and on London’s DLR, the roving attendants perform other duties, including checking tickets (because both systems are open, with proof-of-payment fare collection). London is now closing ticket offices (because virtually nobody uses them any more) and moving staff out into the passenger areas where they can actually be helpful. If, as expected, London moves towards driverless trains on other lines, there will be further staff cost savings, while service to passengers will actually be improved. I, like you, have a love of good public transit. If driverless operation means we can have better service, for less money, then I am for it. Who would not?

    My report for Neptis was primarily about the Metrolinx schemes; I included the “subway modernisation package” mostly for completeness, to show that increasing capacity, especially during off-peak periods, and smarter fares can help achieve the Metrolinx goal of doubling transit ridership. The TTC’s new signalling system will allow about 30% more trains, but unless driver costs are tackled, it will be difficult to justify higher frequency services, which can provide more comfort in the peak, and shorter wait times in the off-peak when demand is most time-sensitive. One-person operation seems very much a halfway solution. I remember when the Provincial Government offices had “elevator operators”. When they installed new automatic elevators in the 1960s, they kept the attendants for years, but they only worked during office hours. After 6PM, late-working civil servants had the thrill of riding unattended elevators. Keeping one driver on board a train when none are required makes about as much sense.

    In the end, I think it all comes down to how many new riders can be attracted onto transit, with the available money. Even Paris, a city with very strong transit unions, is moving to driverless operation, because the benefits to taxpayers and riders are simply too obvious. As you point out, it will be at least a decade until the Bloor Danforth line will be resignalled. So there is plenty of time to redeploy subway drivers into other jobs.

    Please keep reading my report and let me know if you find any other numbers that don’t seem right.

    Steve: How many times do I have to say this? The Sunshine List represents a small fraction of the total operator workforce. It suits your argument to quote a higher figure than me, even though the one I use is actually about 10% higher than the current contracted maximum pay rate (junior operators get less, although they are less likely to be working on the subway). The calculation I performed explicitly recognizes that driving hours and pay hours are not the same thing, and this corrects for things like travel time and paid coffee breaks. I used 6.5 driving hours per operator with a pay hours of 8, or a 23% premium.

    In any event, there is the fundamental point that the vast majority of operator hours on the TTC go to surface, not subway operations. You cannot claim a huge saving against the total operating budget because this simply is not credible. Yes, there is a saving, but there are offsetting capital and operating costs as you note. These were not factored into your claimed saving, and you gave the impression that there were hundreds of millions on the table just for the simple act of getting rid of the drivers. This plays into a strong anti-union bias among some politicians here, not to mention giving the impression that UTO (of whatever technology) offers vast labour savings when this is simply not true. Some savings, probably, but vast, no.

    As for additional service, you also do not factor in the substantial capital costs and increased running costs of the trains. The TTC has yet to budget for the rolling stock, storage space and maintenance costs of a 30% larger fleet, even assuming they could actually operate the shorter headways which I doubt for reasons I have discussed previously. This is not even a “below the line” budget item which is jostling for funding with other projects, it is something the TTC has completely omitted from its projections.

    They are happy to factor in fleet related costs for new lines such as the proposed Relief Line, but they omit the true cost of upgrading the existing system as an alternative thereby making a new line look relatively more expensive.

    Stirring all this together, ATO/UTO is one part of a much more complex discussion which, sitting across the pond, you might not see on a day to day basis.


  22. Hi Steve:

    If you turn to page 52 of the Neptis report, you will see:

    “Adding capacity through the day will require more trains. With the Vaughan extension, TTC will need about 750 cars to operate the current service. Based on recent orders, TTC can purchase additional subway cars at about $3 million each. We allow a cost of $750 million to purchase 250 additional cars. This includes some allowance for additional train sidings that could be built at Wilson where there is vacant land. Adding capacity will also require further capacity improvements at key stations, which could notionally cost a further $500 million, in addition to works that are required in any case to carry ordinary traffic growth.

    Our estimates of the costs and benefits of a “modernization package” are found in Table A8 in the Appendix and include smart pricing, discussed in detail in Chapter 10. Ridership is assumed to grow 10% due to smart pricing, and a further 5% due to the higher frequency of service, with reduced wait times and reduced crowding, for a total uplift of 150,000 passengers per day or about 45 million per year. With an average fare of about $2.00, the revenue increase would be $90 million per year or about $2.1 billion NPV. The net incremental capital and O&M costs for modernization and resignalling are about $3.2 billion, but the net cost, after taking account of incremental revenues, is about $1.1 billion or $5,540 for each incremental daily transit rider.

    We think our estimates are conservative and the additional revenues might well offset all incremental costs.
    As with the GO schemes, traffic on the subway will grow faster with more frequent off-peak services. However, the effect will be less, as the subway already operates a fairly frequent service. We assume a further 33% growth, to 200,000 additional riders due to the “Modernization Package.” The 35% capacity increase will enable the subway to carry underlying growth in demand, but less-crowded off-peak trains will also make the system more attractive to new riders.”

    And on page 119, you will see:

    “Assuming an incremental cost of $20 per train km, the cost to operate 35% more subway train at all times would be about $86 million per year or $1,978 million NPV.”

    Steve: My point was that the TTC has not adequately planned for all of the components of various new and improved services, and they have systematically understated the all-in cost of handling all growth on the existing Yonge-University line. This has caused any new line like the DRL to appear more expensive, relatively, because the additional infrastructure needed to make YUS handle all of the load (assuming this was even possible) was not included in comparative costing. There is a point where we need to stop trying to fit everyone on one line and recognize the value of multiple routes. We must also know what the upper bound for any of the routes actually is rather than blindly drawing maps with no awareness that the network as a whole may not have the presumed capacity. This is my biggest critique of TTC and Metrolinx planning over the past decade.

    Moving beyond the Yonge-DRL debate, I believe that Metrolinx has badly misled the GTAH residents and politicians with claims about reduction in congestion (or limiting its growth) by using overall averages in calculated travel times rather than breaking apart the benefits (or lack thereof) of various potential networks (recognizing that the whole thing will never be built) for different types of travel. The “32 minute” campaign is a complete sham for the many 905-to-905 commuters who might, at best, see some BRTs with limited capacity that might, just might, link their origins and destinations.


Comments are closed.