Train Operator Labor Efficiency
Last summer, I brought up a metric of railroad labor efficiency: annual revenue hours per train driver. Higher numbers mean that train drivers spend a larger proportion of their work schedule driving a revenue train rather than deadheading, driving a non-revenue train, or waiting for their next assignment. As an example, I am told on social media that the LIRR schedules generous crew turnaround times, because the trains aren’t reliably punctual, and by union rules, train drivers get overtime if because their train is late they miss the next shift. Of note, all countries in this post have roughly the same average working hours (and the US has by a small margin the highest), except for France, which means that significant differences in revenue hours per driver are about efficiency rather than overall working hours.
I want to clarify that even when union work rules reduce productivity, low productivity does not equal laziness. Low-frequency lines require longer turnaround times, unless they’re extremely punctual. Peakier lines require more use of split shifts, which require giving workers more time to commute in and out.
The database is smaller than in my posts about construction costs, because it is much harder to find information about how many train operators a subway system or commuter railroad employs than to find information about construction costs. It is often also nontrivial to find information about revenue hours, but those can be estimated from schedules given enough grunt work.
In Helsinki, there is a single subway trunk splitting into two branches, each running one train every 10 minutes all day, every day: see schedules here and here. This works out to 65,000 train-hours a year. There are 75 train drivers according to a 2010 factsheet. 65,000/75 = 867 hours per driver. This is the highest number on this list, and of note, this is on a system without any supplemental peak service, allowing relatively painless scheduling.
In Toronto, there were 80,846,000 revenue car-km on the subway in 2014 (an additional line, the Scarborough Rapid Transit, is driverless). Nearly all subway trains in Toronto have six cars; the Sheppard Line runs four-car trains, but is about 10% of the total route-length and runs lower frequency than the other lines. So this is around 13.5 million revenue train-km. According to both Toronto’s schedule of first and last trains per station and this chart of travel times, average train speed is around 32 km/h between the two main lines, and a bit higher on Sheppard, giving about 420,000 annual service hours. In 2009, there were 393,000 hours. Toronto runs two-person train operation, with an operator (driver) and a guard (conductor); this article from 2014 claims 612 operators and guards, this article from 2009 claims 500 operators alone. 420,000/500 = 840, and, using statistics from 2009, we get 393,000/500 = 786; if the article from 2014 misrepresents things and there are 612 drivers in total, then 420,000/612 = 686. If I had to pick a headline figure, I’d use 786 hours per driver, using the 2009 numbers. Update: the Scarborough RT is not driverless, even though the system could be run driverless; from the same data sources as for the subway, it had 23,000 operating hours in 2014, which adds a few percent to the operating hours per driver statistic.
In London, unlike in North America, the statistics are reported in train-km and not car-km. There are 76.2 million train-km a year, and average train speed is 33 km/h, according to a TfL factsheet; see also PDF-p. 7 of the 2013-4 annual report. In 2012, the last year for which there is actual rather than predicted data, there were 3,193 train drivers, and according to the annual report there were 76 million train-km. 76,000,000/33 = 2,300,000 revenue-hours; 2,300,000/3,193 = 721 hours per driver.
In Tokyo, there used to be publicly available information about the number of employees in each category, at least on Toei, the smaller and less efficient of the city’s two subway systems. As of about 2011, Toei had 700 hours per driver: from Hyperdia‘s schedules, I computed about 390,000 revenue train-hours per year, and as I recall there were 560 drivers, excluding conductors (half of Toei’s lines have conductors, half don’t).
In New York, we can get revenue car-hour statistics from the National Transit Database, which is current as of 2013; the subway is on PDF-p. 13, Metro-North is on PDF-p. 15, and the LIRR is on PDF-p. 18. We can also get payroll numbers from SeeThroughNY. The subway gets 19,000,000 revenue hours per year; most trains have ten cars, but a substantial minority have eight, and a smaller minority have eleven, so figure 2,000,000 train-hours. There were 3,221 train operators on revenue vehicles in 2013, and another 373 at yards. This is 556 hours per driver if the comparable international figure is all drivers, or 621 if it is just revenue vehicle drivers. The LIRR gets 2,100,000 annual revenue car-hours, and usually runs trains of 8 to 12 cars; figure around 210,000. There were 467 engineers on the LIRR in 2013; this is 450 hours per driver. Metro-North gets 1,950,000 annual revenue car-hours, and usually runs 8-car trains; figure about 240,000. It had 413 locomotive engineers in 2013; this is 591 hours per driver.
In Paris, the RER A has 523 train drivers (“conducteurs”). The linked article attacks the short working hours, on average just 2:50 per workday. The timetable is complex, but after adding the travel time for each train, I arrived at a figure of 230,000 train-hours a year. 230,000/523 = 440 hours per driver. There’s a fudge factor, in that the article is from 2009 whereas the timetable is current, but the RER A is at capacity, so it’s unlikely there have been large changes. Note also that in France, workers get six weeks of paid vacation a year, and a full-time workweek is 35 hours rather than 40; adjusting for national working hours makes this equivalent to 534 hours in the US, about the same as the New York subway.
Pedestrian Observations 
It’d be interesting to see this expressed per on-board crew member, or even per dollar of total operational worker compensation (though I don’t know if there’s good enough data on these things).
The problem is that these are really separate issues. You can have conductors, as Toronto and Tokyo do, and still have relatively high efficiency per driver. You can have OPTO, as the RER A does, and still have very low productivity. Eliminating the conductors is one way to improve efficiency; scheduling trains and crews in a way that allows the agency to get more service-hours per driver is another.
Compensation runs into issues like “some cities are way richer than others.”
And some cities are way more expensive to live in. For example, some people love to attack how much BART (SF Bay Area) operators and station agents get paid, but the cost of living in the Bay Area is very high. What would be a princely sum in, say, Cleveland is really just a middle class living in the Bay Area.
Correction to your note on Toronto: The Scarborough Rapid Transit is not driverless. It can operate on its own, but due to union rules, it still has a single operator in the cab (no guard on these trains)
One more point of clarification, since it’s not clear in your write-up, but operators and guards are the same employees at TTC. So, it’s a bit misleading to refer to one statistic as saying “operators alone” and another as saying “operators and guards.” Except for Scarborough RT, the typical subway schedule has an employee operating in one direction and guarding in the other. This eliminates the need for them to walk to the other end of the platform on arrival, which is important because they have very short turn times.
Re operators and guards, the issue here is that these are to some extent separate questions of staffing level vs. crew utilization. Switching operators and conductors due to short turnaround times is also practiced in New York on the 42nd Street Shuttle, but overall, the gains from shortening turnaround times are very small (it takes, what, 2 minutes to walk from one side of the train to the other?). The systems I’m comparing have a mixture of staffing levels – New York has 2 employees per train, usually with separate job titles for drivers, who sit at the front, and conductors, who sit in the middle; London and Helsinki have 1; Toei has 2 on half its trains and 1 on the other half; the RER has 1; the LIRR and Metro-North have between 3 and 7 depending on time of day.
If Toronto actually has just 306 2-person crews on the subway, then that’s truly amazing efficiency, approaching a full-time schedule per employee.
According to this 2014 Toronto Star discussion of TTC automation – which, obviously, is not exactly authoritative – the TTC has 612 subway operators AND guards: http://www.thestar.com/news/gta/transportation/2014/02/07/driverless_subway_trains_not_in_nervous_toronto_says_ttc_head.html.
It’s not an automobile. or even a bus. It has to be shut down and the other end has to be started up. One of the reasons why balloon/loop tracks are valuable.
Adirondacker: And the steam locomotive has to be moved to the other end?
Any guess on SEPTA subway and regional rail productivity? Why wasn’t it included?
It wasn’t included because I don’t know the number of train drivers in US cities outside New York. Any leads would be appreciated.
Here’s a database of 2013 salaries of all public employees in the Bay Area other than San Francisco. It includes BART. I’ve never bothered to crunch the numbers though. http://www.mercurynews.com/salaries/bay-area/2013
MBTA has been one operator per train on the Red, Orange, and Blue lines for about 4 years now, since the Red OPTO rollout finished. That’s with standardized all-day 6-car trains on all 3 lines (pre-OPTO they’d shrink to 4 cars on weekends and far off-peak). Green Line still has operators per every car because of onboard fare collection on all surface routes, although the vehicles do have capability of Proof-of-Payment boarding that could enable one-person ops if the agency ever chooses to pursue full PoP implementation. Almost all Green Line trains are 2-car at any hour. Not many singlets anymore because of the ADA requirement to pair the more plentiful old high-floor cars with at least one newer low-floor car. Triplets are more or less shelved for the time being while >10% of the fleet is out-of-service at any given time for a massive ongoing 5-year midlife rebuild program.
Downside is lots of the bumped Blue/Orange/Red operators were auto-promoted to inspectors, so there’s now a glut of them milling around the stations “assisting” the single operators (who don’t need extra assistance because new platform-side CCTV screens let them video-monitor the entire platform from the front-car starting spot). That staffing skew should self-correct over time with inspector retirements slimming the ranks back to baseline, but they took an indirect short-term hit with overall ops staff efficiency which won’t reflect in the hours-per-driver math.
MBTA commuter rail is 1 conductor per every 2 open cars (i.e. no extra staff needed if a far off-peak train closes cars to passengers) plus the engineer. Efficiency slowly increasing as bi-level cars surpassed a 50% share of total car fleet within the last 18 months and will surge to 100% if they ever fund full replacement of the remaining 200 single-levels in time for their projected 2020-22 retirement. The problem they face is that they’re in a vulnerable spot with car availability due to a lot of failing old equipment on its way out, their first-generation bi-level orders cycling out for months at a time for ongoing mid-life overhaul, and an option order for 75 bi-levels that had to be canceled because their most recent Hyundai-Rotem purchase were a bunch of lemons. So too many odd-numbered consists run carrying the third conductor instead of assembling only even-numbered consists +2 cars at a time. Newly contracted operator Keolis stated in its bid package that it wanted to improve efficiency by going to even-number sets whenever possible (previous operator MBCR just didn’t give a crap) and skewing the bi-levels to the overall fullest trains to max out the riders-per-staffer, but it hasn’t worked out that way due to triaging the acute car shortage. Also still a few too many lightly patronized off-peaks carrying an extra conductor when they should be closing cars, but old abuses die hard and work rules regarding swing shifts still need to be reformed for more staff-as-directed flexibility. That kind of reform is a work-in-progress for the T’s newly assembled Fiscal Control Board.
I assume that your sources are only counting revenue hours or miles, and that layovers, no matter how short, are excluded. When time for recovery, layovers, pre- and post-trip requirements and deadheading are deducted, the maximum amount of time is reduced.
A reasonable expectation (I’m unfamiliar with railroad labor standards, and will loosely apply bus labor standards) would be about 30 minutes for a combination of pre- and post-trip matters, as well an insignificant deadhead (more time would be needed if the train garage was far from the trip starting point, of course). Operators need scheduled time to use the restroom and compensate for the inevitable delays. As a lower bound 10% of trip time is needed for recovery, or 6 minutes every hour; as that is a lower bound a more reasonable estimate may be 8 to 10 minutes per hour, or 12% to 16%. With 7.5 hours available for the shift, 10% to 16.67% for recovery would mean that 6 hours, 15 minutes to 6 hours, 45 minutes is actually available for revenue service.
A midpoint guess of 6.5 actual hours per day, 5 workdays per week, and 52 weeks per year results in 1690 hours theoretically available. Of course, operators get sick, want vacation time, and often have continuing training and other compensated company or union business. 4 to 8 weeks would be a reasonable guess of the actual amount of lost work time for a full-time employee, or 160 to 320 hours.
Considering all the above, the total revenue time per year per employee would theoretically be about 1300 to 1500 hours, assuming perfect schedule efficiency and minimal deadhead. When additional labor rules, longer deadheads, or split shifts are needed in practice, the amount of revenue service per driver would decrease, often substantially, as in your LIRR calculations.
Also, not all people who are listed in the payroll as operators would be scheduled full time. Extra-board operators are nearly universal, and some companies (more likely to be the case at smaller and non-union workplaces) may have employees nominally listed as operators but also have them split their job assignments, such as having them also training new-hires or doing some office work.
In New York, the union has vociferously and successfully resisted part-time work. A few years ago, I sanity-checked by looking at annual wages (and you can do the same, since the dataset includes total compensation), and they were consistent with all or nearly all train drivers working full-time.