Labor and New York Bus and Subway Frequency
In New York, the frequency of a bus or subway service is regularly adjusted every three months to fine-tune crowding. Where Berlin has a fixed clockface timetable in which most trains run every 5 minutes all day, New York prefers to make small changes to the frequency of each service throughout the day based on crowding. The New York approach looks more efficient on paper, but is in fact the opposite. It leads to irregular frequencies whenever trains share tracks with other trains, and weakens the system by leading to long waits. But another problem that I learned about recently is that it is unusually inconvenient for labor, and makes the timetabling of trains too difficult.
How does New York timetable trains?
New York City Transit meets every three months to change the frequency of each named (numbered or lettered) subway service and, I believe, also every bus service. The rule is that, off-peak, train loads should be 125% of seated capacity at the most crowded point of the journey. Of note:
- This is adjusted by time of day – it’s not one fixed frequency for the entire midday off-peak.
- At the peak, the frequency follows the same rule but the guideline allows much more crowding, equal to about 3 times the seated capacity.
- When multiple services share the same trunk, the crowding is based on the service, not the trunk. This matters because sometimes there’s a notable difference, for example the 2 is more crowded than the 3 coming in from the Bronx and Harlem.
- There is no adjustment for the length of the most crowded point: it could be one 1.5-minute interstation, or a long 20-minute stretch.
- The interlining between different services leads to irregular frequencies on each, thus different crowding levels. The frequency guidelines are averaged across different trains of the same service.
- There is a minimum frequency of a train every 10 minutes weekdays, every 12 minutes weekends; late at night, all trains run every 20 minutes.
I wrote in 2015 about the negatives of this approach, focusing on the issue of interlining of different services with different frequencies and the seams this creates. Because the system is not trunk-based, the alternation of (say) 2 and 3 trains on the long trunk that they share is not regular. Thus the frequency is irregular and so is crowding. More recently, in 2019 I wrote about the frequency-ridership spiral. The guidelines are based on thinking from an era when nobody thought ridership was endogenous to frequency; direct commute trips without transfers are long compared with frequency, so in that era, the only perceived purpose of frequency was to provide capacity for a fixed ridership. But in reality, 10 minutes is too infrequent for the subway trips people actually take, which average 13.5 minutes without transfers.
Timetabling and labor
The consequence of the constant fidgeting on frequency is that crew timetables are unpredictable. In one period, the system may need more subway drivers reporting to Coney Island Yard, and in another, it may need more at yards in the Bronx and Queens. Bus depots likewise are located all over the city. Naturally, subway yards and bus depots are at peripheral locations, usually accessible only from one subway line in one direction. Commuting there from most spots in the city is difficult.
Moreover, as is typical in the American unionized public sector, workers at New York City Transit pick their schedules in descending order of seniority. The senior workers can make sure to pick work out of depots near where they live. The junior ones spend years having to work out of the Bronx one day and Southern Brooklyn the next. The commute is so bad that the TWU negotiated paid commute time: workers who have long commutes, forced by erratic timetabling, get paid for commute time, rather than just for time they actually work. Car ownership rates among subway workers are high, which is not typical of New York workers.
The erratic scheduling also means that, even independently of the long commutes for train and bus drivers, there is extensive downtime between runs. A prominent peak in the schedule means that split shifts are unavoidable. Split shifts are undesirable to workers, and therefore shift scheduling always includes some compromises, for example paying workers half-time for time between shifts (as in Boston), or scheduling shorter paid gaps between revenue service. In New York, there are some subway train operators who have three uninterrupted hours of paid work in which they do not drive a revenue train.
As a result, comparing total counts for train operators and service-hours, NYCT gets around 550 hours per train operator. I provided some comparative links in 2016, but they have rotted; Berlin, which runs close to even service on the U-Bahn with very little peaking and little adjustment over time, has 790 drivers and gets 22.1 million annual train-km at an average speed of 30.9 km/h, which is 905 hours per train driver. If you’ve seen me cite lower figures, such as 820 or 829 hours/driver, they come from assuming 20.3 million train-km, which figure is from 2009.
This is not because New York City drivers are lazy or overpaid. The timetabling is forcing unnecessary pain on them, which allows them to demand higher wages, and also leads to inefficiency due to much downtime and paid commutes. NYCT pays bus and train drivers $85,000 a year in base salary per See Through NY, and there aren’t hordes of people knocking on NYCT’s doors demanding those jobs. Boston pays slightly less, around $80,000, and has some retention problems among bus drivers; private bus companies that attempt to pay much less just can’t find qualified workers. The market pay is high, partly because it’s a genuinely physically tough job, but partly because it’s made tougher by erratic scheduling. In Munich, the richest city in Germany, with average per capita incomes comparable to those of New York, S-Bahn drivers get 38,000-45,000€ a year, and one wage comparison site says 40,800€. Berlin pays less, but Berlin is a poorer city than both Munich and New York.
There is another way
New York should timetable its trains differently. Berlin offers a good paradigm, but is not the only one. As far as reasonably practical, frequency should be on a fixed clockface timetable all day. This cannot be exactly 5 minutes in New York, because it needs more capacity at rush hour, but it should aim to run a fixed peak timetable and match off-peak service to peak service.
One possibility is to run all trunks every 2.5 minutes. In some cases, it may be fine to drop a trunk to every 3 minutes or a bit worse: the L train has to run every 3 minutes due to electrical capacity limits, but should run at this frequency all day; the local Broadway Line trains should probably only run every 3 minutes as they have less demand. But I wouldn’t run the 1 train every 3 minutes as it does today, but rather keep it every 2.5, matching the combined trunk of the 2 and 3, and try to time the cross-platform transfers at 96th Street. Train services that share tracks with other services should thus run every 5 minutes, maybe 6. Last year I called this the six-minute city, in which all buses and trains run every (at worst) 6 minutes all day. In the evening this can drop to a train or bus every 10 minutes, and late at night every 20, but this should be done at consistent times, with consistent quantity of service demanded week in, week out.
There may be still some supplemental peak frequency. Taking 3 minutes as the base on every trunk, some trunks may need 2.5 at the peak, or ideally 2 or less with better signaling. It represents a peak-to-base ratio of 1-1.2, or maybe 1.5 in some extreme cases; Berlin, too, has the odd line with 4-minute peak frequency, for a ratio of 1.25. The employee timetabling is unlikely to be onerous with a ratio of 1.25 rather than the present-day ratio of around 2, and while passengers do drop out of riding trains for short distances if they only come every 10-12 minutes, 6 minutes on branches may be tolerable, even if 5 is slightly better.
It’s a large increase in service. That’s fine. Frequency-ridership spirals work in your favor here. Increases in service require small increases in expenditure, even assuming variable costs rise proportionately – but they in fact do not, since regularizing frequency around a consistent number and reducing the peak-to-base ratio make it possible to extract far more hours out of each train driver, as in Berlin. Net of the increase in revenue coming from better service, such a system is unlikely to cost more in public expenditure.
This remains true even assuming no pay cuts for drivers in exchange for better work conditions. Pay cuts are unlikely anyway, but improving the work conditions for workers, especially junior workers, does make it easy to hire more people as necessary. The greater efficiency of workers under consistent timetabling without constant fidgeting doesn’t translate to lower pay, but to much more service, in effect taking those 550 annual hours and turning them into 900 through much higher off-peak frequency. It may well reduce public expenditure: more service and thus greater revenue from passengers on the same labor force.
What it requires is understanding that frequency is not to be constantly messed with. Gone are the days when frequency was naturally so high that it looked to be just a function of capacity. On a system with so many transfers and so much short ridership, ridership is endogenous to it, and therefore high, consistent frequency is a must for passengers. For workers, it is also a must, to avoid imposing 1.5-hour commutes on people without much notice. Modernization in this case is good for everyone.
Looks like the main reason the RATP can get away with varying frequencies on the Métro is that no lines on it share tracks so screwing up frequencies by interlining services isn’t really an issue.
I have made my peace with this. Pray that others may, also.
“ But I wouldn’t run the 1 train every 3 minutes as it does today, but rather keep it every 2.5, matching the combined trunk of the 2 and 3, and try to time the cross-platform transfers at 96th Street.”
At 2.5 minute frequency, is it really necessary (or consistently possible) to time cross-platform transfers? Is there any practical ridership penalty to an average 75 sec transfer wait?
They do at Finchley Road and Wembley Park on the tube.
It’s an entirely different user experience that one you can just walk off and onto another train and go versus having to actually wait at platform. But problem is how can you manage such sub-minute schedule accuracy throughout the day in real life operation, especially considering how passenger movement might disrupt the schedule a bit
My first reaction is padding the schedule so trains can make up time so to speak for disruptions, but unclear this is feasible for lines that are (or were pre COVID) packed at peak time (eg the 4 & 5)
Would that padding not waste more time than the actual transfer?
Hopefully with time fewer people will hold the doors open as the wait for the next train isn’t ten minutes at 8 am.
Yeah that’d be nice, but people still hold the doors open at 9 am even with a three minute wait to the next train
I think it would help with platform crowding because the platform becomes a hallway instead of a waiting room.
The impact on ridership is real, albeit small. 2.5 minutes with timed transfers = 2.5 minutes of worst-case wait time; 3 minutes and then 2.5 without timing = 5.5 minutes. The thing I worry more about is that, because the southbound 1 train dumps its ridership onto the southbound 2/3 at 96th, if the frequencies don’t match there will be large variations in crowding on the southbound 2/3.
To my knowledge there are timed cross platform transfers at Plärrer on the Nuremberg subway even when the U2&U3 overlap to 100 second headways…
NYC is a much higher ridership system though – harder to keep dwell times low
Busier lines than those of New York keep dwells to 30 s.
“The timetabling is forcing unnecessary pain on them, which allows them to demand higher wages, and also leads to inefficiency due to much downtime and paid commutes.”
And yet I keep hearing, even in my Left Coast bubble, that unions are the culprits behind high costs and that Uber is the answer:(
…I mean, vis a viz labor costs, they’re not wrong. However, the unions are a problem that way only insofar as you presuppose that the correct thing to do is just overwork and underpay people, while privatizing utilities like goods delivery and transportation with the aid of regulatory capture and profiting off of inelastic demand curves that are naturally never going to respond well to market forces. See also healthcare and education.
Moving from nominally liberal but really libertarian (read: conservative) New Hampshire to nominally progressive but really liberal (read: conservative) Oregon has taught me a valuable lesson: there are very, very few things Americans are right about.
What Alon is proposing is basically just to maximise the amount of time drivers spend actually driving trains, rather than sitting idle or heading to/from work. This is done by moving close to an all-day peak, with frequencies only ramping down significantly at night. This should leave everyone happy, since:
a) workers are happy because they will be working fewer split shifts and other inconvenient shifts,
b) the MTA is happy because the workers are more productive, running more service for the same number of labour hours,
c) riders are happy because they have higher frequencies throughout the day, and so can lead flexible transport-oriented lives without long waits for trains, and
d) unions should be happy because more frequency means more ridership means more funds means more opportunity to hire workers
It’s a virtuous circle. Not doing so essentially amounts to losing pounds by saving on pennies.
I hope so, but it might not make unions happy. Operators/drivers and conductors might not be happy because some of them at very high in seniority need to work longer to get paid for the same amount (again, there could be some shifts driving trains for significantly shorter period of time within an 8-hour work day while getting paid for 8 hours) or cannot get overtime pay (which I believe 1.5x of hours they worked for overtime).
MTA might not be happy with hire more workers because of the cost to hire more workers (benefits and other items associated with hiring more workers). I heard many transit operators in the U.S. typically have workers working for more overtime when they need to hire more workers though I’m not 100 percent sure if this is the case at the MTA.
They could fly in drivers from Berlin. I’m sure they’d love the pay raise…
They should then move to OPTO and ATO where possible, freeing up staff to run more trains.
On a related note I think the antipathy to unions comes from the perception that unions typically think of themselves as job creators rather than public services – yes, pay to hire when you need to, but organizational change to increase the efficiency of the existing workforce should come first.
I think there is a parallel to NIMBYs here – any change, however good on the net, will be opposed by some union members who oppose any disruption to the current arrangement
If a train every three minutes is as good as it gets, is this really better than a tram that comes every sixty seconds or more often than that?
The capacity is higher than this – New York runs every 2.5 minutes today, and is capable of running every 2 without too much investment in electronics, and those are 180 m long trains. Not to mention, the average speed is twice what a tram with dedicated lanes could do in New York traffic. (Yes, in smaller cities the gap is smaller, but is still there – in Berlin it’s 30 km/h vs. 19 km/h.)
If you’d ever taken the A from Harlem to midtown you would not be saying this.
You’d have to explain local and express first. The service speed is 55mph/81kph which it probably runs at for most of the trip between 125th and 59th. A quick glance at weekday southbound schedules has the A or the D making 125th to 59th in 8 or 9 minutes and the B or the C taking 12.
“paying workers half-time for time between shifts (as in Boston)”
What? No, that’s not how it works here.
It is on commuter rail, not sure about the subway?
On the subway and the bus there’s a notion of a “spread bonus”, where if the total work day – the duration from the beginning of the first part of a split shift to the end of the latter part of the split shift – is more than a certain threshold (I think 12 hours), then the hours past that get some not-quite-overtime-level multiplier.
But an “ordinary” split shift that is 8 hours of work within a 12-hour span is just paid as 8 normal hours.
*The commute is so bad that the TWU negotiated paid commute time: workers who have long commutes, forced by erratic timetabling, get paid for commute time, rather than just for time they actually work.*
Danger! While something called travel time exists, it’s not only used in certain cases. For example, if one starts at Coney Island but finishes at another terminal, you’re entitled to deadhead time back to Coney Island. If you live in CI, but you’re assigned a job that *starts* in the Bronx and *finishes* at the terminal where you started, you’re not entitled to any additional pay.
*Car ownership rates among subway workers are high, which is not typical of New York workers.*
People sign up for the job in the hopes that they can finally afford a car. Non-white people don’t brag about not owning a car in the same way that white core New Yorkers do so.
Coincidentally, for train operators and some maintenance workers, having a clean license at the time of appointment is a requirement.
Otherwise, part of it comes from the fact that a lot of workers end up dumping their OT somewhere, and the easiest place is car which then cuts down on commute times which makes working more OT even more attractive. The other thing that makes driving attractive is the fact that nobody wants to deal with overnight service if one has early reports or late finishes. Also, some choose to drive because they want to get off system property ASAP lest somebody assault them or ask them customer service questions. And yes, there’s a sizable chunk of the employee base that lives outside of the city because it’s where they can afford to buy a home with their wages. New Jersey, Connecticut, the Lower Hudson Valley, and the Poconos are becoming more popular with those trying to stretch their budgets. $400K in the Hudson Valley in a mixed race area with newer housing can be more compelling than a $600K house in Southeastern Queens with meh schools or paying for private schools.
*n Munich, the richest city in Germany, with average per capita incomes comparable to those of New York, S-Bahn drivers get 38,000-45,000€ a year, and one wage comparison site says 40,800€.*
By New York standards, you’ll find somebody to fill the jobs at that income, but you’ll have to work a bit harder given that the reduced income means a reduced income on the pension at retirement. The implication is that these jobs have to pay “buy a house” income, especially when there’s competition with Sanitation, NYPD, and unionized construction for this pool of workers. And increasingly, the younger cohorts signing up have college degrees, so we’re not talking about poorly educated high school dropouts anymore. $55k to operate a train AND work weird hours AND step in fecal matter is less compelling compared to making the same money in a climate controlled office.
*in effect taking those 550 annual hours and turning them into 900 through much higher off-peak frequency*
The problem is that with most lines, adding that additional trip requires paying OT and/or slicing down breaks. Every run starts 15 minutes from the reporting time in case one needs to meet with a supervisor to explain any service diversions. Most two trip lines have runs in the 65-90 minute range. Account for breaks and lunch, and there isn’t much time left over to fit in an extra round trip without breaching overtime.
Berlin doesn’t have lunch and breaks?
As far as revisiting schedules go; does Berlin revise schedules for summer holiday?
At least in New York a good chunk of the quarterly ridership changes are related to school being in or out of session, since NYCDOE has long given up any pretense of trying to serve all million plus schoolkids with dedicated bus services.
High off-peak frequency would address rush from end of the school day in schools close to subway station.
The rush from schools far away from the subway is much difficult problem.
In Girona, Spain city buses run according to fixed schedule at at various times buses are nearly empty. In USA some automobile driver might see this and complain of waste. But, reality is that during peak usage loaded buses more than make up for fuel efficiency relative to same passengers opting to drive car alone. Girona also has various size buses, with largest buses being regional comfort seat type, connecting outlying cities with Girona’s HSR station. But, Girona also has smaller bus vehicles that route into confined medieval Barri Vell and to mountainous neighborhoods of Montjuic and Sant Daniel.
While revising train schedule every 3 months is quite inefficient, as explained in the article, would frequent schedule revision work for the city buses?
No, for the same problem with staffing – if anything, crew is a much larger share of bus operating expenses, so this gets worse. Bus drivers need to be able to arrange their lives around a consistent work location. If they report to (say) East New York every work day, they can choose where to live to avoid too much commute pain. If they have to bounce between different depots until they accrue enough seniority, they can’t. I think a valuable labor reform, useful for all workers but especially junior ones, is to ensure that crew work out of a consistent rail or bus depot, and workers can request a transfer (e.g. if they move) but will not involuntarily be sent to operate out of another.
(And also, I’m informed by A320LGA that it’s not every 3 months, it’s very 6 months.)
*is to ensure that crew work out of a consistent rail or bus depot, and workers can request a transfer*
That assumes that you’re able to secure work from a reporting site that’s convenient to your home. If one lives in Coney Island, but the only open spots are in the Bronx, while the consistency is better, it still doesn’t change the issues that come with a long commute. There have been times where at inductions, there were an equal number of openings in both divisions of the subway, while other times, everybody was forced into a division regardless of one’s residence. They weren’t able to switch out for roughly three to five years.
Then don’t apply.
When NYC first decides to be sensible about assigning people to one start location forever they should offer moving expenses to someone who needs to move to get to a reasonable commute to work. It is an internal transfer, the normal thing to do is offer such transfers to anyone every time change where they work from by an unreasonable commute.
NYC should hire/assign people to a depot. If they want to adjust schedules within one depot that is fine, but people should keep the same start location, or be compensated for the move.
As a supplement to what I wrote above, in effect, the split between the IRT and IND/BMT creates some half-assed attempt at balance, where the Bronx has most of the IRT reports and Brooklyn and Queens the balance of the IND/BMT reports, but if the jobs at the “opposite” end of the world need to be filled, somebody will have to fill them, and sadly that falls to the group of people with the least options.
Admittedly, the IRT reporting sites aren’t too far from each other, but the IND/BMT can have weird swings where you have far outlying terminals like Far Rockaway and Jamaica Centre grouped together with Canarsie for crewing purposes.
Do you think all of the lines’ terminals have the turnback capacity for the service increase? Looking at the vanshnookenragen track map, the switches at South Ferry and Van Cortland Park terminals look like they might be chokepoints. Of course, changing switches probably shouldn’t cost much, even at New York prices.
Would you run the G train every three minutes because of its dependence on connections with other lines, or use lower frequency for it because of its low ridership and lack of branching?
I’d run the G every 3 minutes. Its ridership and usefulness are very high compared to buses. Any riders it can take off buses or the Manhattan subways, with the help of higher frequency, are a significant gain.
I’m talking about raising off-peak frequency. At the peak some of the terminals handle that frequency today already (even difficult ones like Flatbush), and the few that don’t (like Pelham Bay Park) make it work with mid-line turnbacks.
New South Ferry I believe has a stated max capacity of 24TPH.
The only super problematic terminal I can think of at 20TPH might be Astoria-Ditmars, and Jamaica Center. And optimizations to the terminals have never been super expensive even at MTA costs; there were earlier plans to extend from Flatbush 1000 ft that would probably cost in the low hundred milliions today.
The J/Z are never going to be 20 an hour, not if there is going to be an M train. The E train is never going to be 20 an hour either, not if it’s sharing tracks with the F train. And Astoria is never going to get 20 an hour, not if Broadway trains are going to be on Queens Blvd.