Off-Peak Public Transport Usage

Earlier this year, I slowly stumbled across something that I don’t think is well-known in comparative public transportation: European cities have much higher public transport ridership than someone experienced with American patterns would guess from their modal splits. From another direction, Europe has much lower mode share than one would guess from ridership. The key here is that the mode share I’m comparing is for work trips, and overall ridership includes all trip purposes. This strongly suggests that non-work public transportation usage is much higher in European than in American cities even when the usage level for work trips is comparable. Moreover, the reason ought to be better off-peak service in Europe, rather than other factors like land use or culture, since the comparison holds for New York and not only for truly auto-oriented American cities.

Modal shares and ridership levels

My previous post brings up statistics for work trip mode share in England and France. For the purposes of this post, I am going to ignore England and focus on France and wherever I can find data out of Germany and Austria; the reason is that in the secondary cities of England, public transport is dominated by buses, which are hard to find any ridership data for, let alone data that doesn’t have severe double-counting artifacts for transfer passengers. For the same reason, I am not going to look at Canada – too many transfer artifacts.

In contrast, French and German-speaking metro areas with rail-dominated public transport make it relatively convenient to count rail trips per capita, as do the more rail-oriented American metro areas, namely Boston, New York, and Washington. A secondary check involving both bus and rail can be obtained from The Transport Politic, comparing the US with France.

City Population Definition Trips/year Trips/person Mode share
Boston 4,900,000 Subway, commuter rail 204,000,000 42 12%
New York 20,000,000 Subway, PATH, LIRR, MN, NJT Rail 2,050,000,000 103 31%
Washington 6,200,000 Metro, MARC (daily*280), VRE (daily*250) 245,000,000 40 12%
Vienna 3,700,000 U-Bahn, trams, S-Bahn (PDF-p. 44) 822,000,000 222 40%
Berlin 5,000,000 U-Bahn, trams, S-Bahn 1,238,000,000 248 35%
Hamburg 3,100,000 U-Bahn, S-Bahn 531,000,000 171 26%
Stuttgart 2,400,000 Stadtbahn, S-Bahn, Regionalbahn 223,000,000 93 26%
Lyon 2,300,000 Métro, trams, funiculars, 0.5*TER 325,000,000 141 20%
Marseille 1,800,000 Métro, trams (daily*280), 0.5*TER 139,000,000 77 16%
Toulouse 1,300,000 Métro, trams 125,700,000 97 13%
Bordeaux 1,200,000 Light rail 105,500,000 88 13%
Lille 1,200,000 Métro, trams 108,500,000 90 17%


Note that New York, with a 31% mode share, has not much more rail ridership per capita than French metro areas with mode shares in the teens, and is a quarter below Lyon, whose mode share is only 20%. This is not an artifact of transfers: just as the subway dominates ridership in New York, so does the metro dominate Lyon, Toulouse, and Lille, and so does the tram dominate Bordeaux. If anything, it’s Stuttgart, the only European city on this list with comparable ridership per unit of mode share to the US, that should have the most overcounting due to transfers.

Also note that French rail ridership nosedives in the summer, when people go on their 5-week vacations, and I presume that this equally happens in Germany and Austria. The ratio of annual to weekday ridership in France where it is available is fairly low, not because weekend ridership is weak, but because the weekday chosen to represent daily ridership is never in the summer vacation season.


Off-peak public transportation in the United States is quite bad. In New York, 10-minute frequency on most lettered routes is the norm. In Washington, the off-peak frequency is 12 minutes. In Boston, it varies by line; on the Red Line each branch is supposed to come every 12-13 minutes off-peak, but in practice trains don’t run reliably and often leave the terminal bunched, alternating between 3- and 10-minute gaps.

Moreover, commuter trains are so useless except for peak-hour commutes to city center that they might as well not exist. Hourly gaps and even worse are routine, and even the busiest New York commuter lines have at best half-hourly off-peak frequency. These lines are only about 15% of rail ridership in New York and Boston and 6% of rail ridership in Washington, but they contribute a decent volume of commuters who drive for all non-work purposes.

In Berlin, the off-peak frequency on the U-Bahn is a train every 5 minutes most of the day on weekdays. On Sundays it drops to a train every 8 minutes, and in the evening it drops to a train every 10 minutes far too early, leading to overcrowding on the first train after the cut in frequency around 9 pm. The S-Bahn trunks run frequently all day, but the branches in the suburbs only get 10-minute frequency, and the Ring has a 2-hour midday period with 10-minute gaps. The suburban areas with only S-Bahn service get comparable service to neighborhoods on New York subway branches, while closer-in areas get better service. No wonder people use it for more than just work – the train is useful for shopping and socializing at all hours of the day.


The people who manage public transportation in the United States do not have the same profile as most riders. They work traditional hours, that is 9 to 5 on weekdays only, at an office located in city center. Many senior managers do not use their own system. That NYCT President Andy Byford does not own a car or know how to drive and takes the subway and buses to events is unusual for such a senior person, and early media reports noted that some managers looked askance at his not driving.

Growing segments of the American middle class commute by public transportation. In Boston and Washington, transit commuters slightly outearn solo drivers, and in New York they do not but it is close. But those segments have different travel behavior from public-sector planners. For example, lawyers work long hours and depend on the subway at 8 or 9 pm, and programmers work shifted hours and both show up to and leave work hours after the traditional times. But public transportation agencies still work 9 to 5, and thus the middle-class transit-using behavior they are most familiar with is that of the denizen of the segregated suburb, who drives to all destinations but city center.

In such an environment, off-peak service is treated as a luxury. When there is a deficit, agencies cut there first, leading to frequency-ridership spirals in which lower frequency deters riders, justifying further cuts in service until little is left. In New York, there are guidelines for frequency that explicitly state it is to be adjusted based on ridership at the most crowded point of the route, without regard for whether cuts depress ridership further. There is a minimum acceptable frequency in New York, but it is set at 10 minutes on weekdays and 12 on weekends. For a similar reason, the planners tend to split buses between local and limited routes if each can support 10-12 minute headways, at which point the buses are not useful for short trips.

In contrast, in Germany and France, there is a mixture of drivers and public transportation users among managers. German planning stresses consistent schedules throughout the day, so the midday off-peak often gets the same frequency as the peak. French planning does vary frequency, but maintains a higher base frequency even late into the night. The Paris Métro runs every 5 to 7 minutes at 11 pm. The idea of running a big city metro line every 12 minutes is unthinkable.


  1. Robert Jackel

    Even when working Traditional Hours, anyone with a job that has them meeting clients on-site needs good off-peak service. I have to meet clients in Narberth or Conshohocken in the Philly suburbs, and half-hourly (at best) service means I can take the train, I just either cut off meetings early or my 35-40 minute meeting eats my entire day.

  2. Harald

    Can you say how you got to your Stuttgart trip numbers? Based on the source, I see total annual trips including all modes (347 million) and then daily (Mo-Fr) trips by mode. Not sure how this adds up to 223 million?

      • Harald

        Ah, I see, (374/(1.283*365))*(1283-518)*365 = 223 Makes sense.
        Can you just subtract the daily bus trips from the total daily trips, as the total counts trips with more than one mode only once (fn 5 in the report)?

        Another note: From what I can tell, the 26% mode share is only for the City of Stuttgart and only Mo-Fr, not the whole metro region covered by VVS. ( p.9, fig. 2)

  3. Ben Ross

    Don’t blame transit management. In my experience of the Washington area, these priorities are set at the political level and implemented by transit management. The people who set budgets, along with the business lobbies which heavily influence them, think of transit as a way to get cars off the road during rush hour. Transit management makes the best case they can to convince those people to give them more money.

    • Michael Whelan

      As a fellow Washingtonian, I agree in part and disagree in part with Ben Ross. WMATA’s recent cuts to the Metro’s hours of operation on nights and weekends came about because GM Wiedefeld proposed them because maintenance allegedly couldn’t be performed within the existing window (never mind the numerous other transit systems that seem to do maintenance just fine with longer operating hours). The cuts were endorsed by the board on a temporary basis, but since then Wiedefeld has repeatedly said that he wants to make them permanent. The Maryland, Virginia, and Federal board members don’t seem to mind, but the DC members are opposed to permanent cuts and Mayor Bowser raised a lot of fuss about it, precisely because DC is the part of the region where people use the Metro off-peak. I am not a huge fan of Mayor Bowser and I despise Jack Evans, but they seem to understand the value of off-peak service in a way that WMATA management does not.

      • Arlington Traveler (@ArlingtonTravel)

        @Michael Whelan, I think you are misinformed. As the one year Safetrack project due to a close in 2017, Wiedefeld actually proposed in addition to the service hour cuts a cutback in off peak headways. The WMATA Board decided against that directed the GM to prepare a budget with peak period service cuts instead. Now, with peak hour ridership back to pre Safetrack levels (2016 still well below the 2010 peak) rush hour trains feel very crowded. @Ben Ross is absolutely correct in that the big issue is political will. As with bus service operating more frequent off peak service is often not done because of lack of funding. The reason off peak service is cut first, is because it has the lowest ridership.Most “major” American cities have downtowns which have expensive parking which is why some people take transit. Off peak, most cities have both no traffic and either cheaper or available street parking, which is why most people drive. You have to eliminate that free/cheap parking to shift mode share.

        @ Michael Whalen if you are a Washingtonian you likely know about the Wharf development. The parking is crazy expensive:
        0-½ Hour: $5
        ½ – 1 Hour: $10
        1 – 2 Hours: $18
        2 – 3 Hours: $24
        3 – 4 Hours: $29
        More Than 4 Hours: $38

        Weekends and Special Events
        0 – ½ Hour: $5
        ½ – 1 Hour: $10
        1 – 2 Hours: $18
        2 – 3 Hours: $25
        3 – 4 Hours: $35
        More Than 4 Hours: $45

        Guess what very few people drive alone. If all of downtown Washington DC had parking prices like this, then yeah you could have your better off peak service.

        • Arlington Traveler

          Just to clarify, service hours were cut but off peak headways were not touched in the 2017 service cuts. Finally, some cuts in service hours were needed to catch up on maintenance which was deferred for decades. Wiedefeld has proposed restoring some service hours, but at the price of more late night single tracking. Guess what ridership, from 10pm to midnight is higher than 1am to 2am on weekend nights, so it makes no sense to disrupt service when ridership is higher to provide an extra hour of service when ridership even before it was cut, was anemic at best.

  4. Max Wyss

    It is interesting to hear that off-peak service is considered “luxury”, because from the point of view operator, off-peak service is cheap. It is the peak service which is expensive… you have to have vehicles (and staff) available for peak; off-peak, you do not need extra vehicles nor staff (well, maybe some).

    That said, the fixed cost for off-peak service is marginal.

    Another thing is that quality of life means to not depend on an automobile all the time; driving is a question of choice, and not the necessity.

    • rossbleakney

      Yeah, which just shows how ridiculous the approach taken by U. S. transit agencies is. They are focused almost entirely on 9 to 5 workers, even though they are the most expensive to serve. Everything else is considered “for the poor”, which means that basic service (e. g. bus that comes very ten minutes) is considered excessive.

  5. rossbleakney

    I would guess that the biggest difference is in the suburbs. New York City itself is about 8 million people, and that includes various parts of the city that are “out there”. Staten Island, for example, is essentially a suburb. For those places (as well as much of New Jersey, Long Island, etc.) my guess is that transit is almost entirely commute base, or other trips “into the city”. It is a real pain to try and get around with a car inside New York. Even if it is slow or infrequent, it makes sense to use the subway, or buses (or boats, for that matter). But outside the core of the city (and a handful of suburbs), transit is poor, and most everyone owns a car.

    What is true for New York is true for other cities. Washington D. C. itself is only about 10% of the Metro area; San Fransisco is only about 20%. Extend that urban core to include a few (high density) suburbs and you probably find a fair amount of people who take transit on a regular basis, outside of commuting. Yet go a few miles outside of town (where most of the people live) and you will find very few who do. In a handful of other American cities this is the case, but most American cities don’t get to that point. If you ask someone about getting from point A to point B in Seattle, almost everyone will assume you are driving. My guess is that is the biggest difference — it isn’t that the other countries don’t have suburbs, or that ridership doesn’t go down as you go out — it is just that the areas where driving is considered the default mode of transit make up a much bigger part of the city.

    • Lee Ratner

      I don’t know. The LIRR receives relatively heavy use on the weekends because parking in Manhattan can be a big pain and Manhattan is more central to entertainment in the area then other anchor cities would be. You need to go into Manhattan to see a Broady musical, go the museums, etc. Lots of people and/or families find a train trip less stressful than driving in. If the trains ran more frequently and were cheaper, I’d guess more people would venture into Manhattan by rail.

      • rossbleakney

        Yeah, that is why I wrote “other trips ‘into the city'”. I didn’t emphasize it enough, but I meant trips into Manhattan, of the types you describe. The point is, if you live on Long Island, New Jersey, Staten Island or a lot of places in the greater New York area, and want to visit Manhattan, you probably take transit. It doesn’t matter what the reason — my guess is a very high percentage of people take transit for that sort of trip. But there are lots of other trips — such as a trip entirely within one of those areas, or a trip between those areas. Those are trips that are likely to be done via a car.

        • Lee Ratner

          When the LIRR, Metro North, and NJ transit lines were built, it was assumed all traffic would be to and from Manhattan. They aren’t really designed for going between towns.

          • Alon Levy

            It’s even worse than this… these lines were built in the mid-19c as intercity lines, but then they have undergone extensive rebuilding and redesign in the last 110 years to be commuter lines. Some of this rebuilding was great, e.g. when the entirety of the LIRR and nearly all of Metro-North got high platforms; some really wasn’t, and this continues to today – the East Side Access-related track rebuilds (e.g. around Jamaica) are not designed with timetable-infrastructure integration in mind.

          • adirondacker12800

            Integrated with what? At peak there will be a train 20 times an hour to Grand Central and a train 30 times an hour to Penn Station.

          • Alon Levy

            Yeah, that’s not integration, that’s throughput. Integration is “30-minute takt to each of GCT and Penn on each branch,” with schedules printed to the minute years in advance.

          • adirondacker12800

            Like they have now on some branches? You have to decide what is a “branch”.
            They are well trained to “change at Jamaica” or wherever. It’s unclear what they will do in 2038 when East Side Access eventually opens. S’kay if it involves changing at Jamaica or Woodside. They already know how to do it. the especially wily ones will get off a train, in Jamaica, on the outside of the Spanish solution platform, walk through the train on the Spanish solution track and across the other platform to their trains. Very wily and moderately skilled at the concepts involved. Doing it is faster than driving or taking the subway, they’ll do it.

          • Alon Levy

            Yeah, “it is unclear” is how the US rolls, but in Switzerland the proposed schedules are known to the minute years in advance, and of course they’re frequent off-peak too because trains aren’t just about 9-to-5 office workers.

          • adirondacker12800

            It’s every 30 minutes give or take a minute or two 16 hours a day on the Lynbrook to Babylon section. You aren’t proposing they run ever half hour at 3AM are you? I’m sorry your mommy isn’t with you to look at a schedule but it’s every half hour or more frequently during the peaks. I don’t really want to know what the schedule is going to be in 2023 because events in 2023 haven’t been thought of, much less planned in 2019. I’ll look at a schedule when I decide if I want to go. Even if I looked at it now, stuff happens and the schedule may change. And when I get there, since my mommy isn’t with me, I’ll check the departure board to see if there has been any last minute change and if it’s on it’s usual platform. Stuff happens, it may not be.

          • Mike

            The level of schedule integration that Alon is describing relates to not minimizing travel time but to minimize time of transfers. Swiss Federal Railways coordinates their mainline, international, and most regional trains to clock-face scheduling so they’re at main transfer stations at the top of the hour (and oftentimes at the bottom of the hour as well). It simplifies the schedule because most trains run every hour from 5am-1am and fill-in trains have a consistent scheduling pattern to match (e.g. every 1/2 hour, 15 mins). The Swiss then know that most trains will keep the same schedule for years because the trains stop at the stops at X minutes past the hour every time, as everything’s coordinated.

            There isn’t this weird garbage of “peak is every 8/10/12/15 minutes, dropping to 15/20/30/60/non-existent depending on the day and time of day” so often seen in US transit schedules. What good is a fast line if I have to wait 43 minutes for my transfer from LIRR to Amtrak as my fastest trip? (The Swiss are a little more lenient on S-Bahns and local transport matching those schedules since it’s not assigned seating and run frequently as well – but they’ll try and some trains will even wait a minute or two for a late S-Bahn.)

          • Lee Ratner

            I grew up near a station on the Port Washington branch of the LIRR. There was no way for me to travel by train from my hometown to anywhere else not on that line without having to head west into Manhattan first. As far as I can tell, this was true for the entire history of the LIRR even before car culture became a thing. Long Island could have a dense rail network that allows for transport without heading into Manhattan but choose not to do it.

          • adirondacker12800

            How many tracks do you need between Manhattan and Woodside to have 15 trains arrive and depart at the stroke of the hour? Then sit there and do nothing for 25 minutes or so until another 15 arrive and depart at half past?

          • Nilo

            Port Washington Branch is specifically hobbled and handicapped by rich magnates who own estates on the North Shore of Nassau if I remember Caro’s Powerbroker right.

          • Richard Mlynarik

            How many tracks do you need between Manhattan and Woodside to have 15 trains arrive and depart at the stroke of the hour?

            Throttling this infant would significantly raise the quality of pedestrian observations.

          • Alon Levy

            You are not nerd-sniping me into writing down timetables for timed transfers in and around New York. Especially since the traffic density if Amtrak goes from averaging 90-100 km/h and charging $0.30-0.50/p-km to averaging 200-250 km/h and charging $0.10-0.15/p-km will be high enough that nobody needs to time transfers, on the model of the intercity lines that run every 10 minutes in the Netherlands.

            (The trick for transfer timing is to time most of the NJT transfers at Newark and the New Haven Line ones at New Rochelle or Stamford, so on the NJT side you just have M&E. On the LIRR side you have three-way timed transfers at Jamaica, using infrastructure that exists but that’s about to be ripped out for ESA, and then you just need to time two branches, including PW, with Amtrak.)

          • adirondacker12800

            Death threats are so mature. And I’m sure telecommunications companies find them intriguing.

  6. rossbleakney

    OK, looking at this some more, I think leaving out bus service gives misleading numbers. If you include bus service, then New York is at least above Lyon (and every other French City) even though it lags Paris. The numbers for D. C. and Boston would go up substantially as well, although again, they would be lagging their European counterparts.

    • Alon Levy

      The problem is that in all three cities, buses act as rail feeders. A majority of bus ridership in New York, and a large majority of same in Boston, connects to rail. There’s only one big segment of bus-only commuters in New York, in the northwestern suburbs where the Erie lines don’t enter Manhattan.

      • Luke

        ….I mean, ideally, this is how it would work, right? Lower capacity modes in less dense areas into higher capacity modes in denser areas? The bigger issue seems to be dysfunctional organizational and operational practices on the part of US transit agencies, and poor connectivity among and within modes, generally.

      • rossbleakney

        The problem is that in all three cities, buses act as rail feeders.

        Do you have some data to support that claim?

        Just because a bus crosses a subway station doesn’t mean that riders are using it for that purpose. When I look at the New York Subway system, it looks very much much spoke and hub. That is all good and well — and perhaps the best they can do for the amount of rail they have — but it still leaves out a boatload of trips. East-west in Manhattan, for example, for almost the entire length of the (very densely populated) island. Or how about north-south in Brooklyn, or damn near all of Queens. In a city where lots of people don’t own cars (and those that do reluctantly use them) buses are an important way of getting around. When it comes to off peak trips, this is especially important. These are not people headed to work — but people visiting friends, going out to a movie, or a club they heard about. Not all of these destinations are next to a subway station, and even if they are, it is often faster to take the bus.

  7. Gok

    I am not at all convinced 10 minute headways are that big of a deal if there’s no crowding. That’s the peak rate of smaller cities like Oslo and Stockholm which have high transit use. A few more minutes waiting at the station won’t prevent show-up-and-go transit behavior.

    The ridiculous hour+ headways on commuter lines is obviously catastrophic though.

    • Alon Levy

      Stockholm has 10-minute weekend and summer headways on branches, but the midday off-peak headways are at worst like 7 minutes on the Red Line branches (not sure about Blue or Green Line branches).

    • Matthew Hutton

      10 minute headway’s are ok if you’re going to the pub or club and then going home after. But if you’re doing 2-3 chores they are a lot worse than driving.

  8. Andrew in Ezo

    Alon, you may be interested in this data from the Japan MLIT, showing modal share in various Japanese cities, both within the urban agglomerations and the regional ones, on weekdays and weekends/holidays. It is based on surveys from 500 households in each designated city. The data is a bit old, from 2010, but should be fairly representative of current conditions. Note that each mode percentage is the “primary mode”, meaning that, for example in the major urban cities, a single trip designated as (majority) “rail” would likely also include portion(s) that used walking, a bus ride, or bicycling, but the rail portion had the greatest distance covered. The modes reading left to right are railway, bus, automobile, two wheeled motorized vehicles, bicycle, and walking/other. The cities are arranged from top to bottom starting with the northern cities and going south.

  9. Pingback: Rapid transit for Brussels – Métro vélo dodo
  10. Pingback: Mixing and Matching | Pedestrian Observations
  11. Pingback: The French Way of Building Rapid Transit | Pedestrian Observations
  12. Pingback: Regional Rail for Non-Work Trips | Pedestrian Observations

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.