Intercity Rail Frequency and the Perils of Market Segmentation

SNCF loves market segmentation. Run by airline execs, the company loves to create different trains for different classes of people. Not only do individual trains have opaque pricing run on the basis of yield management, in which similar seats on the same train at the same time of day and day of week may have different fares, but also there are separately-branded trains for separate fare classes, the higher-fare InOui and the lower-fare OuiGo. On international trains, SNCF takes it to the limit and thus Eurostar and Thalys charge premium fares (both about twice as high as domestic TGVs per passenger-km) and don’t through-ticket with domestic TGVs. This has gotten so bad that in Belgium, some advocates have proposed a lower-priced service on the legacy Paris-Brussels line, which would have to be subsidized owing to the high cost of low-speed intercity rail service.

But why is market segmentation on rail so bad? The answer has to do with frequency and cost structures that differ from those of airlines. Both ensure that the deadweight loss from market segmentation exceed any gains that could be made from extracting consumer surplus.

The issue of frequency

A segmented market like that of domestic TGVs reduces frequency on each segment. To maintain segmentation, SNCF has to make the segments as difficult to substitute for each other as possible. OuiGo serves Marne-la-Vallée instead of Gare de Lyon and forcing passengers onto a 20-minute RER connection, or even longer if they’re arriving in Paris and the wave of 1,000 TGV riders creates long lines at the ticketing machines; on other LGVs it serves the traditional Parisian station and thus the segments are more substitutable.

The situation of Eurostar and Thalys reduces frequency as well: the high fares discourage ridership and send much of it to intercity buses or suppress travel. Fewer riders, or fewer riders per segment as in the case of domestic TGVs, lead to fewer trains. What’s the impact of this on ridership?

The literature on high-speed rail ridership elasticities has some frequency estimates. In Couto’s thesis (PDF-p. 225), it is stated that passenger rail ridership has an elasticity of 0.53 with respective to overall service provision. There are also multiple papers estimating the elasticity with respect to travel time: in Cascetta-Coppola the elasticity ranges from -1.6 to -2.2, in Börjesson it is -1.12, and in a Civity report it is stated based on other work that it is -0.8 to -2. The lowest values in Börjesson are associated with the premium-fare AVE, while the range for the original TGV, priced at the same level as the slower trains it replaced, is -1.3 to -1.6. The upshot is that halving frequency through market segmentation reduces ridership by a factor of 2^0.53 = 1.44, which is far more than the benefit yield management is claimed to have, which is a 4% increase in revenue per SNCF’s American proposals from 2009.

Why are trains different?

Planes and buses happily use yield management. High-speed trains do not, except for those run by SNCF or RENFE – and ridership in France isn’t really higher than in fixed-fare Northern Europe or East Asia while ridership in Spain is much lower. Why the difference?

The reason has to do with the ratio of waiting time to trip time. Thalys connects Paris and Brussels in 1.5 hours, every half hour at rush hour and every 2 hours midday. At rush hour, frequency is sort of noticeable; off-peak, it dominates travel time. This is nothing like planes – even short-distance trips involve hours of access, waiting, and egress time, and therefore trips are not usually spontaneous, and day trips are rare except for business travelers.

Buses, finally, are so small that a market like New York-Philadelphia supports multiple competitors each running frequently, and passenger behavior is such that different companies are substitutable, so that the effective frequency is multiple buses per hour.

Cost structure and bad incentives

It’s typical to price high-speed rail higher than legacy rail, even when otherwise there is no yield management. This is bad practice. The operating costs of high-speed rail are lower than those of slow trains. The crew is paid per hour; electricity costs are in theory higher at higher speed but in practice greenfield high-speed lines are constant 300 km/h cruises whereas legacy lines have many acceleration and deceleration cycles; high-speed trainsets cost much more than conventional ones (by a factor of about 2 in Europe) but also depreciate by the hour and not by the km and therefore are somewhat cheaper per seat-km.

This is comparable to the bad practice, common in the United States and in developing and newly-industrialized countries, of pricing urban rail higher than a bus. The metro is nicer for consumers than a bus, but it also has far lower operating costs and therefore a wise transit agency will avoid incentivizing passengers to take buses and instead use integrated fares. The same is true for slow and fast trains: the solution proposed by the Belgian advocates is to incentivize passengers to take a high-cost, low-price train over a low-cost, high-price one, and therefore is no solution at all.

Moreover, the cost structure of trains is different from that of planes. Planes don’t pay much for fixed infrastructure; in effect, every plane trip costs money, and then the challenge is to fill all the seats. High-speed railways instead pay a lot for infrastructure, while their above-the-rails costs are a few cents per passenger-km (€0.06/seat-km on the TGV, including trainset costs and a lot of labor inefficiency). Their challenge is how to fill the tracks with trains, not how to fill the trains with passengers. This is why the fixed clockface frequency common in Germany, Switzerland, Austria, and the Netherlands is so powerful: the off-peak trains are less full, but that’s fine, as the marginal operating cost of an off-peak train is low.

Just lower the fares

Bear in mind that frequency is not exogenous – it is set based on demand. This means that anything that affects ridership has its impact magnified by the frequency-ridership spiral. An exogenous shock, such as improvement in trip time or fare reduction, is magnified through the spiral, by a factor of 1/(1-0.53) = 2.13. In other words, every elasticity estimated in isolation must be multiplied by a factor of about 2.

And once this is understood, suddenly the optima for service look very different from what Thalys has settled on. The optimum is to charge fares to pay infrastructure costs but not much more – especially if you’re SNCF and the railway workers’ union will extract all further profit through strikes, as it did 10 years ago. And this means making sure that except at very busy times, known in advance, Paris-Brussels tickets should be 30€, not 50-100€.


  1. Matthew Hutton

    Perhaps better market segmentation is to have more classes onboard the train?

    • Henry Miller

      Maybe – will enough people pay the extra price? How many will not pay and be mad because they don’t get whatever the higher price fare does? For trains it is cheap to add more cars to a train (up to platform length limits), and cheap to add more trains (up to track capacity). If you are anywhere close to track capacity you should be rolling in the cash that you invest in some new line to elsewhere thus taking the load off this line.

      I think most often you are best off advertising “first class legroom at no extra cost!” Which is to say point out how airlines have incentive to reduce legroom, but since your economics mean add more/longer trains as they fill up (if capacity allows) is the best response you shouldn’t be trying to charge more for space.

      Even if you are slightly more expensive than flying, you are still a lot cheaper than flying first class. This is something to push.

      • Phake Nick

        Premium seat service in many railways across Tokyo commuter railway network seems to be really popular among customers.
        I don’t think customers being unhappy about they aren’t getting what others who pay more do is a common concern, and if such sentiment is shared to others then most people will probably see such sentiment as merely the complainer being envious.

        As for close to capacity mean rolling in cash, not necessarily. Especially for markets where demand are significantly concentrated around commuting hours and have very weak mid-day demands.

    • Sascha Claus

      At some point you have to decide how much room to allocate to each class, and swapping out 1st-class cars for 2nd-class ones midday or even enroute is difficult, as well as reordering seats within the cars. And passengers won’t like it to pay 1.3rd Class Premium Special Executive class and then getting only a 1.3rd Class Premium Economy seat because the first one is full. So you want to be careful and not add too many classes.

      • adirondacker12800

        Nah, you tell the frequent travelers they’ve been upgraded and here is you new seat reservation. …. I’ve been upgraded because the flight is overbooked. The script is that they are deeeeply disappointed that coach is quite full and can I be convinced to change my seat to first class. It’s a terrible disappointment but I tolerate it. Nowadays most people have a smartphone welded to their hand, it doesn’t even have to involve staff.

        • Sascha Claus

          A new seat reservation? If I didn’t have an old one? And how does the system know who has their smartphone glued to the hand and would notice the last-minute seat-change, and who has the smartphone tucked away already? And when a car is full after everybody found a seat, how do we get that knowledge back in time to before boarding?

          • adirondacker12800

            If you didn’t give them an address to send text messages to that’s your problem. Take your paper ticket to the very nice customer service agent at the station and see what can be done.

  2. Sassy

    This post obviously has a European focus, but in East Asia the norm of charging more for high speed rail seems to be working (I think?). And specific for Japan, there are higher fares for limited express low speed trains, and for Nozomi vs Hikari/Kodama.

    Also, just from the perspective of running a train, faster trains are cheaper to run, however I’m pretty sure it flips when you consider the infrastructure. A faster train requires more tracks. If every train was local, then you could put way more trains through the same tracks, therefore, the loss of capacity that occurs because you ran a train that skips some stops is really part of the cost of running that train.

    I guess if capacity just isn’t an issue, the capacity argument doesn’t make sense. However, there’s certainly parts of the European rail network facing capacity issues. Sure some of them could be solved with better operations and not more tracks, but the pricing of current fares should still probably reflect the current capacity issue, right?

    In addition, low speed trains in the middle of the day often have to run at relatively low passenger loads, since some minimum frequency is required to have a low speed rail network that is useful for people making shorter trips. Shouldn’t someone willing to put up with using trains intended for short trips, on a long trip, and basically filling an empty seat on a train that must run regardless, be charged less? Obviously not relevant for adding a newer slower cheaper intercity focused train, but definitely relevant for pricing low speed trains that already exist.

    The low speed that must run anyways case is probably even more relevant to Europe than East Asia. East Asia ex-Japan has generally poor legacy low speed suburban/regional rail networks, and Japanese suburban/regional trains are generally more metro-like and less suited for a cross country trip. However, taking an RE train on a trip that ICE is intended for, seems like both something non-students would do to save money, and something that is legitimately filling otherwise empty seats that must run anyways.

    Is the argument really specific to routes comparable to Paris-Brussels which is both low frequency, and doesn’t have some existing low speed service that must run anyways, instead of something more general?

    • Alon Levy

      Nozomi fares are barely higher than Hikari fares – the main purpose of the differentiation is to keep out the tourists on JR Pass.

      The infrastructure costs are higher on a new high-speed line than on a fully amortized legacy line, but once you’ve built an LGV, it’s there and there’s no point in keeping trains off of it. It only gets to be a problem once you hit capacity, but the LGV Nord is operating massively below capacity because the high fares on Thalys and Eurostar discourage ridership and Lille is not large enough to fill trains to Paris by itself. On the LGV Sud-Est, sure, there’s a crunch, but the market segmentation is between two types of train that both use the congested tracks and only reverse-branch as they get to Paris, where there is room on the tracks (and then the non-substitutability of InOui and OuiGo incrementally reduces peak capacity).

      • Sassy

        Obviously what SNCF/RENFE are doing is bad, but what about R/RB/RE vs IC/ICE pricing?

        There’s no point keeping trains off a high speed line, but there’s other, more sensible uses for market segmentation. Even for planes, market segmentation is often a natural result of dealing with capacity issues. (U)LCCs legitimately save money by choosing to not fly out of popular capacity constrained airports like LHR.

      • Eric2

        There is a price which achieves maximum revenue, and it’s a compromise between attracting the most passengers and charging the most money per passenger. (Kind of like a Laffer curve controlled by the monopolist train operator rather than the government.) Why assume that Thalys et al are far from this price?

        Admittedly, the best price for train operators and builders is not the same as the best price for social good – that is another compromise that needs to be struck.

    • Phake Nick

      Shinkansen could have charged two third its current price and still make a profit, which would significantly increase its competitiveness on longer routes by having ticket price below LCC air fare, and making the fare closer to highway buses, however JRs cannot do so because other than the profit of the companies themselves, the revenue from high speed rail also need to be spent on subsidizing slow local services.

      Then if you mean the extra infrastructure cost of building extra passing tracks at station thus that some high speed trains can skip some stations and overtake slower trains, I believe such cost is neglectibly small for the entire system.

      And the “low speed train” mentioned by the blogpost isn’t trains that run at same max speed in km/h but merely serve more stops, rather, it mean conventional express trains on legacy rail line compared to high speed trains on dedicated lines.

  3. Martin

    I think the logic that you should never have market segmentation based on either routes or frequency is solid, to ration high quality services, for higher market prices. One should run frequent trains as allowed by infrastructure all over the day for the approximate reasons highlighted above.

    However, once you do that I think, price will be the only reasonable way to incentivise people to use the train all over the day, and that tickets don’t sell out in advance. Simple demand and supply pricing seems like both a fair and efficient way to do that, requiring very little planning by the train company. Fixed low prices att attractive times us just an invention for market arbitrage, and massively favour those buying tickets months in advance (even though they probably have very weak time preferences, given how early they plan the trip).

    I think this is also exactly the output you will get from a proper travel planning model, that includes both infrastructure costs, and non-market time gains, on the benefit side.

    • Eric2

      We are all agreeing here.

      Varying price by route or train or operator – bad.

      Varying price by time of day in the case of peaks that strain capacity – good

      • Martin

        I agree very much, but not the opening paragraph I would say 🙂

        Of course you can have pre-set prices, where the operator tries to guess demand which would be realtivly similar, but just let it vary by market prices (which meaemns the same seat will have different price, based on when it was brought), seems a more rational way of achieving the same goal. In practice, preset prices will probably always means either lots of empty seats, or sold out trains well in advance.

        • adirondacker12800

          They don’t have to guess. The airlines computerized their ticketing more than 50 years ago. The statisticians and computer programmers have been poking at that since. Or other tickets, rental cars, hotel rooms….. They may pick the wrong forecast but they don’t have to guess.

          • adirondacker12800

            A lot of the numbers that pop out, when extracting information from previous data aren’t, assumptions. It’s fourth grade arithmetic.

          • Tom M

            Doesn’t matter how simple or complex the math is, any prediction of the future contains uncertainty and has elements of “guessing” in it.

          • adirondacker12800

            It’s the pesky pesky people. Someone in management decides at which point it costs too much to service a very small fraction of potential business. When three waves of unpredictable synchronize, that’s just too bad and people are disappointed. They have statistics on how often two waves synchronize and three waves and four waves. No one has capacity to deal with the four wave synchrony. Someone somewhere can tell you it happens, on average every 1,238 days and stuff happens, who knows if the next one will be in 1,141 days or 1,309. And that’s just too bad because it costs too much money to have stuff or staff doing nothing for 1,237 days.
            They aren’t guessing, they decided that it costs to much to service the unpredictable, that is really bad, on average, every 1,238 days. Stuff happens. And, on average, every 102 days the forecast was bad because there was one wave of unpredictable in the morning, two different kinds of odd in the afternoon. Who knows what they were. Stuff happens.

          • Tom M

            To summarise your comment, yes, every forecast has a level of uncertainty.

          • adirondacker12800

            Nobody is guessing at things either. They are pouring a wealth of data, collected from ticket sales, into computers and predicting likely outcomes.

          • Tom M

            Yes, exactly, they are using data to create assumptions to produce forecasts. Hence my original point. In very simplistic and broad terms, this is guessing as per Martin’s phrasing.

          • adirondacker12800

            I’ve done this, using a comptometer, back in the 70s, it’s not guessing.

          • Tom M

            Doesn’t matter if you’re using marbles or the HPE Cray EX235a, you are still forecasting based on assumptions derived from a statistical sample that contains further assumptions and is only a sample, never the full data population. Sure, you’re not plucking a number out of thin air, but you are still creating a prediction containing uncertainty.

          • adirondacker12800

            Nothing is certain. It’s not guessing either. You can go off and consider the difference between charging 38.422935637518 more and 38.422935637517. Someone in management it going to decide to call it 40 percent. Ish. Because they are going to be pricing tickets in whole dollars not fractions of mil.

        • Eric2

          I think preset prices (e.g. 50% surcharge between 7 and 9 AM, actual percentage to be worked out over the course of years by trial and error) would achieve 80% of the benefit at 20% of the inconvenience, and thus are preferred.

          • adirondacker12800

            Then they can’t whack the business traveler who decides in the late afternoon that a trip has to be taken tomorrow morning.

          • Alon Levy

            Yeah, exactly. Intercity rail peak patterns are predictable by time of day, day of week, and time of year. The same predictability is why agencies know to reduce local rail frequency during less busy times. For examples, Stockholm reduces rail frequency to a train every 10 minutes on the T-bana branches during the summer travel season, when commuting traffic is weaker, and Paris does major reconstruction involving line shutdowns in August, when Parisians travel (cf. July, the peak travel season for tourists vacationing in Paris). SNCF also knows how this works and runs higher frequency during the weekend travel peak while taking trains out of service for maintenance during weekdays.

      • Matthew Hutton

        Also there’s no reason in Japan that a Shinkansen from Tokyo to Osaka couldn’t cost ¥10000 and one from Tokyo to Fukouka couldn’t cost ¥10500 or something if there’s more air competition for that.

        • Phake Nick

          You can buy early tickets online at such price according to my understanding but few people use them.
          And, there are also regulatory hurdle since the standard fare of Shinkansen is baked into national system and applied across all the different JRs which are now multiple separate companies, business and political negotiation between all the entities and government will be necessary to change that. And the profit from Shinkansen is also necessary to subsidize local trains, as JR Hokkaido is now in deep financial trouble, but the Japanese government have no plan to save it other than offering short term aids and tell the company they should improve their business performance themselves, and believe that the crisis will resolve by itself automatically once the Shinkansen to Sapporo start entering service.

          • Alon Levy

            The local trains on JR East and West are profitable. JR Hokkaido isn’t, but that’s not where JR East, Central, and West’s profits are going.

          • Phake Nick

            Only if you also take into account limited express trains or Tokyo/Osaka urban rail network, would JR East and West’s local train be profitable. As far as I know, for JR East’s local train network, outside Kanto area, there are zero lines not making losses, even before pandemic. And that is a reason why JR East and JR West are both trying to remove or offload rail infrastructure, because they expect coronavirus caused demand suppression to be at least partially permanent hence they will no longer have that much extra profit from Shinkansen and metropolis commuters to compensate for the losses on those local lines in local area.

          • Matthew Hutton

            Part of the train use being down is because Japan is almost unique in not letting foreign tourists in from outside the country.

          • Phake Nick

            The downward trend have been ongoing even before the pandemic, and are in line with population forecast/stat trend, especially outside major cities.

            And nowadays with the coronavirus further degrading demand, even peak time commute train in Tokyo have to see frequency cut. I doubt there were any significant number of tourists on these commuting trains even when Japan open to tourists.

            Not to mention, overseas tourists contribute to just ~10% of the Japanese tourism market, before this coronavirus happens. Which mean contribution by these passengers would be even less on general rail lines that also take non tourist passengers.

  4. Borners

    Alon, just asking the stupid question here. Are many rail operators fundamentally ignorant of their own operating cost structures and how they compare? Even within the same country in the same industry (UK mainlines vs Tfl, NY vs Chicago etc) let alone with other modes of transport?

  5. df1982

    Deutsche Bahn uses yield management on intercity trains. If I book Berlin-Munich for an out of peak ticket (e.g. early on a Wednesday morning) a few weeks in advance and select the Super Sparpreis ticket, I can get it for as low as 20-30 euros. If I try to book the same leg for a Friday afternoon the night before, my only option is a Flexi ticket for 140 euros. That’s yield management.

    I think their split between Nahverkehr and Fernverkehr makes the most sense. Anything within a daily commutable range has regional trains that are integrated into the local public transport ticketing system, which means zonal fares and no yield management (apart from certain offers like weekend tickets and the like). It’s run as a social service for essential local travel needs and so farebox is willingly subsidised.

    IC/ICE trains, however, are expected to run at a profit or at least cover their operating costs, so tickets are priced at what the market will bear. Yield management allows the more price sensitive customers to plan ahead and travel out of the peak and still get decent deals. Fares are unified so there are no transfer penalties for changing between IC/ICE and RE/RB/SBahn (intra-city transport on U-Bahns, trams or buses requires a small extra payment of a couple of euros).

    In general the division of labour is pretty clear: nobody will get an ICE from Berlin to Potsdam when there are regular S-Bahn and Regionalbahn trains serving that city-pair, just as very few would try to do Stuttgart-Hamburg just using regional trains. There is of course a grey area where ICEs and Regional Expresses tend to compete, which is around the 1.5-2h travel market (e.g. Frankfurt-Kassel) but this is a minor problem.

    • Alon Levy

      DB has advance fares, but it’s a far simpler and more transparent system than that of SNCF, and there’s also nothing like the InOui/OuiGo split or the 19035068.7 different kinds of high-speed train in Spain.

      The Nahverkehr-Fernverkehr split makes a lot of sense, yes, as you note. But that’s not the same as the system of overcharging on international trains and also not having through-ticketing. (When my Eurostar back home was severely delayed because its boarding is so slow it can’t cope with the summer crunch, there was an earlier train to Brussels that I could take to Lille and then transfer, but Eurostar said they could only put me on the train to Lille and I’d have to buy an extra domestic TGV ticket to complete the trip, never mind that Eurostar is majority-owned by SNCF.)

      • df1982

        Yeah, that was my point, they seem to reach a common-sense middle ground between the arcane ticketing systems of SNCF or RENFE (not to mention Britain), while still adopting high-yield on their intercity routes and maintaining subsidised travel on regional trains. There is actually a small difference in fares between ICE and IC (presumably justified in DB’s eyes by the more modern rolling stock) but most travellers wouldn’t even notice this, and it should probably be abolished.

        Your Eurostar problem is a product of international services being offered by bespoke companies running to their own rules, largely as the result of bilateral agreements in the days of national rail monopolies. I can sympathise, as I once had a ticket from Basel to Paris but the only place in Switzerland where I was allowed to print it off was Geneva, so I had to buy a brand new ticket and then get it refunded (this was before QR codes became universal). At least the Eurostar-Thalys merger seems to have gone through, but really they and Lyria should just be absorbed into SNCF’s standard TGV operations.

        • Matthew Hutton

          Eurostar is also harmed by passport control, which could be streamlined but which is politically unviable to drop.

      • Diego

        Wait, I was assuming that the Nah/Fernverkehr split was also detrimental. It’s not a small detail in the Germany given the size and population of the Ruhr and the Rhine valley. Why not allow people to travel from Essen to Düsseldorf or Köln on a fast train with no fare penalty? Of course charge extra for seat reservation. I can an issue with trains filling up but just for the Ruhr bit and then emptying out again. But then why not run Ruhr-only ICEs, if this is where demand is.

        On a personal level, what I hate about yield management and having a seat mandated to be on a specific train is that it effectively increases my travel time, especially in my return trip. In an unfamiliar city, I’m not sure how long it takes me to get to the station, I’m not sure where to find a decent place to eat before leaving and so I always end up at the station way too early. I’d like more anytime or anytime offpeak tickets on HSR (ofc seat reservation is worth it when there’s high demand).

        • Richard Mlynarik

          Wait, I was assuming that the Nah/Fernverkehr split was also detrimental. It’s not a small detail in the Germany given the size and population of the Ruhr and the Rhine valley. Why not allow people to travel from Essen to Düsseldorf or Köln on a fast train with no fare penalty? Of course charge extra for seat reservation. …

          This is the basic Swiss model, and it works just fine.
          A train is a train is a train. If you have a ticket valid between “A” and “B”, then it works on any service. Simple, efficient, not confusing, not a problem.

          Something else to consider: inter-regional services typically “steal” schedule slots from regional traffic. Either they replace a local potential train in the Takt slot, or the local Takt needs arrangement to work around long distance trains. So in some sense it is “only fair” to the locals that the “imposter” trains should have local benefit. A weak argument, sure, but since in practice it doesn’t degrade the inter-regionals, why not go with it? And sure, if some particular trains on some particular routes at some particular times experience problems — real problems, not imaginary ones involving hoi polloi mingling –, then perhaps make particular exceptions to the policy.

          Market segregation and airline-style anything (yield management, security theatre, point-to-point obsession, zero flexibility for schedule changes, etc, etc, …) are just bad ideas promoted by bad people. Defenders of this nonsense (and they’re all over the Anglosphere in particular, where people not only can’t imagine not being abused but masochistically actively support collective population punishment) just don’t seem to be able to even imagine a better world — one that already exists, if they only looked. “Somebody somewhere is getting away with something [riding a “nicer” train than they deserve] and they must be stopped!”

          • Alon Levy

            inter-regional services typically “steal” schedule slots from regional traffic

            It depends on whether they run on tracks that are shared with heavy regional traffic. In Berlin they do not – the S-Bahn runs on almost 100% dedicated track, and they’re quad-tracking the Dresdner Bahn to permit direct intercity rail service next to what war reparations to the USSR reduced to a two-track S-Bahn-only line.

          • Sascha Claus

            This is the basic Swiss model, and it works just fine.
            A train is a train is a train. If you have a ticket valid between “A” and “B”, then it works on any service. Simple, efficient, not confusing, not a problem.

            Note that Switzerland is slightly smaller than Germany; slightly smaller than Lower Saxony even. One fifth larger than Northrhine-Westphalia, where the Rhein-Ruhr region is located, so only international passengers as far as Frankfurt or Munich, because the country ends halfway before.

          • Oreg

            Caveat: the Swiss have no high-speed rail service, i.e., one service class less to differentiate.

        • df1982

          The Swiss also seem to have yield management, at least when I look up tickets for Zürich to Geneva a week from now, the fares range from ca. 20 francs to 44 francs depending on which train I select (still pretty good value for a trip that length), albeit these are all on ICs.

          I can see two arguments for the Nahverker/Fernverkehr split. One is that local transport is considered an essential service, and should be provided at a price point that is basically attainable for the whole population, which means subsidising farebox revenue, while long-distance intercity travel is discretionary and should at least aim to cover its operating costs through fares. If not then you end up subsidising a lot of business travellers and tourists who could afford to pay more and probably should do. High yield ticketing is a way to square this circle: offer discounted travel when the seats aren’t being taken up by anybody else, and market rate when the trains are likely to be packed.

          The other argument is that the “train is a train” mindset can lead to situations where intercity trains are swamped by local commuters at rush hour periods, which would blow out dwell times and lead to delays, as well as making for an unpleasant experience for intercity travellers. The ICE trains are certainly not set up for large numbers of commuters using it to go from Berlin Hauptbahnhof to Südkreuz, even though it stops at both stations. Even Diego mentions the potential need for laying extra Ruhr-only ICEs for local commuters, well if you have to do that, why not just make them REs and have local fares, while ensuring there is enough space on the ICEs for those heading to Frankfurt, Munich, etc.?

          I tend to think it works pretty well in Germany, and it’s not like the situation in the North-East US where you have multiple layers of uncoordinated fares (buses, subways, commuter trains, North-East Regional and Acela). If you have an ICE ticket and it misses your connection, you can get on a Regionalbahn instead (although not vice versa except in extreme circumstances). If you want or need the flexibility to catch whatever train you want you can get a Flexi ticket.

          And I can remember a time before Deutsche Bahn adopted yield management on its ICEs. Berlin-Hamburg was a flat €50 in 2005 euros. I basically never travelled intercity because I couldn’t afford it. You had to wait for special sales offers to come around a couple of times a year.

          • Richard Mlynarik

            The Swiss also seem to have yield management …
            … can lead to situations where intercity trains are swamped …

            I’m not disagreeing.

            Meanwhile, there’s an existence proof out there and it seems to work pretty damned well in practice.

          • Diego

            Hmm, I agree that rush hour commuters could be an issue. One compromise would be to only have the fare integration for the off-peak, charge a premium for the peak. Also works for high-travel holidays such as the one ending today.

          • Sassy

            > The other argument is that the “train is a train” mindset can lead to situations where intercity trains are swamped by local commuters at rush hour periods, which would blow out dwell times and lead to delays, as well as making for an unpleasant experience for intercity travellers.

            Could this be mitigated by better platform layouts?

            If you just put the intercity trains and local trains on consistently separate platforms, people will just show up and wait on the commuter train platform with show up and go frequencies because it’s easier. Sure some people go out of their way to show up on time for the intercity train and intercity platform, but most people surely prefer the convenience of not caring about the schedule.

            Another option could be to just not sell tickets for intercity trains between local station pairs. The price for the same trip could still be the same, but certain trips are just not offered for sale.

          • adirondacker12800

            People who show up and go are going to be too stupid to take the next train, that happens to be the intercity train, that gets there faster, and instead take a later train that takes longer to get there?

          • Krist van Besien

            The way the Swiss do it has its disadvantages too. Because domestic tickets are valid on the TGV Lyria the 7:34 departure is mostly filled with commuters going to Basel. This requires using a second TGV Duplex set, which gets dropped in Basel.
            That is an inefficient usage of an expensive TGV set, and is confusing for the passengers, as Paris bound passengers need to board in the first set only.
            Making reservations compulsory on long distance trains also avoids a common problem that DB suffers, where trains get delayed as a result of being overloaded. This has started to happen on the SBB network as well, on the Gotthard route. I think that Trenitalia has struck the right balance there, with reservations mandatory, but easy to get through multiple channels, and a pricing structure that is transparent.

          • Matthew Hutton

            Surely in this case you run a standard express train 3 minutes before the TGV and ideally retime the TGV so that it misses the Takt points.

          • Alon Levy

            Why would you want the TGV to miss the Takt points? You should want it to make the Takt at Basel SBB so that passengers can connect from other parts of Switzerland.

          • Matthew Hutton

            If the trains arrive for the takt at :25 and leave at :30 I’m suggesting the TGV from Zurich to Paris should arrive at Basel at :33 or something so you can get the TGV from the regional trains but not the other way around.

            Additionally perhaps Zurich to Basel should get a direct train rather than forcing people to change as even with the off timing the TGV has the advantage of being direct.

          • Richard Mlynarik

            The way the Swiss do it has its disadvantages too. …

            What? Some disadvantages too? Say it isn’t so! I want only advantages, always!


            And sure, if some particular trains on some particular routes at some particular times experience problems — real problems, not imaginary ones involving hoi polloi mingling –, then perhaps make particular exceptions to the policy.

            PS “expensive TGV set”? Just how expensive compared to, say, Swiss-domestic RABe 502? Just how many train-km/day compared to RABe 523?

          • adirondacker12800

            Why would a high frequency service have arrival times? The model is departure time, get your sorry ass on the platform a bit before that so you are ready to get on the train when the alighting passengers have alighted. It all happens in well under a minute.

          • Frederick

            That is an inefficient usage of an expensive TGV set, and is confusing for the passengers, as Paris bound passengers need to board in the first set only.

            Don’t worry about those petty misers of SNCF. They are charging 17 CHF for each commuter from Zurich to Basel. This is not inefficient at all; they are raking in Swiss money right there,

            But why did SBB let SNCF into the commuter business? First, you should know that the xx:34 timeslot is actually for IC3 trains, which go from Zurich to Basel non-stop. Giving this timeslot to TGV will free up an InterCity train set, which the SBB probably uses for extra service from Zurich to Bern.

          • Krist van Besien

            > Don’t worry about those petty misers of SNCF. They are charging 17 CHF for each commuter from Zurich to Basel.

            Are they? That train is not operated by SNCF. SBB operates that train from Zurich to Basel, and AFAIK they actually own some of the Lyria sets.

            To me they could easily move the train by about 15 min. Currently it is not connecting with trains to/from Bern at all, so this would actually reduce travel times on my router (Interlaken – Paris).

          • Henry Miller

            >Why would a high frequency service have arrival times?

            Because you often are connecting to something else. Eventually you will connect to something that doesn’t have high frequency service. Much as I want it otherwise and argue ever transit system should go for high frequency service, there are always tails. Cities should turn those into high frequency. I have argued that the expensive investment in good service (good is more than just high frequency, it also means good connections) would save residents a lot of money because most would get rid of their car – but there is no way to ignore that this is extremely expensive even if it is less expensive than cars.

            If you have high frequency service in an entire city you will get a number of farmers just outside your city to take a low frequency transit option into your city. When it is time to go home they need to know when they have to be on the high frequency bus/train by to make their connection to the low frequency final bus home. Most of those farmers will have family inside the city and so the connection to those farms helps get some people to sell their car and thus make the transit investment worth it.

            Remember that cars have a very high fixed cost of ownership, and very small variable costs. It doesn’t matter if you take the bus to work, you still have to pay insurance and taxes on the car that sits in your driveway. As such if your transit system isn’t good enough that people sell their car (not necessarily all cars) you can charge a lot more. If people feel they must have a car for anything you are competing with costs of gas (cheap even at $5/gallon), parking (free in the suburbs!), a little bit of maintenance (The one time large bills make this feel worse than it is) and congestion. If people feel like they don’t have to have a car you are also competing against insurance/ license fees, and the cost of buying the car.

          • adirondacker12800

            If the doors are open for an extraordinarily long time, 30, 40 seconds, the arrival time is the same as the departure time. Putting the same thing on the schedule twice would just clutter it up.

    • Sascha Claus

      I think their split between Nahverkehr and Fernverkehr makes the most sense. Anything within a daily commutable range has regional trains that are integrated into the local public transport ticketing system, […]

      If it makes money, it is kept by DB Fernverkehr, if not, it is dispensed with and has to live with only stopping trains, or the states subsidise it. So you get “Nahverkehr” like Cottbus – Wismar (300 km as the crow flies), Rostock – Elsterwerda (310 km) or Munich – Prague (300 km).

      It’s run as a social service for essential local travel needs and so farebox is willingly subsidised.

      But only as far as the market needs—transit is often priced as high as the market will bear (when compared to patrol prices or the available capacity on the busiest lines).

  6. Diego

    So, just to make it clear, the numbers you’re working with here predict that for a rail service running every hour and charging high fares, cutting fares by half and running half-hour headways would double ridership? With half of the increase coming from lower fares and the other half from increased frequency? Of course this probably only works in a limited regime, there are diminishing returns for reducing headways below 15 min for intercity travel.

    • Alon Levy

      Yes, except that the model for frequency doesn’t tell you what the point of diminishing returns is. On urban transit we know the point because there are estimates that consider both in-vehicle trip time and headways as well as models that connect the two via the waiting penalty, and thus we can just estimate a direct elasticity of ridership with respect to generalized trip time of around -0.8. But on intercity transit we need to make some assumptions – there are transfer penalty estimates but I haven’t seen any waiting penalty estimates; without a waiting penalty (e.g. because passengers don’t just show up and go), waiting an extra half hour on the timetable is equivalent to spending an extra 15 minutes in-vehicle because on average a half hour headway is a 15-minute wait.

      Relatedly, SNCF believes the sweet spot for high-speed rail is a 2-3 hour trip – below 2 hours they don’t think it’s worth bothering with.

      • Diego

        Relatedly, SNCF believes the sweet spot for high-speed rail is a 2-3 hour trip – below 2 hours they don’t think it’s worth bothering with.

        Right, because they don’t run high frequency services even when they could.

  7. adirondacker12800

    Buses, finally, are so small that a market like New York-Philadelphia supports multiple competitors

    There are enough masochists in the market to attract fly-by-night operators.

    Unless it’s a well know brand using the Port Authority Bus Terminal, you have no idea how well maintained the bus is or if the driver has any training. Or gets enough sleep. Some of them you stand on a street corner hoping it arrives. No idea which highway is going to turn into a parking lot and even if it doesn’t, it takes longer than the train. Unless your goal is to, stereotypically, to go from one Chinatown to another. What’s a few decapitated passengers now and then as long as the fares are low?

    passenger behavior is such that different companies are substitutable,

    A cheap bus leaving from Brooklyn isn’t very attractive if I’m in the Bronx. And while I have amused myself looking at fares I don’t want to get on a bus that may or may not have insurance driven by someone who thinks driving a bus is like driving a car, if the blue stuff in the bucket is blue stuff or it’s dollar store windshield washer fluid, or how sketchy the brakes are. There are enough masochists in the market to attract fly-by-night operators.

  8. Krist van Besien

    One thing to understand about the SNCF is that for that company basically only two kinds of persons matter: Railway staff, and Parisians. You see that the moment you try to use the railways to go from somewhere not Paris to somewhere else not Paris.
    So the system gives you fast trips to Paris at times convenient to Parisians wanting to escape, and provincials having some business in Paris.

    You see that for example in the schedule of the trains on the LGV Rhine – Rhone, where the TGV Lyria either stops in Dijon or Belfort, but never both, and there are other trains from Paris to Mulhouse or Besancone that have other stopping patterns. Very few that stop in all stations on that line, which makes the line impractical for commuter.s A sensibel schedule would probably be an hourly Paris – Dijon – Mulhouse – Basel, and an hourly Dijon – Besancon – Belfort – Mulhouse – Colmar – Strasbourg.

    Trenitalia does this better, but they also have a country where Rome isn’t the only city that matters. The high speed line from Milano to Rome and onwards to Naples sees a lot of traffic in both directions the whole day, and Trenitalia runs a logically schedule with both hourly Milano – Bologna – Roma trains, and Milano – Reggio Emilia AV – Bologna – Firenze – Roma trains.
    Reggio Emilia AV even has an nice hourly schedule in both directions. Something none of the French Beetroot stations have…

    And Trenitalia does have a fairly simple tariff structure, and more “on train” market segmentation.

    • Paul

      Sure, but that’s true of France in general. It’s a very centralized country where the political, business, and cultural elite are all based in Paris.

      • Matthew Hutton

        It’s not good though. Same as it’s not good that London is the centre of the universe in England.

        And France being a bit more accommodating of the rest of the country outside the Périphérique would also probably improve the connections from England to the rest of France, Italy and Spain.

        • Borners

          If the rest of England wanted that to be the case then they should do something other than complain and uphold the status quo on land/transport/local government/economic policy.

          France isn’t a perfect comparison since its rust belts have quite similar politics and problems under a more generous welfare state while being small relative to the rest of France. British was most the industrialised society ever and is suffering the hangover of failing to properly move on from that.

          Toulouse is like Bristol an aerospace mid-sized city while Rennes-Nantes are small car cities. Continental geography helps the Périphérique while it screws outer Britain. UK should be compared to East Asian island democracies in terms of urban hierarchies. That still doesn’t make Britain look good, in 1900 they called Osaka “the Manchester of the East” as praise, now its an insult.

          Anyway overcentralisation shouldn’t be a problem the vast majority of the British population lives on a North-South axis and large majority within the NW-SE urban belt which emerged along railway lines. The problem for UK railways is that speedy and cost-effective improvement requires institutional and labour changes that go against deeply held interests in both major parties as well deeply held beliefs in the various factions of the intellectual classes. FFS they just cancelled a 21km connecting branch for HS2 that cost 3 billion pounds and nobody wondered why it was so inanely expensive.

  9. Matthew Hutton

    The other issue with British rail improvements is that our speeds (at least on the north south lines) are pretty good already. Preston-London on a fast train averages 150km/h for example. And even Preston-Glasgow averages 125km/h. Even the latter is faster than Amsterdam to Brussels on a Thalys train which only manages 110km/h.

  10. Frederick

    The reason has to do with the ratio of waiting time to trip time. Thalys connects Paris and Brussels in 1.5 hours, every half hour at rush hour and every 2 hours midday. At rush hour, frequency is sort of noticeable; off-peak, it dominates travel time. This is nothing like planes – even short-distance trips involve hours of access, waiting, and egress time, and therefore trips are not usually spontaneous, and day trips are rare except for business travelers.

    Only Alon Levy would suddenly have the wanderlust to travel from Paris to Brussels, looking at the ticket seller saying, “why does a ticket cost 100 euros? Why is the next train 2 hours away?”

    Normal people who are not Alon Levy always plan for intercity trips. No matter it is a sightseeing trip or a business trip, people plan their trips before they pack their stuff. If people plan early, then they can buy cheaper tickets. And knowing their schedule, people enter the station 10 minutes before the train departs. Therefore, for intercity trips, frequency doesn’t count as waiting time for the passenger, so it doesn’t matter very much.

    Of course, there are people who have urgent need to travel. But these people are few and rail operators shouldn’t plan their network specifically for them. If trains were more frequent just for people with an emergency, the trains would have more vacant seats.

    tl;dr Intercity trips are not spontaneous, no matter it’s a plane trip or a train trip.

    • Krist van Besien

      The first time I took a Thalys I bought the ticket while the train was already standing at the platform. That was in Rotterdam, when the international ticket office there still faced platform 1. I had just missed the IC train half an hour earlier and absolutely had to be in Brussel.
      That was my first experience with the disadvantages of market segmentation.

      I used to travel that route a lot, and before Thalys you had the choice of either the Benelux IC, or one of the Eurocities to Paris. I often travelled on those, as they had a better connection with the IC to Oostende in Antwerp, giving me a half an hour shorter travel time.

      I now no longer live in The Netherlands, so I do not have to experience that several billions of tax payers’ money spend later, the same trip would now actually take me longer than it did then…

      Frequency matters. Even for planned long distance trips. Trips rarely involve one train. Connections are a common and expected feature of train travel. A higher frequency often means a shorter travel time, because you spend less time at a transfer point waiting for your next train. At also reduces the delays when a late train misses your connection.

    • Alon Levy

      Normal people who are not Alon Levy always plan for intercity trips.

      Tell me you don’t know any intercity rail riders in Northern Europe or Japan without telling me you don’t know any intercity rail riders in Northern Europe or Japan. (In Japan, “go to the station and board the next Nozomi” is common behavior; in Germany people look at timetables but the hourly takt is critical for Intercity/ICE ridership.)

      • Richard Mlynarik

        “Our meeting/meal/other trainsportation went
        longer than expected/was delayed/was too enjoyable to leave/was cancelled/was stuck in traffic/etc.
        We’ll just take the next/previous train, since they come every 15/30/60 minutes, no big deal.
        Oh and no need to waste time at the station because we had to be absolutely sure not to miss Our One TIcketed Train and so had to padded out the trip to the station by an extra half hour or so for insurance against a missed connection.”

        It’s just weird how people who can’t imagine Nice Things are dead-set on asserting that Nice Things Don’t Exist Anyway plus Nice Things Oughtn’t be Allowed to Exist and They’re Not Nice Things Anyway You’re Just Confused and in the end are so bonkers gung-ho about ensuring that Other People Don’t Deserve Nice Things Either.

        What do we call this? “Fare Gate Stockholm Syndrome”? “No Nice Things for ANYBODY”?

        • adirondacker12800

          Between New London Connecticut and Newark Delaware you are more than welcome to take advantage of show up and go train fares. Frugal travel writers author articles about the experience quite often.

      • Nilo

        Lots of people in Philadelphia love to go to NYC for the day. People on this site like to deny all sorts of strange things that even happen in America.

        • adirondacker12800

          And vice versa. I know people who get the urge to have a cheesesteak and argue about whether or not to go to Pat’s or Geno’s or both. They swipe or tap a few things on their smartphone and decide to take a Keystone instead of Acela. Or vice versa. One of them has a Club Acela card, they ask the very solicitous customer service agents in the Acela lounge if they might be able to manage a free upgrade. And get one once in a while. Sometimes the answer is that the train is all sold out which is why Amtrak is buying roughly 40 percent more capacity for the marginally “better” trains that people pay premium prices for. Sometimes it can be very very premium.

      • Frederick

        You keep spreading this myth that people just hop on a Nozomi and go. Let me tell you why this is actually a rare occurrence.

        A Nozomi ticket costs about ten thousand yen. Normal Japanese people just don’t spend ten thousands haphazardly; they will search for discounts or at least buy the ticket from ticket shops (kinken shop).

        Is there any Nozomi rider who just tap their IC card at the gate? Yes, either they are rich, or they are spending their employer’s transportation stipend. Mostly the latter.

        You need to spend a large amount of time and money for an intercity trip. That’s why normal people plan for it.

        P.S. it is weird how we went from talking about poor people who thinks 100 euro is too much to rich people who would hop on a Nozomi or Acela without looking at the price tag.

        • adirondacker12800

          People who want simple fare structures will thrash about for all sorts of reasons. The reasons don’t have to coalesce into a coherent whole.

        • John D.

          You’re assuming a hard binary between rigorous planning and complete spontaneity. From experience, many trips fall somewhere in between. If you’re holding a non-reserved ticket and your trip plan permits a flexible arrival time (for example, “before lunch” instead of any more specific timing), then showing up to the station and hopping on the next Nozomi is very much a thing.

        • Alon Levy

          People don’t make the decision to go to Osaka spontaneously, but my understanding from literature and from talking to Japanese rail users is that once they’ve decided to make a trip, they just go and take the next Nozomi rather than timing themselves to a specific train – and that’s in a city where even the subways are timetabled and run on time.

      • Martin

        The just show up nozomi behavior you describe is clearly only possible because nearly all trains have empty seats.

        Put differently lots of other people are taking slower local trains and intercity buses or their car, as the nozomi has so high prices as that nearly always have lots of empty spaces.

        The shadow cost in extra travel time for those (poorer) travellers taking worse modes of transportation is very large, and unfortunate. It is also basically at the convenience of richer travellers.

        • John D.

          “Lots of other people are taking slower local trains and intercity buses”? A typical Nozomi trip like Tokyo to Nagoya would require multiple transfers and quadruple the travel time by Tokaido Main Line locals, so it’s a deeply unpopular option outside of Seishun-18 Ticket bargain hunters. Meanwhile, statistics show the Shinkansen holding around thirty times the market share of expressway buses for Tokyo to Keihanshin travel.

          • Frederick

            Even among Seishun-18 ticket holders, Tokyo-Nagoya is quite an unpopular route, because of the so-called “long-seat hell” in Shizuoka. Middle-distance trains in Japan often have (in Japanese language) “cross-seats”, i.e. seats facing the front or back of train, which are more comfortable. However, most slow trains in Shizuoka have “long-seats”, i.e. the kind of seats you see in metros, facing the sides of train. Those seats are less comfortable, and from Atami to Toyohashi, there is 180km of hell to endure.

        • adirondacker12800

          It’s just awful that in densely populated areas people can drive a private car, take a basic bus, a plush bus, a plane – some of them with three classes of seats, a slow train or fast train. Awful.

        • Frederick

          “lots of other people are taking their car”

          lol. Let me give you an exercise: check the toll of the Tomei Expressway.

          • adirondacker12800

            And you need a car.
            I’m not in the mood to go figure out what the toll is between Manhattan and DC. There are tolls.

    • Henry Miller

      Why shouldn’t intercity trips be spontaneous? For shorter trips they are – for any trip under 2 hours people will make a day trip once in a while to visit someone. Sure if the trip is 5 hours you can’t really get there and back in a day and so more planning is required, but intercity train trips are often just a couple hours, well within the range of a spontaneous trip. It won’t be a regular trip, the distance is too long, but you can still do it a couple times a month.

      Spontaneous trips are mostly about cost. I used to have a friend who worked for an airline – every friday he would go to work with his suitcase packed, when he got off work he would head to the gates and look for a flight that wasn’t full (this was before 9-11, you could walk to any gate, and with his employee discount $20 to get on any plane that wasn’t full). Sometimes he ended up in Stockholm (his brother lived in Sweden so this was the first plane he looked for), sometimes in San Francisco… Obviously someone with a family needs more planning, but it doesn’t take long to pack a suitcase. I’ve had many weekends where my family would have made a spontaneous trip someplace – but we didn’t realize that the weekend would be free far enough in advance.

    • Matthew Hutton

      The fact that planes don’t run frequently can make connecting flights really, really, really, bad. You can find a journey by plane with a 10 hour direct flight taking 20 or even more hours with a connection.

      And ultimately the airlines “get away” with it because it generally only applies to intercontinental trips where there’s no real alternative in terms of time to flying.

      • adirondacker12800

        It applies to flights within the Eastern Time Zone in the U.S. The travel sites, with ads all over everywhere, are really good at offering non-stops or one-stop/change planes at decent prices and more expensive itineraries that take two, three, four times as long and involve multiple airports. Google is good at suggesting the airlines that aren’t listed in the travel sites.

        • Matthew Hutton

          I’m thinking more flying from London to South America – if the connecting flights arrives after the daily flight back to London you have a nearly 24 hour wait 😉.

          • adirondacker12800

            I was just mentioning that the travel sites offer options to masochists who would take the more expensive itinerary that involves changing planes three times and takes 18 hours instead of the itinerary that is two flights and 6 hours. I vaguely remember seeing truly bizarre stuff like 16 hours and four legs to get to someplace I can get non-stops to, from Albany. I’ve been told there are funner ways to indulge masochism. And cheaper, assuming you already have the equipment.

    • Diego

      That’s complete nonsense. Have you never checked the weather forecast for the weekend, seen it’s warm and sunny, and then decided to go to the beach? Plenty of people do that here, given how busy trains to the coast are on sunny days.

      I use this train to the beach example because from Brussels it’s approximately the same travel time to the coast and to Paris, give or take 10 min. The only reason the train to the beach has way more ridership is because of its lower fare and its increased convenience (takt timetable, no need to book in advance).

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 )

Twitter picture

You are commenting using your Twitter 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.