Fares on High-Speed Trains
American journalists sometimes ask me to discuss high-speed rail fares. I’ve written from time to time about how Northeast Corridor fares are significantly higher than French, German, and Japanese high-speed rail fares, but the links to this information are never in the same place. The purpose of this post is to collect all the links together for easy retrieval, with updates to the 2020s whenever possible. Unfortunately, international high-speed rail fares connecting to France are also much higher than domestic ones, which contributes to the poor ridership of those trains relative to city size.
France
ARAFER releases statistics annually. The most recent year for which there is data is 2022; here’s the report in French, and here’s a summary in English. The relevant information is in sections 5-6. The TGV system including international trains averages fare revenue of 6,213M€ for 61 billion passenger-km; in English this is called “non-PSO,” since these are the profitable trains that SNCF runs outside the passenger service obligation system for money-losing slow trains. This works out to 0.102€/p-km. The PPP rate these days is about 1€ = $1.45, making this about $0.15/p-km.
The split between domestic and international trains is large, and the French report has the domestic trains as just 0.093€/p-km, taking a weighted average from pp. 23 and 31. Nominal fares per p-km on domestic TGVs were down 4% from 2019 to 2022, despite 7.5% cumulative inflation over this period.
The international trains, in contrast, are much more expensive: the report doesn’t give exact numbers, but from some weighted averaging and graph eyeballing it looks like it’s around 0.17€/p-km. The all-high-speed international trains – Eurostar and Thalys – are more expensive than the trains running partly at low speed in Germany and Switzerland, like Lyria; this big difference in fares helps their disappointing ridership. Domestic TGVs run from Paris to Lyon 28 times on the 5th of June this year, counting only trains to Lyon Part-Dieu or Perrache, which do not continue onward, and not counting trains that stop at Saint-Exupéry on their way to points south, and those trains are rather full 16-car bilevels. In contrast, on the same day I only see 16 Eurostars from London to Paris. This is despite the fact that London is a far larger city than Lyon, and the in-vehicle travel time is only moderately longer.
Germany
Germany lacks France’s neat separation of low- and high-speed trains. The intercity rail network here is treated as a single system, and increasingly all trains are ICEs even if they spend the majority of the trip on legacy lines at a top speed of 200 km/h.
Overall intercity rail passenger revenue here was 5.1 billion € in 2022; the expression to look for is “SPFV.” Ridership was 42.9 billion p-km per a DB report of 2022-3, PDF-p. 7, averaging 0.119€/p-km, which is $0.17/p-km in PPP US dollars. 2022 was still slightly below 2019 levels, when ridership was 44.7 billion p-km and fares averaged 0.112€/p-km; the three-year increase was less than the cumulative inflation over this period, which was 10.3%.
Japan
Japanese fares are higher than European fares on high-speed rail. JR East’s presentation from 2021, showing depressed ridership during the pandemic (p. 50), reports ¥189.6 billion in Shinkansen revenue on 7.95 billion p-km, or ¥23.8/p-km, and projects recovery to ¥428.9 billion/17.313 billion p-km by 2022, or ¥24.8/p-km. JR Central’s 2020 report says (p. 37) that its Shinkansen service got ¥1.2613 trillion in revenue in the year ending March 2020 on 54.009 billion p-km, or ¥23.4/p-km. JR West’s 2020 factsheets for revenue and ridership show ¥457 billion/21.338 billion p-km in 2019, or ¥21.4/p-km.
The PPP rate for 2020-1 was $1 = ¥100. Taking 9% dollar inflation from 2020 to 2022 into account, this is, in 2022 prices, around $0.25/p-km.
Northeast Corridor
Amtrak publishes monthly performance reports; the fiscal year is October-September, so the September reports, covering an entire fiscal year, are to be preferred. Here are 2022 and 2023; 2022 still shows a considerable corona depression, unlike in France and Germany. The 2023 report shows that Northeast Corridor revenue splits as $495.9 million/581.1 million p-miles Acela, or $0.53/p-km, and $768.2 million/1.6269 billion p-miles Regional, or $0.293/p-km Regional. Altogether, this is $0.356/p-km, which is nearly 50% higher than the Shinkansen, 2.1 times as expensive as the ICE, and 2.4 times as expensive as the TGV.
Discussion
High operating costs on Amtrak are the primary reason for the premium fares. The mainland JRs are all highly profitable; DB Fernverkehr is profitable, as is the TGV network (though SNCF writ large isn’t, the slow intercities falling under the PSO rubric). All five companies pay track access charges for the construction of high-speed rail infrastructure; the ARAFER report goes over these charges in France and a selection of other European countries, designed to prevent state subsidies to intercity rail operations through underpriced track access, since track construction is always done by the state but operations may be done by a private operator or a foreign state railway. The Northeast Corridor is profitable as well – Amtrak doesn’t have to pay track access charges, but the access charges for legacy 19th-century lines would not be significant. However, if Amtrak charged European fares or even Japanese ones, it wouldn’t be. Northeast Corridor rail operations in fiscal 2023 earned $1.266 billion in passenger revenue plus $28.5 million in non-ticket revenue but spent $1.0917 billion, or $0.307/p-km.
A portion of the Amtrak cost premium also comes from adversarial profit maximization, also seen on Thalys and Eurostar. The domestic TGVs and ICEs aim at making a base rate of profit while providing a service for the general public; SNCF doesn’t apply the same logic to Thalys and Eurostar and instead aims at serving only business trips to avoid the possibility of extracting less than maximum fares from international travelers. On Amtrak, the need to subsidize the rest of the system has increased Northeast Corridor fares, though to be clear, in fiscal 2023 the operating margin was small enough that this is at most a secondary factor. Performance reports from the 2000s and 10s showed a larger operating margin, but criticism from advocacy groups centering non-Northeast Corridor passengers alleging that Amtrak accounting was making the Northeast Corridor look better and the night trains look worse led to a recalculation, used in the most recent reports, in which Northeast Corridor operations still turn out to be profitable but not by a large margin.
Pedestrian Observations 
Alon, thank you for this interesting comparison. I’m interested in Italian data, is it available?
Alon, thank you for this interesting comparison. I’m interested in Italian data, is it available?
In 2017, Italo’s average fare was 0.087 Euro/p-km, marginally below the 2016 fares of other European operators (0.105 Euro/p-km on the TGV and DB LD, 0.099 on RENFE AV) and most likely slightly below Trenitalia AV. The significant lowering of tolls (from 12.81 to 8.2 E/train-km) on the Alta Velocita network helped both carriers turn a small profit on their AV operations while offering lower prices than when Trenitalia had a monopoly.
My debit card doesn’t deduct the cost of ticket, from my account, in PPP. It’s deducts the amount at current exchange rates and reasonable charge for the exchange.
Nobody cares what your debit card does. The vast majority of rail tickets (I would hazard north of 99%) are bought by people who live in the country the trains run in. These people earn “wages”, which reflect local economic “conditions”, and for which PPP is the appropriate measurement.
And the vendors these theorectical people buy stuff from don’t care what the PPP is. They want cash.
The supply chains for the US and Europe are basically disjoint for this. Construction labor is local, crew labor is local, the materials are made locally (like concrete), the energy is produced locally, and even the rolling stock is unfortunately not imported from Europe or Asia in the US but made locally at transplant factories.
And it’s very nice to know how purchasers feel about their purchases. If we import wunderwagen trains the vendor isn’t going to adjust the price to PPP. They are going to want cash. We need to know what the current exchange rate was.
Yeah, this, and I also think this is related to the analytical error made by Americans who assume all European HSR works like Eurostar (expensive, low-ridership) – American tourists are disproportionately likely to be wanting to take trains between London and Paris, as opposed to Paris and Lyon, or Paris and Bordeaux, or Cologne and Mannheim, etc.
Why doesn’t anyone compete with Thalys to bring down fares?
Because France and Belgium have gamed the system as to make that very difficult, in order to protect the incumbent.
Good summary although the “adversarial profit maximization ” explanation for higher Eurostar and Thalys fares is simplistic. Lower fares on domestic TGVs mostly reflect the lower cost of home operations with a largely twin deck fleet and higher average seat density.
Chunnel tunnel tolls add about 4 Euro cents/p-km to a Paris-London run. Add higher HS1 tolls and higher station usage fees to cover the cost of custom facilities and Eurostar higher prices are explained.
The TGVs built for continental International services are adapted to multiple electrification and signaling systems. They cost about 50% more to purchase and their complexity implies higher operating costs as well.
I can’t prove it but I have a strong hunch that the Pendelinos have higher operating costs than other British trains.
Now the West Coast Mainline is still very profitable so I suspect the higher speeds tilting unlocks still works out favourably with cheaper staffing and less capital costs because you need fewer trains but still.
It seems very believable that the international trains would have higher running costs though.
The other side of the coin in France in particular is that the international services don’t make efficient use of the stock. The 2:45pm service from Paris to Barcelona could at least allow passengers on board for the presumed ~10pm departure back to Perpignan to recover at least some costs – or alternatively it could continue to Madrid and stable there overnight.
And aside from the first train of the day all of those trains are only making one journey a day which isn’t particularly efficient.
The operations between France and Germany don’t cost extra, though, and I think neither does Lyria. See, for example, the comparison in chapter 4 here: the Three Capitals connections are by far the costliest, while other international connections don’t charge these fares.
Note also that while the services on the LGV Sud-Est are uniformly bilevel, this is less true on the other domestic TGVs and yet fares are barely higher, and also ICEs are single-deck and charge marginally higher fares than TGVs.
The Steer Davies Gleave report you refer to only illustrates another of your talking points, the folly of relying on handsomely paid private consultants. On a report about the cost per km of various train journeys, the authors did not bother looking at actual track distances. While Google-Earth told them that as the crow flies Paris was closer to London than to Lyon, the opposite is true on the rails. The Paris-London and Brussels-London trains have a dog-leg route through new expensive infrastructure owned by private investors (HS1 and the Chunnel tunnel) and the LGV Nord whose cost was at 90% borne by SNCF. Is any one surprised that operators charge more on these trips where they are competitive with air travel than on a Lisboa-Madrid (685 km in 11 hours) or Timisoara-Budapest (340 km in 6 hours)?
” this is less true on the other domestic TGVs and yet fares are barely higher,”
So Sud-Est subsidizes the other lines?
No, the other lines are profitable as well, just by a smaller margin.
By law, SNCF is not allowed to lose money on the LGV programs. It could finance the first 6 lines up to the 2001 LGV Mediterranee because their return on investment was at least 8%. Although ticket sales on these lines have already paid off the investment, the High Speed Trains using these lines are still charged high tolls. These tolls are now largely “Congestion Charges” because these LGVs are indeed busy and do subsidize the maintenance of the loss making conventional network. The more recent LGVs starting with the LGV East were only partially financed by SNCF Reseau because the expected patronage did not allow a sufficient return on investment even after the threshold was lowered to 4.5% to reflect lower interest rates of the time. Until this debt is paid off, these lines only turn a significant profit if patronage is above expectation which seems to be the case nowadays but was not during the pandemic or when oil was cheap.
I ride for first time in NE corridor train from Penn Station to Philadelphia station. Prices are lot higher than Rodalies between Girona and Barcelona. They describe “high speed” but from time table doesn’t go anyway near speed of AVE trains in Spain. I think they still run some diesel burners on standard gauge track in NE.
Alon, could you address airline-style variable fares vs Japanese fixed-price tickets? Also, what are your thoughts on group discounts to get families to ride the train instead of driving?
I know families that routinely drive from DC or Providence to NYC because they only plan the trip a week or two before and Amtrak is extremely expensive, even when accounting for parking costs in NYC (they take the subway once there), gas, and tolls.
Related–there is no reason for price differentiation between the Acela and Northeast Regional, right? Acela is hardly even faster, especially if you are passing through NYC. It severely hurts frequency (and thus depresses ridership) for most riders who can’t afford Acela.
All good questions, and I apologize in advance for not having great answers to all of them.
1. Japanese fixed-price tickets are simpler to implement, and don’t make passengers feel like they’re being suckered for not finding the cheapest fare. They also make it easier to change tickets – Shinkansen tickets are refundable (I believe 90%), there’s just no need to deal with nonrefundable, nonchangeable early bird specials.
2. That said, fares in Japan are higher. I don’t think that it’s related to the fixed regime. German fares are much less airline-like than French ones, but the average is very close.
3. The reason for different prices on the Regional and Acela is just straight price discrimination – Amtrak is allowed to do that without limit, whereas there are legal limits on its ability to do yield management on different seats on the same train. This is akin to the Spanish practice of having multiple different classes of train distinguished by top speed, with the faster trains costing more, and each train class running infrequently; this is why until competition was introduced recently, Madrid-Barcelona ridership was underperforming French city pairs relative to city size.
4. Group discounts are a really difficult subject. I’ve vlogged about it but without definitive answers. They’re clearly important, since trains are less competitive with cars for group travel; and unlike planes, trains have very low marginal cost per rider and decently low marginal cost per extra trainset (in Amtrak accounting the latter is called “frequency-avoidable cost” and is 20% of NEC operating costs). The best idea I’ve come up with is that a child can ride free with a paying adult, maybe just on off-peak trains if that’s otherwise too onerous, so a family of four pays like a couple or two singles.
Why don’t trains just allow groups of up to 4 to travel for the price of 2 off peak?
Business group travel.
Seems like a pretty good solution to that is to only offer it in standard class and not in first class. And if that isn’t sufficient maybe first class should be made more business friendly with compartments etc.
Are compartments business-friendly? I always thought of them as for families… but maybe they’re for group travel in general.
That will depend on the compartment. No business wants a bed – too many issues become possible if sharing rooms. However if the compartment has a nice table and a somewhat large screen that is easy to connect a laptop too – this becomes a meeting room and is very valuable to any group traveling on business. There is no reason a compartment cannot fill both purposes (the bed just isn’t folded down for the business use during the day, put away the business equipment at night for the family), but the logistics probably makes it impractical.
I’m not aware of any train selling a compartment setup for business use, but it seems perfectly doable and something that should be experimented with. The compartment needs some business focused equipment but that should now be available on the market.
The German ICE trains have compartments in 1st class for business use.
Why is it a goal to improve competitiveness for group travel? Relatively full private passenger cars are fairly good from a GHG emissions standpoint, and relatively full private charter buses are great from a GHG emissions standpoint.
Even relatively full private passenger cars still consume a lot of space when driven to and parked in a dense urban area, but the solution isn’t group travel discounts for trains, but rather appropriately priced parking.
A highway full of even relatively full private passenger cars still moves a lot fewer people per footprint than rail, but the solution for that isn’t group travel discounts for trains, but rather appropriately priced highways.
For far-from-full off peak trains where the marginal cost of an extra passenger is extremely low, why should discounts for group travel be favored over off peak fares that apply regardless of how many people you’re traveling with?
Also, Shinkansen tickets are typically refundable minus a small fixed service fee, assuming you didn’t buy through some third party service with a bad refund policy. For Tokaido-Sanyo-Kyushu Shinkansen tickets bought via the first party smartEX service, the fee is 320 JPY, which would a 90% refund for Tokyo-Odawara, but a 98% refund for Tokyo-Shin-osaka. The refund policy is the same even for tickets bought with the (relatively small) early bird discount. If you just don’t tap through the Shinkansen gates that day, the refund is automatic, but a much larger distance variable fee (40%?) is taken.
Interesting points. The trade-offs vary widely from country to country. In Switzerland rail is fully electric and generation is quite CO2-free. That’s hard to beat. With the German energy mix, long-distance trains still emit less than half the CO2 of a full car. Charter buses are indeed not much worse than trains there.
All this doesn’t yet consider the significant negative externalities of cars, some of which you mention. Others include accidents, noise, particulates, or traffic delays.
The case for group discounts may be weaker for places with a worse rail energy mix.
If you get a flexible train ticket in the UK you get a 100% refund if you cancel in my experience.
Group travel should be encouraged for many reasons.
Cars are dangerous, and self driving is not looking to fix that any anytime soon. (time will tell of course). Trains are much safer, even large buses are better than a small group. Thus I want to encourage larger groups where possible.
A large part of the costs of owning a car are fixed – once you have it you have to make the monthly payments and insurance even if you don’t use it (where I live insurance doesn’t give a discount for not using the vehicle – different areas have different rules). So if you have a car for group trips the incremental cost to use it for solo trips is much less. (particularly if you have free parking and no traffic). Thus getting a group on transit makes solo car trips less likely.
Groups often include kids – you want to establish a habit of transit is how you get around when they are young so they don’t think a car is normal. I include anyone in college as a kid – they are in a place where they don’t have much money but do want to travel and so a cheap option can set lifelong habits of transit.
There are people who look only at numbers. Groups mean more people on transit and so those people cannot use the argument nobody uses transit.
Also the expensive bit of the railway is the track and the other fixed costs. So the more passengers the better really.
For local and regional transit, RMV offers a group day-pass for five passengers at the price of two return tickets:
https://www.rmv.de/c/en/tickets/your-ticket/tickets-overview/day-tickets/rmv-group-day-ticket
Don’t know how that works out for the vendor but from a passenger perspective it’s great.
Discount tickets do exist for shinkansen but they are fairly limited. For example JR Central has the plat-kodama which gives a 20-30% discount on the slower kodama train, and JR East Tokudane tickets can give 20-50% early bird discount. I believe both of these have significant late cancellation penalty.
Alon:
Thank you for the helpful presentation of the information. You state that your calculations are based on a conversion rate of 1 euro = 1.45 dollars. The current conversion rate is 1 euro = 1.08 dollars. Even rounding up to 1 euro = 1.1 dollars, the difference in conversion rates should significantly shift the results.
Respectfully,
https://data.oecd.org/conversion/purchasing-power-parities-ppp.htm
Alon:
Thank you for sending the pointer to the PPP manual. I have now done my Penance have gone through it from start to finish. As I remembered, PPP is an excellent method for normalizing complexly Aggregated financial and economic data. It is not appropriate for un or lightly aggregated data; Exchange rates are the appropriate normalizer for unaggregated or lightly aggregated data. The first of several summaries of this are found in table 1–8.
I rather like PPP as a normalizer for aggregated construction costs. However, high-speed rail fares,Broken out by country and further broken by domestic or international, Are Very lightly aggregated and exchange rates are the appropriate normalizer.
I hope this helps.
It doesn’t actually help, because exchange rates fluctuate, without this mattering too much to high-speed rail costs or fares. Nearly the entire cost structure of TGVs and ICEs is in euros, and nearly the entire ridership on these trains earns and spends euros. Changes in the value of the US dollar here are not especially relevant to high-speed rail or any other kind of physical infrastructure; an exchange rate analysis would have European construction costs, operating costs, and fares fall 22% in the year from spring 2014 to spring 2015, or 16% in the year and three months from the beginning of 2021 to the fall of 2022.
Cost/p-km is a great way to evaluate fares (it can be directly correlated to the price of a ticket between two locations a certain distance apart) but is a terrible way to evaluate operating costs. It costs the same to run a train a given distance with one passenger on board as it does to run an identical train the same distance when it is full. Operating-cost/p-km will be dramatically lower in the second case, even though operations are no more efficient. Losing passengers because a low cost airline starts or gaining passengers because gas prices spike can rise lower OpCost/p-km again without any actual change in efficiency or amount spent in operating costs.
With this in mind, is Amtrak’s higher OpCost/p-km a result of poor operating procedures (they spend too much) or an artifact of underserved demand (Acela has only six coaches per set not 14-16, ridership on non-HSR 100-150kph avg lines is less than ridership on 200kph+ avg lines)?
Maybe I am misunderstanding something but Amtraks costs per passenger mile as per page 5 of https://www.amtrak.com/content/dam/projects/dotcom/english/public/documents/corporate/monthlyperformancereports/2023/Amtrak-Monthly-Performance-Report-September-2023.pdf are actually pretty good compared to the costs in the UK adjusted for PPP as per figure 4 on page 8 ofhttps://dataportal.orr.gov.uk/media/algdbizg/rail-industry-finance-uk-statistical-release-202223.pdf
The trains that will be replacing Acela should be in service any year now. 8 revenue cars per train and it can be increased if demand appears.
“It costs the same to run a train a given distance with one passenger on board as it does to run an identical train the same distance when it is full. Operating-cost/p-km will be dramatically lower in the second case, even though operations are no more efficient. “
Yes, but for a route where passenger demand is low, you buy/run fewer trains, so that the trains you do run are still full and the cost per p-km roughly unchanged.
In a sense, this is part of the concept of “operations efficiency”.
To cover the marginal running costs of an extra train you typically don’t need to fill it very full. Most of a railways running costs are fixed – and you need a fair number of trains for the morning weekday peak.
True, but I was speaking about the number of trains you must buy to meet peak demand.
The seat-mile vs. p-mile numbers in the Amtrak 2023 report work out to 65% seat occupancy on both the Regional and the Acela, which compares with 74% for domestic TGVs (see the French ARAFER report, p. 23), and 56% for pre- and 49% for post-corona DB intercity trains (link).
I do think the lower average speed increases operating costs, since crew costs are per-hour and not per-km, and energy costs are higher when you need to change speed frequently as is the case on legacy lines, but evidently in Germany the costs are similar to those of France, even with an average speed that’s regrettably similar to that of the NEC.
Comparison low-speed train:
Oakland Airport BART Connector, completely automated, 5.1 km, $6 fare, 96% fare recovery
6/5.1 = $1.17/p-km
The #1 competition is ride share, especially for families.
That’s not a relevant comparison – stopping, unloading, manning stations and so on are much larger fractions of operating costs for very short lines than long ones.
Also airport connectors are usually very expensive.
What about China? My understanding is that people do use some of the intercity HSR lines for commuting there.
About ‘compartments’ on daytime trains aimed at businessmen. In the 50s/60s PRR’s Congressional (fastest on the NEC 3, later two schedules per day), had a ‘drawing room car’ which did have 7 rooms w/ table (no computers) and a radio telephone booth at the end of the car for staying in touch w/one’s office. In the 90’s Amtrak experimented with a single modified Metroliner car fitted out with a conference room for up to 8 IIRC. Too little equipment so not much usage. No carryover to Acela.
Given wifi etc, the new Acela 2s should have similar spaces, but I doubt they were ordered that way.