American railfans are full of nostalgia for a past era when American trains were great. So much of the discussion among industry insiders, railfans, and advocates is about how to make American railroads great again, how to return to the mid-20th century era of American domination. This is not correct history: while American railroads were in fact in a pretty good position from the 1920s to the 50s, they were not competitive with mass motorization and air travel, and trying to imitate what they were like then has no chance of competing with cars or planes. The story of American railroads has to be understood not as decline but as stagnation: operations, technology, and management stagnated, and this is what led to ridership decline. Instead of indulging in a MAGA fantasy about past greatness, it is important for the United States to implement all the innovations of the last half century that it has missed out on, innovations coming from East Asia and Western Europe.
American railroads were private until Amtrak took over intercity operations and states took over commuter rail operations, which happened well after the terminal decline in ridership. There was intense competition between rival companies, at times leading to physical violence. There was no coordination of operations between different railroads, no coordination with municipal public transportation systems, no attempt at seamless passenger experience. What was the point? This system evolved in the early 20th century, when there was no competition from other modes, only from other railroads.
Over time, most of the rest of the developed world has learned to coordinate different modes of public transportation better, to compete with cars. This usually occurred under nationalized mainline rail companies, but even when companies remained separate, as with the division between municipal subways (e.g. the Berlin U-Bahn) and national railways (e.g. the S-Bahn), or with the separate BLS system around and south of Bern, there has been integration. There is fare and schedule coordination in German cities across rail and bus operators, and even better coordination in the Netherlands and Switzerland.
The US remains fixated on competition, and thus there is no fare integration, but rather relationships between different operators are adversarial. In Chicago, the mayor opposes integration between the municipal L and the regionally-owned Metra commuter rail system, since the city does not own Metra. Every time Amtrak has to share territory with a commuter railroad, one side is screwing the other out of something, whether it’s Amtrak overcharging on electricity or Metro-North arbitrarily slowing Amtrak down. In Boston, there is no integration between city-focused MBTA service, which includes commuter rail, and buses in outlying cities, called RTAs (regional transit authorities); the MBTA is simply uninterested in matching fares or schedules, and is not even integrating its own buses with its commuter trains.
Switzerland has higher rail usage than every place I know of once one controls for city size. Zurich’s modal split may not be as favorable to public transportation as Paris or London’s, but is a world better than that of any French or British city of similar size. Switzerland got to this point through a stingy political process in which planners had to stretch every franc, substituting organizational capacity for money. Thus, construction in the 1990s used the following principles of value engineering:
- Infrastructure, rolling stock, and the timetable should be planned together (the magic triangle), since decisions on each affect the other two.
- Trains should run as fast as necessary for transfer windows, overtakes, meets on single track, etc. Infrastructure should likewise be only as expansive as necessary – if the timetable does not have trains on a given single track meeting, this segment does not need to be doubled.
- Trains should have timed transfers at major cities, to enable everywhere-to-everywhere travel. Connecting buses should be timed with the regional trains at major suburban nodes.
- Electronics before concrete: it’s cheaper to resignal a line to have short headways and high speeds than to add tracks and tunnels.
These principles do not exist in the United States. Worse, too many American activists, even ones who are pretty good on related issues, do not believe it’s even possible to implement them. “This isn’t Switzerland or Japan” is a common refrain. There’s growing understanding among American cycling advocates that 50 years ago the Netherlands wasn’t as bike-friendly as it is today; there sadly isn’t such understanding regarding the state of rail coordination in Switzerland until about the 1990s.
While Switzerland manages to build its Knot System at low cost, leading to sharp increases in rail usage in the 2000s and 2010s, Americans are unable to do the same. Activists propose massive spending, which the political system is unwilling to fund. Nor is the political system interested in adapting low-cost solutions for infrastructure coordination, since the sort of apparatchiks who governors like appointing to head state agencies can’t implement them; we all know what happened last time a foreigner got appointed to a major position and succeeded too much. The way forward is right there, and the entire American political system, from every governor down to most activists, either is ignorant of it or explicitly rejects it.
Amtrak runs slower than it used to on most lines. Trip times on the Northeast Corridor south of New York are if anything slightly slower than they were in the 1970s in the early days of the Metroliner. The corridor and long-distance service outside the Northeast are considerably slower. For example, the Super Chief took 39:45 between Chicago and Los Angeles, whereas its current Amtrak version, the Southwest Chief, takes 43:10. At shorter range, the Chicago-St. Louis trains take 5:20 today, compared with 4:55 in the 1930s. This has led too many Americans to assume that there has been technological regression and that the main focus should be on restoring midcentury service levels rather than on moving forward.
In reality, high speeds in the middle of the 20th century came from the facts that express passenger trains were highly profitable and used by important people and so had priority over all other traffic, and that superelevation was set high for these trains; both of these aspects collapsed as riders and high-value shippers decided driving and flying were better than taking 5-hour train rides, so the profit center shifted to low-value freight. Today, getting high passenger ridership is plausible at high-speed rail speeds, but that requires getting Chicago-St. Louis down to 2 hours, not 5 hours, and having excellent connections to local and regional public transportation at both ends.
Nor was midcentury rolling stock good by current standards. Electric locomotives in Europe weigh around 90 tons. American ones weigh a little more, still in compliance with superseded FRA regulations enacted just after WW2. But the locomotives from just before these regulations weighed far more: the Pennsylvania Railroad’s GG1 weighed 215 metric tons. Europe has achieved weight reductions over generations of innovation since, and Japan has achieved even more impressive reductions; 215 tons would get you 2/3 of the way to a 10-car EMU set in Tokyo.
Worse, even in the middle of the 20th century, the US was no longer at the technological forefront of rail service. The civil service formation following the German Revolution brought forth a new railway law and new technology, such as the tangential switch, since adopted throughout Continental Europe; the US mostly sticks with secant switches built to late-19th century specs. In the 1950s the differences between German and American rail technology weren’t huge, but they were there. Since then they’ve gradually widened – in the 1960s Germany came up with LZB signaling, while the US was at best stuck on 1930s signaling, federal regulations on the matter leading to lower top speeds than to the adoption of automatic train protection.
There seems to be general ignorance of the advances that the US has not been part of. Rail managers ask questions like “does Europe have positive train control?” (yes, ETCS is already a second-generation system, we just call this automatic train protection instead of positive train control) or say “Europe doesn’t have the ADA” (accessibility laws here are comparable to American ones and overall the public transportation networks here are on average more accessible). In technology as in organization, the MAGA mentality for trains refuses to admit that there are innovations abroad to learn from.
The way forward: imitate, don’t innovate
The United States can innovate in public transportation, but only if it imitates better countries first. It needs to learn what works in Japan, France, Germany, Switzerland, Sweden, the Netherlands, Denmark, South Korea, Spain, Italy, Singapore, Belgium, Norway, Taiwan, Finland, Austria. It needs to learn how to plan around cooperation between different agencies and operators, how to integrate infrastructure and technology, how to use 21st-century engineering.
There are great places where such imitation could work. I work a lot on Boston-related issues at TransitMatters; New England has high population density, a wealthy and growing urban core in Boston, ample legacy rail infrastructure, and town centers that work more like Central European suburban sprawl (albeit at lower density) than like structureless Californian or Texan sprawl. But it can’t move forward without rejecting MAGA fantasies and replacing them with a program of learning from what works here and in Japan. There are so many projects under discussion of limited or no value, and some even with negative value, like anything that interacts with the hobby freight railroad Pan Am.
Instead, the tendency in the United States is to do anything to avoid learning from outside North America. Plans for intercity rail improvement outside the Northeast and California are steeped with MAGA language about restoring midcentury rail. Plans in New York spend far too much time on midcentury expansion plans and far too little on understanding cost explosion factors dating to the 1920s. Regional rail plans vaguely nod to European S-Bahns, but are generally filtered through several layers, mainly Philadelphia’s implementation. Anything that touches freight invites kludges that European planners no longer use for cost or maintenance reasons.
This tendency has to end. Meiji Japan didn’t join the first world by closing itself to foreign inventions – quite the opposite. The US needs to understand that the path to a future with better American transportation lies not in America’s past, but in Europe and East Asia’s present. The history isn’t one of American decline and renaissance through rediscovery of ancient learning, but one of American insularity and stagnation, to which the solution is to adapt technologies that work elsewhere.
Is it possible to use a rapid transit-style system to carry light freight, such as parcels? So far no such system exists, and very few semi-relates systems exist (like pneumatic tubes for mail). But it remains an interesting potential technology, provided it is done right. Unfortunately, it is very easy to do it wrong through misunderstanding how freight or how rapid transit works. Therefore, advances in policy in this direction are good but should be done carefully.
Instead of giving people one big takeaway, I’m going to suggest a few good principles for this, motivated by both good and bad proposals.
1. Keep the tracks clear for maintenance at night
Germany’s minister of transport, CSU’s Andreas Scheuer, proposed running freight on the U-Bahn after hours. This is a terrible idea: regular nighttime closures are crucial for maintenance, and without them, maintenance costs go up and daytime reliability tanks. New York’s constant weekend service changes are the result of not shutting down overnight for maintenance nor being able to reliably single-track at nighttime headways. Berlin already runs overnight on weekends and does some daytime maintenance – “Ersatzverkehr mit Bussen” is one of the first ordinary German phrases I learned after moving here. Further encroachment on maintenance windows is not acceptable.
2. Use existing station infrastructure
The main cost in digging urban rail tunnels is the stations – boring tunnels between stations is a solved problem. This means that the main difficulty of urban rail freight is where freight gets on and off the trains. Loading and unloading container-size freight is impossible without massive station digs, all in expensive places. Having a freight car wait on a siding is not possible either – that interface between the customer and the freight railway relies on cheap land and time-insensitive shipping.
Most likely, shipping parcels by rapid transit requires using the existing stations and platforms. There is almost certainly no room at rush hour, when trains are sized to take up the entire platform interface to increase capacity. But in the daytime off-peak, there may be some room for using a portion of a subway station for parcels.
3. Keep up with passenger rail traffic
If freight trains can’t run at night, they have to slot on the same timetable as passenger trains. This isn’t a problem on the tracks – just add an EMU car loaded with parcels rather than passengers. But keeping dwell times under control is critical. Alert reader Mordy K. wrote about this, suggesting a “dynamic Rubik’s cube” that “shifts the packages around in 3D.” This is the real challenge: figure out how parcels get from the train to a designated spot on the platform or from the platform to a designated spot on the train during a 30-second dwell time.
4. Be aware of all interfaces between different systems
There are, at a minimum, five legs to a parcel trip in a city using rapid transit: origin to station, station to train, trip on train, train to station, station to destination. The boarding and alighting steps, so easy for the able-bodied passenger and even for the disabled passenger given rudimentary investment into accessibility, are difficult for a parcel of freight. Tossing a package from a train to the platform is not enough: the package needs to get to the surface for the final leg of the trip. A courier could carry it, but at a high cost – the courier’s modes of transportation for the surface legs, like the e-bike, are bad at getting down to the subway and back up, so the time and physical effort costs are high.
This in turn means that the rail transit freight system needs to be able to put parcels in a freight elevator. Elevators are not free, although they are rarely as expensive as in New York. The problem is that parcels can’t walk across the platform, so the elevator has to face the exact same place every time, which may run into construction difficulties.
5. Don’t wreck passenger rail service
Berlin runs some trains short, especially after hours. Usually the first train after the beginning of short train service is very crowded, because passenger service demand is still too high for a half-train at the typical density of daytime Berlin trains. (I say typical density because I have never seen a Berlin train as crowded as the busiest off-peak trains in Paris or New York, let alone their busiest peak trains.)
What this means is that in practice there isn’t that much space on the train for freight. Running trains every 5 minutes until later at night but then cordoning off half for freight may be feasible – right now headways rise to 10 minutes around 9 in the evening – but it’s still just a few evening and early night hours for delivery.
This principle is equally important at the stations: cordoning off parts of the platform for freight is fine, but only if it does not interfere with passenger capacity or circulation. This may further constrain where freight elevators go: whatever automated system gets parcels from the train to the elevator will have to cross passenger traffic at-grade, and driverless technology can do it but not cheaply or smoothly.
6. Aim to work with a wide range of goods
Pneumatic tube systems for mail work for mail, but modifying them for other goods isn’t trivial. In contrast, a parcel delivery system should aim to be broadly usable by many goods with a high ratio of value added to weight. Subtle differences are important at this level of detail: glass and china goods can’t be thrown on the floor, fresh food spoils if it’s left outside for too long, jewelry and electronics face a high risk of theft. The technology has to have adequate tracking, punctuality, defense from shocks, and so on.
7. Be aware of the competition
Delivery by rapid transit is not the only alternative to trucks for cross-city shipping. Delivery by drone is in active development, both surface drones and flying drones. Surface drones have good synergy with trains, since surface drones are slow and make better first- and last-mile connections. But flying drones are in direct competition, since they work well at a range of a few kilometers rather than a few hundreds of meters. Flying drones so far only work at extremely high value-to-weight ratios, but if they become more widespread, it’s useful to think of how urban rail can compete.
The construction costs of Britain’s just-approved domestic high-speed rail network, High Speed 2, are extreme. The headline costs are, in 2019 figures, £80.7-88.7 billion per the Oakervee review, with one estimate going up to £106.6 billion, all for a system only 530 km in length in mostly flat terrain. This includes rolling stock, but that is less than 10% of the projected cost. At the end of the day, Britain has decided to spend around $200 million per kilometer, a cost comparable to that of base tunnels and mostly-tunneled high-speed lines.
And now the People’s Republic of China has offered to build the entire thing for cheaper with a 5-year timeline, and everyone acts as if it’s a serious offer. So let me dust off my construction costs database and tell you: the PRC won’t save you. There is no alternative to developing good internal cost control. This requires learning from lower-cost countries, but Chinese high-speed rail construction costs are not really low.
Ho Chi Minh City and Hanoi are both building metros. Hanoi uses Chinese financing, HCMC uses Japanese financing. Both have very high construction costs – my database has HCMC’s 13% underground Line 1 at $320 million/km, 82% underground Line 2 phase 1 at $535 million/km, and 84% underground Line 5 phase 1 at $590 million/km, whereas Hanoi’s 74% underground Line 2A is $215 million/km and 32% underground Line 3 is $365 million/km.
The system in Hanoi has been plagued with delays. Line 2A was supposed to be operational by 2016. Construction was only completed in 2018, but the line is yet to open. Testing is ongoing, but Chinese experts couldn’t return to Vietnam after the Chinese New Year holiday because of the coronavirus quarantine. The South China Morning Post has compared the Hanoi project negatively with that of HCMC, which is for the most part on time, if expensive.
Like many developing-world cities, HCMC is paying more for a subway tunnel than Japan pays at home; to get to the cost range of HCMC in Japan, one needs to go to complex regional rail tunnels in Tokyo dipping under multiple older tunnels in city center. In that it is no different from Dhaka or Jakarta. The primary explanation must be that importing Japanese technology means using techniques optimized for a high-skill, high-wage labor force and cheap domestic capital, rather than ones optimized for a low-skill, low-wage labor force and expensive imported capital.
But that does not explain why the Hanoi Metro is so expensive. Chinese metros cost less (though not universally – Shanghai’s construction costs are rising fast): I want to say about $250 million/km on average, about the same as the non-Chinese global median, but the actually big set of data is unpublished so you guys can’t nitpick my sources yet. So what’s going on here? Vietnam is poorer than China, but the difference is not so big. It’s about half as rich as the PRC. It’s comparable to Europe, where Romania and Bulgaria are about half as rich as Western Europe, and they have low construction costs, lower than parts of Eastern Europe closer to Western incomes.
Chinese high-speed rail
The construction costs of high-speed rail in the PRC are fairly high, especially in its richer parts. The costs remain lower than those of tunnel-heavy lines like those of Italy, Japan, and South Korea, but by low-tunnel standards, they are high.
There is a perception that Chinese costs are low, but it comes from using the wrong currency conversion. Here, for example, is a World Bank report on the subject:
[P. 39] Figure 4.1 shows the construction cost of 60 projects. The average cost of a double-track HSR line (including signaling, electrification, and facilities) is about Y 139 million/km (US$20.6 million/km) for a 350 kph HSR line, about Y 114 million (US$16.9 million) for a 250 kph HSR line, and about Y 104 million (US$15.4 million) for a 200 kph HSR line. These costs are at least 40 percent cheaper than construction costs in Europe (European Court of Auditors 2018, 35).
The problem is, the exchange rate of $1 = ¥6.75 is incorrect. The OECD’s PPP conversion factor today is much higher, $1 = ¥3.5; for high-speed lines built a decade ago, it would be even higher, about $1 = ¥3.3, with ten years of American inflation since. Using the correct modern rate, the cost is about $40 million per kilometer, which is not lower than in Europe but rather higher. Beijing-Shanghai, as far as I can tell a ¥220 billion project for 1,318 km of which just 16 km are in tunnel, rises to $50 million per km, and more like $60 million per km in today’s money. It’s still cheaper than High Speed 2, but more expensive than every Continental Europe high-speed line that isn’t predominantly in tunnel, like Bologna-Florence.
There are all these longwinded explanations for why the PRC does things cheaper and faster than the first world, and they are completely false. China is not cheap to build in, especially not high-speed rail. The only reason Chinese costs aren’t even higher is that Eastern China is pretty flat. Even then, China has not taken advantage of this flatness to build tracks at-grade to minimize costs. Instead, it has built long viaducts at high cost, in contrast with the at-grade approach that has kept French LGV costs reasonable.
The PRC doesn’t even build things particularly quickly. Total actual construction time from start to finish per line segment is 4-6 years per Wikipedia’s list, which is comparable to recent LGVs. What is true is that China has been building many lines at once, and each line is long, but this is a matter of throughput, not latency. The limit to throughput is money; the PRC made a political decision to spend a lot of it at once as stimulus in the late 2000s and early 2010s, and by the same token, the UK has just made a political decision to spend just less than £100 billion on High Speed 2, in a trickle so that the system will take 15+ years to complete.
Why are they like this?
The myth of hyper-efficient Chinese construction seems never to die; I’ve seen it from the first days of this blog, e.g. then-US Secretary of Transportation Ray LaHood in 2012. It relates to a mythology that I think is mostly part of Anglo-American culture, of the tension between freedom and efficiency. The English-speaking world in this mythology is the epitome of freedom, with a gradation of less free, more efficient paces: Germany, then Japan, then finally China. It’s a world in which people’s ideas of what totalitarianism looks like come from reading George Orwell and not from hearing about the real-life Soviet Union’s comic incompetence – the gerontocracy, the court politics, the drunk officials, the technologically reactionary party apparatchiks – all of which was happening in real time in Nazi Germany too, which was fighting less efficiently than the UK and US did.
It’s equally a world in which people think rights Germans and Japanese take for granted, like various privacy protections, do not even register as important civil liberties. I dare any reader to try explaining to a British or American transit manager that really, no, you do not need our data, Central Europe manages to plan better than you without smartcards tracking users’ every move and storing the data in servers with infosec that screams “steal me.” Nor do Americans make much of an effort to import policing regimes from democracies with one twentieth their rate of police shootings per capita.
China’s incompetence is now visible to the entire world, in the form of a virus outbreak that local officials flailed about for a month, too afraid to acknowledge mistakes lest they take the fall for them. And yet it’s easier for American and British business leaders and politicians to point to China as an example to emulate than to Pareto-better France or Germany.
If anything, High Speed 2 is low-key overlearning some French lessons, leading to inferior infrastructure planning – but it’s messing up key details leading to cost explosion, such as “don’t build new signature urban train stations.” But my suspicion is that French and German rail experts will point out all those details. To us, if Britain changes some detail in a way that isn’t truly justified by local conditions, we will point it out – and push back when British blowhards try to explain to use that they do things differently because they’re morally superior to us. British people know this – they know they can’t pull rank. Americans are the same, except even less capable of dealing with other nations as equals than the British are.
The way forward
High Speed 2 is a mess, largely because of the cost. To move forward, talking to China about how it’s built high-speed rail may be useful, but it can’t be the primary comparison, not when Continental Europe is right here and does things better and cheaper. For Asian help, Japan has some important lessons about good operations and squeezing maximum use out of limited urban space. A lot of scope can be removed. A lot more can be modified slightly to connect to regional lines better.
More conceptually, Britain has a problem with costs and benefits chasing each other. If benefits are too high, the political system responds with sloppy cost control, for example by lading the project with ancillary side projects that someone wants or by giving in to NIMBY opposition. If the costs are too high, the political system responds with scrounging extra benefits, for example counting the consumer surplus of high-speed rail travelers as a benefit, by which standard every government subsidy to anyone has a benefit-cost ratio of at least 1.
Bringing in the PRC won’t help. It’s value-engineering theater, rather than the hard work required to coordinate infrastructure and timetable planning or to tell Home Counties NIMBYs that the state is not in the business of guaranteeing their views; there is so much tunneling on the proposed line that isn’t really necessary. None of the countries that builds trains cheaply did so by selling its civil service for spare parts; why would Britain be any different?
I wrote a Twitter thread about high-speed rail in the United States that I’d like to expand to a full post, because it illustrates a key network design principle. It comes from Metcalfe’s law: the value of a network is proportional to the square of the number of nodes. The upshot is that once you start a high-speed rail network, the benefits to extending it in every direction are large even if the subsequent cities connected are not nearly so large as on the initial segment. Conversely, isolated networks from the initial segments are of lower value.
The implication for the United States is that, first of all, it should invest in high-speed rail on the entire Northeast Corridor from Boston to Washington, aiming for 3-3.5 hour end-to-end trip times. And as the Corridor is completed, the priority should be extensions in all directions: south to Atlanta, north to Springfield and (by legacy rail) Portland, west to Pittsburgh and Cleveland, northwest to Upstate New York and Toronto.
To quantify the benefits, I’m going to look purely at railroad finances: construction costs go out, annual profits go in. Intercity high-speed rail pretty much universally turns an operating profit, the question is just how it compares with interest on capital construction. For this, in turn, we need to estimate ridership. Here is an illustrative photo of the sophistication of the model I am using:
In the picture: someone who gets on the train without letting you get off first. Credit: William O’Connor.
The theoretical model for ridership is called a gravity model: ridership between two cities of populations Pop_A and Pop_B at distance d is proportional to
However, two complications arise. First of all, there are some diseconomies of scale: the trip time from the train station to one’s ultimate destination is likely to be much higher if the city is as huge as Tokyo or New York than if it is smaller. Empirically, this can be resolved by raising the populations of both cities to an exponent slightly less than 1; on the data I have, which is Japanese (east and west of Tokyo), Spanish (Madrid-Barcelona, Madrid-Seville), and French (see post here – all its sources link-rotted), the best exponent looks like 0.8.
And second, at short distance, the gravity model fails for two reasons: first, access time dominates so in-vehicle time is less important, and second, passengers drive more and take fast trains less. In fact, on the data I’m most certain of the quality of – that from Japan – ridership seems insensitive to distance up to and beyond the distance of Tokyo-Osaka, which is 515 km by Shinkansen. Tokyo-Hiroshima, 821 km and 3:55 by Shinkansen, underperforms Tokyo-Osaka by a factor of about 1.6 if the model is if we lump in air with rail traffic; of course, air travel time is incredibly insensitive to distance over this range, so it may not be fair to do so. French data taken about 3 hours out of Paris overperforms the mid-distance Shinkansen, although that’s partly an artifact of lower fares on the TGV.
To square this circle, I’m going to make the following assumption: the model is,
If the populations of the two metro areas so connected are in millions then the best constant for the model is 75,000: that is, take out the number the formula spits, multiply by to get rid of the denominator at low d, multiply by 0.3, and make that your annual number of passengers in millions.
Finally, operating costs are set at $0.05/seat-km or $0.07/passenger-km, which is somewhat lower than on the TGV but realistic given how overstaffed and peaky the TGV is. This is inclusive of the capital costs of rolling stock, but not of fixed infrastructure. Fares are set at $0.135/passenger-km, a figure chosen to make New York-Boston and New York-Washington exactly $49 each, but on trips longer than 770 km, the fares rise more slowly so that profit is capped at $50/trip. Of note, Shinkansen fares are about $0.23/p-km on average, so training data on Shinkansen fares for a network that’s supposed to charge lower fares yields conservative ridership estimates; I try to be conservative since my model is, as the picture may indicate, not the most reliable.
The model on the Northeast Corridor
The Northeast Corridor connects four metropolitan areas: Boston (8 million people), New York (22), Philadelphia (7), Washington (10). All populations cover combined statistical areas, just as the metropolitan area definitions in Japan are expansive and include faraway exurbs. In the Northeast, the CSAs lump together some independent metro areas, such as Baltimore-Washington, but the largest of the subsidiary metro areas, including Baltimore, Providence, New Haven, and Trenton, are along the Northeast Corridor and would get their own stations.
The distances are 360 km Boston-New York, 140 km New York-Philadelphia, 220 km Philadelphia-Washington. I am not going to take into account subsidiary stations in passenger-km calculations, for simplicity’s sake. Splitting Baltimore apart from Washington would actually raise ridership by a little, first because the 0.8 exponent means that combining metro areas reduces ridership, and second because Boston-bound ridership is higher if we assume the destination is a little bit closer.
The highest-ridership city pair is New York-Washington. Per the formula above, we get
By the same formula, New York-Boston is 18.77 million, New York-Philadelphia is 16.87 million, Washington-Philadelphia is 8.98 million, and Boston-Philadelphia is 7.51 million. All of these are within the 500 km limit in which we assume distance doesn’t matter. Finally, Boston-Washington is
Overall, this is 79.4 million annual passengers, excluding shorter-distance commuter travel like New York-New Haven. Taking distance traveled into account, this is 26.4 billion annual p-km, generating $1.7 billion of operating profit. What I think it should cost to generate this service is investments that, with good value engineering that has been missing from all plans in the last 12 or so years, should cost in the low teens, say $13 billion. If costs can be held to $13 billion, or just less than $20 million per kilometer for a line of which about two-thirds of the physical infrastructure is good enough, then the financial return on investment is 13%. Not bad.
Of note, traffic density is fairly symmetric at the two ends. At the southern end, between Philadelphia and Washington, total traffic density is 36.24 million passengers per year; at the northern end, between New York and Boston, it is 31.1 million. So there should be some extra trains just for New York-Philadelphia, where the expected traffic density is 51.64 million – perhaps ones diverting west to inland Pennsylvania, perhaps interregional trains making an extra stop or two running 5-10 minutes slower than the trains to Washington – but otherwise trains should run on the entire corridor from Boston to Washington.
Also of note, I don’t expect much peakiness on the line – probably none outside the New York-Philadelphia segment. Short-distance lines, including New York-New Haven and New York-Philadelphia, have a rush hour peak in travel. But longer-range intercity lines generate weekend leisure travel and same-day business travel, both of which tend to peak outside the regular rush hour; TGV traffic, heavily weighted toward longer-range city pairs, peaks on Friday and Sunday, with weaker weekday ridership to balance it out. The Northeast Corridor thus benefits from mixing cities at various ranges, with the various peaks mostly canceling each other out. It’s plausible to get away with running service at a regular interval of every 15 minutes all day, with extra trains on New York-Philadelphia.
The Northeast Corridor and Metcalfe’s law
Two examples of Metcalfe’s law in action can be found on the corridor, one for an expansion and one for a contraction.
The contraction would be to ignore Boston and just focus on New York-Washington. The traffic density is higher there, for one. Moreover, no extensive civil infrastructure is required, only some small fixes in Maryland and New Jersey, a rebuild of the catenary, and rebuilds of the station throat interlockings. However, this is less prudent than it seems, because Boston doesn’t just generate traffic on New York-Boston, but also on New York-Washington, on trains bound from points south of New York to Boston.
If we exclude Boston, we have just three city pairs on what is left: New York-Washington, New York-Philadelphia, Washington-Philadelphia. They total 48.3 million passengers per year and 12.4 billion p-km – in other words, slightly less than half the p-km of the entire line including Boston. What’s more, there’s an extra fudge factor, not modeled in my ridership screen, coming from peakiness: a shorter line is one with a more prominent rush hour peak, as the longer trips on Boston-Washington are not included, and this ends up requiring more rush hour-only equipment and increases operating expenses per p-km.
The expansion is, in this section, one that is almost part of the Northeast Corridor today: New Haven-Springfield. The line is unelectrified today despite substantial investment by Connecticut, which like other American states is allergic to rail electrification for reasons that are beyond me. Speeds today are low, even though the right-of-way is straight. However, investment in bypasses and in speedups on the highest-quality legacy segment is possible, and would connect Hartford and Springfield to New York and points south.
The Hartford-Springfield region has 2 million people, and Springfield is 100 km from New Haven and 210 from New York. We apply our usual model and get New York-Springfield ridership of 6.19 million, Philadelphia-Springfield ridership of 2.48 million, and Washington-Springfield ridership of 3.3 million. In passenger-kilometers, these three city pairs amount to 1.3 billion, 620 million, and 1.55 billion respectively, for a total of 3.47 billion, which I will round to 3.5 billion to avoid giving the impression that the model is reliable to 3 significant figures (or even 2, to be honest).
So we have 3.5 billion additional p-km for just 100 km of new construction, or 35 million p-km per km of construction. Note that the expected density on New Haven-Springfield based on the model is just 12 million passengers – the remaining p-km are on the core Northeast Corridor, as passengers from New York and points south travel on a portion of the corridor to get up to the branch to Springfield. So even though the expected traffic is very light, the impact on revenue per kilometer of construction is comparable to that of the base corridor. If costs can be held to $2 billion, which is low-end for an entirely greenfield line but reasonable for service that would partly run on the existing legacy line, then the return on investment is $0.065*3.5/$2 = 11%, almost as high as on the base Northeast Corridor.
To the north, it is valuable to run upgraded legacy trains between Boston and Portland, with a short connection to high-speed trains at South Station. Estimating ridership and revenue there is dicier, because the trains are slower and the data is trained on high-speed trains. We assume 190 km of revenue, as is the current length of the line. But costs and the ridership-suppressing effect of distance are charged at 350 km, roughly scaled for time.
With this in mind, ridership on Boston-Portland is 1.19 million, ridership on New York-Portland is 1.33 million, ridership on Philadelphia-Portland is 0.37 million, and ridership on Washington-Portland is 0.31 million. In total, this is about 1.5 billion p-km, of which 45 million, all from Washington, are beyond the 770 km at which fares are $0.135/km and are charged at the lower rate of $0.07/km. Altogether it’s around $200 million a year in revenue. Costs are around $140 million, including extra costs for service south of Boston. Operating profits are fairly low, but Boston-Portland legacy trains don’t cost per km nearly as much as high-speed rail; electrification and some track work can be done for maybe $600 million, for an ROI of 10%.
Of course, this ROI does not exist without high-speed rail the Northeast Corridor and without the separately-charged North-South Rail Link for local and regional trains. Like other tails, Boston-Portland is valuable once the mainline preexists – it isn’t so great on its own.
The route from Washington to Atlanta has a sequence of cities roughly following the I-85 corridor. They are small and sprawly, but are still valuable to connect thanks to Metcalfe’s law. These include Richmond (1, and 180 km from Washington), Raleigh (2, and 240 km from Richmond), the Piedmont Triad (1.6, 120 km), Charlotte (2.6, 150 km), Greenville (1.4, 160 km), and finally Atlanta (7, 230 km).
The line is long, 50% longer than the Northeast Corridor. With quite sprawly cities in North Carolina and few good rights-of-way, takings and viaducts are needed and would raise construction costs, to perhaps $30 billion. Moreover, there is probably an intercity rail ridership penalty because these cities do not have public transportation; the model does not incorporate such a penalty, which should be regarded as a risk with investments made appropriately. And yet, each city in sequence generates ridership on the line to its north, creating decent ROI if we assume the model applies literally.
Take Richmond. It’s a small city, generating 1.89 million annual riders to Washington, 1.42 million to Philadelphia, 3.05 million to New York, 0.49 million to Boston. But this is 2.9 billion p-km for just 180 km of new construction, and nearly all of these p-km are chargeable at the full rate, giving us a total of $190 million in annual operating profit. If construction can be kept to $5 billion, this is just short of 4% ROI, which is not amazing but is decent for how small Richmond is compared with the cities to its north.
This calculation cascades farther south. We have the following table of ridership levels, in millions of annual passengers as always:
|City N\City S||Richmond||Triangle||Triad||Charlotte||Greenville||Atlanta|
This leads to the following operating profits, in millions of dollars per year:
|City N\City S||Richmond||Triangle||Triad||Charlotte||Greenville||Atlanta|
This totals to $1.8 billion a year, or an ROI of 6%. This is not a safe number – a hefty share of the figure comes from city pairs that trains would connect in 3.5+ hours, like New York-Charlotte, Washington-Atlanta, and even the 5.5-hour New York-Atlanta, in which range the model has essentially two data points (Tokyo-Hiroshima, Paris-Nice). Another noticeable share comes from intra-South connections, in which neither city in the pair has a strong center or a public transport network to connect the station with destinations.
But thankfully, because this line can build itself up by accretion of extensions, starting with Washington-Raleigh and seeing how ridership holds up would not create a white elephant, just missed benefits if the model is in fact correct.
Harrisburg (0.7), Pittsburgh (2.5), and Cleveland (3)
The Keystone corridor is an interesting example of a branch that gets stronger if it is longer. The reason for this is that Harrisburg is pretty small, and Harrisburg-Pittsburgh requires painful tunneling across the Appalachians. Philadelphia-Harrisburg is 170 km and can probably be done for $4 billion; Harrisburg-Pittsburgh is 280 km and, as a pure guess, requires around 40 kilometers of tunnel, let’s say $14 billion. Pittsburgh-Cleveland is 200 km and may require some tunneling near the Pittsburgh end to bypass suburban sprawl without good rights-of-way, but not too much – figure it for $6 billion.
For the benefits, we make a table similar to that for the South, but smaller. Of note, Washington-Harrisburg is 390 km and about 1:45, and costs accordingly to operate, but can only charge for 220 km, or $30, barely more than breakeven rate, because the straight line distance is short and high fares may not be competitive with driving on I-83. The straight line distance is even shorter than 220, about 190 via Baltimore, but Washington-Philadelphia is 220. Trains from Washington are assumed to earn the usual marginal profit west of Harrisburg, $0.065/km up to a maximum of $50, which is not reached even in Cleveland.
Finally, note that Cleveland has a big difference between the population of the core metro area (2 million) and the combined one (3.5), like Boston and Washington. Here we don’t take the bigger population but split the difference, since the biggest subsidiary regions in the combined area, Akron and Canton, could plausibly be on the line – and if they’re not then the line can serve Youngstown (0.7), and then . Note, finally, that Boston-Cleveland is faster via Albany and Buffalo, so the line through Pittsburgh is not considered even if it is built first.
|City E\City W||Harrisburg||Pittsburgh||Cleveland|
And as before, using the special malus for the roundabout Washington-Harrisburg route, we have the following operating revenues in millions of annual dollars:
|City E\City W||Harrisburg||Pittsburgh||Cleveland|
Note that the relatively easy to build segment to Harrisburg only generates $98 million in operating profit on $4 billion in construction costs, just less 2.5% ROI – Harrisburg is almost as big as Richmond, but it’s a branch and not a direct extension. Then Pittsburgh generates $390 million on $14 billion, or 2.8%. But Cleveland, easier to build to and bigger than Pittsburgh, manages to generate $344 million on $6 billion, finally a respectable ROI of 6%.
The northern cross
What may be caled the northern cross or the Albany cross – that is, a cruciform system consisting of lines from Albany to New York, Boston, Montreal, and Toronto – is an interesting case of a system where Metcalfe’s law again applies and encourages going big rather than small.
To apply the model, we make a crucial assumption: the same formula calibrated to domestic travel works internationally. Eurostar severely underperforms it – it has 10 million annual riders, of whom around 7-8 million go between London and Paris, where the formula predicts 18 million. Eurostar fares are very high, and has mandatory security theater and a slow boarding process that breaks down at peak travel time, and this may be enough to explain the low ridership. But then again, domestic TGVs overperform the model.
We also make a secondary assumption: fares charged are for actual distance traveled, even though the New York-Toronto routing isn’t the most direct.
We start with the New York-Toronto leg by itself. It connects New York to Albany (1.2, 220 km), Utica (0.3, 140 km from Albany), Syracuse (0.8, 80 km), Rochester (1.1, 120), Buffalo (1.2, 110), and Toronto (8, 160 km). Toronto’s metro population ranges between 6 million and 9 million depending on definition, and the high-end figure of 8 million is justifiable by the fact that Niagara Falls and Hamilton are on the line.
|City S\City N||Albany||Utica||Syracuse||Rochester||Buffalo||Toronto|
And in operating profit:
|City S\City N||Albany||Utica||Syracuse||Rochester||Buffalo||Toronto|
New York-Albany should cost maybe $5 billion to build and generates $160 million a year in operating profit, just 3.2%. But Albany-Buffalo, for around $11 billion extra, generates $580 million, about 5.2%. And then Buffalo-Toronto, assuming no international penalty, should cost on the order of $3 billion (much of the line in the Toronto suburbs automatically follows from the ongoing electrification project) and generate $670 milion. So the last segment, Buffalo-Toronto, returns around 20% if New York-Buffalo preexists; even if there’s a serious international malus, the ROI is very high. Everything combined is around 7%.
None of this is robust. The model handwaves the forced transfer at Penn Station – through-service from points south to Albany and points north would be excellent given expected traffic levels, but the approaches to both Albany and Philadelphia point west. The model also assumes New York-Toronto fares are in line with rail distance, even though the route is 50% longer by rail than by air. Finally, it assumes no international penalty. A 7% ROI is robust to any one of these assumptions failing, but if all fail, the route drops in profitability.
Or, rather, the base route does. Just as completing Buffalo-Toronto makes New York-Buffalo seem far stronger, so do the two additional legs of the northern cross strengthen the initial Empire Corridor. Here’s the Boston-Albany leg, at 260 km, with Springfield at kp 135, recalling that Hartford and Springfield count as one region of 2 million:
|City W\City E||Boston||Springfield|
And in revenue:
|City W\City E||Boston||Springfield|
$450 million a year, of which nearly half comes out of connecting Toronto to Massachusetts and Hartford, is not a lot, but then constructing 260 km of high-speed rail is not that expensive either; my best guess is around $7 billion, with some tunnels between Springfield and the summit to the west but also some approaches at both ends that would already exist. This is 6.4% ROI, which is better than New York-Toronto gets without the assistance of Philadelphia and Washington even though that route connects to a bigger city and requires less tunneling.
The final leg of the northern cross is to Montreal (4, 370 km from Albany), and is the most speculative. If the model has an international malus, it may well apply here, crossing not just a national border but also a linguistic one. It may apply with no or limited penalty, if the underperformance of Eurostar can be ascribed entirely to fares; but if it applies and is serious, then there is less cushion for mistakes than there is for trains to Toronto. The only intermediate city is Burlington (0.2, kp 220 from Albany), which exists largely for state-level completeness. Note also that Buffalo-Montreal is faster via Toronto and is thus omitted, while Buffalo-Burlington would have third-order impact.
|City ESW\City N||Burlington||Montreal|
And in operating revenue:
|City ESW\City N||Burlington||Montreal|
This is $750 million a yeer, of which Burlington furnishes about 10%, and New York-Montreal about 40%. This isn’t bad ROI – about $10 billion is a reasonable construction cost – but since 90% of it is about Montreal, any serious international or interlinguistic penalty leads to a big drop in profitability. Worse, if traffic drops, there may be a frequency-ridership spiral – I am writing timetables assuming half-hourly frequency, which is just enough for the model’s projected 18 million passengers per year, but if ridership is off by a factor of more than 2, then hourly frequencies start taking a bite out of the nearer markets and trains start running less full.
Lines that do not touch the Northeast Corridor
In the previous sections, I’ve argued in favor of building out a high-speed rail network out of the Northeast Corridor, on the grounds that extensions would be profitable toward Pittsburgh-Cleveland, Montreal, Buffalo-Toronto, and Atlanta. What about other lines?
The answer is that lines that do not feed into the Northeast somehow are a lot weaker. California can get decent ridership out of Los Angeles-San Francisco and thence extensions to Sacramento and San Diego are pretty strong, but the traffic density per the model is both well below California HSR Authority projections and well below the Northeast Corridor.
And it gets worse in parts of the country without a Los Angeles-size city anchoring everything. Texas is currently building a Dallas (8)-Houston (7) high-speed line, using private money by Texas Central, a railroad owned by JR Central using Shinkansen technology. The model predicts 7.5 million annual riders between the two regions, and the system’s public ridership projections for the near term are pretty close. Moreover, construction costs are pretty high, $15 billion for 380 km, despite the flat terrain. If the operating costs and fares are what I’m projecting, the financial ROI is 1.2%.
What’s more, Texas can expect ridership to underperform any model trained on European or Japanese cities. Tokyo has the world’s largest central business district, and maintains high density of destinations at a distance of several kilometers from Tokyo Station as well, and 20-something rapid transit lines depending on how one counts feed this center. Paris is smaller but has a strong center and urban rail connections. The provincial cities in both countries are lower-density and have higher car ownership, but that’s still okay, because people from those cities are not driving into the capital. By the same token, trains to New York should not underperform a model trained on Japan, Spain, or France. But Texas is completely different, with very weak centers, no public transportation to speak of, and no walkable cores near the train stations. The penalty for poor public transport connections is likely to be serious.
The situation in the Midwest is more hopeful than in Texas, but still dicey. Chicago just isn’t that big. Yes, it’s about the same size as Paris, but the cities ringing it don’t form neat lines the way Lyon and Marseille are on the same line out of Paris, just with a short spur rather than through-service.
The funniest thing about the Midwest is that high-speed rail construction there may be justifiable as an accretion of western extensions from the Northeast and Keystone Corridors. Cleveland-Detroit (5) is 280 km long, and would put Detroit 1,070 km and slightly more than 4 hours away from New York. The distance penalty is hefty, but 2.81 million annual users is still a lot over such distance, and the $140 million in operating profits get to around 2.5% ROI on construction costs in flat Midwestern land, without taking any other connections into consideration: Chicago-Cleveland, Chicago-Detroit, Philadelphia-Detroit, Pittsburgh-Detroit, New York-Toledo, Washington-Detroit.
Even New York-Chicago is a fairly solid line per the model: it’s 1,340 km and slightly longer than 5 hours, but there is still a lot of travel volume between the two cities, mostly by air. The model says 3.12 annual high-speed rail riders, somewhat fewer than the current O&D flight volume (4.68 million annualized from 2018 Q2), which is believable by comparison with Paris-Nice’s mode share (70% air, 30% rail, ignoring all other modes). The required mode share is still more favorable to rail, but the airports in New York and Chicago have more congestion and more delays than in Nice, turning what is a little more than an hour in the air into a three-hour flight schedule.
In contrast, just starting from Chicago-Cleveland (550 km)/Detroit (470 km), without any other connections (and without Pittsburgh-Cleveland), would not be financially great. Ignoring Toledo, the three cities generate 13 million annual riders, 6 billion annual p-km, and $400 million in annual operating income, for a system that would take perhaps $13 billion to construct, perhaps slightly more.
What this means for high-speed rail construction
Metcalfe’s law implies that high-speed rail networks get stronger as they add more nodes, even if those nodes are somewhat weaker than the initial ones. But it gives guideines for how to build such networks more broadly:
- Don’t cheap out by only building a short segment.
- Once the initial segment is in place, invest in extending it and building branches off it as soon as possible, in preference to building unconnected segments elsewhere.
- A relatively empty tail may still be financially successful if it fills a trunk line.
- Unless all your cities are on one line, try to build a mesh of lines to allow many origin-destination pairs.
- You’ll always run into a frontier of marginal lines, so value-engineer infrastructure as much as possible to push that frontier forward.
- Be wary of lines for which the analysis involves extrapolation, for example if neither city has a strong center or usable public transport.
High-speed rail is cheap to run when there’s enough scale to fill trains – high speeds ensure that labor and equipment cost per seat-km are fairly low. This means that self-sustaining profits are viable, and once they’re in place, they can generate further borrowing capacity for rapid expansion.
The limit is not the sky. Beyond a certain point, no realistic value engineering can make lines financially viable. Sometimes the cities are just too small or too far apart. But a realistic limit for the United States is still most likely much farther than anyone proposing immediate investment plans thinks: the Northeast Corridor can generate good returns if investment there is ever done competently, and branches and extensions to smaller and less dense cities are still more viable than they look at first glance.
After drawing a map of an integrated timed transfer intercity rail network for the state of New York, people asked me to do other parts of the United States. Here is New England, with trains running every 30 minutes between major cities:
New England is a much friendlier environment for intercity rail growth than Upstate New York, but planning there is much more delicate. The map thus has unavoidable omissions and judgment calls, unlike the New York map, which straightforwardly follows the rule of depicting intercity lines but not suburban lines like the Long Island network. I ask that people not flame me about why I included X but not Y without reading the following explanations.
The tension between S-Bahn and ITT planning
The S-Bahn concept involves interlining suburban rail lines through city center to provide a high-frequency urban trunk line. For example, trains from a number of East Berlin neighborhoods and Brandenburg suburbs interline to form the Stadtbahn: in the suburbs, they run every 10 or 20 minutes, but within the Ring, they combine to form a diameter running regularly every 3:20 minutes.
The integrated transfer timetable concept instead involves connecting different nodes at regular intervals, typically half an hour or an hour, such that trains arrive at every node just before a common time and leave just after, to allow people to transfer. In a number of major Swiss cities, intercity trains arrive a few minutes before the hour every 30 minutes and depart a few minutes after, so that people can connect in a short amount of time.
S-Bahn and ITT planning are both crucial tools for good rail service, but they conflict in major cities. The ITT requires all trains to arrive in a city around the same time, and depart a few minutes later. This forces trains from different cities to have different approach tracks; if they share a trunk, they can still arrive spaced 2-3 minutes apart, but this lengthens the transfer window. The idea of an S-Bahn trunk involves trains serving the trunk evenly, which is not how one runs an ITT.
Normally, this is no problem – ITTs are for intercity trains, S-Bahns are for local service. But this becomes a problem if a city is so big that its S-Bahn network grows to encompass nearby city centers. In New York, the city is so big that its shadow reaches as far as Eastern Long Island, New Haven, Poughkeepsie, and Trenton. Boston is smaller but still casts shadows as far as southern New Hampshire and Cape Cod.
This is why I don’t depict anything on Long Island on my map: it has to be treated as the extension of an S-Bahn system, and cannot be the priority for any intercity ITT. This is also true of Danbury and Waterbury: both are excellent outer ends for an electrified half-hourly regional rail system, but setting up the timed transfers with the New Haven Line (which should be running every 10 minutes) and with high-speed rail (which has no reason to stop at the branch points with either Danbury or Waterbury) is infeasible. In Boston I do depict some lines – see below on the complications of the North-South Rail Link.
The issue of NSRL
The North-South Rail Link is a proposed north-south regional rail tunnel connecting Boston’s North and South Stations. Current plans call for a four-track tunnel extending across the river just north of North Station, about 4.5 km of route; it should cost $4 billion including stations, but Massachusetts is so intent on not building it lies that the cost is $12 billion in 2018 dollars.
In common American fashion, NSRL plans are vague about how service is to run through the tunnel. There are some promises of running intercity trains in addition to regional ones; Amtrak has expressed some interest in running trains through from the Northeast Corridor up to the northern suburbs and thence to Maine. However, we are not engaging in bad American planning for the purposes of this post, but in good Central European planning, and thus we must talk about what trains should run and design the tunnel appropriately.
The rub is that Boston’s location makes NSRL great for local traffic and terrible for intercity traffic. When it comes to local traffic, Boston is right in the middle of its metropolitan region, just offset to the east because of the coast. The populations of the North Side and South Side suburbs are fairly close, as are their commuter volumes into Boston. Current commuter rail ridership is about twice as high on the South Side, but that’s because South Station’s location is more central than North Station’s. NSRL really is a perfect S-Bahn trunk tunnel.
But when it comes to intercity traffic, Boston is in the northeast corner of the United States. There are no major cities north of Boston – the largest such city, Portland, is a metro area of 600,000. In contrast, going south, New York should not be much more than an hour and a half away by high-speed rail. Thus, high-speed rail has no business running through north of Boston – the demand mismatch south and north is too high.
Since NSRL is greatly useful for regional traffic but not intercity traffic, the physical infrastructure should be based on S-Bahn and not ITT principles, even though the regional network connects cities quite far away. For one, the tunnel should require all trains to make all stops (South Station, Aquarium, North Station) for maximum local connectivity. High-speed trains can keep feeding South Station on the surface, while all other traffic uses the tunnel.
But on the North Side, feeding North Station on the surface is not a good idea for intercity trains. The station is still awkwardly just outside city center. It also offers no opportunity to transfer to intercity trains to the most important city of all, New York.
The only resolution is to treat trains to Portland and New Hampshire as regional trains that just go farther than normal. The Nashua-Manchester-Concord corridor is already as economically linked to Boston as Providence and Worcester, and there are plans for commuter rail service there already, which were delayed due to political opposition to spending money on trains from New Hampshire Republicans after their 2010 election victory. Portland is more speculative, but electric trains could connect it with Boston in around an hour and a half to two hours. These trains would be making suburban stops north of Boston that an intercity train shouldn’t normally make, but it’s fine, the Lowell Line has wide stop spacing and the intermediate stops are pretty important post-industrial cities. At Portland, passengers can make a timed connection to trains to Bangor, on the same schedule but with shorter trainsets as the demand north of Portland is much weaker.
On the map, I also depict Boston-Cape Cod trains, which like Boston-Concord trains are really suburban trains but going farther. Potentially, the branch to Cape Cod – the Middleborough branch of the Old Colony Lines – could even run through with the Lowell Line, either the branch to Concord or the Wildcat Branch to Haverhill and Portland. Moreover, the sequencing of the branches should aim to give short connections to Boston-Albany high-speed trains as far as reasonable.
The issue of the Northeast Corridor
The Northeast Corridor wrecks the ITT plan in two ways, one substantial and one graphical.
The snag is that there should be service on legacy track running at a maximum speed of 160-200 km/h in addition to high-speed service on high-speed tracks. There may be some track sharing between New York and New Haven to reduce construction costs, using timed overtakes instead of full track segregation, but east of New Haven the high-speed trains should run on a new line near I-95 to bypass the Shore Line’s curves, and the Shore Line should be running electric regional trains to connect to the intermediate cities.
The graphical problem is that the distance between where the legacy route is and where the high-speed tracks should be is short, especially west of New Haven, and depicting a red line and a blue line together on the map is not easy. I will eventually post something at much higher resolution than 1 pixel = 500 meters. This also affects long-distance regional lines that I’d like to depict on the map but connect only to legacy trains on the Northeast Corridor, that is the Danbury and Waterbury Branches.
For planning purposes, figure that both run every half hour all day, are electric, run through to and beyond New York as branches of the New Haven Line, and are timed to have reasonable connections to high-speed trains to Albany and points north in New York. Figure the same for trains between New Haven and Providence, with some additional runs in the Providence suburbs giving 15-minute urban frequencies to such destinations as Olneyville and Cranston.
The substantial issue is that the Northeast Corridor is far too high-demand for a half-hourly ITT. Intercity trains run between New York and Boston better than hourly today, and that’s taking twice as long as a TGV and charging 2.5-4 times as much. My unspoken assumption when planning how everything should fit together is that there should be a 400-meter long train every 15 minutes on the corridor past New Haven, spaced evenly around Boston to overtake regional trains to Providence at consistent locations. Potentially, there should be more local trains taking around 1:50 and more express trains taking around 1:35, and then all timed transfers should be to the local trains.
On the New Haven Line, too, regional rail demand is much more than a train every half hour. Trains run mostly every half hour today, with management that is flagrantly indifferent to off-peak service, and trip times that are about 50% longer than they should be. Nonetheless, best practice is to set up timed transfers such that various branches all connect to the same train, so that passengers can connect between different branches. This mostly affects Waterbury; it’s useful to ensure that Waterbury trains arrive at Bridgeport with a short transfer to a train toward New Haven that offers a quick connection to trains to points north and east.
Planning HSR around timed connections
Not counting lines that are in the Boston sphere, or the lines around Albany, which I discussed two weeks ago, there are three lines proposed for timed connection to high-speed rail: New London-Norwich, Providence-Worcester-Fitchburg, Springfield-Greenfield.
All three are regional lines, not intercity lines. They are not optimized for intercity speed, but instead make a number of local urban and suburban stops. This is especially true of Springfield-Northampton-Greenfield, a line that Sandy Johnston and I have been talking about since 2014. A Springfield-Greenfield line with 1-2 intermediate stops might be able to do a one-way trip in around 39 minutes, at which point a 45-minute operator schedule may be feasible with a very tight turnaround regime – but there’s enough urban demand along the southern half of the route that adding stops to make it about 50 minutes with a one-hour operator schedule is better.
The Providence-Worcester line is likewise slower than it could be if it were just about Providence and Worcester. The reason is that high-speed rail compresses distances along its route. Providence-Boston by high-speed rail is about 22 minutes nonstop, including schedule contingency. Boston-Worcester is about the same – slower near Boston because of scheduling difficulties along the Turnpike and the inner Worcester Line, faster near the outer end because Worcester has no chance of getting a city center station but rather gets a highway station. Now, passengers have a range of transfer penalties, and to those who are averse to connections and have a high personal penalty, the trip between the two cities is more attractive directly than via Boston. But there are enough passengers who’d make the trip via Boston that the relative importance of intermediate points grows: Pawtucket, Woonsocket, Uxbridge, Millbury. In that situation, the importance of frequency grows (half-hourly is a must, not hourly) and that of raw speed diminishes.
The onward connection to Fitchburg is about three things. First, connecting Providence with Fitchburg. Second, connecting Worcester with Fitchburg. And third, connecting Fitchburg with the high-speed line. This makes investments into higher speed more valuable, since Fitchburg’s importance is high compared with that of points between Worcester and Fitchburg. The transfer between the line and high-speed rail should be timed in the direction of Fitchburg-to-Albany first of all, and Providence-to-Albany second of all, as the connections from the endpoints to Boston are slower than direct commuter trains.
The presence of this connection also forces the Worcester station to be at the intersection with the line to Providence. Without this connection, it may be better to site the station slightly to the west, at 290 rather than 146, as the area already has Auburn Mall.
Finally, the New London-Norwich line is a pure last-mile connector from the New London train station, which is forced to be right underneath the I-95 bridge over the river, to destinations to the north. The northern anchor is Norwich opposite the historic center, but the main destination is probably the Mohegan Sun casino complex. Already there are many buses connecting passengers from New York to the casino. The one-way trip time should be on the order of 21-22 minutes, but with a turnaround it’s a 30-minute schedule, and the extension south to the historic center of New London is for completeness; with a timed connection, trains could get between Penn Station and Norwich in around 1:20 counting connection time, and between Penn Station and Mohegan Sun in maybe 5 minutes less.
What about Vermont?
Vermont’s situation is awkward. Burlington is too far north and too small to justify a connection to high-speed rail by itself. A low-speed connection might work, but the line from Burlington south points toward Rutland and not New York, and connecting it onward requires reversing direction. If Vermont had twice its actual population this might be viable, but it doesn’t.
But Vermont is right between New York and Montreal. I generally don’t show New York-Montreal high-speed rail on my maps. It’s a viable line, but people in both cities severely overrate it, especially compared with New York-Toronto; I have to remind readers this whenever I write about international high-speed trains. In the event such a line does open, Burlington is the only plausible location for a Vermont stop – everything else is too small, even towns that historically did have rail service, like Middlebury. Rutland could get a line running partly on high-speed track and partly on legacy track taking it down to Glens Falls or Saratoga Springs to transfer to onward destinations, or maybe Albany if trains run 2-3 minutes apart in pairs every 30 minutes.
Current plans for Vermont try to connect it directly to Boston via New Hampshire, and that is wrong. The Vermonter route is mountainous from Greenfield to Burlington; trains will never be competitive with driving there. Another route under occasional study going into Boston from the north was even included on a 2009 wishlist of high-speed rail routes, under the traditional American definition of high-speed rail as “train that is faster than a sports bicycle.” That route, crossing mountains in both New Hampshire and Vermont, is even worse. The north-south orientation of the mountains in both states forces east-west routes to either stick to the lowlands or consolidate to strong enough routes that high-speed rail tunnels are worthwhile.
How much does this cost?
As always, I am going to completely omit the Northeast Corridor from this cost analysis; an analysis of that will happen later, and suffice is to say, the benefit-cost ratio if there’s even semi-decent cost control is extremely high.
With that in mind, the central pieces of this program are high-speed lines from Boston to Albany and from New Haven to Springfield, in a T system. The 99 km New Haven-Springfield line, timetabled at 45 minutes including turnaround and maybe 36 minutes in motion, is on the slow side for high-speed rail, as it is short and has a crucial intermediate station in Hartford. It does not need any tunnels or complex viaducts, and property takings are nonzero but light; the cost should not be higher than about $2-2.5 billion, utilizing legacy track for much of the way.
The Boston-Albany line is much costlier. It’s 260 km, and crosses the aforementioned north-south mountains in Western Massachusetts. Tunnels are unavoidable, including a few kilometers of digging required just west of Springfield to avoid a slowdown on suburban curves. At the Boston end, tunneling may also be unavoidable next to the Turnpike. The alternative is sharing a two-track narrows with the MBTA Worcester Line in Newton; it’s possible if the trains run no more than every 15 minutes, which is a reasonable short-term imposition but may be too onerous in the longer term if better service builds up more demand for commuter rail frequency in Newton. My best guess is that without Newton, the line needs around 20 km of tunnel and can piggyback on 35 km of existing lines at both ends, for a total cost in the $6-8 billion range. This figure is sensitive to whether my 20 km estimate is correct, but not too sensitive – at 40 it grows to maybe $9 billion, at 0 it shrinks to $4.5 billion.
Estimating the costs of the blue lines on the map is harder. All of them are, by the standard of high-speed rail, very cheap per kilometer. A track renewal machine on a one-third-in-tunnel German high-speed line can do track rebuilding for about a million euros per single-track-kilometer. All of these lines would also need to be electrified from scratch, for $1.5-3 million per kilometer. Stations would need to be built, for a few million apiece. My first-order estimate is $1 billion for the three blue connector lines and about the same for Boston-Portland-Bangor; the Hyannis and Concord lines would go in a regional rail basket. The NSRL tunnel should be $4 billion or not much more, and not what Massachusetts wants voters to believe it is to justify its decision not to build it.
The reason for the relatively limited map (e.g. no Montreal service) is that these lines are not such slam dunks that they’re worth it at any price. Cost control is paramount, subject to the bare minimum of good service (e.g. electrification and level boarding). For what I think a fair cost is, those lines are still good, providing fast connectivity across New England from most places to most other places. Moreover, the locations of the major nodes, like Worcester and Springfield, allow timing bus interchanges as well, providing further connections to various suburbs and city neighborhoods.
The red high-speed lines are flashy, but the blue ones are important too. That’s the key takeaway from planning in Switzerland, Austria, and the Netherlands, all of which have high rail usage without great geography for intercity rail. Trains should be planned coherently as a network, with all parts designed in tandem to maximize connectivity. This isn’t just about going between Boston and Springfield or Boston and Albany or New Haven and Springfield, but also the long tail of weaker markets using timed connections, like New Haven-Amherst, Brockton-Worcester, Dover-Providence, Stamford-Mohegan Sun, and so on. A robust rail network based on ITT design principles could make all of these and many more connections at reasonable cost and speed.
While researching my previous post about nuclear power, I found various sources about the construction costs of renewable electricity. They all point to the same conclusion: the installation costs of solar and wind power took a nosedive in the 2010s. By now they are down to €1-1.50 per watt for onshore wind, €1.50-2.50 per watt for offshore wind (PDF-p. 24) and around $1.10 per watt for utility-scale solar power (PDF-pp. 7, 50-51). The levelized cost of energy (LCOE) for onshore wind and solar power is, depending on source, 4-7 cents per kWh (see Fraunhofer, Lazard, IRENA), at which point it’s cheaper than new coal and gas without subsidies or carbon taxes and cheaper than existing coal and gas with mild carbon taxes. Intermittency is still an issue, but at continental scale it is much more of an irritant than a serious impediment. Decarbonization of electricity is substantially a solved problem.
The problem is that decarbonization of transportation is not a solved problem.
The world of 2020 is not that of 2000. The greenhouse gas inventory of 2020 is not that of 2000. In developed countries, electricity is down, and transportation is up or at best flat. In developing ones (i.e. China and India) the situation is different, but there too there’s growing public concern with coal power pollution, to the point that one of the premier sites for information about air pollution levels, AQICN, is Beijing-based. Nonetheless, cars remain aspirational in those countries, despite high levels of investment in urban and intercity rail transportation (periodic reminder: about two-thirds of global high-speed rail ridership is in China).
The problem with transportation is cars.
Cars are not getting better. There’s a growing but very small share of the market for new cars that is electric; so far production costs remain high, and there are real long-term issues with rare earth metals used in the batteries. Costs are inching down, but it’s firmly in “more research is needed” territory. And meanwhile, in the carbon tax-free developed world (i.e. the US) the vast majority of cars that are not electric are getting bigger and less fuel-efficient. American transportation emissions peaked in the mid-2000s and fell as fuel prices rose, but now that fuel is cheap they’re rising, faster than population (source, PDF-p. 32).
The problem is partly, but not only, the United States.
Whatever historical causes made Americans this way, American culture of the early 21st century is still one that thinks it’s normal to want every American to have an SUV and deviant to want every American to have an apartment in a big city with a good subway system. This mentality cuts across classes, parties, subcultures, and states, and was recently affirmed in California, particularly Los Angeles. Decarbonizing solutions on the road that may be popular in segments of the tech industry, like electric cars and mobility as a service, are still brushing against a culture that equates the size of one’s car and engine with one’s moral worth. One of two things will happen: this culture will vanish, or hundreds of millions of people in countries Americans can’t find on a map will be inundated.
But it’s not just the United States.
Public transportation usage in Europe is increasing, unlike in the United States – and it’s increasing from an already nontrivial base. But it’s not increasing fast enough. The same motorist culture of the United States exists in a more attenuated form here. There are high fuel taxes and they help a lot – SUVs and pickup trucks are rare, and vehicle-km per capita are maybe half what they are in the US – but the difference between greenhouse gas emissions of 17 and 8 metric tons-CO2 per capita is one of whether catastrophic climate change will happen soon or shortly later.
It’s a priority for Europe specifically, precisely because it’s in uncharted territory.
The United States needs to learn to imitate and get from 17 to 8 on its way to 0, but Europe needs to get from 8 to 0 and to figure out how to do so. Switzerland and the Netherlands are already at the forefront of improving mainline rail, and yet have widespread auto usage in local as well as interregional travel.
There are a number of headwinds to the replacement of cars with public transportation, all of which are politically or technically nontrivial in ways that mass installation of solar and wind power isn’t:
- Public transport is the most convenient in large cities and least convenient in rural areas, but modern nationalism holds the rural to be more authentic and moral. Thus, when rural motorists riot the state is paralyzed with inaction and the media urges understanding of populist anger at elites, whereas when urbanites riot the state immediately engages in mass arrests and the media urges law and order.
- The pace of urban redevelopment is too low, thanks to local NIMBYism, making it hard for people to live in cities where car-free living is already convenient. Local housing activism always focuses on people already present; Berlin passed a new rent control law that is projected to reduce investment by 25%. Even Paris, which is building more housing, is doing so almost exclusively in the suburbs and not in the city proper.
- Local notables tend to drive even controlling for income and social class. One does not become a local notable by working at a city center office with people from many neighborhoods, many of whom are recent migrants to the city, but by staying within one neighborhood and interacting with old-timers. The latter kind of economic and social network is less convenient to travel by train. Thus, the loudest voices in a local discussion are against seizing space from cars and giving it to pedestrians, cyclists, buses, or trams.
- At low levels of public investment, the car will predominate, for two reasons. First, some state action is needed to give buses priority on roads. Second, public transportation has more moving parts that must be integrated – fares, schedules, infrastructure, equipment, development. This makes fiscal austerity a drag on the ability of a developed society to demotorize unless this austerity specifically takes the form of very high taxes on cars and fuel.
- A political process that slows down investment in order to mollify NIMBY opposition makes it very hard to shift priorities on the ground. In this sense, the freeway revolts and the changes they led to are the best thing that ever happened to car culture, even more than the freeways themselves; in the American context, the revolts happened largely only when the freeways intruded on middle-class neighborhoods.
These headwinds are phrased politically, but all have various technical components, like construction costs for new rail lines, public transport network design, interagency cooperation issues that are too far removed from mass politics to be truly political, etc. Is the problem solvable? Most likely, yes. It’s not only in the biggest cities in the world that public transport usage is high; getting Stockholm, Vienna, Zurich, and so on to demotorize is within the realm of possibility, and getting other cities to have what those cities have is as well.
But “within the realm of possibility” is not a statement of utmost confidence. It’s a difficult program, one where failure is regrettably an option. Every aspect is hard: convincing governments that don’t like spending money on mobile people to invest more in rail, raising taxes on fuel and on cars, building more housing in the cities, reclaiming street space from cars, improving the quality of public transport service, improving connections between lines. It all takes money, and though the required subsidies may well fall with better technology and higher usage, the most optimistic view is that public transportation now is like wind and solar power in 2010, when they was still an economic gamble, and not what they are today that the gamble is paying off.
The TGV network put France at the forefront of European intercity rail technology for decades. Early investments, starting in 1981 with high-speed tracks between Paris and Lyon, led to explosive growth in ridership throughout the 1980s, 90s, and 2000s. But since then, usage has stagnated. Domestic ridership in 2009 and 2010 was 100 million; so was domestic ridership in 2016, on a larger network. There was a 10% increase in 2017 when the line to Bordeaux opened, but in 2018 ridership stagnated again. In the late 2000s, there was more ridership on the TGV than on the intercity trains in Germany; now, German intercity trains approach 150 million annual riders, and are not far behind the TGV in passenger-kilometers, Germany running slower trains and thus averaging shorter trips.
I’ve heard a number of different explanations for why TGV ridership has not increased in the last ten years, many of which involve management; I, too, complain about managers who are recruited from the airline industry. But I submit that there’s a deeper, conceptual reason: the TGV is only workable for thick markets, mostly connecting Paris with a major provincial city. Trains run mostly nonstop, and there is no seat turnover. From the 1980s to the late 2000s, ridership rose as more cities were connected to Paris, but then those markets were mostly saturated, and new markets cannot be served adequately.
The TGV hit a wall about ten years ago. This is important, because as the busiest high-speed rail network outside of China and Japan, it has a lot of cachet. Politicians and rail planners propose programs that look much like the TGV network. This is of especial importance in the United Kingdom, which is replicating the TGV’s operating paradigm with the under-construction High-Speed 2 project; in the United States, the geography of the Northeast Corridor has meant that plans look more like the Japanese paradigm, which works better both in general and in the Northeast’s specific context.
In Japan, Germany, and the United States (by which I mean the Northeast Corridor), trains stop at many major cities on one route.
The fastest Shinkansen trains between Tokyo and Shin-Osaka have always stopped at Nagoya and Kyoto. Tokyo-Osaka passengers ride end to end, but many riders go between Tokyo or Osaka and Nagoya, so the seat turns over. Some of these trains continue west to Hakata, with such intermediate stops as Okayama and Hiroshima. The upshot is that the trains don’t just connect these cities to Tokyo, but also to one another. The size of Tokyo means there is demand for very high frequency to Shin-Osaka and decent frequency to Hakata; passengers on intermediate city pairs like Nagoya-Okayama or Kyoto-Hiroshima benefit from infrastructure that those city pairs could never justify on their own.
In Germany, intercity trains generally serve more than two major cities too. Like in France and unlike in Japan and the US, some major cities have stub-end stations, most notably Frankfurt; trains do not skip these cities, but rather serve them, reverse direction in about 5 minutes, and continue. Passengers may reserve seats but do not have to do so, so each seat has an electronic display showing for which portion of the trip it is free for the use of any passenger with an unreserved ticket.
France works by a different principle. Paris, Lyon, and Marseille are collinear, but trains do not serve all three cities. Trains from Paris to Lyon do not continue to Marseille; trains from Paris to Marseille rarely stop at the Lyon airport and never stop at Lyon Part-Dieu, which is on a branch from the Paris-Marseille mainline. There are separate trains between Lyon and Marseille, running generally hourly. Hourly frequency is workable on a line that takes about 1:40 end to end, but is not great.
At least Lyon and Marseille are on the same line coming out of Paris. Trains between Lyon and Lille, 3-3.5 hours apart on opposite sides of Paris, have service gaps of 2-2.5 hours most of the day. Lyon-Strasbourg trains on the LGV Rhin-Rhône lose money – the two cities alone do not have the ridership to fill trains, and there are no transfers with other cities nor larger intermediate cities than Mulhouse.
It’s too late for Paris 21
Berlin Hauptbahnhof is a through-station with service to cities all over Germany; every intercity train to Berlin serves Hauptbahnhof, regardless of which direction it comes from. This is common elsewhere in Germany, too. The second most important stub-end station, Stuttgart, is currently being replaced with an underground through-station at great cost, in a controversial project called Stuttgart 21. The most important, Frankfurt, long had plans for a similar through-station dubbed Frankfurt 21, and recently the federal government announced new plans for such a project.
Paris could have built a Paris 21, or Paris Hauptbahnhof, in the 1970s or 80s. When the city designed the RER, it ripped up Les Halles to build the Chatelet-Les Halles transfer point. The station is palatial: 25 meters underground, with 7 tracks and 4 platforms, 2 of which are 17 meters wide. This was so expensive that the Auber-Nation segment of the RER A, consisting of 6 km of tunnel and the Chatelet-Les Halles and Gare de Lyon RER stations, cost in today’s money around $750 million per km, a record that is yet to be surpassed in a non-English-speaking country.
Planning for the TGV only began in earnest in the late 1970s; the RER was constructed in the late 1960s and 70s, Les Halles opening in 1977. Perhaps the initial omission of intercity tracks was understandable. But the RER D opened in the early 1990s, and by then SNCF should have known it would have a national TGV network. It could have at the very least spent some money on having 2 platforms and 4 tracks at Les Halles dedicated to intercity trains, running through from Gare du Nord to Gare de Lyon. But it didn’t, and now there’s so much regional traffic that repurposing any part of Les Halles for intercity trains is impossible. Moreover, given the cost of the station in the 1970s, a future Paris 21 project would be unaffordable.
The TGV has to live with infrastructure decisions made 30 years ago. Given this reality, some of the kludges of the system today are understandable. And yet, even in outlying areas, there are no scheduled connections with either other TGVs or regional trains. Paris-Nice TGVs are timed to just miss TERs to Monaco and Ventimiglia. The Mâcon TGV station is located at just the wrong place for a transfer to a future extension of the LGV Rhin-Rhône south to Lyon. Other than Part-Dieu and Lille-Europe, major secondary cities do not have urban stations designed for through-service.
The contrast here is partly with German or Japanese practice: Japan built Shin-Osaka to enable through-service from east to west of Osaka without spending too much money tunneling into city center, and Germany serves Kassel at Wilhelmshöhe instead of at Hauptbahnhof since Hauptbahnhof is a stub-end station.
But the contrast is even more with the practice of smaller European countries. Switzerland and the Netherlands do not have anything as voluminous as Paris-Lyon, so they had to design their intercity rail networks around everywhere-to-everywhere travel from the start. Switzerland, too, had much less growth in the 2010s than in the 2000s, but ridership and p-km both grew, and are continuing to grow. What’s more, Switzerland has not tapped out its strongest markets: Lausanne, Luzern, and Geneva are still poorly integrated into the national timed transfer plan.
Getting it right from the start
France boxed itself into a corner. Its high-speed rail infrastructure is designed to connect provincial cities to Paris but not to one another. In some places, it’s possible to retrofit something more usable with the construction of new transfer points and the planning of better timetables. But elsewhere, as in Paris, it is too hard. This suggests that other countries that look to France as a model learn not only from the success of the TGV but also its more recent failures, and get it right from the start.
Any of the following lessons are useful to Britain and to other countries that are building large high-speed rail networks:
- Try to limit branching, to make sure city pairs have adequate frequency. This is especially important on shorter city pairs, such as London-Birmingham, planned to take 38 minutes, and Birmingham-Manchester, planned to take 40 minutes. Adding a few minutes to the trip time of through-trains is fine if it makes the difference between hourly and half-hourly frequencies, or even half-hourly and quarter-hourly frequencies.
- Place stations at good points for transfers to other trains. This includes trains on the same network, for which the best locations are branch points, and legacy trains, for which the best locations are major legacy stations and junctions. For example, the largest cities of the East Midlands – Nottingham, Derby, and Leicester – lie on a Y-shaped system, so it would be valuable to place a hub station at the node of the Y; the currently planned East Midlands Hub is 3.5 km north of the node, not on the leg of any of the three main cities.
- If there is a major city with service going in multiple directions, make sure it has a single through-station, even if constructing one requires a new tunnel. This is less relevant to Britain, since London is at the south end of the network, but is relevant to Italy, which needs to convert multiple urban terminals into through-stations, and Spain, which is doing so at Madrid and Barcelona already, at a fraction of the cost of Stuttgart 21.
- At short range, run trains as fast as necessary – that is, spend a lot of resources on getting trip times between major nodes to be just less than an hour, half an hour, or an hour and a half, but don’t worry too much about 55 vs. 40 minutes in most circumstances. This way, passengers can interchange at major nodes in a short time.
For a generation, the TGV was the envy of the rest of Europe. But it tapped out the strong markets that it was designed around, and now SNCF has its work cut out for it adapting to the needs of other city-to-city travel markets. Other big countries had better take heed and do it right from the start to avoid boxing themselves the way France did.
If Swiss planners were hired to design an intercity rail network for New York State, they might propose something that looks like this:
The trip times depicted on the map are a few minutes longer than intended, especially next to a terminus station like Niagara Falls, Watertown, and Ithaca. The depicted times are inclusive of turnaround time: the 45-minute Buffalo-Niagara Falls line is intended to take around 35 minutes in actual service, with 10-minute turnarounds.
Swiss planning is based on hourly and half-hourly timetables repeating all day on a clockface pattern: if a train leaves your station at 8:24 am, a train will leave your station at xx:24 all day, and if the line runs every half hour then also at xx:54. Moreover, at major nodes, trains are timetabled to arrive a few minutes before the hour and depart a few minutes after, letting passengers connect between different trains with minimal wait. To minimize transfer time and turn time, trains run as fast as necessary – that is, the state invests in higher-speed lines to ensure connections between major cities take a few minutes less than an hour. The Bahn 2000 program set up connections between Zurich, Basel, and Bern taking just less than an hour, with a few further connections elsewhere taking just less than an integer number of half-hours; the Bahn 2030 program aims to do the same with more cities all over the country.
The above map is an adaptation of the concept to New York State. I hope the explanation of how to adapt Switzerland to New York will be of interest to rail advocates elsewhere – the differences between the two geographies matter elsewhere, for example in Germany, France, or Sweden, or for that matter in California or New England.
There is no high-speed rail in Switzerland, unless one counts the mixed passenger and freight rail tunnels through the Alps, which allow 250 km/h passenger trains. The Bahn 2030 planning calls for a 2-hour trip time between Zurich and Lugano, a distance of about 170 km, even with heavy tunneling under all significant mountains; with so much tunneling, 1.5-hour trips are easy and even 1-hour trips are feasible with a bypass around Zug. Clearly, even when higher speeds are allowed, Swiss planning sticks to low- and medium-speed rail, targeting an average speed of about 120 km/h.
This works for Switzerland, a small country in which even Geneva is only 2:45 from Zurich. In New York, it does not. At the speed of upgraded legacy rail, comparable to the Northeast Corridor, the links on the above map along the high-speed spine would take 2 hours each rather than an hour. New York-Buffalo trains would take 6 hours, too long for most travelers, and New York-Rochester would take 5 hours, which is marginal at best. Trains doing New York-Albany in 2 hours could get fairly popular, but even that is long enough that cutting it to just less than an hour is feasible.
Trains are to run every half hour, with the exception of urban lines, namely Buffalo-Niagara Falls, Albany-Troy-Mechanicsville, and Utica-Rome, which run every 15 minutes. The reason for the half-hourly frequency is that all lines need it for either capacity or ridership. The lines either run to New York, which is so big it can easily fill a train every half hour and perhaps even every 15 minutes, or are quite short, so that running only every hour reduces ridership and it’s better to run shorter trains every 30 minutes.
With half-hourly timetables, a stub-end line can take an integer number of quarter-hours and not just half-hours. For example, Syracuse and Albany should have a pulse at :00 and :30 every hour. This in turn means that trains from Albany to Glens Falls can take 1:15, departing Albany just after :00 and :30, arriving at Glens Falls just before :15 and :45, turning back toward Albany just after :15 and :45, and then returning to Albany just before :00 and :30.
The only worry with quarter-hour trip times is that every cycle must sum up to an integer number of half-hours, not quarter-hours. Otherwise, some connections are broken, offset by 15 minutes. Thankfully, the only cycle on this map is New York-Albany-Syracuse-Binghamton-New York, which takes 7 hours.
Syracuse regional rail
Syracuse is depicted as having the most expansive regional rail network in the state, despite being the smallest of Upstate New York’s four major metropolitan areas. The reason is that the goal of the planned network is to provide intercity rather than local service. Rochester has some useful urban lines, for example to Freeport or northwest to the lakefront, but they are so short that they should run every 10 or 15 minutes and not every half hour. However, Rochester has no significant independent towns within an hour or so by rail, and thus there are no timed connections there. In contrast, Syracuse is located right between Watertown, Oswego, Auburn, and Cortland with its connection onward to Ithaca.
The Syracuse system is intended to be fully on the RegionalBahn side of the S-Bahn vs. RegionalBahn divide. The shared segment between Syracuse and the split between the lines to Oswego and Watertown is not meant to overlay to run frequent urban service. Instead, trains should tailgate, followed by a gap of nearly half an hour. Syracuse-bound trains may well call at Liverpool at :20 and :22, arriving at Syracuse at :25 and :27 to exchange passengers with other trains and then continue south, one of Oswego and Watertown paired with Cortland and Binghamton and the other terminating. If north-south S-Bahn service is desired, trains should be slotted in between the intercity trains.
The map depicts greenfield alignments for the high-speed line except on the approaches to New York and Toronto, and legacy alignments for the low-speed lines.
As in Switzerland, the low-speed lines do not necessarily slavishly adhere to legacy alignments. However, the deviations are not the same. Switzerland uses bypasses and tunnels to speed lines up. In New York, the main mechanisms to speed up lines are electrification, track renewal, and higher superelevation. Tunnels are too expensive for the population density of Upstate New York. I can see some bypasses, potentially getting Syracuse-Cortland and Cortland-Binghamton down to 30 minutes each, but none of the Upstate cities off the high-speed line is big enough to justify major civil works.
The one depicted bypass on a blue-colored line is the use of the Boonton Branch in New Jersey to offer an express bypass around the Morristown Line with its dense station spacing. This requires some additional tracks on busy urban regional lines as well as a short tunnel in Paterson, but New York is big enough that investing in faster service to Dover, Delaware Water Gap, and Scranton is worth it.
Upstate, the important deviations involve restoring old tracks, including between Cortland and Ithaca and within some town centers. Corning and Glens Falls both have disused rail alignments serving their centers better than the existing freight lines. But most importantly, Syracuse has an underused freeway running east-west through its center, which I am assuming replaced with a rail line. This is not a new idea – Syracuse is already removing a branch of the freeway, which should be used for a rail connection toward Binghamton, and even the mainline is a vestige of when midcentury planners thought Upstate cities would keep growing. The current Syracuse station is at an inconvenient location, making rail realignment a good use of the right-of-way.
New York State is much more integrated with its neighbors than Switzerland – it’s all the same country. There is extensive interstate travel, and rail planning must accommodate this. Forget the Deutschlandtakt – an Americatakt would be the most complex rail plan in a developed country out of sheer size. Thankfully, the connections depicted on the New York State plan accommodate interstate travel fairly well.
Going east, there are connections to Vermont, Massachusetts, and Connecticut. Albany-Boston can be done in around an hour, which makes for a half-hour takt connection between Albany and Springfield and 45 minutes minus turnaround between Springfield and Boston. Springfield-New Haven is 30 minutes by high-speed rail or 45 minutes by fast legacy rail, both with a stop at Hartford and few to no others; Springfield can then get its own small regional rail line toward Northampton (with some urban overlays for an S-Bahn) and Greenfield. Vermont can get a slow line to Rutland, and/or a fast line to Burlington continuing to Montreal; thence New York-Montreal and Boston-Toronto trains can be timed to connect at Albany, with New York-Toronto trains slotted in between, timed to connect only to the more frequent urban lines like Buffalo-Niagara Falls.
Going south, New York is separated from Pennsylvania by the northern reach of the Appalachians, called the Southern Tier in New York and the Northern Tier in Pennsylvania. This area had many coal mines in the 19th century and as a result has many legacy rail lines, but they are curvy and connect villages. But Scranton is a significant city on a nice line with Allentown and Philadelphia; unfortunately, the Philadelphia-Allentown line stretches via Reading and the Allentown-Scranton line is hilly and curvy, justifying some greenfield construction with some tunneling near the northern end.
Finally, going west, the I-90 route serves Erie and the Midwest. But this is a plausible high-speed rail connection toward Chicago, and so no low-speed interface is needed within the state. Erie could get a line to Youngstown and Pittsburgh, but it would be slower than connecting between high-speed trains in Cleveland; the largest city between Erie and Youngstown is Meadsville, population 13,000.
The cost of the high-speed spine is considerable, but if New York can keep it to the level of France (around $25 million/km), or even Germany (around $35 million/km), the benefits should exceed the costs. New York is huge, and even though nothing in Upstate New York is, the combined populations of Syracuse, Rochester, and Buffalo would add up to a big French or German city. And then there is Toronto at the other end, anchoring everything.
The low-speed lines should be quite cheap. Track renewal in Germany is around $1 million per single-track kilometer; at the frequency envisioned, all the low-speed lines can stay single-track with passing segments. Electrification is maybe $1.5 million per kilometer in Israel, despite a lawsuit that delayed the project by three years.
Is this feasible?
Technically, all of this is feasible. Good transit advocates in the Northeastern United States should push elected officials at the federal and state levels to quickly plan such a system and aim to begin construction early this decade. Bahn 2000 was supposed to take the 1990s to be built, but was delayed to 2004; this is a bigger program but can still happen by 2030 or so.
The trip times, frequencies, and coverage chosen for the map are deliberately conservative. It’s possible to squeeze higher speed at places, and add more branches to smaller towns, like Rochester-Niagara Falls or Buffalo-Jamestown. Bahn 2000 is followed up with Bahn 2030 or Bahn 2035, and likewise rail improvements can accrete in the United States. But as a starter system, this is a solid network connecting all large and nearly all small cities in New York State to one another with maximum convenience and minimum hassle. I hope state planners take heed and plan to invest soon.
In 2011, I wrote a post arguing that international links underperform. I gave examples, using many links nearly all of which have rotted in the 8.5 years since, showing that the ridership on various air and rail city pairs was lower if they were in two different countries than if they were in the same country. The most important example is Eurostar, connecting London and Paris. Eurostar has 11 million passengers per year, of which a growing minority go between London and cities other than Paris, like Brussels and Amsterdam. In contrast, the TGVs from Paris to the southeast have 44.4 million annual passengers; the major secondary cities on the line combine to about half of London’s population. The newly-opened LGV Sud-Europe-Atlantique has 6 million annual passengers on the Paris-Bordeaux city pair alone – and Bordeaux has an order of magnitude fewer inhabitants than London.
My assumption was always that Eurostar’s problem is that it connects two distinct countries, speaking two different languages. Thus, similar international connections, like oft-mooted proposals for high-speed rail between New York or Boston and Montreal or even between New York and Toronto, are likely to severely underperform domestic ones. This is not too relevant to the United States, which is not building high-speed rail of any kind, but is increasingly relevant to Europe, which is slowly building international links. But what if the assumption that the important aspect is the national or linguistic border is incorrect? What if there are other issues on Eurostar and various international air links, which national railroads can choose to solve if they care?
The issue of fares
I fired up Eurostar and SNCF’s sites and looked for tickets departing Tuesday in 13 days. I got 14 trains from London to Paris, charging fares ranging from €52.50 to €144.50. The average is €91.46, and the median is €98.50. From Paris to Lyon, I got 22 regular TGVs (“InOui”), charging €45-97, with an average of €84.63 and a median of €97 – but I also got 5 OuiGo trains, charging €10-25, all but one leaving from Gare de Lyon rather than Marne-la-Vallée with its difficult RER transfer.
On city pairs where SNCF expects more competition than Paris-Lyon, fares are lower, even when trips are longer. Paris-Marseille has 15 regular InOui departures and 8 OuiGos; the InOuis charge €49-79 with an average of €57.33, and the OuiGos charge €10-28, half serving Paris proper and half leaving from Marne-la-Vallée. The OuiGo services overall are unprofitable, but the InOuis aren’t – the Spinetta Report claims the fully-laden cost of TGV service is €0.07/seat-km, and seat utilization is very high (too high, in fact – it’s at the expense of off-peak frequency).
The other international service using the LGV Nord, Thalys, charges high fares as well, if less high than Eurostar. The site shows me 18 departures from Paris to Brussels on the 4th of February; one has a €29 ticket, but the others state that cheap ticket is sold out and offer me €66-99 tickets and one is entirely sold out. Going to Amsterdam, there are 10 departures, charging €98-135. To Cologne, the final of the major cities served by Thalys, there are 5 departures, one with a cheap €35 ticket and the rest charging €76-122. Thalys has 7.5 million annual riders, roughly within the same range relative to metro area population one would expect from Eurostar, depending on what one counts as the metro areas of Rotterdam, Amsterdam, and Cologne.
I compare Eurostar and Thalys with domestic TGVs not just out of convenience. SNCF owns a majority stake in Eurostar and Thalys. The yield management systems are likely similar, making a comparison of trips on the same day reasonable. In contrast, I would not want to do such a comparison with, say, the Shinkansen, which has no yield management at all and charges the same fare for the same class of seat and train speed.
The consequence of high fares
It’s quite likely, then, that the low ridership on Eurostar is connected with its high fares. Once tickets are expensive enough to discourage price-conscious customers, the ridership profile consists of price-insensitive travelers, making it possible to keep escalating fares.
A 2009 study by Christiaan Behrens and Eric Pels on air-rail competition in the London-Paris market finds that in a nested logit model, Eurostar travelers have a price elasticity of -0.14 to -0.15, compared with about -0.43 out of Heathrow on BA for businesses travelers and -0.77 for leisure travelers. The study compares different airlines and airport choices, with most of the market in the 2000s using Heathrow and either BA or Air France, with Air France having higher elasticity. In a mixed logit model, fare elasticities are all much higher, but Eurostar is still much more inelastic than flying, around -0.50 vs. -1 for business flyers and -2.5 for leisure flyers.
The second link in this post mentions growth in American tourist travel as a reason for Eurostar’s recent growth in ridership. It is not surprising that foreign tourists who paid high fares to travel to Europe and are staying in expensive hotels are a significant source of revenue to Eurostar. Presumably American tourist travel on domestic TGVs is up too, but it is far less significant, first because no secondary French city has the tourism of Paris and London except for Nice, 5.5 hours from Paris by train, and second because the domestic market is strong enough that American tourists are barely a blip on the radar.
Regardless, the elevated American tourist numbers present a peril to the state of the American discourse on the subject, even if they generate much-needed revenue for SNCF. Those tourists then come back to the US talking up the convenience of high-speed rail, or at least the version of it with security theater and passport checks, but bemoan the high ticket prices.
We already see what happens when train trips are priced for the top of the market in the United States. Fares per the 2016 annual report, the most recent one to include this data (on PDF-p. 41), average $0.58/p-km on the Acela and $0.30/p-km on the Regional; per an ARAFER report using 2016 data, the corresponding number for domestic TGVs is €0.10/p-km (PDF-pp. 15, 26). With Amtrak’s cheaper trains charging 2.5 times as much as the TGVs, price-conscious travelers decamp for intercity buses – just as price-conscious Europeans ride FlixBus where train travel options are too slow or too expensive. By now, a decade after Megabus and Bolt entered the market, Amtrak is largely only used by people who are price-insensitive or who get motion-sick on buses.
Why are they like this?
If the problem is that international links underperform because they are expensive, then it raises the question, why are fares high to begin with? SNCF charges high fares on Thalys and Eurostar, but not on its domestic trains. This isn’t just about American tourists – I heard too few American accents when I took Eurostar for Americans to be a big enough proportion of revenue. Nor is this about business travelers, because there are many of these traveling between Paris and other French cities.
Rather, my suspicion is that the difference is political. National railroads offering domestic train service face demands from various interests in different directions: the executives themselves as well as the treasury want to maximize revenue, the government writ large wants to give the appearance of successful service, the public wants cheap travel. The major European national railroads seem to have converged on the same solution: intercity trains are not to receive public subsidy for operations or depreciation, but subject to that constraint they should set fares to maximize ridership rather than revenue. The EU even promotes this policy – its directives on passenger rail competition do not allow state subsidies on routes with competition, but do not mandate revenue-maximizing fares.
The political pressure on international rail services is different. The riders are usually foreigners. There is no populist pressure to keep fares low, even on the many French citizens who ride trains to London and Brussels – on the contrary, any inkling of the state not extracting maximum revenue from foreigners may lead to populist pressure to increase fares.
It is possible that more competition will lower fares. This happened in the domestic Italian market, where the entrance of NTV’s Italo service reduced fares on the thicker markets. There is some competition between Paris and points east, such as Frankfurt, where SNCF runs a daily TGV, charging €45 on the 4th of February and DB runs 4 daily ICEs, charging €70-90. Averaged out, it is barely higher than the domestic TGV fare per kilometer.
Which international connections become viable?
European high-speed rail networks are largely domestic. Eurostar stands as the one major exception. What’s more, France, Italy, and Spain have already built the strongest domestic corridors; the only low-hanging domestic fruit are in Germany, where high-speed construction is desirable but is beset by economic austerity, and Britain, where it is beset by very high construction costs. The future of European rail investment is therefore international.
I do not want to claim that charging domestic TGV or ICE fares will automatically lead to ridership density comparable to that of domestic TGVs and ICEs. The language difference probably still matters, just not to the point that Eurostar’s ridership is one quarter that of the LGV Sud-Est.
Moreover, some international routes are clearly a low political priority, so the infrastructure is not optimized for them, leading to low speeds. Trains leaving Brussels going north and east run on a mixture of fast and slow lines, and overall average speeds from Brussels to both Cologne and Amsterdam are within the range for all-legacy upgraded lines. French rail planners, infused by ideas of airline executives who think trains aren’t competitive past three hours, are not trying to optimize the under-construction Mont d’Ambin Base Tunnel for intercity passenger traffic, on the theory that Paris-Turin and Paris-Milan trains would not be competitive either way.
So it’s important to get everything on an international connection right: breakeven rather than revenue-maximizing fares, infrastructure optimized for speed between different cities, sufficient frequency relative to travel time. If these are done right, then city pairs that may look weak may become attractive high-speed rail corridors: Paris-Frankfurt, Paris-Madrid via Bordeaux and Basque Country, Munich-Milan via Innsbruck and the Brenner Base Tunnel, Madrid-Lisbon, Hamburg-Copenhagen, Cologne-Amsterdam.
This is applicable in North America as well, except that there, an additional complication is border controls; the hassle must be reduced to preclearance with short lines (maximum 10-15 minutes), or ideally eliminated with a Schengen-style agreement. This affects Vancouver-Seattle and Toronto-New York, both of which look marginal if we assume international links always underperform. If we accept that New York and Toronto share a language and many cultural features and the weak air travel market is an artifact of high fares, then cross-border trains become an attractive target for investment. In that scenario, New York-Toronto is the strongest North American high-speed rail corridor not touching the Northeast Corridor – it’s like Los Angeles-San Francisco but with stronger connecting public transportation and no mountains to tunnel under.
The upshot is that, given good management, there remains a future for high-speed rail investment, with a plethora of strong lines. The EU can and should take an active role in promoting Union-wide links, ensuring that fares are within the reach of the broad public and that connections between any pair of European cities are reasonable. In North America, two specific links are strong – New York-Toronto and Seattle-Vancouver – and so the federal governments as well as the states and provinces should make sure to invest in them and to charge affordable fares with minimally intrusive border control.
Imagine yourself taking a train somewhere, and imagine the train is big and infrequent. Let’s say it’s the commuter train from New York down the Northeast Corridor to Newark Airport, or perhaps a low-cost OuiGo TGV from Lyon to Paris. Now imagine that you change trains to a small, frequent train, like the AirTrain to Newark Airport, or the RER from the OuiGo stop in the suburbs to Paris itself. What do you think happens?
If your guess is “the train I’m connecting to will be overcrowded,” you are correct. Only a minority of a 200 meter long New Jersey Transit train’s ridership unloads at the Newark Airport station, but this minority is substantial enough to overwhelm the connection to the short AirTrain to the terminals. Normally, the AirTrain operates well below capacity. It uses driverless technology to run small vehicles every 3 minutes, which is more than enough for how many people connect between terminals or go to New York by train. But when a big train that runs every 20-30 minutes arrives, a quantity of passengers who would be easily accommodated if they arrived over 20 minutes all make their way to the monorail at once.
In Paris, the situation is similar, but the details differ. Until recently, OuiGo did not serve Paris at the usual terminal of Gare de Lyon but rather at an outlying station near Eurodisney, Marne-la-Vallée-Chessy, ostensibly to save money by avoiding the Gare de Lyon throat, in reality to immiserate passengers who don’t pay full TGV fare. There, passengers would connect from a 400-meter bilevel TGV on which the entire train ridership would get off to a 220-meter bilevel RER train running every 10 minutes. The worst congestion wasn’t even on the RER itself, but at the ticket machines: enough of the thousand passengers did not have Navigo monthly cards for the RER that long lines formed at the ticket machines, adding 20 minutes to the trip. With the RER connection and the line, the trips would be nearly 3.5 hours, 2 spent on the high-speed train and 1.5 at the Paris end.
I even saw something similar in Shanghai in 2009. I visited Jiaxing, an hour away at the time by train, and when I came back, a mass of people without the Shanghai Public Transportation Card overwhelmed the one working Shanghai Metro ticketing machine. There were three machines at the entrance, but two were out of service. With the 20 minutes of standing in line, I would have gotten back to my hotel faster if I’d walked.
This is a serious problem – the ticketing machine lines alone can add 20 minutes to an otherwise 2.5-hour door-to-door trip. To avoid this problem, railroads and transit agencies need a kit with a number of distinct tools, appropriate for different circumstances.
Run trains more frequently
Commuter trains have to run frequently enough to be useful for short-distance trips. If the RER A consistently fills a train every 10 minutes off-peak between Paris and Marne-la-Vallée, New Jersey Transit can consistently fill a local train every 10 minutes off-peak between Manhattan and New Brunswick. Extra frequency induces extra ridership, but fewer people are going to get off at the Newark Airport stop per train if trains run more often. There are some places where adding frequency induces extra ridership proportionately to the extra service, or even more, but they tend to be shorter-range traffic, for example between Newark and Elizabeth or between Newark and New York.
This tool is useful for urban, suburban, and regional service. A train over a 20 kilometer distance can run frequently enough that transfers to more frequent shuttles are not a problem. Even today, this is mostly a problem with airport connectors, because it’s otherwise uncommon for outlying services to run very frequently. The one non-airport example I am familiar with is in Boston on the Mattapan High-Speed Line, a light rail line that runs every 5 minutes, connecting Mattapan with Ashmont, the terminus of the Red Line subway, on a branch that runs every 8-9 minutes at rush hour and every 12-15 off-peak.
In contrast, this tool is less useful for intercity trains. France should be running TGVs more frequently off-peak, but this means every half hour, not every 10 minutes. The only long-distance European corridors that have any business running an intercity train every 10 minutes are Berlin-Hanover(-Dortmund) and Frankfurt-Cologne, and in both cases it comes from interlining many different branches connecting huge metropolitan areas onto a single trunk.
Eliminate unnecessary transfers
The problem only occurs if there is a transfer to begin with. In some cases, it is feasible to eliminate the transfer and offer a direct trip. SNCF has gradually shifted OuiGo traffic from suburban stations like Marne-la-Vallée and Massy to the regular urban terminals; nowadays, five daily OuiGo trains go from Lyon to Gare de Lyon and only two go to Marne-la-Vallée.
Gare de Lyon is few people’s final destination, but at a major urban station with multiple Métro and RER connections, the infrastructure can handle large crowds better. In that case, the transfer isn’t really from an infrequent vehicle, because a TGV, TER, or Transilien train unloads at Gare de Lyon every few minutes at rush hour. The Métro is still more frequent, but at the resolution of a mainline train every 5 minutes versus a Métro Line 1 or 14 train every 1.5 minutes, this is a non-issue: for one, passengers can easily take 5 minutes just to walk from the far end of the train to the concourse, so effectively they arrive at the Métro at a uniform rate rather than in a short burst.
Of note, Shanghai did this before the high-speed trains opened: the trains served Shanghai Railway Station. The capacity problems occurred mostly because two out of three ticketing machines were broken, a problem that plagued the Shanghai Metro in 2009. Perhaps things are better now, a decade of fast economic growth later; they certainly are better in all first-world cities I’ve taken trains in.
Eliminating unnecessary transfers is also relevant to two urban cases mentioned above: airport people movers, and the Mattapan High-Speed Line. Airport connectors are better when people do not need to take a landside people mover but rather can walk directly from the train station to the terminal. Direct service is more convenient in general, but this is especially true of airport connectors. Tourists are less familiar with the city and may be less willing to transfer; all passengers, tourists and locals, are likely to be traveling with luggage. The upshot is that if an airport connector can be done as an extension of a subway, light rail, or regional rail line, it should; positive examples include the Piccadilly line and soon to be Crossrail in London, the RER B in Paris, and the S-Bahn in Zurich.
The Mattapan High-Speed Line’s peculiar situation as an isolated tramway has likewise led to calls for eliminating the forced transfer. Forces at the MBTA that don’t like providing train service have proposed downgrading it to a bus; forces within the region that do have instead proposed making the necessary investments to turn it into an extension of the Red Line.
Simplify transfer interfaces
The capacity problem at the transfer from an infrequent service to a frequent one is not just inside the frequent but small vehicle, but also at the transfer interface. Permitting a gentler interface can go a long way toward solving the problem.
First, tear down the faregates. There should not be fare barriers between different public transport services, especially not ones where congestion at the transfer point can be expected. Even when everything else is done right, people can overwhelm the gates, as at the Newark Airport train station. The lines aren’t long, but they are stressful. Every mistake (say, if my ticket is invalid, or if someone else tries to ask the stressed station agent a question) slows down a large crowd of people.
And second, sell combined tickets. Intercity train tickets in Germany offer the option of bundling a single-ride city ticket at the destination for the usual price; for the benefit of visitors, this should be expanded to include a bundled multi-ride ticket or short-term pass. New Jersey Transit sells through-tickets to the airport that include the AirTrain transfer, and so there is no congestion at the ticketing machines, only at the faregates and on the train itself.
Both of these options require better integration between different service providers. That said, such integration is clearly possible – New Jersey Transit and Port Authority manage it despite having poor fare and schedule integration elsewhere. In France in particular, there exist sociétés de transport functioning like German Verkehrsverbünde in coordinating regional fares; SNCF and RATP have a long history of managing somehow to work together in and around Paris, so combined TGV + RER tickets, ideally with some kind of mechanism to avoid forcing visitors to deal with the cumbersome process of getting a Navigo pass, should not be a problem.