When Should Cities Separate Short- and Long-Range Commuter Rail?

There’s a big difference between the various regional rail proposals I’ve made for New York and similar examples in Paris and Berlin: the New York maps go a lot further, and incorporate the entirety of regional rail, whereas the RER and the Berlin S-Bahn both focus on shorter-range, higher-frequency lines, with separate trains for longer-range service, generally without through-running. A number of New York-area rail advocates have asked me why do this, often suggesting shorter-range alternatives. Yonah Freemark made a draft proposal many years ago in which through-running trains went as far as New Brunswick, White Plains, and a few other suburbs at that range, on the model of the RER. But I believe my modification of the system used here and in Paris is correct for New York as well as the other American cities I’ve proposed regional rail in.

The reason boils down to a track shortage making it difficult to properly segregate S-Bahn/RER-type service from RegionalBahn/Transilien-type service. These are two different things in Paris, Berlin, Hamburg, and Munich, and Crossrail in London is likewise planned to run separately from longer-range trains, but in Zurich and on Thameslink in London these blend together. Separate operations require four-track mainlines without any two-track narrows at inconvenient places; otherwise, it’s better to blend. And in New York, there are no usable four-track mainlines. Philadelphia and Chicago have them, but not on any corridor where it’s worth running a separate RegionalBahn, which is fundamentally a short-range intercity train, and not a suburban train.

Scale maps

Here is a map of the Berlin S-Bahn (in black) and U-Bahn (in red) overlaid on the New York metropolitan area.

The reach of the S-Bahn here is about comparable to the size of New York City, not that of the metropolitan area. Even taking into account that Berlin is a smaller city, the scope is different. Service to suburbs that are not directly adjacent to Berlin the way Potsdam is is provided by hourly RegionalBahn trains, which do not form a neat network of a frequent north-south and a frequent east-west line through city center.

Here is the same map with the Paris Métro and RER; a branch of the RER D runs off the map but not much, and the RER E branches going east, still within the map box, go further but only every half hour off-peak.

The Parisian Transilien lines are not shown; they all terminate at the legacy stations, and a few have frequent trunks, generally within the scope of the box, but they don’t form axes like the east-west RER A and north-south RER B.

So what I’m proposing is definitely a difference, since I’ve advocated for through-running everything in New York, including trains going from Trenton to New Haven. Why?

Four-track lines and track segregation

In most of Berlin, the infrastructure exists to keep local and longer-range rail traffic separate. The Stadtbahn has four tracks, two for the S-Bahn and two for all other traffic. The North-South Tunnel has only two tracks, dedicated to S-Bahn service; the construction of Berlin Hauptbahnhof involved building new mainline-only tunnels with four tracks. Generally, when the S-Bahn takes over a longer line going out of Berlin, the line has four tracks, or else it is not needed for intercity service. The most glaring exception is the Berlin-Dresden line – the historic line is two-track and given over to the S-Bahn, requiring intercity trains to go around and waste 20 minutes, hence an ongoing project to four-track the line to allow intercity trains to go directly.

In Paris, there are always track paths available. Among the six main intercity terminals, the least amount of infrastructure is four-track approaches, at Gare de Lyon and Gare d’Austerlitz, with two tracks given over to the RER and two to everything else. Of note, the entirety of the Austerlitz network has been given to the RER, as has nearly all of the Lyon network, which is why the lines go so far to the south. The other terminals have more: Saint-Lazare and Nord each have 10 tracks, making segregation very easy. Only subsidiary regional-only stations have two-track approaches, and those are entirely given over to the RER, forming the eastern part of the RER A, the southern part of the RER B, and the western part of the RER C.

New York has a shortage of approach tracks. The reason for this is that historically the mainlines mostly terminated outside Manhattan, so the four-track approaches only went as far as Newark, Jersey City, etc. The LIRR has a four-track mainline into Penn Station from the east, which is why I’ve advocated for some segregation, but even that should eventually involve the express trains via East Side Access through-running to New Jersey; see the second map in this post.

On the New Jersey side there are plans for four tracks with new tunnels across the Hudson, but two tracks have to be shared with intercity trains, and there’s no easy way to neatly separate service into two S-Bahn tracks and two RegionalBahn tracks. In the short run, two of these tracks would have to include trains diverting west to the Morris and Essex Lines, which have a three-track main and therefore cannot segregate their own locals and expresses. In the long run, with the M&E system given its own tunnel across the Hudson, you could theoretically do two local and two express tracks, but that runs into a different issue, which is that east of Penn Station, there are two paths to New Rochelle, both of which have local stops.

The issue of having two paths between the city center station and an important suburban junction, both with local stations, is also a problem in London. North of the Thames, most mainlines are at least four-track, making segregation easy, hence the plans for Crossrail. The only exception is the Lea Valley lines. But in South London, lines are two-track – historically, railways that needed more capacity did not widen one line to four tracks but instead built a parallel two-track lines with its own local stations, often arranging the local stations in a loop. The result is a morass of merging and diverging lines reducing capacity, and London is only slowly disentangling it. In either case, it makes segregation difficult; Thameslink can’t just take over the slow lines the way Crossrail is, and therefore there are Thameslink trains going as far as Bedford and Brighton.

What does this mean?

It’s somewhat unusual for New York to get a regional rail network in which every train, even ones going to distinct cities like New Haven, is part of a central system of through-running. But it’s not unheard of – Thameslink works like this, so does the Zurich S-Bahn, and so does Israel’s national network with its Tel Aviv through-running – and it’s an artifact of a real limitation of the region’s mainline rail system.

But this should not be viewed as a negative. New York really does have suburban sprawl stretching tens of kilometers out. It should have suburban rail accompanying all these suburbs, and wherever possible, it should run on a schedule that is useful to people who are not just 1950s-style 9-to-5 commuters. Moreover, New York lacks either the vast terminals of Paris or the Ringbahn’s mushroom concept, which means trains from outer suburbs have nowhere to go but Manhattan, so they might as well be turned over into a through-running system.

Who Should Bear the Risk in Infrastructure Projects?

The answer to the question is the public sector, always. It’s okay to have private-sector involvement in construction, but the risk must be borne by the public sector, or else the private sector will just want more money to compensate for the extra risk.

The biggest piece of evidence for this is emerging out of our construction costs project, so it will appear in the report and not in a blog post. But for now, I’d like to point out examples from media, the academic literature, and one interview of particular interest.

PPP, Gangnam style

A transportation planner in Korea named Abdirashid Dahir has been giving Eric and me a lot of detailed information about Korean construction costs. We were already aware that Line 9 in Seoul had been built as a PPP, but what we learned was more complicated.

Line 9 is a partnership – the last P in PPP. This means, part of the construction is done by the private sector, and part by the public sector, namely the Seoul Metropolitan Government. The private consortium, led by Hyundai, was responsible for the design and for the construction of the systems, including the tracks, signaling, and rolling stock. SMG was responsible for the civil infrastructure. The total cost of the first phase was 1,167.7 billion won for 25.5 km, split as 492.2 billion in municipal construction and 675.8 billion in private investment.

The importance of this split is that civil infrastructure is the least certain part of underground construction. There are always geotechnical surprises, most small, a few potentially leading to large cost and schedule overruns. These are especially likely during station construction – the tunnels in between tend to be simpler with modern TBMs. Systems, in contrast, are relatively straightforward. Installing rail tracks is the same task regardless of whether it’s in solid rock in an exurban area that has no significant archeology, or through sand that had to be frozen, partly underwater, in the oldest parts of Berlin.

The upshot here is that while low-cost countries do use PPPs, this project keeps the riskiest aspects of construction public and not private. Privatization is fine for less risky, more commoditized situations.

How private bidders respond to risk

Two examples come to mind, both from the United States.

First, in New York, Brian Rosenthal’s seminal New York Times article cited Denise Richardson of the General Contractors’ Association saying that the contractors are barely making any profit and are bidding high because of risks imposed on them by the public sector. I don’t think this is a very high-quality source – it’s extremely biased, for one – but in context, it makes some sense.

Second, we do have more quantifiable data on this, thanks to the work of the Stanford Graduate School of Business economist Shosh Vasserman and Hoover Institute economist Valentin Bolotnyy. They look at highway maintenance contracts in Massachusetts and compare scaling auctions, in which the contracts are itemized, with lump sum auctions, in which they are not. Based on actual differences in price and estimates of contractor risk-aversion, they estimate that itemizing saves 10% of the cost through lower risk.

Supporting structures for public-sector risk assumption

There’s always the problem of moral hazard. Of note, even with this problem, costs are lower with itemized contracts in Massachusetts than with lump-sum contracts. But this does suggest a number of ways to reduce costs through better risk management:

  • Itemized contracts, in enough detail that changes do not need litigation.
  • Fixed profit rates – Spanish contracts are done with a fixed profit rate over the items named in the bid.
  • Public oversight – there needs to be tighter supervision of risky things, which most likely means no PPPs for civil infrastructure.

It is unfortunate that American trends in the last 20 years have been away from those principles and toward greater privatization of the state, and equally unfortunate that American (and British) soft power has led to similar reforms in the wrong direction in the rest of the Anglosphere. But it’s possible to do better and imitate Korean practices to get Korean costs.

Regional Rail for Non-Work Trips

Good public transportation must be useful for all travel needs. This in particular includes trips that are not commutes to work, which are the most typical use case for suburban trains. This is a key difference between American and European public transportation: in some cases the modal split for work trips are similar, for example New York’s split is similar to that of the major German cities, but per capita ridership in the German cities is a lot higher than in New York, because off-peak service is better and people use it more for non-commute trips. In the linked post I went over this as a matter of better off-peak service; in this post I am going to go over the more fundamental question of, what kind of trips can regional public transport provide apart from work trips?

Some data from Germany

Germany periodically conducts surveys of transportation usage, called Mobilität in Deutschland, or MiD. The most recent was in 2017, and the one before it was in 2008. In Berlin, the data I have is from 2008, from the city’s transportation plan for 2030. On PDF-p. 30, it produces the following modal splits by trip type:

  • Work: 39% transit, 40% car
  • School: 31% transit, 31% car
  • Shopping: 19% transit, 30% car
  • Pleasure: 21% transit, 32% car
  • Going back home: 27% transit, 31% car
  • Other, including business: 25% transit, 67% car

Here is Hamburg, also as of 2008 and not 2017, on p. 7, with the city and suburbs listed separately:

  • Work: 33%/48% city, 16%/65% suburbs
  • School: 32%/13% city, 34%/30% suburbs
  • Shopping: 11%/41% city, 4%/65% suburbs
  • Errands: 21%/37% city, 4%/71% suburbs
  • Pleasure: 17%/40% city, 5%/55% suburbs

The Hamburg S-Bahn barely extends past city limits, even less so than the Berlin S-Bahn, so the low suburban modal splits cannot be viewed as a failure of S-Bahn service to be useful for non-work trips. But it’s worth pointing out that in both Berlin and Hamburg (city, not suburbs), work is the activity with the highest modal split for public transport, followed by school.

Long and short trips

In a few posts over the years, I talked about a dichotomy of long and short trips. Long trips include the primary commute but also extraordinary trips such as to the airport; short trips are routine errands, including shopping and short-distance leisure trips.

In most transit cities, short trips are not done on transit but on foot. I usually bring up Asian examples because they concentrate development near the train station, so one walks to the train station either to shop at the department store within the station or to get on an actual train to get to work in city center. But this is equally true of Germany, even with vastly lower extents of transit-oriented development. The intersection points of the Ringbahn with U- and S-Bahn lines are replete with shopping centers.

The upshot is that if people don’t take regional trains to go shopping, because if the system works, then they can walk to retail. This is true even when retail is at big box scale – those Ringbahn stations have Kaufland and Real hypermarkets.

Sporadic long trips

If regional rail is not for short trips, then what non-work trips is it for? The answer cannot be none, because ridership on S-Bahn trains in Berlin and Hamburg is too high compared with the work modal split. The answer can’t just be school either, because there aren’t vast numbers of school commuters.

Rather, it’s better to think in terms of sporadic and usual long trips. When I go play board games, it’s a long trip and not a short trip, because the community is citywide, and people come from Friedrichshain, from Charlottenburg, from far western suburbs, from Lichtenberg, from Pankow, from Schöneberg. The same is true of queer meetups – even when I lived in Neukölln and some of these meetups were walking distance to me, they were not walking distance to most attendees, who would take the U- or S-Bahn.

All of this falls under the rubric of sporadic long trips: long because they are not normally in-neighborhood, and sporadic because they are not one’s usual commute to work or school. But they often involve several roundtrips a week, producing healthy ridership at all hours of day.

Meetups are an example of using the train for socialization at a scale that’s expressly citywide or even regionwide. In that sense, a large city with a good regional transportation network is good for social connections, because it encourages communities to be more specialized to people’s interests. In a city the size of Berlin, even with Germany’s shocking lack of diversity, one can find enough queers for an English-language meetup. In a larger city like New York, without the same language load, one can even find ever more specialized communities – there is no one gaming community but many, and they intersect, for example one can look for queer gamers, or for a Harlem-based group with predominantly black demographics, and so on. This isn’t really any different from the economic advantages of large cities, which offer more division of labor.

The upshot is that even if the sort of travel one can do by regional rail that is not for work does not seem to exist in a city with poor off-peak public transportation service, this does not mean improving off-peak transit is pointless. On the contrary: over time, the existence of such service will let communities form at convenient locations based on the shape of the network. It will make the city bigger in a sense, and this is a good thing economically and socially.

Building Depth and Window Space

How much window space does an apartment need, relative to its area, and how does this affect building style? A fascinating post from about a year ago on Urban Kchoze makes the argument that modern North American buildings are too deep – Simon calls them obese. Simon contrasts the typical building style in major cities in Europe and Asia with modern North American imitators and argues that the North American versions have too much ratio of floor area to exterior window width, which only works with loft-style apartments, which are not fit for families.

Is Simon correct? Not really. There’s an important feature of the block style in Europe that he’s missing. And this leads to an interesting observation by itself about area-to-window-width ratios.

The issue of building depth

Simon shows a bunch of satellite photos of buildings in a style called the euroblock. Here’s one example from Stockholm, in Södermalm:

The block has a width that looks like 14.6 meters. Midblock buildings have front windows overlooking the street and back windows overlooking a central courtyard; corner buildings overlook two streets. Either way, the area-to-frontage ratio is 7.3 meters. In general, buildings in Central Stockholm, urban Berlin, and Paris in average a depth of 13-14 meters, so the above typology would generate a ratio of 6.5-7 meters.

Simon contrasts this with American buildings. The euroblock typology is very uncommon in the US – New York’s typology is much less neat and liberally uses windows that overlook very narrow spaces. But it does exist, generally in higher-end recent developments. For example, here’s the Avalon East Norwalk, a condo project wedged between I-95 and the Northeast Corridor.

It has essentially the same built form as the euroblock. Its development history is of course different: there are no streets on the exterior, only parking lots, and it is a single project surrounding a big plaza with a swimming pool rather than many small buildings that together enclose a courtyard that comprises several separate gardens. But in terms of how the building looks from space, it’s similar. The width is 20 meters, for an area-to-frontage ratio of 10 meters, well above 6.5-7 meters.


Euroblocks are complicated

The above Stockholm pic is a pretty simple building, conceptually: a linear building outlining the edge of a rectangle. This is not the typical euroblock; I had to look around Central Stockholm to find a fitting example. I could equally well use Hamburg or another such city of the same size class.

But in Paris, this form is almost unheard of, and in Berlin it is uncommon, I think mostly denoting postwar reconstruction. Paris and Berlin are larger cities, especially historically – in the Belle Epoque/Wilhelmine era, when this typology flourished, they were two of the largest few cities in the world, Berlin stagnating after World War Two and Paris growing exclusively in the suburbs. So they’d build up more of each lot and leave less unbuilt space between buildings. Instead, here is what a traditional Berlin block looks like, in this case in Neukölln:

These buildings enclose a central courtyard, as in Stockholm, but there the similarity ends. The courtyard is small, and there are several to a block. All these wings have internal corners with limited window space. Moreover, the wings that do not make it all the way to enclosing the courtyard, like the ones on the buildings north of Laubestrasse, have blank walls facing northeast, because they were built expecting the wing of another building to directly abut them. The wing of the building at the Laubestrasse/Elbestrasse eastern corner likewise has blank northeast-facing walls, and from space looks awkward, like a half-building. All of this was designed for more buildings, but some were never built or were knocked down.

If the euroblock has one big courtyard for the entire block as in the Stockholm and Norwalk examples, then the area-to-frontage ratio equals exactly half the building depth. But as soon as there are multiple courtyards, the ratio grows. The dimensions of the C-shaped building on Sonnenallee (one block south of Laubestrasse) just west of the corner building with which it shares the courtyard are 18 meters of street frontage by 38 of lot depth minus a half-courtyard of 11.5*12. This works out to 546 m^2/71 m, for a total ratio of 7.7 m, even though technically the building is never deeper than 13 m.

The blocks can get even more fractured. Here’s Prenzlauer Berg, in an area wedged between the former Wall and the Ringbahn:

The dimensions of the buildings fronting Korsörer Strasse on the north are pretty consistent. They all have an overall lot depth of about 32 meters, consisting of 14 meters of building, 11 meters of courtyard, and 7 meters of half-building with blank north-facing walls. The side wings are pretty consistently 7 meters deep each as half-buildings. Taking the pair of buildings flanking the second courtyard from the east as an example, they together are 35*32 minus 21*11, for 889 m^2/99 m = 9 m.

In Paris, building forms vary. But here is an example with wings, in the 17th:

The courtyards are smaller than in Berlin. Taking the second building from the west, we get 35*25 – 11*13, or 732 m^2/98 m = 7.5 m. When the courtyard is only about as wide as the building is deep, the above typology, similar to the image from Neukölln, generates a ratio equal to 5/8 the building depth, and not 1/2 as in the Stockholm example. The Prenzlauer Berg typology generates an even higher ratio, a full 2/3 of building depth if the courtyard is a square of side equal to the building depth.

And this matters. Buildings with simpler sides do get deeper in Paris. For example, this building in the 16th, wedged between two streets:

The depth of these buildings is 18 meters, so the area-to-frontage ratio is 9 m.

What does this mean?

My choice of the 16th and 17th in Paris for my examples is not random. Western Paris has been rich from the moment it urbanized – families of means choose to live this way. In general, within the family of euroblocks, the more desirable areas seem to have buildings with a slightly larger depth – the more working-class parts, such as Eastern Paris, have shallower buildings. Rich people would all else being equal prefer more window frontage space, but all else is not equal, and they prefer bigger apartments.

There is a definite limit on how deep buildings can be and how large the ratio of area to window frontage can be, but it is not as low as Simon posits. Ratios in the 8-9 region are not unheard of in old European buildings, and it stands to reason that euroblocks built in an environment of more prosperity, such as that of the early 21st century, could go slightly higher.

Moreover, the Norwalk example of a deeper building without wings is generally preferable to the traditional Berlin and Paris form of shallower buildings with wings. In Berlin, the apartments with street-facing windows are the most desirable. Historically, the wings were for the working class, which had to make do with narrow courtyards – sometimes narrower than today, the original statutory limit being less than 6 m wide due to 19th-century fire regulations. So the evolution of the euroblock is likely to be toward its American condo form.

More on Station Costs

Talking to Marco Chitti about the history of Italian construction always fills me with hope. He’s been gathering data about metro construction in Milan and Rome, and told Elif, Eric, and me about the issue of building through constrained areas. Historic city centers are constrained because tunneling can damage buildings – the first two lines in Milan, built in the 1950s and 60s at abnormally low costs, caused some damage to buildings, since they involved cut-and-cover under streets only 12-15 meters wide. The good news is that tunneling with a tunnel boring machine is fine now. Stations remain an enormous challenge – but the conversation did fill me with optimism about future construction in cities that were not global imperial capitals 2,000 years ago.

TBM technology

Tunnel-boring machines have advanced to the point of being archeology-safe. Italian heritage protection laws from the 2000s forbid any risk to historic buildings and historic sites, but TBM technology at this point allows preserving artifacts. It involves injecting a gel ahead of the cutting edge, which is not supposed to be a cost-raiser.

The result is that tunneling is cheap. This is not a matter of low wages – in fact, Marco cites higher wages for Italian skilled workers who staff TBMs, up to €4,500 a month net, which rises to about €9,000 gross with social contributions. These are based on a nationwide scale that only weakly varies with location, which helps explain why Naples costs are not low despite the region’s low incomes.

Station construction

Station construction costs vary immensely by location. In Rome, on the same project, stations in a suburban part of the city might be €60-70 million. This does not mean construction is trivially easy: Rome’s suburbs still often host historic sites, having been home to patrician villas in Antiquity, and in fact the word suburb dates to that era. However, it’s relatively safe, and I don’t think Line C ran into such sites.

Then in the most constrained parts of the city, things are different. The extension plans for Line C deeper into city center have station costs in the €400-600 million range. This is not what things cost everything within Rome, or even everywhere within the densely-built parts of the city. But the Line C extension passes through the most historic sites. An already-under construction segment will go to the Colosseum, and a planned extension deeper into city center is to go to Piazza Venezia, at the Wedding Cake, and it is that station that is projected to cost €600 million.

The reason for the high cost is that it is not possible to do archeology- and building-safe cut-and-cover. Piazza Venezia doesn’t quite have enough room for a cut-and-cover dig of a full-length station. It is fed by a wide street, the Via Fori Imperali, and I asked Marco why not build cut-and-cover there, but he pointed out that the street goes through the historic Forum. It is in fact elevated over the ruins; any cut-and-cover there would endanger the Forum, and is not acceptable.

Without cut-and-cover, the only alternative is to mine the stations. Rome investigated the option of large-diameter TBMs on the Barcelona L9 model and found it infeasible, since the tunnels are so big they might themselves cause some building damage. Once the stations are mined from a small shaft, their costs explode. Second Avenue Subway built stations using the same method, and had similar per-station construction costs.

The good news

Mined station construction is in practically all cases not necessary. New Yorkers talk about the city’s high built-up density as a reason why costs are high. But in terms of actual stuff in the way of a tunnel, there’s less in New York than in Rome or Istanbul, which has even lower construction costs.

In fact, there is a line in Rome that is rather similar in urban geography to Second Avenue Subway: the Line B1 branch. It runs under a 27 meter wide street flanked by modern buildings that are about 9 stories tall above ground but also have underground parking, Italy having such a car culture that the middle class expects to own cars even in Rome. The cost: €527 million for 3.9 km, in 2010-15.

Moreover, the hard rock in New York should make it easier to build stations while maintaining building safety. Manhattan’s schist is brittle and therefore requires concrete lining, unlike the more uniform gneiss of Stockholm, famously forming natural arches that are pretty to look at from within the tunnels. However, it is still better soil for construction than the sand of Berlin’s U5 extension, to be opened next month, or the alluvial soil of Amsterdam.

The explanation Marco gives concerning station construction is physical and not institutional. This means it should transplant well into another setting – which it does!

In Berlin, the city-center U5 extension, including U55, is in today’s money around €240 million/km. The stations look like cut-and-cover to me, and if they’re not then it comes from severe NIMBYism since the line goes under the very wide Unter den Linden, but one of the stations is basically under the river and another is under U6 and involves moving the U6 station, and the sandy soil is genuinely bad to tunnel through. Suburban extensions in Berlin, with easy cut-and-cover stations, are consistently in the €100-150 million/km range, which is barely higher than the non-Forum Italian range. So Berlin looks fine, and just needs to invest resources into U- and S-Bahn extensions and not into extending the A 100 motorway.

Can New York have what Italy has?

Almost certainly! Second Avenue is not an old or narrow street by Italian standards. Nor are any of the other streets slated for subway construction in New York, such as Nostrand, Utica, and even 125th. Importing construction techniques from Italy and Germany should be feasible. There may be problems with local politics – New Yorkers absolutely hate admitting that another city may be better than theirs in any way, and this makes learning harder. But it is not impossible, and so far there do not seem to be any physical or economic obstacles to doing so.

The German Way of Building Rapid Transit

Continuing my series on different traditions of building urban rapid transit, today it’s time for Germany and Austria, following the posts on the US, the Soviet bloc, Britain, and France. Germany had a small maritime empire by British and French standards and lost it all after World War 1, but has been tremendously influential on its immediate neighbors as a continental power. This is equally true of rapid transit: Germany and Austria’s rail traditions have evolved in a similar direction, influential also in Switzerland, Denmark, the Netherlands, and Belgium to varying extents.

S-Bahns and U-Bahns

Germany is one of the origins of urban regional rail, called S-Bahn here in contrast with the U-Bahn subway. The first frequent urban rail service in the world appeared in London in 1836, but trains ran every 20 minutes and the stop spacing was only borderline urban. Berlin in contrast innovated when it opened the east-west elevated Stadtbahn in 1882, running frequent steam trains with local spacing.

As elevated steam-powered urban rail, the Stadtbahn was not particularly innovative. New York had already been running such service on its own els going back to 1872. But the Stadtbahn differed in being integrated into the mainline rail system from the start. Berlin already had the Ringbahn circling the city’s then-built up area to permit freight trains to go around, but it still built the Stadtbahn with four tracks, two dedicated to local traffic and two to intercity traffic. Moreover, it was built to mainline rail standards, and was upgraded over time as these standards changed with the new national rail regulation of 1925. This more than anything was the origin of the concept of regional rail or S-Bahn today.

Vienna built such a system as well, inspired by many sources, including Berlin, opening in 1898. Hamburg further built a mainline urban rail connection between Hauptbahnhof and Altona, electrifying it in 1907 to become the first electrified S-Bahn in the world. Copenhagen, today not particularly German in its transportation system, built an S-Bahn in the 1930s, naming it S-tog after the German term.

However, German cities that built such S-Bahn systems would also build separate U-Bahn systems. U-Bahns in Germany have short stop spacing and tend to mostly serve inner areas: for example, on this map of Munich, the U-Bahn is in blue, and the trams are in red. Berlin has some farther-reaching U-Bahn lines, especially U7, a Cold War line built when the West got the U-Bahn and the East got the S-Bahn; had the city not been divided, it’s unlikely it would have been built at all.

Some of the early U-Bahns were even elevated, similarly to New York subway lines and a few Paris Métro lines. Hamburg’s operator is even called Hochbahn in recognition of the elevated characteristic of much of its system. Like Paris and unlike New York, those elevated segments are on concrete viaducts and not steel structures, and therefore the trains above are not very noisy, generally quieter than the cars at street level.

Light rail and Stadtbahns

The early els of Berlin and Vienna were called Stadtbahn when built in the 19th century, but since the 1960s, the term has been used to refer to mixed subway-surface systems.

Germany had long been a world leader in streetcar systems – the first electric streetcar in the world opened in Berlin in 1881. But after World War Two, streetcars began to be viewed as old-fashioned and just getting in the way of cars. West German cities generally tore out their streetcars in their centers, but unlike American or French cities, they replaced those streetcars with Stadtbahn tunnels and retained the historic streetcar alignments in outer neighborhoods feeding those tunnels.

The closure of the streetcars was not universal. Munich and Vienna retained the majority of their tram route-length, though they did close lines parallel to the fully grade-separated U-Bahn systems both cities built postwar. Many smaller cities retained their trams, like Augsburg and Salzburg, though this was generally more consistent in the Eastern Bloc, which built very little rapid transit (East Berlin) or severed itself from the German planning tradition and Sovietized (Prague, Budapest).

The Stadtbahn concept is also extensively used in Belgium, where it is called pre-metro; the Vienna U-Bahn and even the generally un-German Stockholm T-bana both have pre-metro history, only later transitioning to full grade separation. Mixed rapid transit-streetcar operations also exist in the Netherlands, but not in the consistent fashion of either the fast-in-the-center-slow-outside Stadtbahn or its fast-outside-slow-in-the-center inverse, the Karlsruhe model of the tram-train.

Network design

Rail network design in German-speaking cities is highly coordinated between modes but is not very systematic or coherent.

The coordination means that different lines work together, even across modes. In the post about France, I noted that the Paris Métro benefited from coordinated planning from the start, so that on the current network, there is only one place where two lines cross without a transfer. This is true, but there are unfortunately many places where a Métro line and an RER line cross without a transfer; the central RER B+D tunnel alone crosses three east-west Métro lines without a transfer. In Berlin, in contrast, there are no missed connections on the U-Bahn and the S-Bahn, and only one between the U-Bahn and S-Bahn, which S21 plans do aim to fix. Hamburg has two missed connections on the U-Bahn and one between the U- and S-Bahn. Munich has no missed connections at all.

But while the lines work well as a graph, they are not very coherent in the sense of having a clear design paradigm. Berlin is the most obvious example of this, having an U-Bahn that is neither radial like London or Moscow nor a grid like Paris. This is not even a Cold War artifact: U6 and U8 are parallel north-south lines, and have been since they opened in the 1920s and early 20s. Hamburg and Vienna are haphazard too. Munich is more coherent – its U-Bahn has three trunk lines meeting in a Soviet triangle – but its branching structure is weird, with two rush hour-only reverse-branches running as U7 and U8. The larger Stadtbahn networks, especially Cologne, are a hodgepodge of mergers and splits.


The German planning tradition has distinguishing characteristics that are rare in other traditions, particularly when it comes to fare payment – in many other respects, the Berlin U-Bahn looks similar to the Paris Métro, especially if one ignores regional rail.

Proof of payment: stations have no fare barriers, and the fare is enforced entirely with proof of payment inspections. This is common globally on light rail (itself partly a German import in North America) and on European commuter rail networks, but in Germany this system is used even on U-Bahns and on very busy S-Bahn trunks like Munich and Berlin’s; in Paris there’s POP on the RER but only in the suburbs, not in the city.

Unstaffed stations: because there are no fare barriers, stations do not require station agents, which reduces operating expenses. In Berlin, most U-Bahn stations have a consistent layout: an island platform with a stairway exit at each end. This is also common in the rest of the German-speaking world. Because there is no need for fare barriers, it is easy to make the stations barrier-free – only one elevator is needed per station, and thus Berlin is approaching fully wheelchair accessibility at low cost, even though it’s contemporary with New York (only 25% accessible) and Paris (only 3% accessible, the lowest among major world metros).

Fare integration: fares are mode-neutral, so riding an express regional train within the city costs the same as the U-Bahn or the bus, and transfers are free. This is such an important component of good transit that it’s spreading across Europe, but Germany is the origin, and this is really part of the coordination of planning between U- and S-Bahn service.

Zonal fares: fares are in zones, rather than depending more granularly on distance as is common in Asia. Zones can be concentric and highly non-granular as in Berlin, concentric and granular as in Munich, or non-concentric as in Zurich.

Monthly and annual discounts: there is a large discount for unlimited monthly tickets, in order to encourage people to prepay and not forget the fare when they ride the train. There are even annual tickets, with further discounts.

No smartcards: the German-speaking world has resisted the nearly global trend of smartcards. Passengers can use paper tickets, or pay by app. This feature, unlike many others, has not really been exported – proof-of-payment is common enough in much of Northern and Central Europe, but there is a smartcard and the fare inspectors have handheld card readers.

Verkehrsverbund: the Verkehrsverbund is an association of transport operators within a region, coordinating fares first of all, and often also timetables. This way, S-Bahn services operated by DB or a concessionaire and U-Bahn and bus services operated by a municipal corporation can share revenue. The first Verkehrsverbund was Hamburg’s, set up in 1965, and now nearly all of Germany is covered by Verkehrsverbünde. This concept has spread as a matter of fare integration and coordinated planning, and now Paris and Lyon have such bodies as well, as does Stockholm.

Germany has no head

The American, Soviet, British, and French traditions all rely on exports of ideas from one head megacity: New York, Moscow, London, Paris. This is not at all true of the German tradition. Berlin was the richest German city up until World War 2, and did influence planning elsewhere, inspiring the Vienna Stadtbahn and the re-electrification of the Hamburg S-Bahn with third rail in the late 1930s. But it was never dominant; Hamburg electrified its S-Bahn 20 years earlier, and the Rhine-Ruhr region was planning express regional service connecting its main cities as early as the 1920s.

Instead, German transportation knowledge has evolved in a more polycentric fashion. Hamburg invented the Verkehrsverbund. Munich invented the postwar S-Bahn, with innovations like scheduling a clockface timetable (“Takt”) around single-track branches. Cologne and Frankfurt opened the first German Stadtbahn tunnels (Boston had done so generations earlier, but this fell out of the American planning paradigm). Karlsruhe is so identified with the tram-train that this technology is called the Karlsruhe model. Nuremberg atypically built a fully segregated U-Bahn, and even more atypically was a pioneer of driverless operations, even beating Paris to be the first city in the world to automate a previously-manual subway, doing so in 2010 vs. 2012 for Paris.

There’s even significant learning from the periphery, or at least from the periphery that Germany deigns acknowledge, that is its immediate neighbors, but not anything non-European. Plans for the Deutschlandtakt are based on the success of intercity rail takt planning in Switzerland, Austria, and the Netherlands, and aim to build the same system at grander scale in a larger country.

The same polycentric, headless geography is also apparent in intercity rail. It’s not just Germany and Switzerland that build an everywhere-to-everywhere intercity rail system, in lieu of the French focus on connecting the capital with specific secondary cities. It’s Austria too, even though Vienna is a dominant capital. For that matter, the metropolitan area of Zurich too is around a fifth of the population of Switzerland, and yet the Swiss integrated timed transfer concept is polycentric.

Does this work?

On the most ridiculously wide definition of its metropolitan area, Vienna has 3.7 million people, consisting of the city proper and of Lower Austria. In 2012, it had 922 million rail trips (source, PDF-p. 44); the weighted average work trip modal split in these two states is 40% (source, PDF-p. 39). In reality, Vienna is smaller and its modal split is higher. Zurich, an even smaller and richer city, has a 30% modal split. Mode shares in Germany are somewhat lower – nationwide Austria’s is 20%, Germany’s is 16% – but still healthy for how small German cities are. Hamburg and Stuttgart both have metropolitan public transport modal splits of 26%, and neither is a very large city – their metro areas are about 3.1 and 2.6 million, respectively. Munich is within that range as well.

In fact, in the developed world, one doesn’t really find larger modal splits than these in the 2 million size class. Stockholm is very high as well, as are 1.5th-world Prague and Budapest, but one sees certain German influences in all three, even though for the most part Stockholm is its own thing and the other two are Soviet. Significantly higher rates of public transport usage exist in very large Asian cities and in Paris, and Germany does deserve demerits for its NIMBYism, but NIMBYism is not why Munich is a smaller city than Taipei.

To the extent there’s any criticism of the German rapid transit planning tradition, it’s that construction costs lately have been high by Continental European standards, stymieing plans for needed expansion. Märkisches Viertel has been waiting for an extension of U8 for 50 years and it might finally get it this decade.

The activist sphere in Germany is especially remarkable for not caring very much about U-Bahn expansion. One occasionally finds dedicated transport activists, like Zukunft Mobilität, but the main of green urbanist activism here is bike lanes and trams. People perceive U- and S-Bahn expansion as a center-right pro-car plot to remove public transit from the streets in order to make more room for cars.

The high construction costs in Germany and the slow, NIMBY-infused process are both big drags on Germany’s ability to provide better public transportation in the future. It’s plausible that YIMBYer countries will overtake it – that Korean and Taiwanese cities of the same size as Munich and Hamburg will have higher modal splits than Munich and Hamburg thanks to better transit-oriented development. But in the present, the systems in Munich and Zurich are more or less at the technological frontier of urban public transportation for cities of their size class, and not for nothing, much of Europe is slowly Germanizing its public transport planning paradigm.

Governance in Rich Liberal American Cities

Matt Yglesias has a blog post called Make Blue America Great Again, about governance in rich liberal states like New York and California. He talks about various good government issues, and he pays a lot of attention specifically to TransitMatters and our Regional Rail project for the Boston region, so I feel obliged to comment more on this.

The basic point Matt makes is that the quality of governance in rich liberal American states is poor, and as a result, people do not associate them with wealth very consistently. He brings up examples about the quality of schools and health care, but his main focus is land use and transportation: the transportation infrastructure built in New York, California, etc. is expensive and not of high quality, and tight zoning regulations choke housing production.

That said, I think there’s a really important screwup in those states and cities that Matt misses: the problem isn’t (just) high costs, but mostly total unwillingness to do anything. Do-nothing leaders like Charlie Baker, Andrew Cuomo, Gavin Newsom, and Bill de Blasio aren’t particularly interested in optimizing for costs, even the first two, who project an image of moderation and reason.

The Regional Rail proposal’s political obstacles are not exactly a matter of cost. It’s not that this should cost $4 billion (without the North-South Rail Link) but it was estimated at $15 billion and therefore there’s no will to do it. No: the Baker administration seems completely uninterested in governing, and has published two fraudulent studies making up high costs for both the North-South Rail Link and rail electrification, as well as a more recent piece of fraud making up high costs for Boston-Springfield intercity rail. The no comes first, and the high costs come second.

This history – no first, then high costs – is also the case for New York’s subway accessibility program. The MTA does not want it; the political system does not care either. Therefore, when disability rights advocates do force some investment, the MTA makes up high costs, often through bundling unnecessary investments that it does want, like rebuilding station interiors, and charging these projects to the accessibility account. A judge can force an agency to build something, but not to build it competently and without siphoning money.

I want to emphasize that this does not cover all cases of high American costs. Second Avenue Subway, for example, is not the result of such a sandbag: everyone wants it built, but the people in charge in New York are not competent enough to build it affordably. This does accord with Matt’s explanation of poor Northeastern and West Coast governance. But not everything does, and it’s important to recognize what’s going on.

The other important point is that these do-nothing leaders are popular. Baker is near-tied for the most popular governor in the United States with another do-nothing Northeastern moderate Republican, Maryland’s Larry Hogan. Andrew Cuomo’s approval rate has soared since he got 43,000 people in the state killed in the corona crisis.

People who live in New York may joke that the city has trash on the street and cockroaches in apartments, but they’re pretty desensitized to it. They politically identify as Democrats, and once corona happened they blamed Trump, as did many people elsewhere in the United States, and forgave Democrats who mismanaged the crisis like Cuomo. Baker and Hogan are of course Republicans, but they perform a not-like-the-other-Republicans persona, complete with open opposition to Trump, and therefore Massachusetts Democrats who have a strong partisan identity in federal elections are still okay with do-nothing Republicans. People who really can’t stand the low quality of public services leave.

Construction cost reform is pretty drastic policy, requiring the destruction of pretty powerful political forces – the system of political appointments, state legislators and mayors with a local rather than national-partisan identity, NIMBYs, politically-connected managers, the building trades, various equity consultants (such as many Los Angeles-area urbanists). They are not legally strong, and a governor with a modicum of courage could disempower them, but to be a moderate in the United States means to be extremely timid and technologically conservative. Matt himself understands that last point, and has pointed this out in connection with moderates who hold the balance of power in the Senate, like Joe Manchin and Susan Collins, but use it only to slightly shrink proposed changes and never to push a positive agenda of their own.

So yes, this is a construction cost crisis, but it’s not purely that. A lot of it is a broader crisis of political cowardice, in which non-leftist forces think government doesn’t work and then get elected and prove it (and leftists think real change comes from bottom-up action and the state is purely for sinecures, courtesy of the New Left). I warned in the spring that corona is not WW2 – the crisis is big enough to get people to close ranks behind leaders, but not to get them to change anything important. These states are rich; comfortable people are not going to agitate for the destruction of just about every local political power structure just to get better infrastructure.

Surplus Extraction

Ever since reading Brooks-Liscow on the growth in American road construction costs since the 1960s, I’ve been interested in the surplus extraction theory of costs. The authors call their main theory citizen voice, in which local groups can use litigation to extract the social surplus generated by infrastructure construction. I’d like to go more deeply into what this theory is and what it implies.

What is surplus?

Normally, a competitive market has no surplus. The owner of a restaurant, the developer of a building in an unconstrained area like suburban Texas, the seller of cloth masks on Etsy, the freelance web developer – none of them is making a killing. People enter the market until profits are driven down to levels low enough to essentially be the owner-manager’s wage. Companies can only make a large profit if they operate at enormous scale, which takes a long time to develop – the profit margins on a single Walmart or Carrefour or Lidl are small, but the profit margins on 10,000 stores add up to a couple billion dollars a year.

Infrastructure is not a competitive market, for a number of different reasons:

  • The construction of transportation infrastructure has strong positive externalities, through enabling agglomeration. In a country with cars, the construction of public transportation also helps mitigate the negative externalities of cars.
  • Infrastructure is not meaningfully competitive. The largest city in the world, Tokyo, has around two competing rail operators per suburban region. In Tokyo, it’s a natural duopoly; in just about every smaller city, it’s a natural monopoly.
  • The barriers to entry are so steep that some kind of price regulation is obligatory. The result is extensive consumer surplus for riders who are not poor.
  • Government involvement means that regulations that make it easier or harder to build infrastructure have large impact, which can create or destroy social surplus.

The upshot is that at non-New York costs, infrastructure construction in New York generates enormous social surplus. I could break this down by component, but for brevity I won’t, and just cite what looks like the upper limit of what the publics in the United States and Europe are willing to pay for urban and regional rail: around $50,000 per projected weekday trip. Lines teetering on the edge of cancellation, like M18 in Paris, Second Avenue Subway Phase 2 in New York, and Crossrail 2 in London, all cluster around this figure.

If we take $50,000/rider as the lowest possible benefit-cost ratio that gets a project built, around 1.2-1.3 in countries that conduct such analyses, then Second Avenue Subway Phase 2, currently projected around $60,000/rider, is 1. But at the median global cost, which exists in France and Germany, it would cost $700 million, or $7,000/rider, for a benefit-cost ratio of 8.5. At costs that exist in Southern Europe, Scandinavia, Switzerland, and Korea, make it $400 million, or $4,000/rider, for a benefit-cost ratio of 15. That’s a big net profit for New York City Transit (or, it would be if its operating costs were not abnormally high too), and a huge net social surplus for New York. Every group that wants a piece of that surplus then has an incentive to make noise and raise costs.

How can surplus be extracted?

People who wish to seize public resources have a variety of methods with which to do so. Some are net transfers of surplus from society to one special interest, but most are net destruction of value in the sense that the loss of social surplus exceeds the gain to the special interest, usually by a large margin.

The technique for surplus extraction is usually the threat of a lawsuit, but in some cases it can be direct political lobbying. The actual lawsuit is almost never important – in the US and Germany, at least, the state usually wins these suits, and the impact of litigation is to delay and to deny political capital.

However, surplus can also vanish into the ether through poor planning. Consultants who are not under pressure to save money may well propose oversize infrastructure just because that’s what they are used to, or to avoid sharing right-of-way across railroads; this has led to unusual cost premiums in the United States for everything that touches mainline rail, whereas the subway and light rail premiums are, outside New York, bad but less onerous.

The demands made by special interests that extract surplus vary. They include any of the following:

  • Gratuitous tunneling instead of above-ground construction. This is usually a demand made of high-speed rail, but there are some gratuitous tunnels in suburban rail as well, for example Crossrail 2. The surplus here is that NIMBYs do not like to see trains from their houses; the emotional value of their views is naturally a fraction of that of the cost of tunneling.
  • Compromise alignments that either increase costs or reduce benefits. This is usually about avoiding specific places; Brooks-Liscow give an example of a Detroit highway swerving around a Jewish community center. But sometimes it can be the opposite – in fact, early US freeway builders expected that communities would lobby for highways near them, not far from them. Los Angeles County’s advocacy for a high-speed rail detour through Palmdale is one such example.
  • Extortion of community benefits to activists, for example demands for larger stations to act as neighborhood centers. A large degree of the cost explosion of the Green Line Extension in Boston came from the policy of accommodating local demands, leading to oversize stations. But such overbuilding can also occur absent extortion – the surplus can vanish into poor practices, representing incompetence rather than malice, as in the oversize viaducts of California High-Speed Rail.
  • Contracts to favored companies. This led to cost explosion in Italy in the 1970s and 80s, especially in Rome but also Milan; unlike the other items on this list, this is generally illegal, and costs in Italy came down after crackdowns on corruption in the 1990s. However, legal versions exist – sometimes the government is just used to doing business with a company with a poor track record, for example the “the devil we know” attitude in California toward Tutor Perini. The surplus in the latter case vanishes not quite into someone’s pockets but more into the state’s unwillingness to oversee contractors more tightly.
  • Labor demands. If the demands are purely about wages then the surplus is distributed without being destroyed. However, these demands are in all cases I know of also about other things. For example, the sandhogs in New York opposed the use of more efficient tunnel boring instead of more dangerous but more labor-intensive dynamite. Protectionism also leads to inferior equipment in addition to higher costs.

Who can extract surplus?

Surplus extraction works through informal mechanisms. The purpose of the nuisance lawsuit is not to win, but to extract a settlement. The threat is delay and loss of political favor for the project rather than outright cancellation. The NIMBY lawsuit in Silicon Valley against California High-Speed Rail was right on the technical merit – the Pacheco Pass route, which would pass through the richest suburbs was technically inferior to the Altamont Pass route, which wouldn’t – still lost; Pacheco was favored due to another kind of surplus extraction, namely Rod Diridon’s desire for shorter Los Angeles-San Jose trip times.

Because surplus extraction works through politics and not clear rules, it benefits those with the most political power. In this way, the rise in NIMBYism in the 1960s and 70s, for example the freeway revolts, contrasts with the contemporary free speech movement, which used formal lawsuits with the intent of winning to expand the boundaries of free speech in America.

The free speech movement celebrated protections for communist Berkeley professors and for pornographers; people with normative professions and normative political views were already protected. In contrast, NIMBYism was most powerful in already rich areas, like Jane Jacobs’ Greenwich Village, or Boston’s South End. Baltimore’s racially integrated freeway revolt was exceptional. New York built freeways through working-class neighborhoods easily, and only encountered political obstacles in the Village, which was by the 1950s gentrified (Jacobs was a journalist with some college education, married to an architect, and her father was a doctor), a new development that hadn’t happened in urban history before and thus the city elites had missed it. Moreover, Jacobs’ remedy of creating and empowering community boards has ensured that only powerful people and powerful communities could change city decisions.

Even more recent attempts to create equity have failed. Slowing down the state and empowering community is always bad for equity, because the community is where inegalitarian traditions live. Black leaders now can derail transit plans just as white leaders can; non-leaders have no voice in neighborhood politics, and it’s those non-leaders who work outside the neighborhood who rely on public transit.

Surplus extraction remains the domain of people with political and cultural cachet. One can fight redevelopment in San Francisco on behalf of a mural to Cesar Chavez; fighting it on behalf of pornographers is harder. Similarly, the unions that have been the best at extracting surplus are traditional ones, doing jobs that existed 100 years ago, at productivity levels that remain stuck in that era, mainly the trades.

Conclusion: saying no

Surplus extraction theory does not say it is impossible to reduce costs. Italy’s sharp fall in costs in the 1990s and Turkey’s gentle fall in the 2010s both suggest that cost reduction is possible. What it does say is that the role of the state is to safeguard surplus and keep it socialized, against demands from many special interests, which should be disempowered through legal changes making lawsuits harder and reducing the ability of consultants and unions to drive up costs.

In that sense, the role of the planner is to say no – and moreover, to say no to charismatic groups representing much-romanticized people. No, dear mother with children, we will not build you a noise wall just because you think 140 km/h electric trains will reduce your quality of life. No, dear tradesman much-profiled as a non-college white voter, we will not hire you for $110/hour when there exist people who will do your job better than you can at $35/hour. No, dear third-generation business owner, we will not listen to what you think about traffic as we replace parking spots with bus lanes. No, dear anti-gentrification activist, we will not pay you as an equity consultant, we will just build the subway in the city. No, dear white flight homeowner, we will not build you a tunnel just to avoid taking a few houses through eminent domain. No, dear deindustrialized city leader, we will not require companies to set up factories in your city at high cost when we can get cheaper imports. It’s never going to be no, dear criminal, or no, dear Nazi, because criminals and Nazis are not used to making such requests and having people listen.

It’s optimistic in a sense, because much cost control comes just from knowing that it’s possible and having the nerve to say no to people who are used to hearing yes. The engineering factors that lead to low costs are important, but first of all, it’s necessary to believe that they are feasible, over local objections.

More on Consultants and Design-Build

A few months ago, there appeared an article comparing construction costs for subways in the US and Europe. It has a little table, not PPP adjusted, with cases from elsewhere, but the bulk of the reporting covers differences between the US and Europe. It’s interesting and I urge everyone to read it – but read it critically. It has a long list of bullet points naming various differences, some already covered here, some new but still within reason.

One aspect that seems especially apt is this:

The construction cost [in the US] represents slightly more than 50% of the overall program cost, while soft costs and stakeholders’ commitments at 45% are significantly higher in comparison with other types of major projects or similar projects in other global regions.

Labor cost and construction schedule are the most important factors affecting the construction cost. Labor cost is often driven by labor union rules which vary significantly among states and cities. One of the highest labor costs of tunnel construction workers is the Sandhogs in New York which can be as high as $110/hr and on an overtime basis, it can reach over two to three times this value. Their rates are higher than other tunnel workers in the country and significantly higher than European or Asian workers rates. Also, the number of workers assigned in the tunnel in New York is significantly more than other parts of the country and as much as 4 times more than tunnel workers assigned to comparable projects in Europe. Tunneling being linear structures, the opportunities to accelerate the construction schedule in order to reduce overall labor cost are limited.

That said, I’d like to caution about fully accepting everything the article says. The key issue is that the authors’ experience is as consultants – they work for AECOM. This means that to at least some extent, their expertise is informed by their work as outside consultants, which means that they are the most familiar with projects that at some point invite consultants in.

This is important, because this may be an important difference between low- and medium-cost countries. I am not sure – I’m trying to investigate those differences more carefully, but this involves listening to German complaints about NIMBYism and trying to figure out how relevant it is that NIMBYs are far less empowered in Southern Europe, counting Turkey as part of that region since it acts much like a peripheral European country in construction. I don’t think that low-cost countries in Southern Europe use international consultants – Milan and Madrid at least don’t, and Istanbul used Italian consultants at one point but nowadays seems mostly to design things itself.

What’s more, AECOM’s experience is not just in countries that use AECOM’s advice regularly, but also in specific projects that bought its services. This is relevant to the claim that,

European owners spend less time and money on planning, studies, conceptual developments, and detailed design. Most projects are implemented using the Design-Build model with the detailed design provided by the contractor during construction to suit his means and methods; this results in efficiency and eliminates repeating of design work.

There’s the rub: design-build does exist in Continental Europe. Turkey uses it, and France is glancing in that direction. But it’s uncommon – Italy and Spain do not use this method, and France largely does not either and I think neither do Germany or the Nordic countries. Moreover, design-build in Turkey means there is extensive in-house oversight, much more so than in American or British design-build projects.

French design-build is even more tightly overseen, because its purpose is not to forgo public planning. Rather, France traditionally maintains the separation of public planning, private design, and private construction, in order to fight corruption and guarantee fair procurement. This separation leads to problems when projects require redesign in case they are very complex, and as a result, Grand Paris Express exists as a large public-sector planning agency to facilitate coordination between the design and construction teams. Technically this can be called design-build, but it has approximately nothing to do with American design-build projects that pay Skanska or Dragados a large sum of money to dig a subway and have extensive public regulations and red tape but little public engineering. The role of the public sector in American, British, and increasingly rest-of-Anglosphere eyes is to make sure companies follow capricious rules but not to actively build infrastructure or, perhaps, change the rules to be more favorable to swift action.

Regrettably, in the coda the authors buy into this mentality that the public sector cannot change the rules. They list various action items that can be undertaken to reduce costs, all of which are very good – those items include streamlining regulations, improving risk sharing mechanisms, and offloading some peripheral costs, among others, rather than expanding design-build. They’re missing a few things that we’re learning from the low-cost world – for example, Istanbul makes an effort to site stations in parks in order to be able to build them more easily and reduce their costs, which I believe is also true of Milan. But for the most part, the list of things that the US needs to do to have what France and Germany have cannot be too dissimilar to that produced by the authors.

But then the authors throw it all away and say it’s unlikely that the US could match European costs. They give a bare-bones explanation that boils down to saying “these recommendations won’t really be implemented.” I agree to some extent – it’s plausible, though not yet certain, that New York will need to union-bust the sandhogs and probably also the other trades, and these are politically powerful unions that know very well that they earn several times what their labor is worth and fight to preserve this. But, first of all, not every recommendation is that fraught; questions of risk sharing, public planning, and procurement do not lend themselves to political populism and remain unreformed mostly because the Northeastern US has timid, reactive governance.

And second, the authors say it’s unlikely the US could match European costs even if their recommendations are followed. They don’t explain why – there are few intangibles in the article, and they mostly seem peripheral to the main question, for example the fact that the US is an auto-oriented society. I can’t tell if it’s just uncertainty, which does not appear in the body of the piece, or if there’s more to it. It could just be a writing artifact and what they meant to say was that their recommendations could help New York match Parisian costs but they’re skeptical their recommendations are politically palatable to New York.

I emphasize the criticism, even though it’s generally a good overview, because all of the experts we talk to have biases. These could be consultant biases, or political biases (Turkey is far more polarized than any mature democracy), or engineering biases, or language biases. Even reading my blog is to some extent a bias – people who read me and think well of my analysis might well look for reasons in their own domain why design-build is bad, which means that to be certain I am correct in my prescription against it, we need to cleanroom this, for example by interviewing people who do not know me directly (or at all) and asking how engineering is done where they are.


I recently covered the Stadtbahn, a mode of rail transportation running as rapid transit (almost always a subway) in city center and as a tramway farther out. The tram-train is the opposite kind of system: it runs as a tramway within the city, but as rapid transit farther out. There’s a Human Transit blog post about this from 2009, describing how it works in Karlsruhe, which invented this kind of service pattern. Jarrett is bearish on the tram-train in most contexts, giving a list of required patterns that he says is uncommon elsewhere. It’s worth revising this question, because while the tram-train is not very useful in an American context, it is in countries with discontinuous built-up areas, including Germany and the Netherlands but also Israel. Israeli readers may be especially interested in how this technology fits the rail network away from the Tel Aviv region.

What is a tram-train?

Let’s dredge the 2*2 table from the Stadtbahn post:

Slow in centerFast in center
Slow in outlying areasTramwayStadtbahn
Fast in outlying areasTram-trainRapid transit

The terms fast and slow are again relative to general traffic. The Paris Métro averages 25 km/h, less than some mixed-traffic buses in small cities, but it still counts as fast because the speed in destinations accessed per hour is very high.

Be aware that I am using the terms Stadtbahn and tram-train to denote two different things, but in Karlsruhe the system is locally called Stadtbahn. German cities use the term Stadtbahn to mean “a tramway that doesn’t suck,” much as American cities call a dazzling variety of distinct things light rail, including lines in all four cells of the above table. Nonetheless, in this post I am keeping my terminology distinct, using the advantage of switching between different languages and dialects.

Tram-trains and regional rail

The Karlsruhe model involves trains running on mainline track alongside mainline trains, diverging to dedicated tramway tracks in the city, to connect Karlsruhe Hauptbahnhof with city center around Marktplatz. This also includes lines that do not touch the mainline, like S2, but still run with higher-quality right-of-way separation outside city center; but most lines run on mainline rail part of the way.

North American light rail lines, with the exception of the Boston, Philadelphia, and San Francisco Stadtbahn systems, tend to run as tram-trains, but never have this regional rail tie-in. They run on entirely dedicated tracks, which has two important effects, both negative. First, it increases construction costs. And second, it means that the shape of the network is much more a skeletal tramway map than the more complicated combination of an S-Bahn and a tramway that one sees in Karlsruhe. San Diego has a short segment sharing tracks with freight with time separation, but the shape of the network isn’t any different from that of other American post-1970s light rail systems, and there’s an ongoing extension parallel to a mainline railroad that nonetheless constructs a new right-of-way.

In this sense, the Karlsruhe model can be likened to a cheaper S-Bahn. S-Bahn systems carve new right-of-way under city center to provide through-service whenever the historic city station is a terminus, such as in Frankfurt, Stuttgart, Munich, or German-inspired Philadelphia. They can also build new lines for more expansive service, higher capacity, or a better connection to city center, like the second S-Bahn trunk in Hamburg; Karlsruhe itself is building a combined road and rail tunnel, the Kombilösung, after a generation of at-grade operation. The tram-train is then a way to achieve some of the same desirable attributes but without spending money on a tunnel.

It follows that the tram-train is best when it can run on actual regional rail tracks, with good integration with the mainline system. It is a lower-speed, lower-cost version of a regional rail tunnel, whereas the North American version running on dedicated tracks is a lower-cost version of a subway. Note also that regional rail can be run at different scales, the shorter and higher-frequency end deserving the moniker S-Bahn; the Karlsruhe version is long-range, with S1 and S11 reaching 30 km south of city center and S5 reaching 70 km east.

Where is a tram-train appropriate?

Jarrett’s 2009 post lays down three criteria for when tram-trains work:

  • The travel market must be small enough that an S-Bahn tunnel is not justified.
  • The destination to be served isn’t right next to the rail mainline.
  • The destination to be served away from the mainline is so dominant that it’s worthwhile running at tramway speeds just to get there and there aren’t too many people riding the line beyond it.

The center of Karlsruhe satisfies the second and third criteria. It is borderline for the first – the region has maybe a million people, depending on definitions, and the city proper has 312,000 people; the Kombilösung is only under-construction now and was not built generations ago, unlike S-Bahn tunnels in larger cities like Munich.

Jarrett points out that in the urban world he’s most familiar with, consisting of the United States, Canada, Australia, and New Zealand, it is not common for cities to satisfy these criteria. He does list exceptions, for example Long Beach, where the Blue Line runs in tramway mode before heading into Los Angeles on a mostly grade-separated right-of-way, whereupon it goes back into the surface in Downtown LA before heading into an under-construction tunnel. But overall, this is not common. City centers tend to be near the train station, and in the United States there’s such job sprawl that just serving one downtown destination is not good enough.

That said, the Long Beach example is instructive, because it is not the primary city in its region – Los Angeles is. I went over the issue of outlying S-Bahn tunnels a year ago, specifying some places where they are appropriate in Israel. The tram-train must be a key tool in the planner’s box as a cheaper, lower-capacity, lower-speed version of the same concept, diverging from the mainline in tramway mode in order to serve a secondary center. Karlsruhe itself is a primary urban center – the only time it’s the secondary node is when it connects to Mannheim, and that train doesn’t use the tramway tracks – but a secondary tram-train connection is being built in outlying areas there, namely Heilbronn.

Different models of urban geography

In the American model of urban geography, cities are contiguous blobs. Stare at, for example, Chicago – you’ll see an enormous blob of gray stretching in all directions. Parkland is mostly patches of green in between the gray, or sometimes wedges of green alternating with wedges of gray, the gray following commuter railroads and the green lying in between. Boundaries between municipalities look completely arbitrary on a satellite map.

In the German model of urban geography, it’s different. Look at Cologne, Frankfurt, Mannheim, or Stuttgart – the built-up area is surrounded by green, and then there are various suburban towns with parkland or farmland in between. This goes even beyond the greenbelt around London – there’s real effort at keeping all these municipalities distinct.

I don’t want to give the impression that the United States is the weird one. The contiguous model in the United States is also common in France – Ile-de-France is one contiguous built-up area. That’s how despite being clearly a smaller metropolitan region than London, Paris has the larger contiguous population – see here, WUP 2007, and see also how small the German and Dutch urban areas look on that table. Urban agglomeration in democratic East Asia is contiguous as in the US and France. Canada looks rather American to me too, especially Vancouver, the city both Jarrett and I are the most familiar with, while Toronto has a greenbelt.

This distinction moreover has to be viewed as a spectrum rather than as absolutes. Boston, for example, has some of the German model in it – there’s continuous urbanization with inner suburbs like Cambridge and Newton, but beyond Route 128, there are many small secondary cities with low density between them and the primary center. Conversely, Berlin is mostly American or French; the few suburbs it has outside city limits are mostly contiguous with the city’s built-up area, with the major exception of Potsdam.

The relevance of this distinction is that in the German or Dutch model of urban geography, it’s likely that a railway will pass through a small city rather far from its center, fulfilling the second criterion in Jarrett’s post. Moreover, this model of independent podlike cities means that there is likely to be a significant core, which fulfills the third criterion. The first criterion is fulfilled whenever this is not the center of a large metropolitan area.

It’s not surprising, then, that the Karlsruhe model has spread to the Netherlands. This is not a matter of similarity in transport models: the Netherlands differs from the German-speaking world, for examples it does not have monocentric S-Bahns or S-Bahn tunnels and it builds train stations with bike parking where Germany lets people bring bikes on trains. Nonetheless, the shared model of distinct municipalities makes tram-train technology attractive in South Holland.

Israel and tram-trains

In Israel, there are very few historic railways. A large share of construction is new, and therefore has to either swerve around cities or tunnel to enter them, or in a handful of cases run on elevated alignments. Israel Railways and local NIMBYs have generally preferred swerving.

Moreover, the urban layout in Israel is very podlike. There do exist contiguous areas of adjacent cities; Tel Aviv in particular forms a single blob of gray with Ramat Gan, Givatayim, Bni Brak, Petah Tikva, Bat Yam, and Holon, with a total population of 1.5 million. But for the most part, adjacent cities are buffered with undeveloped areas, and the cities jealously fight to stay this way despite extensive developer pressure.

The final important piece in Israel’s situation is that despite considerable population growth, there is very little rail-adjacent transit-oriented development. The railway was an afterthought until the Ayalon Railway opened in 1993, and even then it took until last decade for mainline rail to be a significant regional mode of transport. The state aggressively builds new pod-towns without any attempt to expand existing towns toward the railway.

The upshot is that all three of Jarrett’s criteria for tram-trains are satisfied in Israel, everywhere except in and around Tel Aviv. Tel Aviv is large enough for a fully grade-separated route, i.e. the already-existing Ayalon Railway. Moreover, because Tel Aviv needs full-size trains, anything that is planned to run through to Tel Aviv, even as far as Netanya and Ashdod, has to be rapid transit, using short tunnels and els to reach city centers where needed. A tram-train through Ashdod may look like a prudent investment, but if the result is that it feeds a 45 meter long light rail vehicle through the Ayalon Railway then it’s a waste of precious capacity.

But Outside Tel Aviv, the case for tram-trains is strong. One of my mutuals on Twitter brings up the Beer Sheva region as an example. The mainline going north has a station called Lehavim-Rahat, vaguely tangent to Lehavim, a ways away from Rahat. It could get two tramway branches, one diverging to the built-up area of Lehavim, a small suburb that is one of Israel’s richest municipalities, and the other to Rahat, one of Israel’s poorest. There are also interesting options of divergence going south and east, but they suffer from being so far from the mainline the network would look scarcely different from an ordinary tramway.

Beer Sheva itself would benefit from tramways with train through-service as well. The commercial center of the city is close to the train station, but the university and the hospital aren’t, and are not even that close to the subsidiary Beer Sheva North station. The station is also awkwardly off-center, lying southeast of the city’s geographic center, which means that feeding buses into it with timed transfers screws internal connections. So tramway tracks on Rager Boulevard, cutting off Beer Sheva North for regional trains, would do a lot to improve regional connectivity in Beer Sheva; intercity trains should naturally keep using the existing line.

In the North, there are similar examples. Haifa is not going to need the capacity of full-size trains anytime soon, which makes the case for various branches diverging into smaller cities to provide closer service in tramway mode strong. Unlike in Beer Sheva, the case for doing so in the primary center is weak. Haifa’s topography is the stuff of nightmares, up a steep hill with switchback streets. The mainline already serves the Lower City well, and climbing the hill is not possible.

This creates an interesting situation, in which the technology of the tram-train in the North can be used to serve secondary cities like Kiryat Ata and Tirat Carmel and maybe enter the Old City of Acre, but the operational pattern is really that of a Stadtbahn – fast through Haifa and up most of the Krayot, slow through smaller suburbs.