Little Things That Matter: Interchange Siting

I’ve written a lot about the importance of radial network design for urban metros, for examples here, here, here, here, and here. In short, an urban rail network should look something like the following diagram:

That is, every two radial routes should intersect exactly once, with a transfer. In this post I am going to zoom in on a specific feature of importance: the location of the intersection points. In most cities, the intersection points should be as close as possible to the center, first in order to serve the most intensely developed location by all lines, and second in order to avoid backtracking.

The situation in Berlin

Here is the map of the central parts of Berlin’s U- and S-Bahn network, with my apartment in green and three places I frequently go to in red:

(Larger image can be found here.)

The Ring is severed this month due to construction: trains do not run between Ostkreuz, at its intersection with the Stadtbahn, and Frankfurter Allee, one stop to the north at the intersection with U5. As a result, going to the locations of the two northern red dots requires detours, namely walking longer from Warschauer Strasse to the central dot, and making a complex trip via U7, U8, and U2 to the northern dot.

But even when the Ring is operational, the Ring-to-U2 trip to the northern dot in Prenzlauer Berg is circuitous, and as a result I have not made it as often as I’d have liked; the restaurants in Prenzlauer Berg are much better than in Neukölln, but I can’t go there as often now. The real problem is not just that the Ring is interrupted due to construction, but that the U7-U2 connection is at the wrong place for the city’s current geography: it is too far west.

As with all of my criticism of Berlin’s U-Bahn network layout, there is a method to the madness: most of the route of U7 was built during the Cold War, and if you assumed that Berlin would be divided forever, the alignment would make sense. Today, it does not: U7 comes very close to U2 in Kreuzberg but then turns southwest to connect with the North-South Tunnel, which at the time was part of the Western S-Bahn network, running nonstop in the center underneath Mitte, then part of the East.

On hindsight, a better radial design for U7 would have made it a northwest-southeast line through the center. West of the U6 connection at Mehringdamm it would have connected to the North-South Tunnel at Anhalter Bahnhof and to U2 at Mendelssohn Park, and then continued west toward the Zoo. That area between U1/U2 and Tiergarten Park is densely developed, with its northern part containing the Cold War-era Kulturforum, and in the Cold War the commercial center of West Berlin was the Zoo, well to the east of the route of U7.

Avoiding three-seat rides

If the interchange points between lines are all within city center, then the optimal route between any two points is at worst a two-seat ride. This is important: transfers are pretty onerous, so transit planners should minimize them when it is reasonably practical. Two-seat rides are unavoidable, but three-seat rides aren’t.

The two-seat ride rule should be followed to the spirit, not the letter. If there are two existing lines with a somewhat awkward transfer, and a third line is built that makes a three-seat ride better than connecting between those two lines, then the third line is not by itself a problem, and it should be built if its projected ridership is sufficient. The problem is that the transfer was at the wrong location, or maybe at the right location but with too long a walk between the platforms.

Berlin’s awkward U-Bahn network is such that people say that the travel time between any two points within the Ring is about 30 minutes, no matter what. When I tried pushing back, citing a few 20-minute trips, my interlocutors noted that with walking time to the station, the inevitable wait times, and transfers, my 20-minute trips were exceptional, and most were about 30 or slightly longer.

The value of an untimed transfer rises with frequency. Berlin runs the U-Bahn every 5 minutes during the daytime on weekdays and the S-Bahn mostly every 5 minutes (or slightly better) as well; wait times are shorter in a city like Paris, where much of the Metro runs every 3 minutes off-peak, and only drops to 5 or 6 minutes late in the evening, when Berlin runs trains every 10 minutes. However, Parisian train frequencies are only supportable in huge cities like Paris, London, and Tokyo, all of which have very complex transfers, as the cities are so intensely built that the only good locations for train platforms require long walks between lines.

New York of course has the worst of all worlds: a highly non-radial subway network with dozens of missed connections, disappointing off-peak frequencies, and long transfer corridors in Midtown. In New York, three-seat rides are ubiquitous, which may contribute to weak off-peak ridership. Who wants to take three separate subway lines, each coming every 10 minutes, to go 10 kilometers between some residential Brooklyn neighborhood and a social event in Queens?

Free Public Transportation

Note: this may turn into a long series of posts about public transportation fare systems and payments.

From time to time, people propose free public transport. Supporters have a variety of motivations, including an attempt to mirror cars (“do state roads charge tolls?”), ideological socialism, positive externalities, and the efficiencies of getting rid of fare collection.

In reality, making service free at the point of use means spending money on subsidies from other sources – money that could be spent on other things than zeroing out the fares. There are opportunity costs, and robust public transportation networks do not gain much efficiency from being free. If there is money to make service free, there is money to spend on service improvements, including more metro lines, higher frequency, and wheelchair accessibility where it isn’t already present.

Literature review

A tweetstorm from two days ago includes references to a number of studies on this issue:

Proof of payment

One argument for free transit is that it simplifies operations because no fare collection is needed. Front-door boarding and paying the drivers slow down bus boarding – each passenger takes 2.6-3 seconds to board (source, PDF-p. 20). Rapid transit systems also suffer from the complexity of fare collection infrastructure: batteries of faregates create chokepoints and require maintenance, and usually rapid transit agencies also have to hire station agents to watch the gates.

However, proof-of-payment fare enforcement, or POP, gets around most of these issues. If passengers do not need to pay at entry, everything becomes much simpler: they can board buses from any door, and get onto the train without crossing faregates. Berlin has all-door boarding and open, unstaffed U-Bahn stations. There are fare-vending machines, which are not free, but they are cheap. There are fare inspectors working on consignment – they get paid by catching non-paying riders.

Better uses for money

New York City Transit has $9.1 billion in operating and maintenance expenses as of 2016, and $4.3 billion in fare revenue (source). Ile-de-France Mobilités has a total of about €10 billion in annual operating and capital expenses, with about 10% of this being capital and the rest operating, and €2.8 billion in fare revenue. As of 2015, BVG had a total transport income of €1.344 billion (PDF-p. 7) and an additional subsidy of €620 million (PDF-p. 21).

In all of these cities, if there is money for fare elimination, there is money for further improvements in service. A disability rights advocacy group in Paris estimates the cost of making the Metro accessible at €4-6 billion, or 1.5-2 years’ worth of fares. Parisian construction costs for further Metro extensions are such that the budget for free fares could instead be spent on adding around 14 annual kilometers of new tunnels. In Berlin, a third S-Bahn trunk line running northwest-southeast would require about a year and a half’s worth of present-day fares to construct; adding service to guarantee 5-minute frequency on all trunk lines even on weekends and evenings would require a small increase in operating expenses.

New York’s construction costs are much higher than those of Paris and Berlin, and even its operating costs are elevated, but then it also charges higher fares. If there is $4.3 billion a year for free fares, there is much less $4.3 billion a year for boosting off-peak frequency on every named route (2, 4, A, etc.) to at worst 6 minutes, with 2- and 3-minute off-peak frequencies on interlined trunk lines. As with Paris, there is also a dire need for wheelchair accessibility; thanks to very high costs, full installation would not cost just 1.5-2 years’ worth of fare revenue, but more like 3 years’ worth.

Cities with and without public transport

The above discussion centers where the vast majority of public transportation takes place – that is, in cities with serious public transportation systems. The argument changes completely in smaller cities, which run the occasional bus but not at the required speed, coverage, or frequency for it to count as a real public transport network.

In Germany, there is no free transit, but the difference between big-city and small-city fare enforcement is telling: only relatively big cities have POP systems. Small-town Germany makes bus passengers pay the fare to the driver, and runs trains with conductors checking tickets. The reason is that roving inspectors only work on systems with enough frequency and coverage, or else they can’t efficiently ride the buses and trains and check tickets.

If POP is not possible, then the cost of collecting fares rises: buses are slowed down by every additional passenger, and trains require a second crew member. Such systems often have very low farebox recovery in the first place, and a very low-income rider profile, since everyone who can afford to drive drives rather than waits 25 minutes for the bus. In Los Angeles, total fare revenue on Metro (which includes most buses) is $350 million a year and total operating and maintenance expenses amount to $1.57 billion, and the average public transport commuter has about half the average income of the average solo driver. In that specific use case, making public transport free may be justified.

The one caveat is that if the plan is to convert a city from one without public transportation to speak of to one with a good system, for example in Los Angeles, then in the future, revenue will become more important. Even if free public transport is a good idea in the conditions of 2019, it may not be such a good idea in those of 2035, at least if grandiose transit city plans materialize (and I don’t think they will – the state of American local governance just isn’t good enough for cities to follow through).

Costs are Rising, US Highway Edition

There’s a preliminary paper circulating at Brookings, looking at American infrastructure construction costs. Authors Leah Brooks and Zachary Liscow have tabulated the real costs of the American Interstate program over time, from the 1950s to the 1990s, and find that they increased from $5.3 million per km ($8.5 million/mile) in 1958-63 to $21.3 million/km ($34.25 million/mile) in 1988-93.

Moreover, they have some controls for road difficulty, expressed in slope (though not, I believe, in tunnel quantity), urbanization, and river and wetland crossings, and those barely change the overall picture. They go over several different explanations for high American infrastructure costs, and find most of them either directly contradicted by their results or at best not affirmed by them.

I urge readers to read the entire paper. It is long, but very readable, and it is easy to skip the statistical model and go over the narrative, including favored and disfavored explanations, and then poke at the graphs and tables. I’m going to summarize some of their explanations, but add some important context from cross-national comparisons.

Why costs (probably) aren’t rising

The authors identify four hypotheses they rule out using their research, in pp. 19-23 (they say five but only list four):

Difficult segments postponed and built later – they have some controls for that, as mentioned above. The controls are imperfect, but the maps depicted on pp. 59-61 for the Interstate network’s buildout by decade don’t scream “the segments built after 1970 were harder than those built before.”

Time-invariant features – these include cross-national comparisons, since the United States has always been the United States. I will discuss this in a subsequent section, because two separate refinements of what I’ve seen from cross-national comparisons deal with this issue specifically.

Input prices – this is by far the longest explanation the authors deal with. Anecdotally, it’s the one I hear most often: “labor costs are rising.” What the authors show is that labor and materials costs did not rise much over the period in question. Construction worker wages actually peaked in real terms in 1973 and fell thereafter; materials costs jumped in the aftermath of the oil crisis, but came down later, and were back at pre-crisis levels by the 1990s (p. 48). Land costs did rise and have kept rising, but over the entire period, only 17.7% of total costs were preliminary engineering and land acquisition, and the rest were in construction.

Higher standards – the authors looked and did not find changes in standards leading to more extensive construction.

There are several more incorrect explanations that jump from the data. I was surprised to learn that throughout the 1970s and 80s, completion time remained mostly steady at 3-3.5 years of construction; thus, delays in construction cannot be the explanation, though delays in planning and engineering can be.

The authors themselves list additional explanations that have limited evidence but are not ruled out completely from their data, on pp. 32-35. Construction industry market concentration may be an explanation, but so far data is lacking. Government fragmentation, measured in total number of governments per capita, has no effect on the result (for example, California has high costs and not much municipal fragmentation); I’ll add that Europe’s most municipally fragmented country, France, has middle-of-the-road subway construction costs. State government quality, as measured by corruption convictions, has little explanatory power – and as with fragmentation, I’ll add that in Europe we do not see higher costs in states with well-known problems of clientelism and corruption, like Italy and Greece. Work rules requiring the addition of more workers may be relevant, but unionization and left-right politics are not explanatory variables (and this also holds for rail costs).

Economies of scale look irrelevant as well: there is negative correlation between costs and construction, but the causality could well go the other way. Finally, soft budget constraints are unlikely, as the federal government can punish states that mismanage projects and take more money; it’s possible that as the Interstate program ended states felt less constrained because there wouldn’t be money in the future either way (“end of repeated game”), but the fact that costs keep rising in subway construction suggests this is not relevant.

Favored explanations

Two explanations stand out to the authors. The first is that nearly the entire increase in construction costs over time can be attributed to a mix of higher real incomes and higher house prices. While the construction workers themselves did not see their wages rise in the late 1970s and 80s, a richer population may demand more highways, no matter the cost.

Higher real estate costs could have an impact disproportionate to the share of land acquisition in overall costs by forcing various mitigations that the paper does not control for, such as sound walls and tunnels, or by sending roads over higher-cost alignments.

The second explanation is what the authors call citizen voice. Regulatory changes in the 1960s and early 70s gave organized local groups greater ability to raise objections to planning and force changes, reducing community impact at the cost of higher monetary expenditures. The authors give an example from suburban Detroit, where a highway segment that disrupted a Jewish community center took 25 years to be built as a result of litigation.

The authors don’t say this explicitly, but the two explanations interact well together. The citizen voice is very locally NIMBY but is also pro-road outside a handful of rich urban neighborhoods. Higher incomes may have led to public acceptance of higher costs, but local empowerment through citizen voice is the mechanism through which people can express their preference for higher costs over construction inconvenience.

How time-invariant are national features, anyway?

The authors contrast two proposed explanations – higher incomes and property values, and stronger NIMBY empowerment – with what they call time-invariant features, which could not explain an increase in costs. But can’t they?

I spent years plugging the theory that common law correlates with high subway construction costs, and it does in the developed world, but upon looking at more data from developing countries as well as from before the last 25 years, I stopped believing in that theory. It started when I saw a datapoint for Indonesia, a civil-law country, but even then it took me a few more years to look systematically enough, not to mention to wait for more civil-law third-world countries to build subways, like Vietnam. By last year I was giving counterexamples, including Montreal, low rail electrification costs in some common law countries, and the lack of a London cost premium over Paris until the late 20th century.

In lieu of common law, what I use to explain high costs in the US relative to the rest of the world, and to some extent also in most first-world common law countries as well as third-world former colonies, is weak civil service. In the developed world, the theory behind this is called adversarial legalism, as analyzed by Robert Kagan. Adversarial legalism enforces the law through litigation, leading to a web of consent decrees. Some are naked power grabs: for example, in Los Angeles, a union sued a rolling stock vendor for environmental remediation and agreed to drop the lawsuit in exchange for a pledge that its factory be unionized, which may play a role in why the trains cost around 50% more than equivalent European products.

American litigiousness developed specifically in the 1970s – it’s exactly how what the authors of the paper call citizen voice is enforced. In contrast, on this side of the Channel, and to some extent even generally on this side of the Pond, laws are enforced by regulators, tripartite labor-business-government meetings, ombudsmen, or street protests. French riotousness is legendary, but its ability to systematically change infrastructure is limited, since rioting imposes a real cost on the activist, namely the risk of arrest and backlash; in contrast, it is impossible to retaliate against people who launch frivolous lawsuits.

I bring up the fact that I said most of this last year, and the rest at the beginning of this year, whereas I was not aware of the paper under discussion until it was released a few hours ago, to make it clear that I’m not overfitting. This is something that I’ve been talking about for around a year now, and a jump in American construction costs in the 1970s and 80s – something that also looks to be the case in subway construction – is fully compatible with this theory.

West Station is an Overbuilt Mess

Boston has been on a commuter rail infill binge lately; it has opened four stations on the Fairmount Line this decade, with general success, and is now eying the Worcester Line, where the MBTA has already opened a single in-city station called Boston Landing. The next station to be opened is called West Station, serving Allston, a middle-class urban neighborhood home to Boston University. Unfortunately, the West Station project has suffered from budget and schedule overruns: the current projection is $90 million, where past stations in the area have opened for about $15-25 million each, and construction will start next decade and only wrap up by 2040.

The cause of the extreme cost is poor design. The station as currently proposed is an overbuilt mess. It is development-oriented transit, sited next to an area that Harvard wishes to redevelop as a new campus, and the compromises made between good rail service, intermodal bus-rail connections, and encouraging development make the project fail at all of its objectives. The idea of an infill station in Allston is solid and the MBTA should keep working on the project, but it should do it right – that is, maximize passenger utility while also slashing the budget by a factor of about 4.

I encourage readers to look at a presentation about the status of the project from May, and at another presentation from June, which was sent to members of the media and neighborhood.

Intermodal integration done wrong

The West Station site is roughly in the center of the new development. Unfortunately, it is poorly-located relative to the street network. With its hierarchy of major and minor streets, Boston is not forgiving to wrong station siting: buses would have to meander to reach the site.

The busiest bus in the area, and among the busiest in the region, is the 66. See image below:

The Red and Green Lines of the subway are in their respective colors (and the Green Line’s branches are surface light rail), the Worcester Line is in purple with its existing stations marked alongside the proposed West Station site, and the 66 bus is in black. The dashed purple line is the disused Grand Junction Railroad – see below for more explanation.

North of the West Station site, the bus could still reach the platforms relatively easily, as the plan includes mapping new streets over the entire site. But to the south, the streets are narrow and practically unusable. All north-south through-traffic is funneled through Harvard Avenue – anything else would meander at speeds not much higher than that of walking.

What’s more, the zigzag in the image above comes from a detour to the center of Allston, called Union Square. The West Station site would move service farther away from Union Square, forcing it to either abandon its single busiest stop or have a more circuitous route. Serving both West Station and Union Square requires running two separate north-south bus routes sharing much of their southern legs, which is bad for frequency. Already the 66 runs every 10 minutes off-peak in one direction and every 14 in the other; this is worse than the minimum acceptable on such a key route, and any further reduction in frequency through route splitting is unacceptable.

Finally, the station design as shown in the presentations includes ample room for bus bays, so that buses can terminate at the station. Such a layout may be appropriate at the center of a small town with timed bus-rail transfers; in the middle of the city, it is pointless. The 66 crosses the rail tracks and has no use for terminal berths. Nor is there any need for terminating buses running parallel to the tracks – passengers could walk to another train station on the Worcester Line or on the Green Line.

The MBTA has never released any public plan for a bus redesign around West Station. It talks about intermodal transfers but refuses to give any details, and it’s likely these details don’t even exist yet. There are occasional excuses, such as intercity buses (why would they terminate there instead of continuing to South Station?), buses to Kendall Square (they don’t need bus bays either), and buses to Longwood (Longwood is south of the Worcester Line and would be better-served by a commuter rail-to-Green Line transfer near Fenway Park).

Track design for maximum conflict

The latest option for West Station is called the flip option. The diagrams below are from the June presentation, pp. 8-10, going west to east:

There are to be two bypass tracks (“WML Express”), located where the current mainline is. There are also to be three tracks with station access, both on the other side of the railyard. The tracks serving the platforms cross the bypass tracks in a flat junction, forcing dependency between the inbound and outbound schedule. The flat junction is not especially quick, either – it is a long ladder track, requiring inbound local trains to South Station to make two slow diverging moves in succession.

The MBTA is planning to spend tens of millions of dollars on station platforms in Newton turning the line into full double-track all the way from Boston to Worcester, freeing the schedule from such dependency, but at the same time it’s planning to add new conflicts.

While the diagrams label two tracks as freight tracks, there is little to no freight on that portion of the line. A freight rail spur in the area, serving Houghton Chemical, was just removed in preparation for the project. The line can and should be designed exclusively around the needs of regional passenger trains, for which the most important thing is continuous operation of double track, preferably with no flat junctions with oncoming traffic, and not any ancillary frills.

The Grand Junction tangential

The MBTA has grandiose plans to use the Grand Junction Railroad to allow trains from Allston and points west to avoid South Station entirely. The Grand Junction provides a bypass to the west of Downtown Boston, which currently sees no passenger service but is used for non-revenue moves between the South Station and North Station networks. There are periodic plans to reactive service so as to enable trains from the west to serve Cambridge and North Station instead. In the flip option, all local trains are required to go to the Grand Junction or switch back to the mainline using the ladder track.

Consult the following table, sourced to OnTheMap, for the number of jobs accessible within walking distance of the various station sites:

Station Walkshed boundaries Jobs
South Station Essex, Tremont, State, the harbor 119,191
Back Bay Hereford, Belvidere, Columbus, Arlington, Storrow 62,513
Kendall Binney, Third, Wadsworth, Memorial, Mass Ave, Windsor, Bristol 29,248
North Station Blossom, Cambridge, State, Prince, the river 33,232

Jobs accessible on the existing mainline outnumber ones accessible via the Grand Junction by a factor of about three. It is not technically sound to avoid city center on an urban rail line, much less a suburban one. Only if the line is a consistent circumferential line is there a good reason to go around the center.

A far-future subway duplicating the 66 route may succeed. The same may be true of a shuttle using the Grand Junction, but such shuttle may well need extensive new track – West Station is not necessarily the best south-of-Charles footprint (turning east toward BU to form a loop with a future North-South Rail Link is better). In contrast, the current plan for diversion of Newton trains toward a secondary job center and away from Downtown Boston has no chance of getting substantial ridership.

The railyard as an obstacle

For a project so focused on redevelopment, West Station does not do a good job encouraging construction in the area. It plans to keep the railyard in the middle, and even forces local and express trains to go on opposite sides of it. But the railyard is an obstacle not only to sound railway operations but also to redevelopment.

Building anything over rail tracks is complicated. New York supplies a few such examples: the link mentions the difficulties of Atlantic Yards, and to that I will add that the construction of the Hudson Yards towers cost around $12,000/m^2, compared with $3,000-6,000 for Manhattan supertall office towers on firma. Hudson Yards has managed to be financially successful, albeit with tax breaks, but it’s located right outside Midtown Manhattan. Allston’s location is not so favored. The cost penalty of building over railyards is likely to make air rights unviable.

There is still an extensive portion of the site that’s on firma. However, if the point is to maximize redevelopment potential, the city and the state must discard any plans for air rights. The railyard should go in order to increase the buildable area.

In lieu of parking at a railyard in a desirable near-center location, trains should circulate back and forth between Boston and Worcester. The MBTA keeps saddling itself with capital costs because it likes running trains one-way to Downtown Boston in the morning and then back to the suburbs in the afternoon, parking them near South Station midday. This is bad practice – trains are not just for suburban salarymen’s commutes. Urban infill stations in particular benefit from high all-day frequency and symmetric service. If the MBTA needs space for train parking, it should sell the railyard in Allston and charge Allston land prices, and instead buy space in Framingham and Worcester for Framingham and Worcester land prices.

West Station, done right

Thanks to delays and cost overruns, West Station is still in preliminary design. There is plenty of time to discard the flip option as well as the original plan in favor of a route that maximizes intermodal connections at minimum cost. A better West Station should have all of the following features:

  • A simple four-track design, either with two stopping tracks and two bypass tracks or four stopping tracks and two island platforms, depending on long-term plans for train timetables
  • High design speed, as high as the rest of the line for nonstop trains, as the tracks are straight and do not require any speed restriction
  • Retention of double-track rail service throughout construction, even at the cost of more disruption to the Massachusetts Turnpike
  • No at-grade conflicts in opposing directions: tracks should go slow-fast-fast-slow or fast-slow-slow-fast rather than slow-slow-fast-fast
  • No bus bays: crosstown buses (that is, the 66) should stop on the street crossing the station right above the tracks, with vertical circulation directly from the bus stop to the platform in order to minimize transferring time
  • Subject to site availability, platforms reaching Cambridge Street for a connection to the present-day 66 and a shorter walk to Union Square
  • Elimination of the railyard to make more room for development, and if the line needs more yard space, then the state should find cheaper land for it in Framingham and Worcester

There is no reason for such a project to cost more than past infill stations built in Boston, which have cost around $15-25 million, about the same range as Berlin. By removing unnecessary scope, the MBTA can make West Station not only cheaper and easier to build but also more useful for passengers. The idea of an infill commuter rail station in Allston is good and I commend the MBTA for it, but the current plan is overbuilt and interferes with good rail and bus operations and needs to be changed immediately, in advance of engineering and construction.

Climate Urgency

Every year that passes, climate change becomes a more urgent problem to solve: every year that emissions do not fall means that future emissions will have to fall even faster to avoid catastrophic global warming and ocean level rise. This aspect makes climate change different as an issue from air pollution, health care, education, etc., all of which can be solved tomorrow in approximately the same way as today.

Transportation is an increasingly important aspect of climate change. In the 1990s activists could focus on electricity generation, due to the prevalence of coal power in developed countries. Today, when coal has terminally declined in most of the developed world, and is controversial in China and India because of its severe air pollution emissions, the share of transportation in greenhouse gas emissions is higher, and still rising (see e.g. US data on PDF-p. 32 and UK data).

Fast decisionmaking

As the biggest challenge of urbanism and transportation shifts from local public health to global climate change, the need for mechanisms that enable rapid demotorization and reurbanization becomes more urgent. I wrote a lot about consensus urbanism in 2011, and a lot of what I said still works if the aim is long-term improvement of democratic decisionmaking through inclusion; in essence, the consensus process spends time on buying goodwill from various groups instead of money (through open or de facto bribes) or political capital (through controversial coercion). But if the goal is to prevent catastrophic climate change, then the value of time is high and will grow as the years go by and no action is taken, and thus the consensus process loses a lot of its appeal.

In lieu of slow attempts at consensus, there are two ways to implement policy fast: market pricing, and top-down coercion. In cultural theory terms, consensus is egalitarian, market pricing individualist, and coercion hierarchical; the fourth cultural bias, fatalism, is not really associated with any system, but rather with the government by exception that characterizes populism, and does not proceed in a particular direction.

The upshot is that governments should aim to spend money and political capital instead of time, and use governing mechanisms that facilitate rapid change. In areas where the market supports green decisions, for example urban real estate construction, it is necessary to remove restrictions on market activity. Where it cannot, for example any question of infrastructure, it is necessary to reduce delays, for example by removing the ability of individuals to sue over environmental reviews – decisions about environmental impact should be taken internally through a civil service.

Learn to say no

One of the biggest loci of opposition to the green transition is a culture war by an old guard that clings to a postwar vision of the good life that centers car ownership and either the suburbs (in the US and parts of Europe) or a small town that turned into a suburb (in the other parts of Europe). Waiting for the old guard to die off or otherwise slowing down the process of change to make it more palatable may work for other goals, such as reducing urban housing costs, curbing air pollution, and providing better mobility for people who already don’t drive. It does not work for climate change.

The upshot is that there are two valid strategies to deal with literally hundreds of millions of first-world citizens who stand to lose income, wealth, or social or cultural status from the green transition. The first is to buy them off, or at least buy off those who can be bought off without bankrupting the state. The second is to tell them no. No, we are not going to accommodate you: saving the planet is too important a goal, and turning your 20-minute car commute into a one-hour three-seat ride by a bus because you kept voting against trains is a price we are willing to pay, and even if you’re not willing to pay it, we don’t need you to vote for us.

This is easier in Europe than in the United States; Canada is somewhere in between. If NATO-Europe gets into a war with Russia tomorrow and bans personal car use the next day to conserve fuel for tanks, people will for the most part be able to adapt; the trains will get more crowded, but outside Paris and London, the main constraint on train capacity is rolling stock, which is cheap to make more of even in an environment of total mobilization. If the United States gets into a shooting war, it will not be able to do so – at most it may be able to organize car-sharing clubs as in World War Two, but even then, many weak-centered cities would cease to function.

Climate change is urgent but less urgent than a total war starting tomorrow, which gives some time for expansion of transit. There’s about a generation’s worth of time; in the same timeframe, Vancouver has turned itself from a postwar suburban hellscape into something resembling a transit city. However, two important caveats make a public works-only green transition impossible. First, there is political opposition to transit, especially cost-effective transit (for example, buses taking freeway lanes from cars rather than adding lanes to freeways). And second, without some combination of transit-oriented development and coercive taxes on fuel, public transport remains underutilized – a number of American cities have built ample urban rail but have far lower ridership than comparable European and Canadian examples. Rail expansion makes confrontational green politics more palatable; it does not remove the need for confrontation.

The one saving grace of this need for confrontational, risk-taking politics is that the status-anxious opposition is the same to everything: to urban redevelopment, to public transportation, to raising taxes on cars, and often even to a consensus-based process if this process empowers the wrong social classes or ethnic groups. Quite often this opposition is exceptionally loud and connected, but running against it, while risky, is not political suicide. California voted against expansion of rent control last year, congestion pricing proved popular in London and Stockholm after the initial controversy of implementation, carbon taxes in Sweden keep going up and emissions keep going down, the German Christian Democrats’ road warrior tendency is conservative rather than reactionary. The green movement should expect to lose battles; it should not expect to lose the war.

How France builds high-speed rail and how Spain builds subways

France and Spain have opposite approaches to cost containment. France spends time rather than money: informal political opposition in rural areas is hard to break – what the state will let the police do to suburban Arabs and blacks who protest brutality it won’t dare let it do to rural whites who protest trains despoiling their romantic Provence views – so the state painstakingly negotiates with the landowners. The resulting construction costs are reasonable: the 106 km LGV Est phase 2, with 4 km in tunnel, cost €2.01 billion euros in 2008 prices. However, the process takes a long time: in Provence, where placating the NIMBYs proved impossible, the resulting alignment is tunnel-heavy and expensive, and even though public debate goes back to 2005, the line will likely open well into the 2030s.

Spain takes the opposite approach. In the view of Manuel Melis Maynar, time is money, and the faster a project is completed, the cheaper it will be, as there will be less time for problems to accumulate. Madrid Metro awards contracts based on how fast construction can be completed as well as on the budget, and its internal planning process is designed around fast decisionmaking.

Spain builds infrastructure more cheaply than France, but that by itself is not enough to argue in favor of the Spanish approach. Spain does many things to curb costs that France does not, and the question of whether time and money are substitutes or complements occurs in many industries with different answers. In tech, there may well arise situations in which code can be written cheaply or quickly and ones in which delays add costs within the same project.

That the time or money question is delicate means that infrastructure builders need to cultivate enough expertise to be able to know when it’s one or the other and when it’s both or neither. However, that, by itself, has nothing to do with urgency; “work on building infrastructure more cheaply” is a good principle regardless of whether everything needs to be in place in 10 years or in 100.

What the urgency of climate change does mean is that there should be a bias against delays. In situations in which it is certain that time and money are substitutes, agencies should prefer to spend money, for example by buying off property owners and paying above market rates. In situations in which it is unclear, agencies should act as if time is money and aim to complete projects quickly even at the cost of budget overruns, rather than to complete them on a prescribed budget even at the cost of schedule slips.

That Spain has lower construction costs than France suggests that acting as if Spain is right and France is wrong is not likely to have too many drawbacks. It may require some internal cultural changes in how infrastructure builders think, and possibly regulatory changes streamlining environmental reviews, but it’s likely to either save money in the long run or only cost a little more.

Costs are Rising

This is a partial data dump from an in-progress database I’m compiling for subway construction costs around the world. The key point is that costs are rising: in cities with enough historical data points we can see a secular increase in construction costs. The difference between expensive cities like New York and London is that their costs have been high for a while, whereas cheap ones like Madrid and Seoul are seeing construction cost growth from a very low basis.

As a note of caution, while growth in costs seems universal, the rate of growth is not the same everywhere. Some cities, most notably Singapore and Toronto, have seen a cost explosion in the last 15-20 years; others, most notably Seoul, have seen only a moderate increase in costs.

I also urge readers to look at some 20th century historical costs here, as in this post I am going to focus on the very end of the 20th century and the 21st century.

Paris: the original Metro Line 14 cost €1.174 billion for 9.2 km (link), built between the 1990s and 2007; deflated to 2012 euros, the baseline year used for ongoing extensions, this is around €160 million per km. More recent projects in Paris cost around the same, including the Line 1 extension and Grand Paris Express – but those are mostly suburban extensions, whereas M14 had to go underneath central Paris.

Toronto: Jonathan English, who has been working with me on Canadian construction costs issues, notes that the Sheppard subway, opened in 2002, cost C$1 billion for 5.5 km, but ongoing projects are far more expensive. The one-stop 6.2 km Scarborough subway is projected to cost around C$3.5 billion not including extra items for interfacing with the existing rapid transit line along the same alignment, which is to be dismantled. Not only is the nominal cost 3 times higher – and the real cost is still around twice as high – but also the Scarborough subway has just one station to be constructed, which makes it a simpler project.

Montreal: Jonathan equally looks at the cost explosion in Montreal. The Laval extension was built in the 2000s and opened in 2007, costing C$742 million for 5.2 km and 3 stations, or C$143 million per km, crossing under the Rivière des Prairies. In contrast, an extension of the Blue Line planned for next decade is to cost C$3.9 billion for 5.8 km and 5 stations, or C$672 million per km – and even adjusting for inflation only reduces the cost differential to a factor of about 3.

(In case Canadian readers wonder why I’m not covering Vancouver, even though the escalations on the Broadway subway have pushed its per-km cost well beyond that of the Canada Line, the reason is that the Canada Line was built cut-and-cover whereas the Broadway subway will be bored.)

Singapore: Singapore’s cost overrun history in the 21st century has been unusually severe. Built mainly in the 2000s, the original Circle Line cost S$10 billion for 33.3 km, or S$300m/km, an overrun of 50% over the original budget. Subsequently, the Downtown MRT Line, built from 2008 to 2017, cost S$21 billion for 42 km, and the Thomson Line S$24 billion for 43 km. The Thomson Line has a complex interchange at Orchard, but also long segments in easy suburban areas – Upper Thomson Road, after which it is named, is very wide and borders modernist housing projects on one side and a forest on the other. Moreover, the last stage of the Circle Line, completing the circle, is to cost S$4.85 billion for 4 km and 3 stations – depending on PPP rates, it may be the first line outside New York to cross the US$1 billion/km line.

Seoul: South Korean costs are fairly stable. JRTR has data for the Seoul Metro going back to its start in the 1970s. After adjusting for inflation, costs were initially about $70 million per km, and rose gently to $80-90 million. The cost increases are continuing, albeit at a slow pace. As best as I can tell, the 2020s’ expansion program is budgeted at about $110 million per km in PPP terms.

Madrid: in the 1995-2003 period the city built tunnels for very low costs. The 1995-8 program cost $55 million per km, all underground, and the 1999-2003 program cost €3.147 billion for 74.7 km, 77% underground, around $52 million per km based on the era’s PPP conversion rate. In the conditions of 2010 this would be roughly $65-70 million per km – but the Line 2 extension, built 2008-11, cost €315 million for 4.6 km and 4 stations, and the Line 9 extension, built 2009-15, cost €191 million for 3 km and 2 stations, about $80-90 million per km.

Update: since people have asked for high-speed rail data, it confirms the same story. Ferropedia has costs in Spain, which have risen from €4.88 million per kilometer in 2001 terms for Madrid-Seville, which opened in 1992, to about €15-20 million per kilometer in 2006-7 terms for subsequent lines from the 2000s and 2010s. France displays the same history of escalation: built in the early 1980s, the LGV Sud-Est cost €5.5 million per km, much less than the late 1980s and early 1990s’ LGVs Nord and Atlantique (which cost €10 million), let alone this decade’s LGV Est (which cost €16 million for Phase 1 and €19 million for Phase 2); all of these lines are through comparable terrain, with very little to no tunneling.

Scooters

Three weeks ago, the consultancy 6t released a study about dockless e-scooters in France. The study is available only in French but there is an executive summary in English. It has convenient demographic profiles of e-scooter riders in Paris, Lyon, and Marseille, and generated some media controversy over the fact that scooters are barely displacing car trips – rather, they’re replacing trips by foot or public transit. This is on top of calls for greater regulations of the mode in multiple countries, not just by NIMBYs but also by serious urbanists like Streestblog’s Angie Schmitt; I was alerted to the study in the first place by Jonathan Rosin, who proposes regulations requiring geofencing to prevent riding on the sidewalk.

And yet, there’s something interesting about scooters and transit in the study, which suggests to me scooters have a positive role to play in a transit city. On p. 80, figure 48 shows combination of scooters with other modes. Out of about 4,000 respondents, 886 say they used scooters in combination with another mode – and of the latter, 66% used it in combination with public transportation.

How worried should we be about rider behavior?

Not really. There is an American discourse concerning dockless transportation that complains about clutter, scooter-pedestrian conflict, and nuisance scooters or bikes left on the sidewalk. The study itself discusses the regulations of e-scooters in various countries. In Britain and Italy e-scooters are legally classified as motor vehicles, which is effectively a ban, and in the US there are onerous regulations such as a requirement for a driver’s license and a minimum age of 18, such as in California. In France, Germany, Switzerland, Austria, Sweden, and Denmark, regulations are laxer, e.g. in Germany the minimum age is 14 and the e-scooter is treated as a bike.

The sort of clutter that Americans complain about was not evident to me in Paris, one of the largest markets in the world for e-scooters as well as bike share (it still has the largest bike share program in the world outside China). As far as I could tell, scooters were mostly used around Nation for recreational trips – at the very least, people did not preferentially leave them right at the Metro and RER station, and I did see a fair number of dockless vehicles (I forget if just bikes or also scooters) at the Bois de Boulogne. Central Paris had a higher density of scooters, but they too did not seem to clutter on the street, and I don’t remember ever having had to dodge a scooter even though people did ride on the sidewalk.

At least at eye level, the lax regulations France does have – the minimum age is 8, cities may choose to permit or prohibit riding on the sidewalk, riding on all streets with speed limit up to 50 km/h is required – appear sufficient. The American, British, and Italian approaches are too draconian and only serve to discourage this mode of transportation.

Gaps in the transit system

Pp. 111-4 have tables describing mode switching. Few of the scooter users would have traveled by car if the scooters hadn’t been available, only 8% including taxis and TNCs (“VTC” in French). In contrast, 46% would have walked and 32% would have taken public transportation.

But is this even a problem? The same tables have the average transit-to-scooter switcher gaining 5 minutes, taking 19 minutes instead of 24. On short trips, scooters are useful for filling little gaps in the regional public transport network. Maybe the origin and destination are not well-connected by Metro, as is for example the case for Nation and much of the Left Bank, so that a transit trip would require transfers. On a poll suggesting non-mutually-exclusive options for why people choose scooters over transit, 68% say it’s nicer, but 44% say it’s faster and 39% say it’s direct. From the perspective of the transit agencies, a mode that makes certain crosstown trips easier without changing trains at Chatelet is a net positive, as it decongests the station as well as other complex transfer points.

According to Owen Gutfreund’s book 20th Century Sprawl, in the 1900s and 1910s the American railroads were supportive of road expansion. To the railroads, cars were a natural complement to trains, extending their range beyond that of a horse or bicycle. Of course, soon the cars turned into competitors, once roads improved to the point of allowing longer-distance travel. But scooters, limited to 25 km/h, do not have that capability. The mode of transportation most comparable to the e-scooter, the bicycle, coexists with a solid regional and intercity rail network in the Netherlands.

Ban cars

The ultimate goal of the green movement in general and of public transit activism in particular should be to ban cars, or else get as close as possible to banning them. Modes of transportation that are not cars that provide alternative functionality to cars are almost always a good idea in this scheme.

Trains are an excellent alternative for long trips, that is out-of-neighborhood trips for such purposes as work, school, citywide social events, and intercity travel. Shorter trips are dominated by walking in transit cities. However, there are two important caveats for the idea of doing short trips on foot. First, there is a genuine in-between region in the 2-4 km range. And second, people with disabilities may not be able to walk long distances, which lowers the upper limit from the 1-2 km range to a much shorter point, perhaps 500 meters – and if their disabilities do not require the use of a wheelchair, then they may well find scooters an acceptable alternative.

In Paris itself, which dominates the survey, scooters are not replacing cars, for a simple reason: few trips in Paris are done by car in the first place. But a robust scooter network can expand out of the city into suburbs with higher present-day car usage, and those suburbs can then become ever more walkable thanks to the displacement of cars by greener modes of travel.

Assume Nordic Costs: London Edition

A month ago I made maps proposing some subway and regional rail extensions in New York and noting what they would cost if New York could build as cheaply as the Scandinavian capitals. Here is the same concept, but with London rather than New York. Here is everything in a single large map:

A full-size (74 MB) map can be viewed here.

Solid lines are existing or under construction, that is Crossrail and the Battersea extension; proposed lines are dashed. Commuter rail lines, that is Thameslink, the soon-to-open Crossrail, and four additional Crossrail tunnels labeled 2 through 5, are always depicted as having separate stations from the other modes, to avoid confusion where one Crossrail station has connections to two adjacent Tube stations (such as Farringdon-Barbican and Moorgate-Liverpool Street). It has many additional interchanges between lines and branches, including some that were left out on purpose, like a Crossrail 1 connection to Oxford Circus, omitted from the under-construction line to discourage riders from using the oversubscribed Victoria line; with four more cross-city lines, the capacity problems would be lessened substantially.

The overall picture is sparser than my New York map. The total projected cost of all of these projects, including some allocated for redoing stations on commuter branches to be given to Tube lines, is £6.8 billion, compared with $37 billion for the New York maps. The reason is that unlike New York, London already has excellent coverage thanks to extensive branching – what it needs is core capacity, which consists of city center tunnels that have high cost per kilometer but need not be long.

There is considerable overbuilding planned in London. Crossrail 2 as depicted on my map is a 6.5 km tunnel between the approach to Victoria Station and the approach to Kings Cross. But as planned, Crossrail 2 extends to a long tunnel parallel to the South West Main Line, a four-track line in a right-of-way that could if truly necessary accommodate six, as well as a long tunnel going north to take over the Lea Valley Lines, which on my map go into Crossrail 5. With gratuitous suburban tunnels and extremely high British construction costs, the budget for Crossrail 2 is around £30 billion, about 20 times what Scandinavia might spend on such a project. Even allowing for the possibility that crossing under three lines at once at Bank is more complex than crossing under two at T-Centralen, this is a difference of a full order of magnitude, counting both total required tunnel length and cost per km.

In addition, there is network simplification. On the Tube this consists of segregating the Northern line’s Bank and Charing Cross branches (already in planning pending the Battersea extension and reconstruction of Camden Town) and through breaking the Circle line into separate Metropolitan and District lines. The latter was estimated by a British blogger to cost £5 billion, based on a rubric in which the Met/District transfer at Aldgate (or Tower Hill) should by itself cost £1 billion; Crossrail and Second Avenue Subway stations cost around half that much, and the more complex T-Centralen and Odenplan stations on Citybanan cost less.

On mainline rail, the service plan is supposed to be deinterlined, as is Transport for London’s long-term goal. The slow tracks of the various mainlines feeding into Central London turn into Crossrail branches, or occasionally Underground extensions, such as Hayes and the Hounslow Loop. The fast tracks stay on the surface to avoid interfering with high-frequency regional metro service. For historic reasons Thameslink mostly stays as-is, with a combination of fast and stopping services, but the curve toward London Bridge should not be used – instead, passengers should have access to Crossrail 3 plus interchanges to the City at London Bridge and a new infill station at Southwark.

London owes it to itself to understand why its construction costs are so high that instead of solving its transport capacity problems with multiple cross-city tunnels in a decade, it’s taking multiple generations to build out such a system. There’s a lot of ongoing discussion about the last-minute delays and cost overruns on Crossrail, but the absolute costs even before the overrun were very high, the highest in the world outside New York City – and Crossrail 2 is set to break that record by a margin.

Teacher Housing and the Rot

I recently saw that San Francisco is considering fast-tracking residential development dedicated to teacher housing. There are quibbles between the moderate mayor and the progressives on city council (“Board of Supervisors”) over the exact structure of the housing subsidies, but both sides agree at least in theory that it should be easier to build housing for teachers; for more background, see article here and Twitter back-and-forth here. I bring this up because it’s an example of bad governance at the local level in the US, one that sends everyone the message, “you should get more clout to bribe politicians.”

The basic problem is that market-rate housing in San Francisco is extremely expensive; in the Mission, a two-bedroom apartment rents for about $5,000 or $5,500 a month. There’s rent control, but it requires one to have lived in the city for a very long time – friends who have lived in the city since the mid-2000s pay around $2,700, which is borderline on a teacher’s salary. Usually the city’s local notables don’t have to care about whether housing is affordable to people in intermediate professions, since our rent is their property values, but “teachers can’t afford housing here” could be a rallying cry for more housing. Thus, they feel like making an exception.

Making an exception is the hallmark of populist governance. In a system with not much rule of law and no trust that there will ever be rule of law, people don’t ask for better rules but to benefit from exceptions. That various exporters threatened to leave Britain over Brexit did not faze Theresa May – every time a company people didn’t hate made such a threat, she offered special subsidies to stay no matter what would happen with the trade agreement with the rump-EU.

The problem with populism is that it sends the message, invest in political marketing and not in productivity. A company that sees that San Francisco is subsidizing housing for teachers in preference to other workers with similar pay and skill level – clerical workers, social workers, lab techs – gets a clear incentive to give its workers more political prestige through political contributions, sponsorships of events the local politicians are interested in, etc. It faces less pressure to invest in its productivity and pay its workers better, since housing is not allocated by market pricing but by political whims.

Under liberal governance, if San Francisco wishes to give its teachers perks, it can pay them better. Programmers get paid $110,000 a year plus benefits (stock options, good health insurance, free food), and the city can if it wants raise taxes and pay teachers similarly; if it can commit to maintaining such high pay indefinitely it can ensure the profession will get more prestige and attract people who otherwise would be writing code for how to sell user data to advertisers slightly more efficiently.

However, a tax hike might fall on the local homeowners and on other rich people who have invested a lot of time and money in obtaining political influence. To avoid burdening the powerful, the city can’t do this – it has to come up with some one-off bespoke deal for teacher housing, rather than permitting more housing across the board and also raising salaries to be competitive with those of the private sector.

Improving the quality of governance requires making it harder for politicians to create such deals. The original YIMBY praxis of state preemption laws is one way to do this: it completely takes local notables out of the loop. While the YIMBY groups on the ground in California don’t go further with this, their favorite state politician, Scott Wiener, who represents San Francisco in the State Senate, is consciously trying to form an informal state party with some ideological coherence based on relevant state issues, led by the question of housing.

It may be prudent to refine this preemption doctrine by interfering with local rules that favor some groups over others in housing. Thus the state should pass a preemption law that forbids dedicated housing for teachers, cops, or other charismatic professions, and requires all housing to be allocated by market pricing, or, failing that, by a clear process of rent control, such as waitlists or income limits. Private actors may continue to buy and sell housing based on their wishes, subject to the usual anti-discrimination law, but municipalities may not use incentives such as subsidies, tax breaks, access to public land, or special fast-tracking of approvals. Such a law may well succeed in the state legislature – unlike the SB 50 process preempting zoning restrictions, this law would not be nakedly offensive to the privileged group of suburban homeowners who managed to scuttle SB 50.

It is not really possible to develop rule of law in an environment in which powerful people can easily circumvent the rules. A city that can offer a way out of an onerous permitting regime to people who make it attractive offers – that is, bribery – has no incentive to make the permitting regime easier, and a powerful incentive to keep it as it is. If building housing becomes easier, politicians lose the ability to extort community benefits by threatening to withhold permits. And if there is a way out for socioeconomic classes that demagogues can’t dismiss as gentrifiers, transients, and rootless cosmopolitans, then politicians gain the ability to threaten everyone else, while employers as well as nonprofits get a powerful message that they should pay more bribes. It’s a win-win for everyone except the hapless residents governed by such corruption.

The question is whether area YIMBYs are willing to leverage the one point of power they do have – namely, their connection to nationwide ideological networks that the local notables of these cities pay lip service to. Out of four New York Times op-ed writers who online liberals like, two (Paul Krugman, Jamelle Bouie) have openly called for more housing in cities, and two (Charles Blow, Michelle Goldberg) have never opined on this issue; NIMBYs have ample local power but little national clout. YIMBYs have this advantage and need to press it to completely sideline machine politics and personality politics – that is, to form a coherent, identifiable political party in California (or New York, or Massachusetts) contesting state and local elections, and if winning local elections without assimilating to the local rot is not possible then work to delegitimize government below the state level as irredeemably corrupt.

Job Centralization in the City

The table below collates job centralization not by CBD as in this post but by central city. Parisian data comes from INSEE, here and here; American data comes from Wikipedia for population and OnTheMap for job counts. In general, I tried making the central city definition about 18% of the metro area to be comparable with Paris, but there is still a lot of variation, so this table should absolutely not be read as a ranking of metro areas by job centralization.

Metro area Population Jobs Central city Central pop’n Central jobs Central job share
Ile-de-France 12,082,144 5,682,048 Paris 2,206,488 1,797,745 31.6%
New York 19,979,477 8,364,410 Manhattan, Brooklyn 4,313,498 2,905,675 34.7%
Los Angeles 13,291,486 5,372,008 Downtown LA to Santa Monica ~1,500,000 1,051,648 19.6%
Chicago 9,498,716 4,142,542 Chicago ex-O’Hare 2,705,994 1,198,562 28.9%
Dallas 7,539,711 3,146,973 Dallas 1,345,047 809,077 25.7%
Houston 6,997,384 2,791,647 Inside 610 + Uptown ~650,000 749,661 26.9%
Washington 6,249,950 2,717,790 District, Arlington, Alexandria 1,100,496 859,751 31.6%
Miami 6,198,782 2,308,048 Miami, Miami Beach 563,221 324,260 14%
Philadelphia 6,096,372 2,570,460 Philadelphia 1,584,138 628,423 24.4%
Atlanta 5,949,951 2,374,233 Fulton County 1,050,114 780,259 32.9%
Boston 4,875,390 2,381,555 Boston, Cambridge 805,234 687,237 28.9%
Bay Area 4,729,484 2,121,580 San Francisco 883,305 642,375 30.3%

The Sunbelt

There appears to be a fair amount of job centralization in the Sunbelt cities, right? In Metro Atlanta, Fulton County has a slightly higher proportion of regional jobs than Paris with a slightly lower share of residential population.

But actually, no. Absolute densities matter in addition to relative centralization of jobs versus residences. In Houston and Los Angeles the central areas are drawn to encircle the downtown and near-downtown job centers – both cities preferentially annex suburban job sites so using municipal boundaries is not useful. A hefty share of area jobs are in these centers, especially in Houston. But ultimately it’s still not a lot of jobs in a very large land area, around 300 square kilometers for both, compared with 100 for the city of Paris or for San Francisco. Fulton County is vast, and the jobs are distributed all over Atlanta and its northern suburbs within the county.

Houston is a particularly good example of monocentrism with a weak center. There are not a lot of strong suburban job centers in Houston – nothing like Silicon Valley, Downtown Newark, the Route 128 corridor, La Defense, Burbank, or Tysons Corner. The city itself has about two thirds of area jobs, thanks to selective annexations. But the share of the CBD in area jobs is low, just 150,000 jobs in the 45/69/10 beltway, or 5.3% of area jobs. Outside the CBD job density plummets, as the outlying job centers making the difference between 5.3% and 26.9% are located at haphazard locations all over 610.

Older American cities

The extent of centralization in the Northeast, Chicago, and San Francisco is greater. New York in particular is a lot like Paris, with about a third of area jobs in a high-density contiguous blob consisting of less than one fifth of regional population. It has nothing like La Defense in the suburbs, but its suburban job centers, while much smaller, include some recognizably dense ones, especially Newark and the Jersey City waterfront. One needs to go well into suburbia to see the difference between Paris, where the suburbs have a structure of density with mid- and high-rise residential development as well as offices next to train stations, and New York, where the job centers in farther-out suburbia, like Central Jersey, have no such structure and are located exclusively based on auto access.

Boston, Washington, and San Francisco all have varying degrees of centralization. I mentioned last year that Boston is increasingly an example of European-style job sprawl, in which jobs spill over from the CBD to nearby areas rather than to faraway office parks. New York has long had such spillover – Long Island City is such a job center, and may at this point have more jobs than Downtown Brooklyn; the Jersey City waterfront is another such example, as is the growth of the Meatpacking District around Google. In Boston the equivalents are Kendall Square and the Seaport; in San Francisco it’s SoMa; in Washington it’s jobs in Arlington around the Orange Line, where older TOD was residential.

Chicago and Philadelphia are the least centralized. Chicago has a well-defined supertall skyline with about 500,000 people working in or near the Loop. But outside that central area, job density craters. Chicago’s share of metro area jobs is about 1.5% higher than its share of metro population, and if we remove the airport, surrounded by suburbia, this difference drops to 0.5%. Philadelphia’s share of metro area jobs is actually lower than its share of metro area population by 1.5%. In these regions, if you’re not working in city center, you’re working at an office park in a middle-class-to-rich suburb built without regard for the area’s vast legacy mainline rail network.