Dispersal of Urban Networks is Bad

The debate over upzoning has reached Paul Krugman, who is a strong supporter of liberalization (and an opponent of rent control), on the grounds that rich cities like New York and San Francisco are hotbeds of productivity and people should be allowed to move to them in greater numbers. Per Krugman, zoning rules in rich cities force people out, so instead they live in environments where they are less productive and thus earn lower wages, such as the Southern US. Dietrich Vollrath, an economist studying economic growth, makes a different suggestion:

Of course, there is an equivalent solution – move everyone in San Francisco to Houston or Atlanta. The reason SF is the most productive city is not because of some fixed, inherent quality of the location at 37.78 degrees North, 122.41 degrees West. It’s certainly not because of it’s fantastic summer climate. San Fran is the most productive city because it so happened that a unique collection of nerds coalesced there starting in the 1960’s. More nerds were attracted to the bright, shiny things that the original nerds were making, and now I have an iPhone. But here’s the thing about nerds – they are easy to move. You can easily strap one to a dolly and wheel them anywhere you want.

This is the economic equivalent of proposals for population dispersal used in discussions of poverty: urban renewal tends to involve such dispersal, with negative effects on community life, social support networks, and crime. (See for example what I wrote of proposals on the Israel left to disperse black refugees away from South Tel Aviv; while I fingered just one political party, the others seem to believe the same today.) Of course, the people Dietrich characterizes as nerds are not oppressed and are not going to turn to crime because of lack of opportunity, but they will not be as productive in Houston as they are in San Francisco, for similar reasons.

The key to the Bay Area’s success in the tech sector is not that it has people who came from all over the world, who could equally congregate elsewhere. On the contrary, as per a Wired infographic, tech giants tend to hire locally: the top universities feeding Silicon Valley firms are in the Bay Area (including San Jose State, and not just Stanford and Berkeley), and the top university feeding Microsoft is the University of Washington. The Bay Area, and to a lesser extent Boston and Seattle, has a culture that propels people with interests in science and engineering toward programming. New York’s culture is different, and propels them to finance. In addition to different regional cultures, there are also university cultures: Harvard may be in Cambridge, but is far less important as a tech feeder than MIT, with fewer than half as many grads per capita going to Silicon Valley.

Dispersing people away from the Bay Area means dispersing them toward regions in which the business and social networks do not favor the same activities, and do not reward them as much. Houston has a core of nerds working for NASA, who may be interested in working for private tech firms that find themselves priced out of Boston and San Francisco. But those nerds are used to what is presumably a totally different business culture. If these private tech firms are started by local Houstonians, then they will have a business culture familiar to Houstonians, and alien to any San Franciscan they hire.

People in the software sector have specific ideas about how to do things, reinforced by what works in their industry; as a result, their ideas regarding public transit, a mature industry in which immense capital requirements and routinized tasks make the modern startup model inapplicable, are often painful and wrong, as I’ve ranted here, here, and here, and as Jarrett Walker has ranted here, here, here, and here. This also goes in the other direction – a corporate culture built around mature technology is unlikely to create innovative smartphone apps. If Uber were a Washington firm, it would be better at lobbying for regulations that would retroactively legalize it and give it favorable insurance requirements, but then it wouldn’t have invented a new way of hailing cabs in the first place. This is historically related to the growth of the Bay Area as a tech hub in the first place, as explained in Regional Advantage: Boston got there first, but its traditional corporate hierarchies couldn’t innovate at the same rate as the flatter networks of the Bay Area. One of the candidate US Sunbelt cities for poaching the tech cluster in Dietrich’s proposal, Dallas, is the home of Texas Instruments, where the integrated circuit was invented, but it is today a tertiary tech cluster because of this problem of corporate culture.

But all this assumes the tech cluster would even exist in whichever low-cost city it moved to. There is no real reason for it to do so. If high prices lead to an exodus of tech firms from the Bay Area, the community will dissipate rather than relocate. The richest members of it – Google, Apple, Facebook, the major venture capital firms – have the money to stay in the Bay Area, and to pay employees extra to cover rent in San Francisco. It’s the weaker members, typically startups, who are in danger of being priced out, and they are probably going to move to many different cities, depending on personal ties.

Once away from the Bay Area, they’d have to not only contend with a new urban culture, but also deal with it as a small minority. The same factors that cause unassimilated minorities to stay in their ethnic enclaves, even when discrimination is not a factor (ultra-Orthodox Jews own much of the housing in Brooklyn even outside their enclaves), favor clustering of industries. A hundred thousand Bay Area nerds could possibly remake parts of Houston in a way that’s favorable for their economic production; ten thousand could not. They’d have to rely on local venture capital firms, which are almost certainly looking for different business models. They’d have to recruit new workers from universities with student populations with different interests, expectations, and summer internships. In analogy with forced assimilation of ethnic whites in early- and mid-20th century America, they’d assimilate, to a nationwide economy with much lower per capita income than is normal in their sector.

The second assumption is that, if Houston became the next San Francisco, it would eventually accommodate a larger pool of tech workers. This is not necessarily true: Houston has a liberal process for permitting new construction, including of apartments, but only subject to onerous parking minimums and setbacks, which it doesn’t call zoning but which appear on the zoning codes of cities that do have zoning. It makes it easy to build new sprawl, but not so much new density, and eventually, the sprawl is going to lead to long commutes, producing the same rising prices in the deed-restricted and de facto zoned center that are seen in San Francisco and other coastal US cities. Fast-growing exurbs exist at the edge of metro areas everywhere in the US; the reason there’s not much growth in the expensive metro areas is that these exurbs are so far from the center that the commutes are too long for people to bother.

This is worse outside the US. The exact same problems of high costs coming from high rents exist in most other developed countries, only they don’t have fast-growing cities as large as Dallas, Atlanta, or Houston. Stockholm is rent-controlled and, judging by the almost complete absence of high-rises, tightly zoned; there are likely many people who’d move here if market-rate rents were in line with construction costs, but instead they have to live in Norrland, Malmö, and other peripheral areas. Houston’s metro area is not much smaller than the Bay Area’s, but in Europe, the cheaper cities are far smaller than their respective countries’ more expensive cities (often the capitals). The business networks formed in those cities would have to be smaller and less specialized. This is similar to the situation in the US involving New York’s great size, except that smaller Boston, San Francisco, and Washington achieve equivalent or higher incomes, so Houston should not be penalized for not being a hypercity.

The only problem is that Europe has no Houston. Its cheap larger cities, such as Naples and Berlin, have high unemployment and low incomes. Browse per capita income net of rent (see definitions here) by regions of European countries here, and per capita income by US metro area or county here. Houston and Dallas are both richer than the US average, and Atlanta is about 10% poorer. Berlin is 20% poorer than the German average, and 40% poorer than Munich’s region, Upper Bavaria, part of a general pattern of East-West inequality, driving a flight from the former East Germany to the West. In Italy, with its north-south divide, the southern regions, including Naples’ Campania, are about half as rich as Milan’s Lombardy, leading to a similar pattern of more immigration to the north. To tell people to move there and start their own social networks is, in American terms, like to tell people to move to Mississippi. In the smaller European countries we do not see such large income gaps, but we also do not see large metro areas with affordable housing.

Now, those rich capital cities (or non-capital ones, in the case of Milan) usually have rent control, which is how they achieve such high levels of per capita income even after subtracting rents. The people from the provinces who might have moved to them if they were cheaper do not benefit from this rent control, and have to either wait years for an apartment to open up or pay exorbitant rents. This reduces interregional mobility, and is a predictable side effect of a system in which housing is allocated to people based on how long they’ve lived in the city.

The idea of making do with tight zoning restrictions in some cities by bypassing them and developing alternative networks around the Houstons of the world is attractive, but fatally flawed. San Francisco – and Munich, and Stockholm, and Paris, and the major cities of Switzerland – is productive for reasons that go beyond individual denizens, who can be moved elsewhere freely. This, ironically, goes against the grain of San Francisco’s tech culture, which uses technology to overcome regulatory failures: NextBus and similar apps try to overcome byzantine bus schedules, Uber and Lyft try to avoid taxi medallion restrictions, AirBnB tries to overcome hotel regulations. The tech sector’s thinking is often that bad regulations should be subverted rather than reformed. It works in some cases and fails in others; in the case of zoning, it’s doomed to failure, on every level, from trying to shrink dwelling sizes to fit more people in, to recreating business clusters in cities with less awful zoning rules, as Dietrich proposes. Like Dietrich, I am pessimistic about the ability of the US and most European countries to reform their zoning rules to allow more intense urban development (or about Japan’s ability to allow more immigration), but on the other hand I also don’t think there’s any workaround avoiding a massive political fight about it.

Zoning and Market Pricing of Housing

The question of the effects of the supply restrictions in zoning on housing prices has erupted among leftist urbanist bloggers again. On the side saying that US urban housing prices are rising because of zoning, see anything by Daniel Kay Hertz, but most recently his article in the Washington Post on the subject. On the side saying that zoning doesn’t matter and the problem is demand (and by implication demand needs to be curbed), see the article Daniel is responding to in Gawker, and anything recent by Jim Russell of Burgh Diaspora, e.g. this link set and his Pacific Standard article on the subject.

This is not a post about why rising prices really are a matter of supply. I will briefly explain why they are, but the bulk of this post is about why, given that this is the case, cities need to apportion the bulk of their housing via market pricing and not rent controls, as a matter of good political economy. Few do, which is also explainable in terms of political economy.

But first, let us look at the anti-supply articles. Gawker claims that San Francisco prices are rising despite a building boom. We’ll come back to this point later, but let me note that in reality, growth in housing supply has been sluggish: Gawker links to a SPUR article about San Francisco’s housing growth, which shows there was high growth in 2012, but anemic growth in previous years. The Census put the city’s annual housing unit growth last decade at 0.8%. In New York, annual growth was 0.5%, as per a London study comparing London, Paris, New York, and Tokyo. In contrast, Tokyo, where zoning is relatively lax, growth was 2%, and rents have sharply fallen. The myth that there is a building boom in cities with very low housing unit growth is an important aspect of the non-market-priced system.

Jim’s arguments are more interesting. He quotes a Fed study showing that housing vacancies in the most expensive US cities have not fallen, as we’d expect if price hikes came from lack of supply. (In San Francisco, vacancies went up last decade, at least if you believe that the Census did not miss anyone.) This is too not completely right, because in Los Angeles County, as noted on PDF-page 18 here, vacancies did recently fall. But broadly, it’s correct that e.g. New York’s vacancy rate has been 3% since the late 1990s, as per its housing surveys. But I do not think it’s devastating to the supply position at all. The best way to think about it is in analogy with natural rates of unemployment.

Briefly: it’s understood in both Keynesian and neo-classical macroeconomics that an economy with zero employment will have high and rising inflation, because to get new workers, employers have to hire them away from existing jobs by offering higher wages. There is a minimum rate of unemployment consistent with stable inflation, below which even stable unemployment will trigger accelerating inflation. In the US, this is to my understanding about 4%; whether the recession caused structural changes that raised it is of course a critical question for macroeconomic policy. A similar concept can be borrowed into the more microeconomic concept of the housing market.

There’s also the issue of friction, again borrowed from unemployment. There’s a minimum frictional vacancy, in which all vacant apartments are briefly between tenants, and if people move between apartments more, it rises. For what it’s worth, the breakdown of 2011 New York vacancies on pages 3-4 by borough and type of apartment suggests friction is at play. First, the lowest vacancy by borough is 2.61%, in Brooklyn, not far below city average. Second, the only type of apartment with much lower vacancy than the city average is the public housing sector, with 1.4% vacancy, where presumably people stay for decades so that friction is very low; rent-stabilized units have lower vacancy than market-rate units, 2.6% vs. 4.4%, which accords with what I would guess about how often people move.

So if high rents are the result of supply restrictions, and it appears that they are, the way to reduce them should be to relax zoning restrictions. If this is done, then this allows living even in currently expensive areas without spending much on rent. Urban construction costs are lower than people think: New York’s condo average is $2,300 per square meter, and London’s is not much higher, entirely eaten by PPP conversions; Payton Chung notes the much higher cost of high-rises than that of low-rises, but the cost of high-rise apartment buildings is still only about $2,650/m^2 in Washington, and (using the same tool) about $3,100 in New York, and at least based on the same tool, mid-rises are barely any cheaper. For US-wide single-family houses, construction costs are 61.7% of sale prices, but the $3,100 figure already includes overheads and profit. Excluding land costs, which are someone else’s profit, construction, profit, and overheads are 92.5%; so let’s take our $3,100/m^2 New York high-rise and add the rest to get about $3,300, which is already more than most non-supertall office skyscrapers I have found data for in other major cities. The metro area appears to have a price-to-rent ratio of about 25, and with the caveat that this may go down slightly if the city gets more affordable, this corresponds to a monthly rent of $11 per square meter, at which point, a 100-m^2 apartment, sized for a middle-class family of four, becomes affordable, without subsidies, to families making about $44,000 a year and up, about twice the poverty line and well below the median for a family of that size. If we allow some compromises on construction costs – perhaps slightly smaller apartments, perhaps somewhat lower-end construction – we could cover most of the gap between this and the poverty line.

But given that demand for housing at prices that match construction costs, there has to be a way of allocating apartments. Under market pricing, they’re allocated to the highest bidder. If there is a perfectly rigid supply of 2 million housing units and a demand for 4 million at construction costs, the top 2 million bidders get housing, at the rent that the 2 millionth bidder is willing to pay.

I do not know of any expensive city with low home ownership that uses market pricing: too many existing residents would lose their homes. High home ownership has the opposite effect, of course – Tel Aviv may have rising rents, and high price-to-income ratios, but since home ownership is high, the local middle class is profiting rather than being squeezed, or at least its older and slightly richer members are.

Instead, cities give preference to people who have lived in them for the longest time. Rent control, which limits the increase in annual rent, is one way to do this. City-states, i.e. Singapore and Monaco, have citizenship preference for public housing to keep rents down for their citizens. Other cities use regulations, including rent control but also assorted protections for tenants from eviction, to establish this preference. Instead of market pricing allocation, there is allocation based on a social hierarchy, depending on political connections and how long one has lived in the city. People who moved to San Francisco eight years ago, at age 23, organize to make it harder for other people to move to the city at this age today.

Going to market pricing, which means weakening rent controls over the next few years until they’re dead letter, is the only way to also ensure there is upzoning. Although rent control and upzoning both seem to be different policies aimed at affordability, they’re diametrically opposed to each other: one makes it easy for people to move in, one makes it hard. As I mentioned years ago, rent-controlled cities tend to have parallel markets: one is protected for long-timers, and for the rest there is a market that’s unregulated and, because so much of the city’s housing supply is taken off it, very expensive. In exchange-rate dollars, I pay $1,000 for a studio of 30 square meters, of which maybe 20 are usable, the rest having low sloped ceilings. In PPP dollars it’s $730, still very high for the size of the unit. If I put my name on a waiting list, I could get a similar apartment for a fraction of the price; to nearly all residents, rents are far lower than what I pay, because of tight rent controls. Stockholm at least has a relatively short waiting list for rent-controlled apartments, 1.5 years, for international visitors at my university; American cities (or perhaps American universities) never do foreigners such favors.

The problem here is entirely political. Cities have the power to zone. Thus, supply depends entirely on whether local community leaders accept more housing. This housing, almost invariably, goes to outsiders, who would dilute the community’s politics, forming alternative social networks and possibly caring about different political issues. It’s somewhat telling that ultra-Orthodox Jews in the New York areas support aggressive upzoning, since the new residents are their children and not outsiders; Stephen Smith has written before about the Brooklyn Satmars’ support for upzoning, and the resulting relatively low prices. In the vast majority of the first world, with its at- or below-replacement birth rates, this is not the case, and communities tend to oppose making it easier to build more housing.

There is a certain privilege to being organized here. We see the pattern when we compare how US minorities vote on zoning to what minority community leaders say. In San Francisco specifically, activists who oppose additional development have made appeals to white gentrification in nonwhite neighborhoods, primarily the Mission District. Actual votes on the subject reveal the exact opposite: see the discussion on PDF-pp. 13-15 of this history of Houston land use controls, which notes that low-income blacks voted against zoning by an overwhelming margin because of scare tactics employed by the zoning opponents. (Middle-income blacks voted for zoning, by a fairly large margin.) Polling can provide us with additional data, less dependent on voter turnout and mobilization, and in Santa Monica, Hispanics again favor new hotel development more than whites. In areas where being low-income or nonwhite means one is not organized, low-income minorities are not going to support restrictions that benefit community leaders.

The result is that organized communities are going to instead favor zoning, because it gives them more power, as long as they are insulated from the effect of rising prices. In suburbs with high home ownership, they actually want higher prices: my rents are their property values. In cities with low home ownership, rent controls provide the crucial insulation, ensuring that established factions do not have to pay higher rents. Zoning also ensures that, since the developers who do get variances can make great profits, community groups can extort them into providing amenities. This is of course the worst in high-income areas: every abuse of power is worse when committed by people who are already powerful. But the poor can learn to do it just the same, and this is what happens in San Francisco; TechCrunch has a comprehensive article about various abuses, by San Franciscans of all social classes, culminating in the violent protests against the Google shuttles, and in many cases, the key to the abuse was the community’s ability to veto private developments.

The risk, of course, is displacement. As the gap between the regulated and market rent grows, landlords have a greater incentive to harass regulated tenants into leaving. This is routine in New York and San Francisco. Community groups respond by attacking such harassment individually, which amounts to supporting additional tenant protections. In California, this is the debate over the Ellis Act. The present housing shortages are such that supporting measures that would lower the market rent has no visible short-term benefits, and may even backfire, if a small rent-controlled building is replaced by a large unregulated building.

So with rent controls, community groups have every incentive to support restrictive zoning, and none to support additional development. With market pricing, the opposite is the case. What of low-income city residents’ access to housing, then? Daniel mentions housing subsidies as a necessity for the poor. To be honest, I don’t see the purpose, outside land-constrained cities like Hong Kong and Singapore. If it is possible through supply saturation to cut rents to levels that are affordable to families making not much more than the poverty line, say 133% of the US poverty line, the Medicaid threshold, then direct cash benefits are better. In the ongoing debate over a guaranteed minimum income, the minimum should be slightly higher than the US poverty line, which is lower as a proportion of GDP per capita than most other developed countries’ poverty lines, as seen in the government programs with slightly higher limits, led by Medicaid.

Leftists have spent decades arguing for state involvement in health care and education – not just cash benefits, but either state provision, or state subsidies combined with some measure of cost control. There are many arguments, but the way I understand them, none applies to housing:

1. Positive externalities: Ed Glaeser has noted that if some people in a metro area get more education then there is higher income growth even for other people in the area. In health care, there are issues like herd immunity.

2. Very long-term benefits: if college is as expensive as it is in the US today, it takes many years for graduates’ extra incomes to be worth the debt. With health care, the equivalent is preventive care. When benefits take so much time to accrue, first some people face poverty traps and don’t have the disposable income today to invest in their own health and education, and second, the assumptions of rational behavior in classical economics are less true.

3. Natural monopolies outside large cities: hospitals, schools, and universities have high fixed capital costs, so there can only be sufficient competition in very large cities. The same is of course true of rail transit.

4. Asymmetric information: students and parents can’t know easily whether a school is effective, and patients face the same problem with doctors; short-term satisfaction surveys, such as student evaluations, may miss long-term benefits, and are as a result very unpopular in academia.

With housing, we instead have competitive builder markets everywhere, no appreciable benefits to having your neighbor get a bigger or better apartment, and properties that can be evaluated by viewing them.

The only question is what to do in the transition from the present situation to market pricing. This is where a limited amount of protection can be useful. For example, rent controls could be relaxed into a steady annual gain in the maximum allowed real rent. While market-rate housing remains expensive, public housing is a stopgap solution, and although it should be awarded primarily based on need rather than how long one has lived in the city, a small proportion should be set aside to people in rent-controlled small buildings that were replaced by new towers. None of this should be a long-term solution, but in the short run, this may guarantee the most vulnerable tenants a soft landing.

What this is not, however, is a workable compromise. Community organizations are not going to accept any zoning reform that lets in people who are members of out-groups. They have no real reason to negotiate in good faith; they can negotiate in bad faith as a delaying tactic, which has much the same effect as present zoning regimes. What they want is not just specific amenities, but also the power to demand more in the future; it’s precisely this power that ensures the neighborhoods that are desirable to outsiders are unaffordable to them. What they want is a system in which their political connections and social networks are real resources. A city that welcomes newcomers is the exact opposite. Expensive housing is ultimately not a market failure; it’s a political failure.

Coordinated Planning and High-Speed Rail

High-speed rail and rapid transit both change economic geography, in that they compress distances along the lines built, emphasizing connections along the lines at the expense of ones perpendicular to them. I’ve written about this before, giving the example of the division of Uptown Manhattan into East and West Sides. In contrast to the similar implications for economic geography, we see different political treatment of transportation planning: rapid transit is usually planned centrally within a city, together with lower-capacity perpendicular forms of public transit, but there is less centralized planning of high-speed rail and connecting legacy lines.

It’s against this background that I’ve read two recent posts on Itinerant Urbanist, one advocating Northeast-wide intercity rail planning, and one expressing skepticism of plans to run trains from New York to Pittsfield along the Housatonic Railroad, whose southern end hosts the Danbury Branch. In the second post, Sandy shows how, even today, it is faster to get from New York to Pittsfield via Albany, along existing Amtrak routes, than it could be via the curvy Housatonic. The trains from New York to Albany are not HSR, but are some of the fastest in the US outside the Northeast Corridor, and that’s enough to obviate the need for some adjacent lines. But we can extend this analysis further, looking at potential HSR routes and identifying the effect on other regional and intercity lines mentioned in Sandy’s first post.

For our main example, consider Providence-Worcester. There is a direct line, the Providence and Worcester mainline, which hosts no passenger trains. I have previously called for running passenger service on the southern 25 km of the line, from Providence to Woonsocket, and integrating the schedules with MBTA trains to Boston and future HSR; in 2009, the Providence Foundation made a similar proposal, finding that it was possible to slot a reasonable frequency of in-state regional trains between the Providence and Worcester freight trains. Superficially, one might think that trains should not turn at Woonsocket, but go all the way to Worcester, a distance of 69 km, providing a key crosstown link in a New England-wide rail network.

The problem is that the presence of HSR makes the line completely useless for end-to-end traffic. HSR averages between 180 and 260 km/h, whereas regional trains average between 50 and 90, with a few trains overlapping with intercity rail going up to 120. This makes it worthwhile to go two to three times as long as the most direct route, if this can be done on high-speed lines.

It’s 70 km from Providence to Boston; from Boston to Worcester, it’s 71 along the present Worcester Line, while an HSR line following I-90 would be about 65, serving Worcester at an outlying station at the intersection with Route 122 (and the Providence and Worcester line), 6 km outside the legacy station. My attempt to work out a schedule for Providence-Boston gives about 20.5 minutes for nonstop HSR; Boston-Worcester is probably similar, giving 41 minutes plus a short transfer time. (Trains with intermediate stops would stop at Back Bay, and if the transfer can happen there, then it saves about 3 minutes total.) Let’s say the transfers at Boston are not optimized, and the total travel time is 50 minutes.

It is not easy to achieve this travel time on the legacy Providence and Worcester line: 69 km in 50 minutes is 83 km/h, and 63 km (from Providence to I-90 and Route 122) is 76. The latter speed is very ambitious, and the former even more so. While there are regional lines in New England that could approach 100, this is not one of them. The line hosts some freight traffic, so it requires additional sidings if passenger trains go at intercity rail speeds and not at regional rail speeds, which are similar to freight speeds. There is a significant commuter market at the Providence end, requiring more stops in Providence and its inner suburbs: the end-to-end travel time in the schedule I constructed for Providence-Woonsocket is 26 minutes, an average speed of 59 km/h. To get to I-90 in 50 minutes, trains would need to average 94 km/h north of Woonsocket; achieving this makes it almost impossible to stop anywhere in Massachusetts except Worcester, which defeats the purpose of the line. Worcester-Woonsocket is not important enough a travel market to reopen a passenger rail line for. For the same reason, there is no hope of achieving sufficient speed by including a mix of local and express trains: there’s not enough demand to support multiple service patterns.

The Providence-Worcester example is somewhat unfair in that it’s unlikely such a line could be activated without interstate cooperation in intercity rail planning. The same cooperation that could restore service on the Providence and Worcester line would first push for faster intercity trains on the Northeast Corridor, which would be the first step in obviating this direct line. I bring this up because it’s a very clean example of how the presence of HSR allows for circuitous routings on some city pairs, and how this should be reflected in rail planning. There are less clean examples, pitting a unified system with HSR as a trunk and branches feeding the trunk against potential in-state projects and priorities:

1. Unless HSR fares are designed to discourage this, the fastest way to get to New York from suburbs far out along the New Haven Line, and to a lesser extent the Northeast Corridor Line in New Jersey, would be to take commuter rail to New Haven or Trenton and then backtrack on HSR. This changes the optimal service patterns, away from express trains to New York and toward local trains in the outer service area, and this in turn influences planning for capacity improvement. For example, fitting HSR and commuter trains on existing tracks in New Jersey probably requires giving up express service south of Rahway, but at the outer end of the line, around Princeton Junction, going out to Trenton and backtracking on HSR would make this not as onerous as commuters may initially think. On the level of station design, the presence of backtracking means that stations may need to be reconfigured to have more access points from northbound to southbound platforms, to make transfers easier.

2. New Jersey Transit has plans from last decade to reactivate passenger rail service along the West Trenton Line. The presence of HSR makes West Trenton a less useful commuter rail station, to either Philadelphia or New York. In Philadelphia it remains useful if one wants to go to destinations on the Reading side of SEPTA, such as Temple University, or even Market East, but in New York, the nearest job center to West Trenton is Newark, which is on the Northeast Corridor. This means that better transit service from West Trenton to Trenton becomes a greater priority than direct rail service from West Trenton to New York.

3. There is a secondary rail line from New London to Norwich, passing next to Mohegan Sun. It is not very useful if intercity trains remain as they are, but the presence of HSR makes it a good feeder, and also allows trains to beat express buses for trips from New York to the casino.

4. It is vanishingly unlikely Pennsylvania will try to build in-state rail service to Erie. However, if it does, Erie-Pittsburgh service would be similar to Providence-Worcester service, with Cleveland fulfilling the same function as Boston in New England.

Difficult Transit

Many people have heard that certain regions are well-suited for these projects, for example the Northeast Corridor is unusually good for HSR because it links four major cities and several medium-size ones on a single line. By implication, there has to be a flip side, i.e. regions that are poorly-suited for HSR and cities that are poorly-suited for new rapid transit. If there weren’t – if every region were like the Northeast Corridor – then the ridership models would just have higher first-order estimates. Several proposals I’ve seen in comments and on my blogroll in the last few days are in areas where the urban geography makes it harder to justify such projects. These and a few others are the examples I will use in this post.

As usual, there’s a caveat that difficult does not equal bad. Some of these ideas are worth pursuing, but have more challenges that their easier counterparts do not, and if those challenges are solved, then they can perform well. One of the biggest success stories of modern rail investment, the TGV, is in an urban geography that’s not particularly conducive to rail: France’s secondary cities surround Paris in all directions (although Lyon and Marseille are collinear with Paris), the stub-end layout of stations in Paris and many other cities forces awkward branching, Lyon needed a business district to be built from scratch around Part-Dieu. France made this work, and it’s possible some of the projects on this list can be made to work in similar vein.

High-Speed Rail in Sweden

Project: greenfield HSR lines connecting Stockholm with Sweden’s major secondary cities, Gothenburg and Malmö.

The problem: Stockholm, Gothenburg, and Malmö do not lie on a straight line. The three cities are quite small by the standards of more populated countries: Stockholm has a bit more than 2 million people, Gothenburg has a bit less than a million, Malmö has 700,000. A line connecting just two of them, or even a Y-shaped line, is unlikely to get enough ridership to justify the construction costs of full HSR. There are no large intermediate cities: the largest, Linköping, has about 100,000 people. As noted above, French urban geography is not great for HSR, either, but at least the LGV Sud-Est could serve both Lyon and Marseille, and France’s greater population ensures that its secondary cities are large enough to generate enough traffic to fill an HSR line.

As a silver lining, Malmö is adjacent to Copenhagen, and the difficult part, bridging the Øresund, has already been done. While international lines tend to underperform, the tight cultural and economic connections between the Scandinavian countries make it likely that international projects within Scandinavia would be exceptions to the rule. Copenhagen would add another 2 million people at the end of the line. However, even that is unlikely to generate enough ridership to pay for 500-odd kilometers of greenfield HSR (plus a connection to Gothenburg).

Because of its poor urban geography for conventional HSR, Sweden has investigated cheaper solutions, allowing higher speeds on legacy track or on greenfield tracks built to lower standards. As a result, there is research into the possibility of high-speed tilting trains, running faster than the 250 km/h Pendolino. This research is likely to be useful in the UK and US, where the urban geography is better-suited for HSR but fully greenfield construction is obstructed by suburban development near the rights-of-way and by high construction costs, but the original context was faster speeds within Sweden.

High-Speed Rail in the Pacific Northwest

Project: greenfield HSR connecting Portland, Seattle, and Vancouver. This is not officially proposed anywhere that I know; current plans focus on incremental improvements to the Amtrak Cascades. However, every American HSR fantasy map I’ve seen (including the ones I’ve drawn) includes this link, since at least superficially based on city populations it would succeed.

The problem: getting out of the major cities involves a slog on curvy legacy track in areas where it’s hard to straighten the right-of-way. Heading north of Seattle, the route goes along the water, in terrain that is too hilly for an easy inland cutoff all the way to Everett, 50 km north. Getting out of Vancouver is also hard, because of suburban development in Surrey, and becomes even harder if one wants the Vancouver station to be Waterfront rather than Amtrak’s current stop, the less centrally located Pacific Central. The Northeast Corridor is said to have slowdowns near the major stations, leading to proposals to bypass them with new tunnels, but at no point are there 50 nearly-continuous km of low curve radii; the New Haven Line does not look as curvy, while the Shore Line farther east is easy to bypass on I-95.

The Seattle-Portland segment is much easier: the route heading south of Seattle is not constrained, and north of Portland it is possible to run alongside I-5. However, the most important intermediate cities, Tacoma and Olympia, can only be served with exurban stations, since getting into their centers would require the mainline to detour on curvy alignments.

Through-Run Commuter Rail in Chicago

Project: there are many proposals by transit activists to construct new infrastructure to enable through-running on Metra, analogous to Crossrail, SEPTA Regional Rail, the Paris RER, and multiple S-Bahns. Details differ, but other than the lines through Union Station, through-running generally means connecting Metra Electric to some of the lines feeding into Union Station from the north or the Union Pacific lines; UP-North is especially notable for serving dense neighborhoods and not having any freight traffic.

The problem: the layout of the lines entering the Chicago central business district makes it hard to build a coherent network. What I mean by coherent is that commuter lines can make multiple CBD stops to serve different CBDs, or different parts of the same CBD: in New York, a Penn Station-Grand Central connection would let trains serve both the West Side and the East Side. Look at the map proposed by Sandy Johnston, in the second link above: there is no station on the Near North Side, there is no connection from the West Loop stations to the Loop, and effectively lines are still going to be split between lines bound for the West Loop and lines bound for the Loop in the through-run system.

None of this is the fault of any of the people drawing these maps. To serve both the West Loop and the Loop, a line would have to go east-west in the vicinity of Union Station, where there is no legacy line pointing in the right direction. The options boil down to a long greenfield east-west subway, and an awkward transition to the preexisting east-west lines, BNSF (which is too far south) and UP-West (which is too far north), which to add another complication carry heavy freight traffic.

A system prioritizing north-south connections runs into different dilemmas, concerning the tradeoff between service to the Near North Side and easier connections to the rest of the North Side Metra lines. A north-south line connecting UP-North to Metra Electric through the Near North Side would be beautiful, and miss all other Metra lines and most L lines. Sandy’s proposal has Metra Electric swerving west to meet UP-North just north of its terminus at Ogilvie Transportation Center, meeting all L lines and potentially the North Side Metra lines but missing the job centers in the West Loop and Near North Side.

Rail to LaGuardia

Project: construct some rail extension to LaGuardia Airport. Which rail extension varies based on the proposal. The most mainstream proposal, in the sense that it was supported by Giuliani until it was torpedoed by neighborhood opposition, would have extended the Astoria Line east to airport grounds. More recent proposals from various activists have included not just the Astoria Line extension, but also a Northeast Corridor spur, an AirTrain from the Astoria Line, an AirTrain from Jamaica with JFK connections, a subway shuttle under Junction, and a subway running from the airport to 125th Street along the route of the M60 bus.

The problem: all of the above ideas face the same pair of problems. At the airport end, the airport competes with other urban destinations, rather than complementing them by lying on the same straight line with them. An extension from the west, such as the Astoria Line extension, needs to choose between serving the airport and serving the Astoria Boulevard corridor, which has high residential density and no nearby subway service; Astoria Boulevard itself is so wide that as with Queens Boulevard, an elevated line in its middle would be an improvement. Farther east, there is nothing that a LaGuardia extension could be continued to, because of Flushing Bay. An extension across the bay going to Flushing or College Point could be useful, but an extension of the 7 to College Point would be even more useful and avoid underwater tunneling. The bay, and more generally the Long Island Sound, dooms any proposal for a loop returning to the mainline, in the manner of Zurich Airport, while a spur would again compete for capacity with more important lines. Compare this with LAX, which, going along the Harbor Subdivision, is collinear with Inglewood, the Slauson corridor, and Union Station, and would have an easy connection to El Segundo.

At the other end, the question with every airport extension is, what does it connect the airport to? The answer for LaGuardia has to be the Upper East Side, where as I remember most riders originate; but there is no good way of connecting to the Upper East Side, which has no east-west subway line, and shouldn’t, as there are perhaps a hundred kilometers of higher-priority tunnels in the region. A connection to 125th Street is ruled out by the fact that Second Avenue Subway has an even better connection to 125th. The Astoria Line serves the Midtown hotel cluster well, and has a connection to the Lexington trains to the Upper East Side, but I doubt that it can beat a taxi across the bridge in non-rush-hour traffic.

Providence East Side Tunnel

Project: restore rail service through the East Side Rail Tunnel, with a new connection to Downcity at the western end and connections to new or restored rail lines in and beyond East Providence. In Jef Nickerson’s version, the trains are light rail and drop to the surface at the Downcity end. In mine, they continue elevated through Downcity, with a new station replacing Providence Station for both commuter and intercity rail. All versions include a stop at Thayer Street for Brown University service, should one be constructable at reasonable cost.

The problem: there’s no real need for local or regional service from the east along the tunnel (intercity service could be sped up by about half a minute to a minute by avoiding curves in Pawtucket). Light rail service would run into the problem of incredibly spread-out suburbanization east of Providence. Commuter rail would run into separate problems: the legacy lines go along the water in East Providence and don’t serve the town itself well; beyond East Providence, the line going north serves the same suburbs as the existing Providence Line minus Pawtucket, while the line going south would need extensive and costly restoration work to get to Fall River, and only passes through small and low-density intermediate points.

Cutting off Providence Station to move the city’s main station to the south is useful, but the only rail from Providence to Pawtucket and Woonsocket goes due north of Downcity and would be left out of this system. Shoehorning it to the same station that leads to the East Side Tunnel would produce every adverse impact of viaducts on cities: heavy visual impact coming from elevated-over-elevated grade separation, squeal coming from low curve radii, takings of condo buildings near the existing Providence Station.

California HSR Should Not Have Been Funded This Way

Last month, California made a budget deal for the formula that would be used to distribute its cap-and-trade revenues. The state’s cap-and-trade bill does not deed the money to the general budget but to a separate account, to be distributed based on a variety of goals including subsidies to programs that reduce greenhouse gas emissions. The recent deal is to give most of the money to transportation (including transit-oriented development): this year the budget gives $600 out of $850 million to transportation (see PDF-p. 6 here), of which $250 million will go to high-speed rail, and according to an informational hearing the long-term deal gives 80% of revenues to transportation, including 15% to high-speed rail. Transit bloggers who are not in the process of moving across oceans covered the issue last month as the deal was made: Streetsblog wrote about the plan, Robert Cruickshank wrote multiple times in support of the decision, and Bruce McFarling explained how HSR’s projected emissions reductions should entitle it to a share of the cap-and-trade proceeds.

In reality, although it’s a good thing that California HSR is getting funded, it’s a bad way of funding it, betraying both environmental incompetence and political mistrust. The basic problem is that the HSR project is not going to reduce emissions enough to justify 15% of the pot, nor is transportation such a big share of California’s emissions inventory to deserve 80%: it accounts for only 37% of statewide emissions. Electricity, and related sources of emissions such as building heating and industrial emissions, get far less than their share of emissions.

Bruce’s post runs the numbers on HSR, notes that the projections are currently $250-400 in construction costs per ton of CO2 reduction, and proposes that if cap-and-trade results in a carbon cost of $75 per ton then this justifies using the revenues for 20-35% of the cost of HSR. The projected revenue from cap-and-trade is a range whose top end is $5 billion statewide, corresponding to about $11 per metric ton; at this level, assuming HSR saves $250/t-CO2 means it should get 4.4% of its funding from emissions reduction, or (at the current cost of $53 billion in constant dollars) about $2.3 billion over the lifetime of the program. If the revenue is indeed $5 billion a year, this spending level is projected to be reached in 3 years.

For some evidence of what the state is really doing, consider how the deal comments on each share of the funding. The informational hearing details the investment strategy as follows:

25% for a permanent source of funding for transit operations, distributed based on greenhouse gas criteria.
20% for affordable housing and miscellaneous urban planning goals (including TOD), of which at least half must be for affordable housing (including TOD, again); the money is to be distributed based on “competitive GHG performance.”
15% low-carbon transportation, based on both long-term clean air and GHG goals.
13% energy, including electricity and building efficiency.
7% natural resources, waste diversion, and water projects.
15% HSR.
5% “new or existing” intercity rail, based on GHG criteria.

Note that internally to four categories, comprising 65% of the total funds, the hearing mentions greenhouse gas criteria. In three out of the four, comprising half of the funds, the hearing implies that the decision of how to distribute the funds will be based on competitive grants according to which project reduces emissions the most.

The key point here is that the state has effectively said what the best way is to ensure the spending side of cap-and-trade will reduce emissions optimally: projects will compete for scarce funding based on greenhouse gas criteria. Once it has made the political decision to distribute funds by a formula that disproportionately goes to transportation, it has no objection to using greenhouse gas criteria internally to each category. The problem is that the transportation projects in general and HSR in particular would never make it out of a grant process based on such criteria if they were not shielded from competition with non-transportation priorities.

There are two legitimate ways to distribute funds coming out of an externality tax, which is what cap-and-trade really is. One is to let the tax side do the work of reducing impact, and put the money into the general budget. This is common practice for most developed countries’ fuel taxes (though not the US’s). In this approach, HSR would compete with all of the state’s other budget priorities. If the state wanted to reduce other taxes against the cap-and-trade funds rather than raise spending, it could. If it wanted to spend the money on unrelated things, such as education, it could as well. There already is a more or less open and democratic budget process for this.

The other way is to reduce all political discretion, and distribute the funds based entirely on greenhouse gas criteria, without breaking the money into categories. The state seems to prefer this way, judging by its use of this process within each category. With other externality taxes there is another option, of giving the money directly to victims of the externality, e.g. spending cigarette taxes on lung cancer treatment; however, the bulk of damage caused by climate change is to developing countries, and spending cap-and-trade revenues on targeted aid to vulnerable developing countries is politically unacceptable.

The state’s hybrid approach is effectively a slush fund. High-level politicians, including Governor Jerry Brown, want to build a visible legacy, and HSR is far more visible than making household appliances consume less electricity. Emissions reductions are secondary to this concern. They’ll be happy to make their legacy a project that reduces greenhouse gas emissions, but they have no quantitative preference for projects that reduce emissions more than others. On the contrary, when they pull strings, they might even make decisions that make these projects less environmentally beneficial: the decision to connect Los Angeles to Bakersfield via Palmdale rather than directly has no technical merit, and judging by LA County’s support appears to be motivated by concerns for development in the Palmdale area. As the incremental cost of going through Palmdale is about $5 billion, nearly 10% of the HSR cost, the result is that the state is going to spend a substantial amount of cap-and-trade money on spurring more development in the High Desert exurbs.

Needless to say, when the cap-and-trade bill was passed, it did not state or even imply that the state could use the money to spur more development in the exurbs. The bill did not adopt a GHG-only approach, but listed several additional goals, none of which included transportation. Chapter 1, Part 2, paragraph h states,

It is the intent of the Legislature that the State Air Resources Board design emissions reduction measures to meet the state wide emissions limits for greenhouse gases established pursuant to this division in a manner that minimizes costs and maximizes benefits for California’s economy, improves and modernizes California’s energy infrastructure and maintains electric system reliability, maximizes additional environmental and economic co-benefits for California, and complements the state’s efforts to improve air quality.

There is an explicit mention of air quality, and explicit mentions of energy and electricity, which are only getting 13% of the funding despite accounting for 54% of emissions. Elsewhere the list of legislative intents includes vague terms such as technological leadership, but the only explicit mention of transportation in the bill is in paragraph c, which says that historically California provided leadership on several environmental issues, including emissions limits on cars as well as energy efficiency and renewable energy.

However, the cap-and-trade bill is older than the current administration, and the political priorities have changed. Since a regular budget process giving HSR the money it needs would run into opposition from competing priorities, it’s best to raid a new source of revenue, one without legislative inertia or established supporters directing the money to more useful purposes.

Hence, a slush fund.

Brief Personal Update

I no longer live in Vancouver. I am leaving the city tonight, heading to New York and New Haven until July 8th (email me if you want to meet). I will then travel to Europe: I got a two-year position at the Royal Institute of Technology (KTH). As Stockholm has tight rent control, expect a post on that subject once I’m settled in there and have gotten the chance to talk to people.

Stockholm also has a subway system that’s the opposite of the paradigm used in Vancouver or Copenhagen: it runs full-length trains, and branches heavily, with low frequencies on the branches. KTH is unfortunately on a branch, served every 10 minutes off-peak. (However, unlike in New York, in Stockholm branches are consistent, in that two branches of separate trunk lines do not combine to form another trunk line farther out the way the E and F branches combine to form the Queens Boulevard Line.)

What Elites Do Instead of Providing Services

I realized last year that even when they face a problem that is evidently about city services, city governments prefer to go for monuments that glorify their leadership. The most blatant example then was Cornell NYC Tech, the city-backed university whose campus construction alone is several times as expensive as the CUNY system per student. Since then I’ve tried to collect examples of power brokers proposing similar schemes, of which the worst is Larry Summers’ proposal to solve US inequality by spending public money on airport improvements. These are, to be frank, analogs of what American transit activists have to deal with routinely, with agencies preferring expensive iconic stations to ordinary capital and operating improvements in service.

The argument for Cornell NYC Tech is that New York needs tech entrepreneurs of the kind that Silicon Valley has, and that for that it needs its own Stanford. Instead of investing in STEM education across the CUNY system, or in its dedicated technological campus at the New York City College of Technology, it decided to start a private university from scratch, inviting other universities to bid on it. The city wanted Stanford to win the bid, but instead the winning bid was a joint effort by Cornell and the Technion, Israel’s technological university. The Technion was never run this way; it was started as a German-style technical university and is now a public university, funded and run on the same terms as the other Israeli public universities.

For Cornell NYC Tech, the city has lined up $2 billion in public and private funds for campus construction, expecting 2,000 students in 2037, which at 4% interest is $40,000 per student-year; annual capital and operating spending together, from all sources including tuition, is $16,000 per full-time equivalent student at the CUNY senior colleges and $11,000 at the CUNY community colleges (see PDF-page 65 of the budget request). This is the educational equivalent of airport connectors, which cities routinely spend several times per rider on as they would on ordinary subway extensions.

Summers’ proposal for airport improvements is in a way more frustrating, and more telling. He did not propose it as part of an independent infrastructure plan, but as a way to build public works to reduce US inequality, on the grounds that JFK is “an embarrassment as an entry point” and “the wealthiest, by flying privately, largely escape its depredations.” The proportion of people who fly privately is tiny; an income level at the bottom of the US top 1%, $400,000 per year, will buy you a lot of intercontinental business-class travel or some first-class travel, while affording a late-model Learjet requires an annual income of many tens of millions of dollars. Since poor people don’t fly as much as rich people, the users of JFK skew richer than the general city public.

My frustration comes from the fact that Summers is not trying to derail the conversation: he previously wrote about inequality as a problem and proposed standard center-left solutions, including raising taxes on capital gains and inheritances, supporting unionization, and (by implication) investment in public education. He clearly cares about the problem. He just seems to think that airport investment benefits the poor more than the rich. Most likely, this comes out of years of insider schmoozing with people so rich that they do own private jets, and generalizing to the considerably broader class of rich people.

In both cases, even on its stated merits, the proposal misses key facts about the situation. Silicon Valley began around Stanford, but once the initial tech cluster formed, it became independent of the university, so that even companies formed by people with no affiliation with Stanford or the Bay Area, such as Facebook, relocated to the area. New York is not going to grow its tech industry to the proportion of Silicon Valley’s by building an enterprise university any more than the Bay Area can become a world financial center by building affiliate universities for Columbia and NYU, from which many finance workers are recruited. As for JFK, like many of its users, when I arrive my first experience is the immigration line, a humiliating experience that involves fingerprinting and standing in line possibly for hours, depending on what terminal I use and what time I arrive. Public works will not solve that.

The problem with making even the merit-based argument is that public monuments are never truly merit-based projects. The decision-making process goes in the other direction: first the city elites (or, in case Summers’ proposal makes it into a national jobs bill, national elites) decide on something they want to see built, usually with the adjective world-class thrown in: a world-class university, a world-class airport, a world-class train station, a world-class office tower. The image of a world-class monument is more important than whether it works at its stated goal, such as improving education or transportation or fulfilling a need for class A office space.

Witness all the problems involving World Trade Center, which is being built entirely for prestige value, at enormous cost. The associated PATH station is $4 billion, almost as much as Second Avenue Subway, and about the same as 20 kilometers of subway in an average first-world city. One World Trade Center cost about $12,000 per square meter. I am not aware of any office tower in the world that is this expensive outside the WTC area and Hudson Yards; the tallest recent tower built in New York excluding 1 WTC, Bank of America Tower, cost about $5,500 per square meter in 2012 dollars, while the range I have seen for office towers in the 200+ meter range is about $2,500-6,000. Meanwhile, the WTC site struggles to find tenants: 1 WTC is almost half empty.

The sentiments after 9/11 ensured WTC would be rebuilt taller, regardless of actual demand in Lower Manhattan. Viewed through this lens, 1 WTC is not really about office space, but about proving a point about the power of US and New York to come back and not surrender to terrorism. This is why the transit spending went mainly to the PATH station and not to bringing the LIRR to Lower Manhattan, as proposed by the Regional Plan Association and studied officially in subsequent years: the LIRR project would’ve been about Lower Manhattan in general, without enhancing the specific prestige of WTC, while the billions poured into the WTC site and its PATH stations are all about the prestige.

Those other projects – various overrated transit schemes such as airport connectors, but also Cornell NYC Tech and Summers’ JFK proposals – are the same. They are not about what people living in, working in, or visiting the city need. They are not even about what they want. Whereas there was a citywide impulse to rebuild WTC taller after 9/11, there is no equivalent impulse to build an exclusive technical university, except among the power brokers. They are entirely about being able to say, “we have our own ___” and “I got that built.” It looks like development, but at best provides a fraction of the advertised value, and at worst provides nothing.

Whenever an urban project is proposed, the most important question to be asked is “what problem is this solving?”. Often, the problem is real, but there are much cheaper and less glamorous solutions. At other times, the project is a solution in search of a problem, and this is often detectable when proponents tout many unrelated benefits, almost as if the project can solve every major problem.

Compare this with solid public transit projects. Consider the lines I think North American cities should be focusing on, and the lines proposed in comments, especially as the Vermont subway in Los Angeles. In every single case, there are strong arguments for why the ridership of those lines would be high relative to the cost, and why existing subway lines (if any) and surface transit options are inadequate. The problem being solved is underserved neighborhoods with high transit demand, or in the case of the crosstown lines underserved origin-destination pairs in high demand. For other lines, not listed, there might be a separate argument regarding transit-oriented development: American cities tend to oversell TOD, as the problems with Hudson Yards show, but there do exist cases in which extending a subway line can allow dense development, or the construction of a new business district. But this involves figuring out where the development comes from – for example, the housing market may be very expensive, signaling high demand, or there may be projections of high future metropolitan population growth.

Usually, support for prestige projects to the exclusion of providing public services is the hallmark of moderates, along a broad arc from the center-left to the center-right. In the last few years, Republicans too far right to be called center-right have prioritized cutting taxes and spending and weakening the unions; signature projects conflict with their opposition to government spending. Conversely, urban leftist activists tend to oppose these prestige projects, on such grounds as gentrification, displacement, and private-sector involvement in public services.

The people in between those two ends are the ones most guilty of this kind of thinking. They are usually neo-liberal enough that they believe the government should champion market solutions and oppose industrial policy, and yet what they do is in many cases exactly industrial policy: Cornell NYC Tech is an attempt to curry favor with the technology industry. They are not so conservative as to believe government is always the problem, but the role they envision for government is to partner with the private sector to build public projects, which they tend to choose on grounds of what looks good rather than what provides the best public service. They know the buzzwords of urban politics well: for example, they’ll happily argue climate change to push a desired agenda that is usually only partly related to the problem, but lack the urgency of actual environmentalist activists and often also build roads and other dirty projects.

As with most bad things in politics, it’s a result of weak democratic institutions on the local level. American mayors tend to be elected dictators, and the opposition to them tends to be based on personality rather than ideology. In this non-ideological framework, the role of government is not to balance market and state solutions based on the voters’ preferences, but to aggrandize the leaders. Signature initiatives must appeal to the broad spectrum of non-ideological voters, so they can’t involve merely increasing spending on a chosen priority like education or transportation. Doing nothing is not an option – something has to be passed to remind people that the government still exists and has a purpose. The political incentives are against any incremental improvements that lead to tangible results, and for white elephants.

Airport Connectors

The most interesting transit planner in the world:

jkuw0dd

This principle is true primarily for large international airports. As I will explain, this is less true of smaller airports. But before going on, I would like to clarify a distinction between bad and overrated. Airport connectors, as I have argued many times, are overrated: city elites tend to like them disproportionately to their transit usage, as do many urban boosters, who think a comfortable airport connector is a necessary feature of a great global city. The result of this thinking (and also the main evidence we have that this thinking exists) is that airport connectors are built at much higher costs per rider than other transit projects: the JFK and Newark AirTrains cost more than $100,000 per weekday rider, much more than other recent rail projects in New York; even the far over-budget East Side Access, at current estimates, is about $60,000.

However, overrated does not mean bad. There exist airport connector projects with reasonable cost per rider. They’re still overrated, which means they’ll be built concurrently with even more cost-effective non-airport projects, but they’re good enough by themselves. As an example, take the Canada Line. The total cost was about $2 billion, and the latest ridership figure I have, from 2011, is 136,000 per weekday, ahead of projections. At $15,000 per rider, this is reasonable by European standards and very good by North American ones. Let us now look at the two branches of the line, to Richmond and the airport. Lacking separate cost data for them, I am going to estimate them at about $300 million each, as they are entirely above-ground; the airport branch is 4 km and the Richmond branch is 3 km, but the Richmond branch has an urban el and the airport branch doesn’t. For ridership data, we have this set of figures per station (which results in a Canada Line total of only 113,000). Boardings and alightings sum to 19,000 on the airport branch and 34,000 on the Richmond branch; we’re double counting intra-branch trips, but there presumably are very few of these. As we see, the Richmond branch is more cost-effective, but the airport branch holds its own – since the per-station data has a lower overall Canada Line ridership, the airport branch’s presumed cost per extra rider generated is less than that of the entire line! (This sometimes happens, even with branches that generate less ridership than the trunk.) Clearly, despite the fact that airport connectors are overrated, this is an example of a good project.

The importance of the overrated vs. bad distinction is then that good transit advocates need to be wary, since airport connectors that don’t work well might get funded anyway, ahead of more deserving projects. But there remain good airport connectors, and therefore we should discuss what features they might have. The answer given by city elites is typically “nonstop connection to the CBD,” often with a premium fare. But the good transit answer is more complicated, and the graphic at the top of the post is only a partial answer.

There is a difference between short- and long-distance air travel. In many cities it doesn’t matter much because there’s a single dominant airport – Beijing, Frankfurt, Zurich, Atlanta, Toronto – but in others there are multiple airports, with different roles. Often there will be a smaller, closer-in, older airport, serving mostly domestic flights, and a larger, farther away, newer international airport. Paris has Orly and Charles-de-Gaulle, Chicago has Midway and O’Hare, New York has LaGuardia and JFK (Newark is intermediate in its role, even if it’s the oldest), Los Angeles has Burbank and LAX (the other airports are somewhat outside this division), Dallas has Love Field and DFW, Tokyo has Haneda and Narita, Seoul has Gimpo and Incheon. Because those airports have different functions, they require different kinds of transportation links.

First, let us consider departing passengers. If they travel to another continent, their options are quite restricted: for example, if they live within driving distance of Atlanta, they’re flying out of Atlanta. Even if there are closer secondary airports (such as Greenville-Spartanburg and Chattanooga), they don’t offer such service – at most, they offer a connecting puddle jumper flight to the primary airport. In contrast, if they travel shorter distances, and live far from the primary airport, they could fly out of a secondary airport, or might just drive instead of flying: a 2-hour drive to the airport is comparatively more tolerable for an 8-hour intercontinental flight than for a 1.5-hour short-hop flight. For example, when I lived in Providence, my air trips were all to the West Coast or Europe, so I flew out of Boston or even New York; but when my sister visited, she chained trips and also visited her boyfriend, who at the time lived in North Carolina, and for the domestic leg of the trip she flew out of T. F. Green.

The result is that primary international airports draw their departing passengers from a much wider shed than mainly domestic airports. In metro areas with such separation of airports, the international airports – Charles de Gaulle, JFK, DFW, Incheon, etc. – draw riders from faraway suburbs and even from adjacent small metro areas, whereas the domestic airports draw riders primarily from the city and its nearby suburbs.

Now, let us consider arriving passengers. Destinations are more centralized than origins, but this is especially true for international trips than for domestic ones. Tourism trips are heavily centralized around a few attractions, which in most cities are in the CBD, or in specific locations: if you’re flying to the Paris region for tourism, your destination is either Paris proper or Eurodisney, rather than an average suburb. Business trips are also heavily centralized around the CBD and a few edge cities. Personal visits have no such concentration, and these are much more common for short-distance domestic flights than for long-distance international flights. I am unusual in that I live on a different continent from my parents; usually, people live within ground transportation or short-distance flying distance from family and friends, depending on the country they live in (short-distance flying distance is more common in the US). The result here is that arriving passengers at domestic airports are typically interested in visiting the CBD but often also the rest of the metro area, whereas arriving passengers at international airports are much more CBD- or tourist attraction-centric.

Some evidence for this difference can be found in looking at the Consumer Airfare Report, which has domestic O&D traffic counts between airport pairs. The primary international airport usually has a smaller percentage of its domestic O&D traffic going to shorter-distance cities. For example, at LAX, 13% of traffic is within California, and another 6% is to Las Vegas, Phoenix, and Tucson, within a 3-hour high-speed rail range. At Burbank, the corresponding figures are 42% and 21% respectively. The same pattern can be observed for O’Hare (8.6% of traffic is internal to the Midwest) and Midway (14.6%), and DFW (3% of traffic is internal to the Texas Triangle) and Love Field (27%).

The mode of transportation that best suits the needs of international airports is then mainline rail. On the one hand, it tends to be better than urban transit at serving trips that are dedicated to CBD service, since commuter rail is more radial than urban transit, and the stop spacing is typically also longer (although dedicated premium connectors are still often wastes of money). On the other hand, it can extend deep into the suburbs and to adjacent metro areas, and expand the airport’s draw. People can ride intercity (often high-speed) trains direct to the terminal at Frankfurt, Zurich, and Charles-de-Gaulle, and this allows those airports to be the primary international airports for metro areas in a wide radius: SNCF code-shares with airlines to connect people from Charles-de-Gaulle to Lyon, 400 km and 2 hours away by TGV.

This is not true of small domestic airports. A TGV connection to Orly would’ve been much less beneficial than the current connection to Charles-de-Gaulle: most of Orly’s traffic is short-distance, often competing with the TGV rather than complementing it.

With this distinction in mind, we should look at the situation at the major American airports. In California, the current plan is to have California High-Speed Rail serve both SFO (at Millbrae) and Burbank Airport; the original plan served Downtown Burbank instead of the airport, but the HSR Authority seems to have shifted its focus, and wants Burbank to be the southern terminus of the line, pending construction to LA Union Station. This is bad planning. Nearly two-thirds of Burbank’s traffic competes either with California HSR or with future tie-ins. People from Bakersfield and Fresno are unlikely to take a train to the airport to connect to a flight, since they can take a train the whole way, or drive directly to Las Vegas or Phoenix. People in Bakersfield and Fresno would be more interested in a connection to LAX, whose traffic complements rather than competes with intercity rail.

Los Angeles could build a connection to LAX, running both frequent electric commuter trains and high-speed trains on it. The Harbor Subdivision has existing tracks from Union Station almost the entire way to the airport, although the route is at-grade, with a large portion of it running next to Slauson Avenue, and most likely a major project like this would require viaducts. Only a short greenfield segment, elevated over Century, is required to reach the proposed Terminal 0 location, and that is only necessary if, as in Zurich and Frankfurt, LAX wishes to avoid a landside people mover. It is both bad transit and bad politics to build this only for nonstop trains: the route passes through reasonably dense urban neighborhoods, and should have 10-12 stops along the way, with some trains running local and others making only 1-3 stops, at major nodes such as Inglewood or the intersection with the Blue Line. There is room for passing sidings at the line’s midpoint, but the low top speed and the short length of the line is such that overtakes are only necessary if there are nonstop and local trains every 10 minutes. Such an airport connector would serve many different trips at once: HSR trips from Central Valley cities to LAX, arriving trips from LAX to Downtown LA (and, via transfers at intermediate stops, to the Westside), and local trips on the Slauson corridor. It’s a flexibility that modernized regional rail has, and that other modes of transportation, which can’t mix local and intercity traffic as well, lack.

Leaving California, let us look at New York. There are perennial proposals for a new connection to LaGuardia (via an extension of the N) and an additional connection to JFK (usually using the Rockaway Cutoff). There is also a new proposal for a Newark connection via PATH. With the distinction between short-distance domestic and long-distance international airports (Newark is intermediate between the two), we can analyze these proposals. Newark is the easiest to dispose of: the cost is extreme, $1.5 billion for 4 km above ground. It also has several design flaws: unlike the LAX connector I outlined above, this proposal is nonstop from Newark Penn, skipping the former South Street railroad station; the lack of intercity service improvement and the poor service to the Midtown hotel clusters doom it as a CBD connector.

The JFK proposals are problematic as well. The AirTrain connection to Jamaica is quite useful, since it lets people from all over Long Island connect to the airport. Improving JFK access hinges on improving service to Jamaica, then: through-service from New Jersey, higher off-peak LIRR frequencies, reelectrification with catenary to permit Amtrak send Northeast Corridor trains that aren’t needed for Boston service to Jamaica. East Side Access improves JFK access as well, since it allows LIRR trains to serve Grand Central, which is closer to the Midtown hotel clusters than Penn Station. Ideally there wouldn’t be an AirTrain connection, but it’s the best that can be done given existing infrastructure and given Jamaica’s importance. A Rockaway Cutoff connection, which branches from the LIRR Main Line west of Jamaica, would not help Long Islanders go to JFK; it would also not be able to carry intercity trains, since Amtrak trains to Jamaica can serve both airport riders and Long Island riders, each of which groups alone is too small to justify intercity trains on its own.

In contrast, LaGuardia proposals are better, since for a close-in, domestic airport, service to the entire city is more important. I remain somewhat skeptical – airport connectors are still overrated – but less dismissive than of Newark and JFK proposals. LaGuardia travelers from the Upper East Side, which as far as I remember supplies a majority of its departing traffic, would have to transfer at 59th Street; but they have to detour through 59th or 125th via taxi already, and the subway would not get stuck in Manhattan traffic. Conversely, there is much less need to connect the airport with the suburbs and with neighboring metro areas than there is with JFK, which means that there is no point in constructing people movers to the LIRR.

Finally, let us look at Chicago. O’Hare has the airport connection of a domestic airport rather than that of an international airport. There are plans for an express link to the Loop, but these do nothing for departing passengers from neighboring areas. While airport connectors tend to be overrated, express premium-fare links are especially overrated, since they give business travelers dedicated trains, on which they always find seats, without needing to commingle with lower-income riders.

However, some of the Midwestern high-speed rail proposals include a connection to O’Hare from the outlying metro areas, and this is good planning, assuming the cost is not excessive. SNCF’s proposal includes a bypass of Chicago that serves O’Hare, similar to the Interconnexion Est. A second step, if such a connection is built, is to attempt to connect regional lines to it, if they are electrified. This includes both inward connections, i.e. a frequent commuter rail connection to the Loop or West Loop with good connections (ideally, through-service) to other commuter lines, and outward connections, i.e. low-speed short-distance intercity lines, such as to Rockford.

In all of these cases, the common thread is that the connection to the airport does not need to be a premium service, marketed only to the business traveler. These services are never the majority of airport transit ridership: see Hong Kong, Tokyo, and London numbers on PDF-p. 28 here. However, it does need to provide service to both departing and arriving passengers, and for a major international airport, this requires good service to the suburbs and to adjoining metro areas. The optimal technologies are often bundled together with premium fares – high-speed rail is in many countries, mainline rail is in North America – but the benefits come from features of the technology and service pattern, rather than of the branding. Good transit projects connecting to airports will make sure to have the correct service reach, while at the same time not excluding local riders.

Suburban Geography and Transit Modes

A post on Let’s Go LA from last year, about different suburban development patterns in different regions of the US, praises Los Angeles’s suburbs for having an arterial grid that allows some density and permits frequent bus service. The Northeast, in contrast, has a hierarchical system, of town centers surrounded by fractured streets and cul-de-sacs, at much lower density. This is how Los Angeles’s urban area has the highest standard density in the US, and one of the highest weighted densities, nearly tying San Francisco for second place after New York. It sounds like a point in favor of Los Angeles, but missing from the post is an analysis of how Rust Belt suburban development patterns reinforce prewar transit. Briefly, Western US grids are ideal for arterial buses, Northeastern town centers are ideal for commuter rail, which used to serve every town.

For a Northeastern example, the post brings up Attleboro as a historic town center. Look at the image and notice the walkable grid and development near the train station, although one quadrant of the station radius is taken up by parking. Attleboro is in fact the town with the oldest development on the Northeast Corridor between Boston and the Providence conurbation, and the only one that, when taking the train between Boston and Providence, I’d be able to see development in from the train. Sharon and Mansfield, both developed decades later, do not have as strong town centers. But conversely, many town centers similar to Attleboro’s exist in the Northeast: Framingham, Norwalk, Tarrytown/Sleepy Hollow, Huntington, Morristown, Paoli.

Now, a careful look at the specific examples of Norwalk and Huntington will show that the most walkable development is not necessarily at the train station. In both suburbs, the old town center is where the original road goes – Northern Boulevard and its eastern extensions in Long Island, the Boston Post Road in Connecticut. Huntington has a second center around the LIRR station; Norwalk has a much smaller second center around the South Norwalk Metro-North station. For the most part, the railroads went close enough to the older roads that the town center is the same, as is the case especially in Attleboro, Tarrytown, and Paoli, and in those cases, commuter rail can at least in principle serve jobs at the suburban town center.

This boils down to the difference between optimal bus and rail networks. Buses love grids: they typically serve the scale of a single city and its inner suburbs, and there it’s feasible to provide everywhere-to-everywhere service, which grids are optimal for. For the suburbs, this breaks down. Buses on uncongested arterial roads are still surface transit; an average speed of 30 km/h is aspirational, and that is for suburbs, not dense urban neighborhoods. On a road where the bus can average 30, cars can average 50, and cars can also use expressways without splitting frequency between different suburban destinations, speeding their journeys up greatly. Meanwhile, commuter rail can, depending on stop spacing, average 50-60 km/h easily, and an aggressive timetable can cross 80 if the stop spacing is relatively express.

There is no such thing as a rapid transit grid. Subway networks almost invariably look like a central mesh, often containing a circumferential line, with spokes radiating out of it in all directions. Mexico City has a larger mesh, approximating a subway grid, but its outer ends again look hub-and-spoke. Counting commuter rail, the hub-and-spoke system is as far as I can tell universal, with the exception of highly polycentric metro areas like the Ruhr. The spokes are rarely clean: they often cross each other (see for example the London Underground to scale). But looking at a city’s rail transit map, you’ll almost always be able to tell where the CBD is, where the inner-urban neighborhoods are, and where the outer-urban and suburban areas are.

At this distance, then, having a bus-friendly grid doesn’t matter much. What matters is having a good network of historical rights-of-way that can be used for regional rail, and a preexisting pattern of development following these lines and their junctions. In the US, the older cities have this, whereas the newer ones do not. In a suburb like Attleboro, good transit means good regional rail, with high all-day frequency, and a network of feeder buses timed to meet the trains. Grids aren’t especially useful for that.

And this is why, despite being so dense, Los Angeles has so little transit usage. Its street network is set up for bare-bones public transit, usable by people who can commute two hours in each direction and will never get cars. Because it was a medium-size city when its car ownership exploded, it doesn’t have as many town centers; its density is uniform. It has a higher weighted density than the Rust Belt outside New York, but its weighted-to-standard density ratio is much lower than those of Philadelphia, Boston, and Chicago. (It barely trails Washington, which has fewer town-center suburbs than the Rust Belt, but made an effort to actually build them around Metro; its Tarrytowns have Metro service rather than infrequent commuter rail.)

The optimal urban geography for urban transit is not the same as that for suburban transit, and the optimal street network for surface transit is not the same as that for rapid transit. Los Angeles could potentially excel at surface urban transit, but there’s only so much surface transit can provide the backbone of public transportation in a city. It has a handful of strong lines for rapid transit, and that’s a serious problem, which a grid won’t really solve.

State Boundaries and the Northeast (Hoisted from Comments)

Aaron Renn’s repost on US states mattering more than some people imagine made me think about the difference in attitudes toward state lines in different US regions. Aaron’s examples of state lines mattering come from the Midwest, specifically Indianapolis and Columbus. My usual examples of state lines not mattering come from the Northeast. And those two regions treat states very differently.

Imagine a thought experiment in which Congress allows states to redraw their own boundaries – to split, merge, or change borders on their own accord. Let’s ignore the Senate – perhaps it still uses the old boundaries. Let’s also assume that this is not a completely de novo redrawing, akin to the creation of the French departments, in which states are drawn to be of specific size or population.

In such a case, in most of the US, there would be small changes only. Indiana would lose the suburbs of Chicago, Cincinnati, and Louisville, but otherwise remain intact. Virginia would lose the DC suburbs and gain the North Carolina suburbs of Hampton Roads. Tennessee would gain the Memphis suburbs, and maybe possibly lose Chattanooga. Oregon and Washington would merge. California, Texas, and Florida could either survive more or less intact or split based on metro area spheres of influence. I do not know Florida well enough, but my understanding is that Texas and California have strong enough state identity that in a referendum, their major regions would vote against a split. Ohio might have cleaved if it had had only Cleveland and Cincinnati, but I believe the presence of Columbus would make it survive more or less intact. The only Midwestern state that would be completely dismembered is Missouri, which has no equivalent of Columbus between St. Louis and Kansas City.

The opposite is true of the Northeast. From talking to people from both Upstate and Downstate New York, I believe a referendum would result in both sides voting for a split. New Jersey exists as a coherent entity only in jokes about the state made by people from other states. Pennsylvania has at least three regions that do not identify with one another. But at the same time, a coherent Northeast region exists: there are strong migration ties, not only among the four main coastal metro areas but also to and from Pittsburgh and the Upstate metro areas, which have stronger migration ties to New York and Philadelphia than to Cleveland and Chicago. Along the coast, there is also suburb-sharing, which has led to the formation of just four combined statistical areas; there’s even a chain of suburb sharing connecting New York, Allentown (now in the New York CSA), and Philadelphia, and there may soon be direct sharing between New York and Philadelphia.

Unlike in the Midwest or Texas or California, the Northeast does not have the same university-enforced state boundaries, which are probably a major reason why Columbus specifically has migration ties with the entire state but not much with areas just outside the state. In much of the Northeast, a huge number of students go to private universities. In Massachusetts as far as I can tell there are more students at private universities than at public four-year colleges. New York has a very large public university system, but the SUNY/CUNY distinction reinforces the state’s internal divisions rather than erasing them the way Ohio State does.

In terms of a national rail plan, the Northeast practically is a single state (as is the Pacific Northwest, but that’s just two states), from Portland to Quantico. In California, Texas, and Florida, and even Ohio and South Carolina, there are potentially strong in-state intercity rail routes. New York and Pennsylvania have those as well, but both have even stronger routes that cross many states. The Midwest is full of routes that cross states, but usually those connect one or two states to Chicago; the main exception, Chicago-Detroit via Toledo, is indeed not pursued, in favor of the inferior I-94 route that mainly serves Michigan.

Regional rail is similar. It is possible to come up with a plan that’s at least theoretically coherent for regional rail in most parts of the US, to be run by a state agency (or in borderline cases a bi-state agency), or by a local agency with powers delegated by the state. In the Northeast, it’s completely impossible. It’s not even possible to cleanly cleave the region into separate states for the four primary coastal metro areas, because commuter rail services on the Northeast Corridor need to share track with intercity trains at least part of the way, and building infrastructure to avoid such track-sharing is needlessly expensive.

I do not know of a transport association that crosses so many boundaries of subnational entities in Europe. French services are run by the regions; they sometimes cross boundaries, but only in the Midwestern sense of a region bordering Ile-de-France running some of its regional trains to Paris. In Germany, Berlin and Brandenburg have the same transport association, and for all intents and purposes are a single state when it comes to rail network planning. Swiss services cross cantonal borders, but at least the Zurich regional rail network is again French or Midwestern, in that there’s a core of services funded by the ZVV, and services in bordering cantons that run through. In the Northeast, there are good reasons to have commuter services run through from Philadelphia to New York along the Northeast Corridor and maybe also the West Trenton Line; even metro area boundaries are not hard, let alone state boundaries.

Stepping back from the thought experiment, let’s think of how to organize transportation planning in the US. In the Midwest and the South, states are coherent entities. In the West, the areas where states really do not mean much are deep in the Interior West, where there’s no point in building additional ground transportation infrastructure in the first place. But in the Northeast, there may have to be a special exemption treating all of it, including Northern Virginia, as a single state for planning purposes. It can’t be run as tightly as a single state because of its size and its natural division into several metro areas, but some joint service between its various divisions is unavoidable.

More in general, the Ohio example showcases how coherent state identities can be manufactured by the presence of state institutions. On maps that center Cleveland and Cincinnati, such as maps of which baseball and football teams people support, Ohio looks completely dismembered. And yet, the presence of Columbus and Ohio State changes everything when it comes to economic ties such as migration: suddenly, the otherwise-artificial state border means something in terms of social services.

This is not something Northeastern states can really do, nor should they. Pennsylvania has Penn State at State College, but it’s in a small, faraway town, and people who can instead go to Penn or Carnegie Mellon will. New York can expand CUNY and SUNY, but there are too many campuses to provide the same social function of Ohio State. Of course states should expand public higher education, in terms of both opening new campuses where needed and subsidizing tuition, but there’s no room to create a new Columbus; such expansion would provide a necessary service to state residents, but not change economic geography the way it did in Ohio.