Missing the Point on High-Speed Rail

Ed Glaeser is a fantastic economist. He has done magnificent work analyzing the economics of urban growth and written indispensable papers on the connection between housing regulations and migration.

But when the man picks up his pen to write a piece for public consumption, he tends to take complete leave of his senses. I realize that this is a common affliction among economists, but Glaeser suffers from a severe case of the syndrome.

In a Friday piece in the Boston Globe, Glaeser takes on the administration’s push to fund construction of high-speed rail corridors around the country. In doing so, he combines the cognitive failures of every amateur train hater with a serious lapse in critical thinking.

He begins by making two very valid points: transit agencies are currently suffering serious and unfortunate shortfalls, and transportation funding generally is allocated where it’s least needed — to states with low levels of population, population density, and congestion.

But then he rapidly goes off the rails. Glaeser writes:

Now the administration wants Americans to envision high-speed rail lines in the wide-open spaces of Texas.

For most workers in America’s sprawling metropolitan areas, no train is
going to drop them within walking distance of their home or job. In
Greater Houston, only 11.6 percent of jobs are within three miles of an
area’s center and more than 55 percent of jobs are more than 10 miles
away from the city center.

Of course, Texas has four of the nation’s fastest growing metropolitan areas, all within a few hundred miles of each other — an ideal distance for high-speed rail. Austin, Dallas, Houston, and San Antonio are currently home to some 16 million people, and those metropolitan areas have added 3 million people since 2000 alone. Congestion is an issue within those metropolitan areas and will continue to worsen as they grow.

Not only is it entirely appropriate to build transportation infrastructure with future growth in mind, it’s imperative. America’s current sprawling growth pattern resulted in no small part from the mass construction of interstates and highways, which drew suburbanites to previously unsettled areas.

Moreover, Texan metropolitan areas are working to accommodate future growth in a denser fashion by building miles of metropolitan transit systems. Transit and rail are complementary technologies, each of which will increase the return on investment of the other.

Glaeser’s errors continue.
He says:

There is a reason why 48 percent of Amtrak’s passengers travel on only
two routes: the Northeast Corridor and the Los Angeles-San Diego line.
For travelers in the less-dense areas between the coasts, cars beat
trains for modest distances and planes win over long hauls.

But of course, so many people use these lines because they are effective, which is not only a function of congestion. Rail carries passengers between New York and Washington in fewer than three hours, while drivers couldn’t manage the trip in fewer than four even without congestion of any sort.

In other words, causation between mode choice, density, and service quality moves in several directions. Better service would lead to higher rates of travel and increased density. In the pitifully funded corridors where intercity services average 45 miles per hour or less, it’s no wonder that ridership lags.

He goes on:

The national high-speed rail agenda is being pushed with claims that
these trains will jump-start economic growth. No serious evidence
supports such claims. When new transportation does affect local
economies, it generally does so by moving activity from one place to
another, not by creating nationwide benefits.

This is a strikingly economically ignorant paragraph. Because Glaeser doesn’t mention specific claims about the ways in which rail construction might affect growth, we can’t argue against his first statement.

But as Glaeser almost certainly knows, transportation investment has multiple effects. Some of these shift the relative attractiveness of one area relative to another. Other effects improve absolute economic performance, by facilitating agglomeration, increasing potential scale, and allowing for greater specialization and intercity trade (of goods and services).

Glaeser’s assertion doesn’t pass a basic sense check. Has America enjoyed net economic benefits from construction of its roads, rails, canals, ports, and airports? Unquestionably.

Glaeser then channels his inner Randal O’Toole:

The case for subsidizing urban mass
transit, like the MBTA, is certainly debatable, but it is much stronger
than the case for subsidizing rail links between non-coastal cities.

MBTA’s core problem is that its operating expenses have always been
double its operating costs. To function, the system needs about $900
million a year in subsidies from taxes and local assessments, which
works out to slightly more than $2 a trip.

Amtrak also regularly faces a $1 billion gap between revenues and
expenses, including depreciation, but since Amtrak carries only 29
million passengers each year, the per-trip subsidy tops $30.

Glaeser makes several key mistakes. First, the case for subsidizing transit is really pretty clear — it is economically efficient to do so.

Second, it’s absurd to leave out a discussion of highway spending. All modes are heavily subsidised by the government, and roads come nowhere close to paying for themselves. Reducing those subsidies would increase demand for transit and rail.

Further, congestion on highways indicates that roads are seriously underpriced — another subsidy which cost Americans $78 billion in 2007.

And finally, Glaeser’s Amtrak statistic is extraordinarily misleading, and bordering on dishonest. Amtrak routes include fast, competitive, and successful lines — like the Northeastern corridor — as well as congressionally mandated, slow, and poorly traveled ones. Amtrak makes an operational profit on its top corridors, but loses a lot of money on slow routes.

Glaeser would have you believe that all routes are money losers, when in fact it seems clear that high-speed, reliable service is extremely cost-effective.

Glaeser makes the same mistake in complaining that Amtrak is only 17 percent more fuel efficient than airlines. But efficiency varies with ridership; a full train on a medium distance run is much more efficient than a plane ride over a similar distance. In energy terms, it makes little sense to run a mostly empty train across the country at slow speeds, but that has nothing to do with the potential advantages of a quality high-speed rail network.

He closes by saying:

A rational transportation program would target money to the areas that
have the most congestion. A smart transportation policy would recognize
the wisdom of using our existing infrastructure more efficiently, with
the help of congestion pricing, rather than building more roads.
Unfortunately, wisdom seems to take wing whenever politicians start
envisioning the shining splendor of fast trains.

Glaeser is correct that a good place to begin addressing our transportation failures is by pricing congested highway and air routes more effectively.

But we have every indication that doing so would significantly increase demand for rail services, while also raising tens of billions of dollars every year that could be used to construct a rail system that would be cleaner and faster than driving or flying. Contra Glaeser, pricing our existing infrastructure would make it painfully clear just how badly we need an effective intercity rail system.

In environmental and economic terms, the case for major investment in high-speed rail is quite strong. Unfortunately, wisdom seems to take wing whenever economists start writing about public spending.

  • Dee

    This is an interesting, and very thoughtful commentary. Glaeser’s piece really irked me too, after reading about it on another blog this morning. I know you focus on the economic side of things here, but after reading another piece it made me think that Glaeser also ignores the need for rail systems in not-so-densely populated areas. I think your point on Texas is spot-on, (isn’t Dallas/Forth Worth the best performing metropolitan economy right now), but I also think we need to increase access to transit in decaying rustbelt cities too. It wouldn’t respond to any economic need or congestion need, but it certainly would fulfill a social need in these isolated cities. Anyways, these ideas came from another post, complementary to yours in many ways: http://socialsciencelite.blogspot.com/2009/07/congestion-isolation-and-goal-of-mass.html

  • You can find in Glaeser’s piece an underlying case for expanding rail in both modes at once. I used to rely on Atlanta’s public transit and it was very useful if you were going to the airport from somewhere near a MARTA station or to travel between two places both near MARTA stations. I think there is a case to be made that while train service and pedestrianism within Texas cities are inconvenient or a vale of sorrows, high-speed rail between them will be less attractive so less economically beneficial than they might be.

    One interesting datum to find: San Diego has a fairly small and dense commercial area while LA deserves it’s notoriety as an automotive city. I wonder how many of the LA-SD train trips originate in Los Angeles and how many San Diego. I imagine it is much more attractive to train south than north.

  • Good post. However, we really shouldn’t be debating HSR until the nation is plastered with Commuter Rail and Light Rail. Attacking car ridership at its most dense levels of penetration is where electrified transport can be most effective first, in capturing efficiencies. And, in positive externalities.

    The larger point is that the Obama Administration has exactly the wrong approach. They have invested massively in cars, car companies, biofuels, and roads. Then when they address Rail, they go to the last place first: HSR. From a political standpoint, this is also a mistake. For it opens them to exactly the kind of criticism coming from Glaeser–even if it too is infused with wrong-headed analysis.


  • bdbd

    Dee, thanks for providing another instance of the near universal truth that in any document (actual or virtual) of any length that concerns the Dallas/Ft Worth area, there will be at least one occurrence of “Forth Worth”

  • bdbd

    The FAA has been trying to include authority to price access to airspace, including congested airspace, more rationally, and Congress has decided not to go along. The Obama folks have indicated that they intend to include a shift from aviation ticket taxes and fees to an operations based user fee system in future proposals.

  • Andrew

    With the long distance trains sleeping car operations do operate at a small profit, they also serve areas that lack travel options. Though there are even long distance trains on the NEC too, they’re symbiotic.

  • Why should we be subsidizing rail? I understand that autos are subsidized via the highways, but isn’t the answer there just to raise the gas tax and other user fees so that this is no longer the case? I am not sure how one dumb policy — subsidizing highways — justifies a second.

  • Very well said. I critiques this article myself on our website.


    I was astonished to find such a one-sided viewpoint. The role of mass transit should be to alleviate the problems of sustainability and inequity across the U.S., not just in a particular region.

  • i’d have to say this is one of the worst posts i’ve ever read on streetsblog. depending on how generous one is feeling, it falls somewhere in between incompetent, and dishonest.

    let’s take just one case:

    Of course, Texas has four of the nation’s fastest growing metropolitan areas, all within a few hundred miles of each other — an ideal distance for high-speed rail.

    what does this have to do with Glaeser’s argument? nothing. it’s totally irrelevant. we might as well talk about the weather.

    in fact, a couple of paragraphs later, you make Glaeser’s case for him when you write:

    Moreover, Texan metropolitan areas are working to accommodate future growth in a denser fashion by building miles of metropolitan transit systems.

    Glaeser just spent an entire article telling us to ‘put transit where the people are,’ and presumably you have a problem with this, but then back up Glaeser’s point by telling us that Texan metro areas are, indeed, ‘putting transit where people are.’ just bizarre.

    it’s not like Glaeser is the first to call high-speed rail a joke–pretty much anyone with any common sense has been saying it, including Kunstler and Chomsky.

  • John Thacker

    “Glaeser makes the same mistake in complaining that Amtrak is only 17 percent more fuel efficient than airlines. But efficiency varies with ridership; a full train on a medium distance run is much more efficient than a plane ride over a similar distance. In energy terms, it makes little sense to run a mostly empty train across the country at slow speeds, but that has nothing to do with the potential advantages of a quality high-speed rail network.”

    Sure, efficiency is different with load. That’s why inner city buses at peak times are more energy efficient than trains, but off peak buses are some of the least efficient transportation there is.

    However, there are several points you’re ignoring.

    1) It is impossible to avoid partially empty trains. United Airlines just announced that their load factor across all planes was 85.9% in June 2009; such a level is nearly impossible for a train system. Planes fly point-to-point; a train must pass by each station. Anyone who has ridden or attempted to buy tickets for Amtrak’s Carolinan knows that it is much more full on the portion on the Northeast Corridor. When it sells out between two points, say Richmond and DC, it is impossible to buy tickets from Raleigh to DC. The same is true of any Amtrak long distance train. Increasing their load factor is *the* story about increasing airline energy efficiency over the last 20 years. Amtrak’s overall load factor is 50%, even in the last two years with higher ridership (see section A2, Summary Metrics, of any of the recent reports), and the Acela total load factor, weighted by miles traveled, is not much better at around 60% (63% last year, 57% this year.) (Look at section C of any of Amtrak’s recent monthly reports, and divide contribution per seat mile by contribution per passenger mile to get passenger miles per seat miles, or load factor.)

    These numbers are typical for trains worldwide. Amtrak’s load factor for its peak legs are higher, but any train route with multiple stations is going to have peak busier legs, and thus more empty inefficient legs to compensate. (Unless the train sheds cars at one stop, of course, but this happens much less efficiently than the ability of an airline to upgauge or downgauge aircraft size on various point-to-point routes.)

    Yes, hypothetically if trains could achieve the load factor that sophisticated advances in route scheduling have allowed airlines to reach recently, they would look better. But it’s not possible. Feel free to search for information on other successful high speed rail systems, like the shinkansen, or Taiwan’s high speed rail. Load factor for an entire system does not exceed 50-65% anywhere, though again obviously on peak routes it’s better.

    2) While train stations can be located in central cities, there is environmental and energy costs in laying and maintaining track. Rail has to be maintained in order to stay at the level necessary to maintain high speed service. That cost is not counted in these numbers presented, and I would strongly suspect that the energy cost of rail maintenance and construction exceeds that of airport maintenance and construction. The energy numbers presented are thus biased in favor of rail.

  • Ed Glaeser is a fantastic economist.

    Maybe, but Paul Krugman is a fantastic economist who gets urbanism. And he rides the train.

  • NikolasM

    But a train with 400 seats could conceivably service 1200 people over a route. Is that a load factor of 300%? I’m sure airlines would love to cram that many in.

  • TimW

    John, there’s a few key things that you’re ignoring. There already are cases where Amtrak adjusts capacity en route; for example on Amtrak’s Chicago-Seattle Empire Builder, a car is cut off at Minneapolis from the westbound train to be attached to the next day’s eastbound train. Amtrak could stand to do this more frequently, of course, but one can’t write off the whole mode of transport just because it’s not being done as often as it should. If Amtrak weren’t so cash-starved, they would have extra equipment and could add a car between your hypothetical example cities of Richmond and DC. It’s actually a lot easier to attach a car on a train than it is to change the model of aircraft used, remap the seat assignments, find a new crew qualified for that model, etc.

    Secondly, it’s not easy to assess the load factor of a train. Do you do it with a snapshot at some moment in time, or do you measure it over the whole run? In the latter case, numerous routes are running over 100% capacity because seats turn over more than once. Again, citing the Empire Builder as an example, only 5% of the train’s passengers are going between the endpoints of Chicago and Portland/Seattle (the train splits in Spokane). That’s some major turnover. Even something simple like initial load is actually rather difficult to assess. Just about every Amtrak long-distance train makes a suburban stop right after it originates and right before it terminates. This is a great convenience to certain passengers who don’t want to travel all the way downtown only to have to come back out to the suburbs, but it means that the train is going to have more empty seats for part of its journey than it otherwise would.

    It’s also Amtrak’s policy to not overbook reserved trains. I’m sure if they wanted to deal with the fallout of having overbooking policy, they could sell more tickets than seats and push up those numbers somewhat. As a frequent rail passenger, I’m glad they don’t. Sold out trains are not uncommon during the summer months, believe it or not.

    Finally, I disagree with the assessment that the energy costs associated with maintaining the infrastructure are higher for railroads than they are for aircraft. A single airport will have thousands of lights, hundreds of acres of paved surfaces that need to be patched or plowed, and several high-frequency navigation radios that operate all hours of the day and night. It requires a lot of electricity and diesel to tend to that.

    Contrast this with a rail line which has signals that are illuminated only when a train is near, radios that activate only when they’re being used, and a narrow strip of land that rarely even needs to be plowed for a train to make it through. Recent advances in regenerative breaking will only help to lessen the energy usage of trains, but there’s never going to be a way around the fact that you need to convert gobs of petro energy to kinetic energy to potential energy just to get a plane off the ground. And that energy doesn’t really convert back.

  • Eric H

    I took Glaeser’s main point to be that local mass transit in congested commuter corridors provides more bang for the buck than any inter-urban HSR project, in terms of reducing travel time, improving fuel economy, etc. This is largely because inter-urban air (and existing rail or even vehicle) travel is relatively cheap and convenient (for now) in comparison to local auto-centric commutes for most people. While inter-urban rail (including HSR) may be viable in some places, limited funding should be spent on moving commuters around and alleviating congestion at the local level.

    Glaeser picked a lousy example to trash HSR–the Texas cities are among the best potential nodes for regional HSR if and when it’s ever economically competitive with airlines or highway travel (294 daily one-way non-stops between these 4 cities on just 3 airlines)–but if you’re comparing apples to oranges the apples win out.

    I’d agree with his premise that priority funding for local mass transit to relieve congestion (and coincidentally increase development and population dnsity along the routes) is the way to go for now–whether you’re in Boston or Austin.

  • Adam K

    There are lots of dimensions missing from the discussion about passenger rail:

    #1 — Passenger miles. “Passengers” or “trips” is not as useful a metric as “Passenger miles.” Passenger miles is the “work” done by a transportation system. If you want to normalize transportation subsidies, it is useful to look at passenger miles in addition to trips. Taxpayers may spend $2.00 to move someone from home to the gocery store on a city bus, and they may spend $15 to move someone 100 miles on Amtrak. These two “trips” are not comparable, so comparing the subsidy in this way is useless.

    #2 — Long Distance train economics are often not as bad as short-distance economics. Even when passenger rail systems were private in the USA, commuter trains lost money faster and sooner than the long-distance trains (you can refer to historical material and economics articles from the 1950s for a discussion of that). Government stepped in far sooner to subsidize commuter rail than it did to rescue the intercity passenger train. The key factors in passenger train economics are costs (e.g. variable costs like labor, as well as fixed costs) and revenue.

    On the revenue side, you earn more money with higher fares per mile and a higher load factor (percentage of seats occupied for each mile of service). Long Distance trains offer amenities and serve a market that can command a reasonable fare. Long Distance trains often have a higher load factor than short distance trains for reasons of geography — it isn’t serving a “tide” of commuters who only fill certain segments or certain times of day. Also, the long distance train runs at night, so equipment is “earning” revenue more hours of the day. Short distance trains get parked every night.

    On the “costs” side of the ledger sheet, the long distance trains share most of the capital costs (the infrastructure) with freight railroads. Since Amtrak pays an access charge as opposed to having to deploy capital to actually build and maintain the rail infrastructure, they are, in fact, converting capital costs to marginal operating cost, which can be advantageous to Amtrak. In contrast, when you actually own the railroad, and when your railroad is a very old but very fast line (and thus has high maintenance costs) in an urban area, the costs of ownership can be very significant. The Northeast Corridor only works because it reaches New York City, and the level of infrastructure necessary to reach Manhattan is staggering. Even now, as digging under the rivers and the island continues for improved rail access, massive sums are being deployed to improve train travel into New York. If those capital sums were charged against passenger service, the subsidy would be plain.

    To conclude: intercity passenger rail makes sense on the merits of the intercity travel market. Too often we compare apples to organges– comparing commuter trains with intercity trains — when each has a role to play in effective transportation.


Glaeser Goes Out With a Whimper

For those just tuning in, economist Ed Glaeser has been writing a four-part series on the potential costs and benefits of high-speed rail at the New York Times‘ Economix blog. He began three weeks ago with an introduction. The following week he addressed direct costs and benefits from a hypothetical line, and last week he […]

Glaeser Takes an Unserious Look at High-Speed Rail

Ed Glaeser is a very good economist, and his papers are indispensable reading for those interested in the workings of urban areas. But he is also a strident conservative, whose popular writings frequently challenge conventional progressive wisdom (and my own views). Harvard University economist Ed Glaeser (Photo: NPR) I was interested, then, to read that […]

High-Speed Rail: Still a Good Idea

You may remember, back in August, economist Ed Glaeser’s series on high-speed rail at the New York Times’ Economix blog. Glaeser put together a back-of-the-envelope cost-benefit analysis of a hypothetical Houston-Dallas line, which purported to show that rail was a poor investment. You may also remember me responding in detail, and generally pointing out how […]

In Defense of High-Speed Rail

Today on the Streetsblog Network, we’ve got a post from Yonah Freemark at The Transport Politic on the importance of funding both intercity and intracity rail, despite limits on the amount of money available. Freemark takes on the argument that investment in transportation within cities should trump the construction of more efficient rail connections between […]

The Washington Post Features Rail Hack Job

This is the big problem with Ed Glaeser’s New York Times posts purporting to analyze the costs and benefits of a high speed rail system. Despite Glaeser’s acknowledgment that his "back-of-the-envelope calculation" doesn’t "[represent] a complete evaluation of any actual proposed route," the posts are sure to be read and regurgitated by rail opponents uninterested […]

Ed Glaeser’s Rail Fail

The story so far: Ed Glaeser recently began an effort to assess the costs and benefits of constructing high-speed rail lines at the New York Times’ Economix blog. Last week, he posted his first substantive take on the issue, an attempt to estimate direct costs and benefits from a hypothetical line between Houston and Dallas. […]