Maybe Pigs Can Fly

If I am forced to bet on who is correct in an argument, and all that I know is that Paul Krugman is on one side of it, and Robert Samuelson is on the other, I will pick Krugman. I am guessing that the net expected value of an even money $1 bet would be something like $0.95.

I also happen to like trains. For the semester I commuted from Washington to Philadelphia, I was very grateful that I had Amtrak, for driving up I-95 several times a week would have been stressful. When I lived in Washington, Metro allowed me to keep my car at home on all but the rarest occasions. When I go to Atlanta, I use MARTA to get to downtown from the airport; when I go to San Francisco, I take BART, etc. I even try to take transit one day a week to get to work here in LA, because I feel that I should (although it takes a lot more time than driving).

All that said, I think Samuelson is correct that high-speed rail in the United States would a be wasteful expenditure. He makes the argument that because the US is less dense the Europe and Asia, rail makes less sense for intercity travel. I am sympathetic to this argument, as I have made it myself. Krugman replies that the Northeastern United States has density similar to Europe, and that Samuelson’s comparison is therefore spurious. I noted this as well.

Krugman’s argument has two problems. First, Obama’s high speed rail plan includes several places outside of the Northeast, including the South, Texas, Chicago-St. Louis-Kansas City, where densities are considerably lower than in Europe, let along Asia. Second, even in the Northeast, cities are considerably less dense than in Europe. The graph below is from Alain Bertaud.

An advantage rail has over air in Europe is that large numbers of people live close to rail stations, while airports are generally on urban peripheries (lots more people live near the Gare de Lyon than Charles de Gaulle airport). No matter where a rail station is placed in an American city, it will be far from a lot of people, at which point the speed of an airplane overtakes the speed of a high-speed train, in terms of determining total trip length.

Finally, even taking external costs into account, air costs much less than rail in a place as dense as California. Let’s turn it over to David Levinson, David Gillen, Adib Kanafani and Jean-Michel Mathieu from UC-Berkeley [correction–the paper was posted in a UC-Berkeley series. Levinson is from the University of Minnesota]:

The Full Cost of Intercity Transportation Page ES-1 This study evaluates
the full cost of three modes of intercity transportation: air, highway, and high
speed rail. The evaluation is done within the context of the California Corridor,
connecting the Los Angeles Basin and the San Francisco Bay Area. The purpose
of evaluating full cost is to compare the economic implications of investment in,
or expansion of, any of these three modes. The scope of the analysis is full
transportation cost. Full transportation costs includes external, or social cost,
in addition to the internal costs of construction, operation and maintenance. In
this study we include estimates of four types of external, social costs: accidents,
congestion, noise, and air pollution. The 677 kilometer corridor for which these
estimates are computed represents one of the alignments of a proposed high
speed rail system between Los Angeles and San Francisco. The methodology
used is to construct cost functions that relate costs to levels of output, as measured by passenger-kms. or vehicle-kms.

Different types of costs are estimated as permitted by available data. These include short run costs, in which the physical capacity is held fixed; and long run functions in which capacity is allowed to expand to meet higher levels of demand. Average and marginal costs are computed for highway and for air transportation. But given the absence of high speed rail systems in California only average costs are estimated. The highway and air cost models are developed from basic principles and are estimated
with actual data and system design characteristics observed in the California
corridor. Rail costs are estimated with models that have been adapted from
estimates for the French high speed rail system, the TGV, using available data
for their estimation. Based on the results summarized in Chapter 7 and shown
in Table 7.1, we find that the full cost of air transportation for the California
Corridor ($0.1315 per passenger-kilometer traveled (pkt)) is significantly less
costly than the other two modes. The full cost of high speed rail and highway
transportation cost approximately the same; rail costs $0.2350/pkt and highway
costs $0.2302/pkt. The internal, or private, monetary costs comprising
infrastructure, carrier, and vehicle operating costs are clearly highest for rail
($0.19/pkt), followed by air ($0.11/pkt) and then highway ($0.10/pkt). And as
is to be expected, user time costs are highest for the slowest mode, the highway
system, followed by rail and then air. Adding user travel time costs to the
monetary costs results in the total internal system costs per passenger-km. of
$0.124 for air; $0.233 for rail; and $0.198 for highway.

In other words, if we disregard external costs then we find that high speed rail is nearly twice as costly as air and that the highway is not far behind. However, if we look at social costs alone – congestion, air pollution, noise, and accidents – we find that high speed rail is clearly less costly than the other modes. In this research the only measurable social cost of high speed rail is that of noise, which at $0.002/pkt, is
significantly lower than that of air at $0.0043/pkt and highway at $0.0045/pkt.
Highway transportation, on the other hand, has a relatively high cost in terms
of air pollution and accidents, two externalities which are virtually absent in
high speed rail. In this study, we consider that the pollution resulting from the
electric power generation used to drive a train is to be allocated to the energy,
and not the transportation sector. Thus, any pollution externality associated
with high speed rail should be already internalized in a higher price for electricity.

Similarly, a 100% safe system, such as high speed rail, implies higher capital
costs due to construction of grade separations, more intelligent systems, etc…
Hence, the avoidance of accidents by high speed trains is not “free”. Therefore,
high speed rail, while more costly than highway transportation in terms of internal
costs, primarily due to its high capital cost, is significantly less costly than
highway in terms of social costs. This comparison is illustrated in the following
figure, where full costs are broken down into three categories: internal, travel
time, and external.

Note that the full cost of rail (.23/pkt) is considerably less than air (.13/pkt) even though any pollution created by trains is attributed to the power plant that provides the electricity, rather than the train itself.

A couple of other points. I have yet to meet a transportation economist (and I talk to a fair number) who is thrilled by high-speed rail as a technology. John Kain was among the most rigorous and influential transportation economist of the past 50 years, and he was very skeptical about rail. I also think that we yet again have evidence that we don’t come even close to internalizing the social costs of automobiles, but I see no political will for really reducing our dependence on the auto, in part because most people love their cars (this is hardly unique to America). High speed rail also seems to me to be a way to redistribute income from lower income Americans to higher income Americans, because lower-income Americans will choose Southwest Airlines (which will be cheaper) when they need to travel from one city to another.

Finally, I see that when Ed Glaeser expressed skepticism about high speed rail, he was accused of being a “hack,” so I guess I will join his company. But while Paul Krugman is indeed much smarter than Robert Samuelson, I would have a very hard time choosing (not that I would want to) between Glaeser and Krugman.