You are using an outdated browser.
Please upgrade your browser
and improve your visit to our site.
Skip Navigation

High Speed Rail: Getting the Assumptions Right

With a just-released Brookings report suggesting that high speed rail (HSR) could mitigate excessive congestion at airports, it would be nice to know if and where rail is a viable investment. The Harvard economist Ed Glaeser sparked a useful debate on the merits of high speed rail with a few recent posts at Economix over at The New York Times.

Glaeser concluded that HSR was not likely to be an efficient investment, but, as with any analysis, examining the assumptions is key.

Though Glaeser’s core approach is sound, if we tweak two of his assumptions ever so slightly--in a more realistic direction--we see that the net benefits go from negative to positive. I’ll focus on just those aspects of Glaeser’s analysis which merit criticism: the interest rate and the costs per passenger.

To begin with, he presented a simple formula for estimating the net benefits of a hypothetical high speed rail (HSR) project:

Number of Riders times (Benefit per Rider minus Variable Costs per Rider) minus Fixed Costs.

First, within that calculation he picked an arbitrary interest rate of 5 percent, which, as he points out, makes an enormous difference in the final calculation. But there is a well-functioning market that generates daily the interest rate of a long-term investment, accounting for inflation in addition to all available economic opportunities. As of September 24, 2009, the 20-year inflation-indexed interest rate on U.S. treasury bills was 2.13 percent (the nominal 30-year interest rate was 4.17 percent). The bottom line is that Glaeser’s interest rate of five percent overstates what the market believes an alternative long-term investment is worth.

The second major issue is that Glaeser sets the net operating costs of rail at $0.30 per passenger per mile. Yet, Amtrak’s monthly reports show a more favorable figure. For the year leading up to July of 2008, the operating losses per passenger mile were just $0.11 (see page 65 or C2). Even during the midst of the recession, over the year leading up to July of 2009, the losses were just $0.13 per passenger mile.

Moreover, not all corridors operate at a loss. The Northeast corridor trains brought in a profit of $0.08 per passenger mile for YTD July 2008. Still, to be conservative, I’ll use the minus $0.11 figure in my revised calculations.

The table below shows the results of re-running Glaeser’s formula while changing just two of the variables. I also add the L.A.-Vegas route, which has the second most passengers for corridors under 400 miles, and thus should be a good candidate for HSR. To avoid a recession-induced bias, I use 2008 metro-to-metro air passenger data from that report (this makes the Dallas-Houston passenger total 3.2 million and the LA-Vegas corridor 4.3 million). Finally, note that the calculations below ignore the environmental benefit of rail (which Glaeser puts at $22 million per year for a corridor like Dallas-Houston).

Benefits of High Speed Rail Under Variations to Glaeser's Assumptions

(in $ Millions)

Notes: Nominal Fed Rate refers to the interest rate on 30-year treasury bills (4.17%, in this case); Real rate refers to 20-year inflation indexed bills (2.10%). "Amtrak costs" refer to net costs per passenger mile for all Amtrak routes for year-to-date July of 2008, according to Amtrak's monthly report (I used -$0.11). 2008 passenger data and mileage are taken from the Brookings report by Adie Tomer and Rob Puentes, "Expect Delays: An Analysis of Air Travel Trends in the United States" (Washington: Brookings Institution, 2009). All other data are from Glaeser.

On the far left hand side are the net benefits under Glaeser’s assumptions. Moving from left to right, more realistic assumptions are introduced one by one, first using a nominal interest rate and then an inflation-adjusted one. It doesn’t take much to change the minus sign to plus for HSR, especially for the potentially strong L.A.-Vegas line.