Tuesday, April 15, 2014

Marginal costs and revenues for Amtrak trains

With the release of the combined FY2014 Budget and Business Plan, FY2015 Budget Request Justification, and FY2014 – 2018 Five Year Financial Plan by Amtrak, we're able to gain a little insight into the marginal expenses of running Amtrak trains according to whether they are long distance, state supported corridor trains, or part of the Northeast Corridor. This counts only fuel, T&E and OBS crew and labor, host railroad maintenance of way, host railroad performance incentives, commissary, car and locomotive maintenance and turnaround, and maintenance of way support. These figures can be found on pages 54 (for NEC), 61 (for state supported corridors), and 69 (for long distance trains). Train mile figures are in the appendix, pages 79-80. It goes without saying that calculations based on these figures are of necessity somewhat crude. These numbers do not, of course, account for expenses which are marginal at the level of the route itself, such as stations, nor for maintenance of engineering support or other such costs.

Train miles Marginal cost per train mile Fuel and power T&E crew labor OBS crew labor and support Host RR MoW Host RR performance incentives Commissary provisions and management Car & Locomotive Maintenance and Turnaround MoW support
Long distance

Northeast Corridor

State supported


It is completely unsurprising that the long distance trains have a vastly increased onboard services cost compared to the other train services. Staffing on the Northeast Corridor trains is minimal, with only a single OBS crew member on most trains, and not all state supported trains have jobs for OBS personnel. This is in stark contrast to the long distance trains which generally operate with ten or more such crew, except for the Palmetto. Nor is it surprising that the Northeast Corridor, filled with various tunnels, bridges, and half a century of deferred maintenance, has significantly increased maintenance of way expenses. What is surprising is that the fuel cost of the Northeast Corridor, running on rather cheap electricity, is still higher than state supported trains. However, we know that Amtrak's current trains in the Northeast Corridor are incredibly energy inefficient and one source notes that "the very bad aerodynamics of the Amtrak high speed trains (one AEM-7 locomotive with six Amfleet coaches) means that, at higher speeds, the TGV or X-2000 (Swedish technology) actually use half or less the energy per p-km used by Amtrak in their Northeast Corridor" (page 25). With the introduction of the ACS-64 and the NextGen HSR trains, we should see a more appropriate fuel and power cost for the Northeast Corridor.

A grain of salt should be taken when looking at the fuel & power and car & locomotive maintenance figures for the long distance trains; these numbers will be slightly skewed upwards by the Auto Train carrying a lengthy consist consisting mainly of freight cars. Somewhat surprisingly, state supported train costs appear to be lower than those for Metrolink's commuter rail, despite the fact that the Metrolink costs did not include maintenance of way expenses. This may be due to higher operational costs from Metrolink's dispatching center, costs that were not offset in that calculation by the appropriate revenues from freight and Amtrak, as well as the fact that Metrolink has a significantly higher fuel cost. The failure prone old age of the locomotives may also be a major contribution to the higher cost.

With a view of the marginal costs, one that is admittedly quite rough, we can look to the question of which, if any, trains cover their marginal costs. Those trains which cover their marginal costs of operation benefit from increased frequencies since they will help bring down the total required subsidy. This is not to say that routes below that number do not benefit from increased frequencies or that they should not be subsidized (though subsidization expenses should not exceed the social benefits).

Unfortunately, Amtrak does not provide a breakdown of train miles by route, only by the business segment. Thankfully the FRA does provide those numbers, though only to the extent of the nearest ten thousand train-miles. These figures are available in the Quarterly Rail Service Metrics and Performance Reports, specifically in Table 11, located on page 16. Using these figures and Amtrak's 2013 ticket revenue figures, I was able to come up with a rough figure for the revenue of each Amtrak route per train mile.

Revenue per train mile Ticket revenue Train miles

Northeast Regional
—Richmond/Newport News
—All Other Northeast Regional

New York-Albany
Pacific Surfliner
San Joaquins
Pere Marquette
Blue Water
Capitol Corridor
Maple Leaf
Lincoln Service
Ethan Allen Express
Carl Sandburg/Illinois Zephyr
Kansas City-St. Louis
Heartland Flyer
Hoosier State
New York-Niagara Falls


Auto Train
Coast Starlight
Lake Shore Limited
Silver Meteor
Capitol Limited
Empire Builder
Silver Star
City of New Orleans
Texas Eagle
California Zephyr
Southwest Chief
Sunset Limited

A few explanatory notes are in order. First, the Springfield-New Haven Shuttles are not counted because I could find no separate figure for their train miles in the FRA data. It's possible that they are included with the Northeast Regionals, but, lacking knowledge, I excluded their revenue. Second, though the FRA includes all of the train miles for the Amtrak Virginia trains, such as from Lynchburg to Boston, Amtrak only counts the revenue from passengers who board or depart in Virginia; their actual performance is better than this indicates. Third, some non-Regional trains on the NEC, such as the Vermonter, have their revenue, expenses, and passenger numbers accounted for as Regionals until they leave the NEC; while a perfectly legitimate accounting measure, if their train-miles as reported to the FRA are not similarly adjusted, they will appear worse than reality. Finally, Empire Service trains to Buffalo apparently have their revenue counted under the Maple Leaf; I have added their train miles to those of the Maple Leaf's as a result.

It's worth noting that though the Northeast Corridor and several state supported trains cover their marginal costs of operation, the long distance trains do not, with the probable exception of the Auto Train. The Starlight comes the closest, but has additional expenses relating to the Pacific Parlour Car, I expect that the Lake Shore Limited comes the closest to meeting its actual marginal costs, rather than the rough estimate I can provide. It is also completely unsurprising that the California Zephyr, Southwest Chief, and Sunset Limited do as terribly as they do. While the other long distance trains have a variety of solid markets that they are able to serve across a significant portion of their route, frequently at either end, and many sharing that route with multiple other trains (Starlight with Surfliner, Capitol Corridor, and Cascades; Texas Eagle with Lincoln Service, etc.), these particular trains just have a whole heap of nothing for most of their routes, nothing that is often served at poor times as well, contributing to an outsized percentage of riders who take the trains from end point to end point.


  1. I hit Preview and my Comment disappeared. No wonder nobody posts Comments here. LOL.

    1. Huh. Haven't heard of that problem before and it's not reoccurring for me. Maybe your browser is interacting weirdly with Google?

  2. Note that fuel costs are per train-mile, and what the actual trains look like varies between the NEC and the state-supported corridors. In particular, NEC trains are pretty uniformly an AEM-7/HHP-8 with 8 Amfleets, while the state trains are often shorter, and thus require less energy to move.

  3. Are these numbers measured by route directly or are they estimated by aggregating total expenses and dividing them by route miles?
    The latter method is Amtrak's traditional system and successfully shifts much of the expense of short routes, such as the NEC, to longer routes such as the Sunset Limited.
    An outside audit of the actual fiscal performance of each route would be very useful for making economic decisions, particularly if a train itself is actually making a profit on operation, if not for excessive overhead assigned to it.

  4. By way of explanation of one of the oddities you encountered, under Amtrak accounting the Maple Leaf is just another state-supported NY-Niagara Falls train; everything west of Niagara Falls is VIA's business.

    So west of Niagara Falls, Amtrak doesn't pay the costs and Amtrak doesn't get the revenue; VIA gets both halves. What I don't know is whether VIA actually pays anything to Amtrak for wear-and-tear on the cars, and what the financial arrangement is for fuel, etc.