Monday, April 2, 2012

Amtrak costs per train-mile per route

After doing a bit of research with my last post and being both intrigued and perplexed by some of the disparities, I thought I'd go through and expand this to the entire Amtrak network and see how it broke down per individual route. The methodology involved is using the October 2008 schedule and assuming that it holds true for the entire year as a simplification while using Amtrak's FY 2009 annual report for cost information. Weekly one directional runs were calculated and it was assumed that the other direction would mirror image it. Total train miles were calculated by multiplying the weekly run by 52. There was no attempt to take note of special days of extra running or not running and it was assumed every single train ran and was not annulled. This was actually a fairly accurate simplification as it lead to a total of 37,943,672 calculated train miles against a reported 37,439,000 train miles. Additionally, the train miles and costs of the Northeast Regional, Washington-Newport News, and New Haven-Springfield were combined together into one accounting.

Northeast Corridor
Train Name Train Miles $ per train-mile
Acela Express
Northeast Regional

Corridor Trains
Train Name Train Miles $ per train-mile
Hiawathas429,312 $79.66
Maple Leaf 396,760 $71.33
Pacific Surfliner 1,632,800 $63.02
Capitol Corridor 1,201,928 $59.65
Blue Water 232,232 $50.81
Wolverine 663,936 $49.25
Pere Marquette 128,128 $49.17
San Joaquins 1,326,416 $47.57
Hoosier State 81,536 $45.38
Cascades 1,086,176 $44.28
Empire Service 1,350,544 $43.98
Chicago-St. Louis 827,008 $43.77
Heartland Flyer 149,968 $43.34
Illinois Zephyr/Carl Sandburg 375,648 $43.13
Saluki/Illini 451,360 $40.99
Pennsylvanian 323,232 $40.83
Adirondack 277,368 $38.58
Keystone 1,642,680 $37.5
Carolinian 512,512 $33.76
Kansas City-St. Louis 412,048 $27.91
Downeaster 422,240 $27.7
Piedmont 125,944 $26.2
Ethan Allen Express 175,448 $25.65
Vermonter 444,808 $17.76

The average Corridor cost per train-mile comes out to $46.68 per mile.

Long Distance
Train Name Train Miles $ per train-mile
Auto Train 622,440 $124.19
Coast Starlight 1,002,456 $75.81
Sunset Limited 622,440 $71.81
Silver Meteor 1,011,192 $71.8
Silver Star 1,011,192 $71.6
Crescent 1,002,456 $67.04
Capitol Limited 567,840 $66.39
Lake Shore Limited 843,024 $66.31
Cardinal 357,864 $60.35
Empire Builder 1,897,168 $56.87
Southwest Chief 1,642,368 $56.87
California Zephyr 1,774,864 $53.19
City of New Orleans 674,128 $51.47
Texas Eagle 950,040 $47.89
Palmetto 603,512 $42.92

The average Long Distance cost per train-mile amounts to $63.58 per mile.

The Hiawathas extreme expense per mile can, I believe, be explained by their short duration, only 86 miles per direction. This would overrepresent any fixed costs relative to operating costs. The Maple Leaf, while possessing a much longer run, will have additional costs due to operating as an international train; I would not be surprised if the costs for the Cascades are elevated in later years due to their connections to Vancouver, BC.

I am, however, rather puzzled as to why the California trains are so expensive. If it were just the Capitol Corridors, which invest rather heavily for high on-time performance, I could understand it, but the Surfliners are even more expensive and for no real reason that I can understand. There's nothing exceptional about the service that ought to result in an extra thirty million dollars a year to run it. By the same token, the Vermonter's extremely low cost is also rather puzzling and I do not know how to explain it.

What leaves me truly flabbergasted however is the cost of the Acela and the Northeast Regional. There is absolutely no justification for these services being more than double the price of a Corridor train and more expensive than every single long distance train bar the Auto Train. It becomes even more absurd when considering that the Keystone, which operates entirely on Amtrak-owned trackage, half of which is the NEC, and operates the same electric train consists as the Northeast Regional, yet costs only $37.5 dollars per train-mile to operate.

Try as I might, I can't think of any reason why the Regionals and the Acela would cost so much except for the imputed costs of maintaining the Northeast Corridor and those being attributed to them and apparently solely them rather than any of the other trains which share the corridor. This is, in my opinion, another reason why ownership of the Northeast Corridor ought to be removed from Amtrak and given to an infrastructure company, restoring Amtrak to its original role. This results in a more equitable and open understanding of the costs as well as the possibility of alternative funding rather than charging the various NEC commuter agencies and Amtrak for trackage rights (for instance, a multistate carbon tax or gasoline tax is one possibility). Preferentially I would suggest charging on a ton-mile basis; this has the added benefit of greatly encouraging Amtrak and commuter agencies to invest in, and lobby for, light-weight locomotives amenable to high acceleration.

If the Regional and Acela were to have costs equal to the average of Corridor and Long Distance trains, respectively, as might be the reasonable result of transferring ownership, the resulting savings would pay for 75% of the operating loss on Amtrak's long distance trains or even possibly the entire amount should Amtrak engage in the activities urged by passenger groups for sometime, such as buying more sleepers and converting the Cardinal and Sunset Limited to daily service, with the lower losses that would entail. Losing the NEC may very well be the best way to save Amtrak as a result.


  1. Thanks for this and some of the other math you've been doing lately. I've done some of my own calculating in the past -- estimating passenger-miles and average trip lengths, for instance -- and it's interesting to see how these compare. My spreadsheet is based on FY2010 data, however.

  2. You're probably right about the NEC costs being attributed to the Acela/Northeast Regional.

    Besides the NEC, the only other significant stretch of track that Amtrak owns is in Michigan. Surprise surprise, the Michigan Services are near the top of the list.

    Another big chunk of the Chicago-Detroit corridor is about to fall into public ownership, although the owner will be the state of Michigan instead of Amtrak.

  3. How would seperating the NEC from Amtrak help?

    It would just require track access charges to be spread over every train rather than over the two most capable of supporting it, further weakening the business case for LD services which use the NEC.

  4. It wouldn't be spread over just Amtrak would be the result, but also the commuter rail agencies which have vastly more trains operating over the infrastructure (though for less miles), as well as allowing for alternate means of paying for track maintenance. Right now the apparent toll is $55 per mile or $34 dollars per kilometer. That's more than double what RFF charges on Paris-Lyons (€14 per train-kilometer) which, to my knowledge, is the highest in the world, and for much better track than the NEC. Sharing it with commuters would cut it significantly, by about half (although still raising the question of why so expensive compared to rest of the world).

    Meanwhile, Amtrak only pays ~$5.27 per train-mile with its host railroads (it's the section just above "State Support").

  5. I spent some time with these numbers (and overall contribution/loss, and contribution/loss per seat-mile) over the weekend and found that there's a strong correlation between cost per train mile and the number of seats per train. However, Amtrak seems to be fudging the number of seats per train a little in some cases and a lot in others. The Maple Leaf and Vermonter are prime examples -- despite the fact that I haven't been able to find any instances of a Maple Leaf more than 5 Amfleets long, Amtrak seems to be assigning it a value of 554 seats per train on average. The Vermonter, which also runs a 5-car Amfleet consist in every case I can find, is apparently being recorded with an average of 87 seats per train.


Note: Only a member of this blog may post a comment.