Monday, March 11, 2013

Long distance trains and direct costs

In CEO Boardman's recent testimony to Congress, there is a rather interesting graph showing the contribution or loss from the long distance trains. According to this, prior to the allocation of overhead, the Palmetto and Silver Meteor both return a small surplus and the Auto Train breaks even.

Backtracking from revenues and expenses via eyeballing the graph and adjusting from 2012 revenues, rounding to the nearest dollar:
Palmetto is implied at a direct cost of ~$23 per train mile.
Silver Meteor is implied $41
Auto Train at $112
Lake Shore Limited is $51
City of New Orleans $40
Capitol Limited $48
Silver Star $43
Cardinal $40
Crescent $42
Empire Builder $44
Texas Eagle $30
Coast Starlight $62
Sunset Limited $57
Southwest Chief $45
California Zephyr $47

At first glance, there appears to be a simple explanation for the various costs. The Palmetto is the only long distance service without sleeper cars or diner service and costs are appropriately lower with significantly fewer crew, cars, or related expenses; the Auto Train carries a significant number of freight cars; the Texas Eagle sends through cars on the Sunset Limited and may simply be miscalculated as a result; and the Coast Starlight carries the unique, and maintenance intensive, Pacific Parlour Car.

However, I do not entirely trust these numbers. The Capitol Corridor reports a FY2012-2013 standard of $45.59 in train expenses per train mile (Appendix C) and the Pacific Surfliner had a 2007-2008 standard of $47.74 per train mile (page 21). The San Joaquin was somewhat lower with a 2008-2009 standard of $34.82 (page 32), but still higher than the Palmetto or Texas Eagle. Out east, the Keystone Corridor reported 2007-2008 expenses of $34.59 per train mile, rising to $37.01 in 2008-2009 (page 127). The Keystone report does an additional service in showing the number of car miles (as does an older report, which also shows the state subsidy per car mile, on page 54); with an expense of $7.40 per car mile in 2008-2009.

One possible explanation is that the state supported corridors, unlike the long distance trains, are burdened by the costs of track access fees while the long distance trains pay merely an incidental expense per train-mile. However, the Keystone Corridor is not subjected to such costs and of the three California trains, the Surfliner is the one that I would expect to have the lowest costs as a result of such a discrepancy, rather than the highest, since three of its frequencies were considered part of the national train system and thus potentially liable to the same incidental rates.

The information does provide for a bit of analysis on the ability of sleepers to pay for themselves however. If the numbers do hold true, and a train with six coaches and a baggage car is approximately $20 per train-mile cheaper to run than a typical overnight long distance train with three coaches, three sleepers, and a diner, the imputed cost of running long distance sleeper services is about $10.50 per car mile. With 45 "seats" in a Superliner II sleeper car, this comes out to $0.233 per seat-mile; since one of the roomettes is taken up by the train attendant, this actually rises to $0.244 per seat-mile. For a Viewliner, which can only accommodate 30 passengers (28 if a roomette is taken up by an attendant, but I do not know if this is the practice on single level trains), this rises to $0.35/0.375 per seat mile.

Unfortunately, Amtrak does not seem to be pricing their sleepers in such a fashion as to recoup the costs of operating them. Using information from the Performance Improvement Plan for the top five performing long distance trains, as well as the September 2011 monthly performance report, I put together this brief spreadsheet.

# of Sleeper Passengers
Average sleeper trip length
Sleeper Revenue
Revenue per sleeper passenger
Revenue per sleeper passenger-mile
City of New Orleans
Southwest Chief
Auto Train
Coast Starlight
Empire Builder

The City of New Orleans (the sole single level train), Southwest Chief, and Empire Builder do not even attempt to sell sleeper service at a price which could recoup the cost of providing it and while the Starlight and Auto Train are above the putative costs per seat-mile, it is doubtful that they match expenses after occupancy factors are accounted for. Additionally, these two trains have costs significantly in excess of the others. It is one thing to simply fail to attract sufficient patronage as to recoup your costs; it is quite another to deliberately price your trips in such a fashion that it is impossible to recoup costs.

I am aware that the National Association of Railroad Passengers has previously claimed that sleepers provide an incremental profit. Unfortunately, I do not possess the in depth access to Amtrak financials which NARP appears to be blessed with. However, with sleeper tickets priced at only a 1.6 multiple of coach tickets, per passenger mile, yet with the unique exception of the Auto Train, coach providing 2.4-4.1 times the number of passenger miles, I am strongly doubtful of the validity of NARP's analysis. Now, at this point, a caveat may be made. Because the cost of a sleeper is charged in addition to a base rail fare (the lowest bucket coach fare), it is possible that Amtrak, for reasons beyond the ken of mortal man, does not include the cost of that base rail fare in the sleeper revenue, which would make the current sleeper performance look better. Some quick trip pricing on Amtrak's website indicates that this is not the case however. The Auto Train, it must be noted, runs an abnormally high cost per passenger mile due to the necessity of paying for hauling a vehicle; it appears that coach passengers may be disproportionately likely to be traveling alone.

# of Coach Passengers
Average coach passenger trip length
Coach Revenue
Revenue per coach passenger
Revenue per coach passenger-mile
City of New Orleans
Southwest Chief
Auto Train
Coast Starlight
Empire Builder

Interestingly, assuming that the direct costs are indeed accurate, sleeper cars are, in a very real sense, the bane of long distance train travel. For the cost and equipment involved in running a single average long distance train, two Palmetto-style all coach trains could be run. I think it doubtful that any losses in revenue and passengers would not be made up, and indeed quite exceeded, by a second, opposed frequency, with the result of reducing much of the burden which the overhead imposes on the current long distance trains. Of course, this would likely run into issues with the Class Is demanding track access fees, however, Amtrak may be able to negotiate with the various states for them to pick up such extra costs in order to run the trains.

With advocacy organizations such as RailPAC continually calling for lengthening long distance trains and pointing to sold out sleepers as evidence in their favor, they would do well to also call on Amtrak to price sleepers appropriately. Sufficient demand for an additional car is also sufficient demand for a higher priced ticket and, as the long distance trains are insanely capital intensive, a luxury service such as sleeper cars ought not to be expanded, nor called for expansion, until it shows that it pays for itself at least on the margins.


  1. I'm a little confused. City of New Orleans doesn't come close to breaking even, since it's only making double level car revenue on single car stock, but aren't Southwest Chief and Empire Builder coming really close to breaking even at only 1 or 2 cents per passenger mile deficit?

    That looks like it adds up to about a million dollar deficit per year on the Empire Builder, which is well within the margin of error on your calculations, right?

    I find passenger car revenue on the EB of particualar interest, since I went from St Paul to Portland in a sleeper car for my honeymoon, and I was hoping not to cost the taxpayer too much.

    1. Southwest Chief and Empire Builder have direct cost losses of ~27 million and ~11 million each. The deficit is exacerbated by occupancy issues and the length of the route. A few cents per mile, with an average mileage of 1300-1400 miles, adds up to quite a bit. These don't account for the entire loss, but they do impose a significant burden.

  2. Paul, sleeper reservations currently have to be booked a year in advance. This suggests significant unmet demand relative to supply, which (since Amtrak is effectively a monopoly) means that Amtrak certainly can raise prices quite a bit. Hm.

    1. It's worth looking at the price difference with VIA Rail. The Canadian, from Vancouver to Toronto, is 2,775 miles, making it a fair match for the Southwest Chief or Empire Builder. The economy coach fare for a whimsically chosen departure of July 2nd is $552 pre-tax, $0.199 per passenger mile. With an upper berth that rises to $1,364.00 ($0.492), and a cabin for 2, which might substitute for Amtrak's bedroom, is $3,086.00 ($1.112). Admittedly Canadian dollars, but they're pretty much at parity lately. On the same day, a train from Los Angeles to Chicago costs a maximum of $1434 for a single person traveling in a family bedroom (no rooms sold out yet on this trip) at a price of only $0.633 (dropping to $1051/0.464 with a two person bedroom at single occupancy). The price differential is even larger if you include the taxes which must be paid on Canadian rail fare.

      While VIA Rail certainly has problems, a general unwillingness to patronize their sleepers is not one which I've heard of (though I certainly welcome any corrections on that).

    2. VIA Rail has had sufficiently low patronage on its sleepers that it keeps running its trains on fewer and fewer days per week. They also have a big pile of sidelined cars. What does THAT tell you?

      Anyway, I've been told before that there's a legal limit on the maximum difference between "top price" and "bottom price" which Amtrak can charge. No such thing for VIA.

    3. "Anyway, I've been told before that there's a legal limit on the maximum difference between "top price" and "bottom price" which Amtrak can charge. No such thing for VIA."

      Well then, that's something that needs to change.

  3. Sleepers, at the margin, pay for themselves. I have seen, but can't put my hands on the reference right now, an analysis of adding an additional sleeper to the Capitol Limited for the six months there's demand to support it. Incremental revenue would be $535K; incremental costs -- attendant, additional food service worker for the peak 90 days, routine maintenance on an additional car, additional fuel to drag the extra car, etc. -- would be $425K. Notice that the incremental gain, in the context of long distance operations, is minimal.

    The big difference in train-mile costs for the Palmetto is surely from its lack of a diner. Diners cost. They require lots of on-board staff and they also require a whole commissary infrastructure to deliver food and beverages to them in the middle of their run. Mica made fun of the $16 hamburger, but the big losses on food service come from the long distance trains and are due to their commissary structure.

    One should also note that the Palmetto is a fast train. It's the only long distance train that can cover more than the requisite 750 miles in 15 hours. Staff costs dominate Amtrak costs and staff are paid per hour, while you're measuring costs per mile (and revenue comes from mileage). There aren't many long distance routes where an all coach day train is even feasible. The City of New Orleans, pre-Amtrak, was one, but I think the route has changed by now.

  4. Paul,
    I'll be brief, but I see two errors you've made:

    (1) The easy one is that the CONO is bilevel. It's a Superliner train, and though there's been talk of switching it back to single-level, it's still bilevel as of now.

    (2) I'd ask if you accounted for the fact that the numbers you're using for fare analysis are about 2-3 fare hikes out of date when checking fares on Amtrak (or Amsnag). At the very least, the average sleeper fare on the Silvers ($275.25 for the Meteor and $245.16 on the Star) ring true to what I'm used to handing over for my accommodation charges (which can crack $500 in a roomette or $1100 in a bedroom to go to Florida from DC) and feels too low for the two charges together (especially as coach is going to pick up a /lot/ of short-hop, often discounted intrastate business in FL). Yes, I know you get a lot of intermediate business, but especially with the Meteor, most of the top markets are NEC-Florida (the average sleeper distances are 888 miles on the Star and 944 on the Meteor...which in the case of the Meteor is longer than WAS-ORL).

    So I do think that, to quote your line, Amtrak may be acting beyond the ken of mortal men with their accounting practices...but as someone who's spent the last few years following arguments over cost allocations and whatnot, this is nothing new to me.

  5. BZZZT. You made a major error in calculation. (Jim pointed this out too.) You assumed that removing the sleepers would allow the removal of the diner, and so you attributed the diner expenses to the sleeper.

    Unfortunately, the fact is you need to provide a diner on any train of sufficiently long runtime, whether it has sleepers or not. You lose a very large amount of patronage otherwise.

    It's very hard to cover diner costs from food prices, with 5 employees, and an entire car with no revenue seating. On the ground, diners pay their employees very low wages; but railroad diner employees have to be qualified on various railroad things and are paid quite well. On the ground, restaurants pay rent or property tax; on the railroad, there's fuel and rolling stock maintenance as well. And maintenance should be absurdly high on the Heritage diners.

    The alternatives to having a diner are (1) meal stops, which slow down the train enough to hurt ridership and revenue, and (2) run the trains faster so that they don't run through multiple mealtimes. (2) is better. The Palmetto is barely fast enough to do without a diner, and really manages it only because most passengers aren't travelling end-to-end.

    If the Twilight Shoreliner was still running with a sleeper, it would become a lot more clear to you that sleepers are profit, while diners aren't.

    1. FWIW, it is quite imaginable that some of the "long-distance" trains east of Chicago could be sped up enough to not require diners.

      The CONO wouldn't require a diner for people travelling from Chicago to Memphis, really. Chicago to New Orleans is just about two hours too long to comfortably go without a diner (and the train's not quite on the right schedule, either). You could probably knock an hour off the schedule just with improvements within the city limits of Chicago and New Orleans. So you see that if the trains can be made faster, in ways which *aren't that expensive*, the diners can be dispensed with.

  6. So, your conclusion is totally, utterly wrong due to your gross oversight.

    *DINER CARS* are, in a very real sense, the bane of long-distance train travel.

    Sleepers are gravy, while diners are loss leaders. You can't cover the costs of a diner without *huge* volume, which is not reached even by filling up one diner. I'm not sure how many tables a diner can actually handle (though certainly one diner can handle two cars full of tables, since it happens on the Auto Train).

  7. Hi, love the blog. But I do have a question. Amtrak fare prices are already on parity with most plane fares and require a significant time investment to take the trip. As a rational consumer, it makes ZERO sense for me to take a train when I could fly for faster and parity if not cheaper. It costs 160 around both ways for me to take my Amtrak to Chicago and back. It's 150 to fly and takes an EIGHTH of the time. I love passenger trains, but I simply can't get around this. How then can you justify RAISING the cost of the fare to make the train pay for itself? Wouldn't it make more sense to decrease train operating costs by either reducing services such as diners and sleepers and/or reducing Labor costs through renegotiation of contracts to be able to offer service that is less than flying? Forgive me but until Amtrak is less expensive it probably won't be increasing Ridership anytime soon.

  8. What's the hedonic value of taking a train over a plane? It must be positive, or else time would indeed be the sole variable.

    I'm all for raising fares, especially for sleepers and luxury services. Those could be profitably operated, I'm sure of it.

    As far as labor renegotiation- it really ought to happen.