Alon Levy has a nice, question-posing article up at Urbanize.LA where he discusses the operating costs reported by major rail operators across the US and the EU. He has a legitimate question: why are operating costs in LA so high? We can partially attribute the higher costs to unionized labor, which shows up generally in cost-per-hour figures (See BART). Levy reports that the per-mile or per-kilometer figures, but it’s not clear exactly what the “per-hour” numbers are: I’m guessing this is vehicle-hours because that’s a good way to measure for cost allocation how much it costs to put service on the road.

This super-duper for me because we just went over cost allocation in my transit class, and I can show them this article to prove they are learning the cutting-edge, y’all!

The cost-per-vehicle-mile, for me, is most relevant internally to the agency itself as a managerial metric. You want to see where your money is going in terms of running your fleet. For buses, a manager might ask if we are spending a lot on tires per kilometer? If you are, you might be operating on crap roads (a big problem for LA Metro) or you might want to examine your supplier there, or maybe there needs to be a change in the maintenance practices or scheduling. Of course the public might want to scrutinize these figures, but it’s not really an answer to value-for-money questions that the public usually has. The NTD reports them because agencies look at how their peers are doing and ask questions from there.

I think a better measure for audiences external to the agencies is cost per passenger mile. The costs per vehicle mile you incur, in general, because you are operating vehicles across a distance, and you have to pay your operators, maintenance, etc regardless of whether you have any butts in seats or not. It’s a measure of cost-effectiveness; cost per passenger mile is a measure of system work or productivity. One of my big concerns for LA Metro is that the last time I ran estimates, they were having to spend more for each additional passenger-mile they were attracting (I will do that again and post some numbers here) and that’s not good. It may be attributable to their large capital improvements program, but bad capital investments can just put agencies on the hook for operating unproductive lines and large amounts of debt service. Any and all new rail supply is not always good news. (But it can be really hard to know in advance what is a good investment: nobody seemed all that excited about the Expo Line in the beginning of the planning stages, and now it’s everybody’s darling. As it should be. I love you, little train that takes me to work!)

Back to Alon’s numbers. This is a pretty difficult exercise that he’s undertaken here, simply because I don’t think you can really trust that the numbers have been reported to the sources in a consistent way. Budget documents vary quite a bit in how they are presented, and cost allocation practices internally vary as well with agency context.

I am suspicious of the European numbers because they are so much lower and relatively uniform. It may be a denominator question; perhaps they really are offering way more miles of service than US agencies are, but I don’t think so. I think it may have to do with how internal accounting practices allocate administrative costs in budgets and whether European agencies only report costs that vary by hour or whether they allocate out salaried employees. If they do not put salaried employees into the figure, that’s a pretty big difference because US agencies often do so. Accounting, even though accountants are as fiercely scientifical as economists, can be pretty subjective. Think about it this way: where do you allocate the costs for your route planning staff? Do you leave them out because they don’t actually wear and tear with vehicle service, but you do need them to provide service? Where should we charge Phil Washington’s salary? Where do charge the service contract out to the LA Sheriff’s office for security?

The cost allocation model I present to my students follows from Eric Bruun:

This is a pretty simple cost allocation model, but it does the job. The point is that there are things that vary with distance, things that vary with hours, things that vary with both, and things that vary because you have to acquire and keep rolling stock (and a whole bunch of things, like landscaping and yard maintenance, security, etc etc). And your asset commitment.

You could, by way of policy, assign percentages of Phil Washington’s salary across all these categories because he’s responsible for the whole ship, but does that tell us about whether we are spending more on landscaping than we want?

By way of my guess, I think the commenters who have attributed LA’s larger numbers to frequencies are probably right. The lower service frequency, the smaller the denominator in calculating the unit coefficient, the bigger that coefficient will be. But I would need to grub around more with the data to be sure.