David Levinson, the Transportationist, responds to Yonah Freemark’s commentary on David’s breakdown of California High Speed Rail Authority’s ridership forecasts. Freemark suggests that we can’t really argue that the existing ridership forecasts are “unreasonable.”

We discussed this a bit in Richard Green’s class today. Here’s the deal:

1) It is pretty well-established that ridership forecasts suffer from optimism bias. With simple forecasting error, when we plot actual versus projected ridership, we should just see white noise. Some systems come out of early slumps in ridership to hit and exceed targets. These forecasts are really tricky because you have to think about about demand over time. Demand during everybody’s favorite topic–the bubble–and demand during everybody’s second favorite topic–the recession–are now two different contexts, both of which have a potentially big effect on ridership in general. And–and this is a big and–the post-bubble world for changing land use and development patterns to boost ridership is REALLY different post-bubble.

Those are information problems that affect just about everybody in predicting markets for new services or products.

2) Then there are the endogeneity problems, and Richard and I talked at some length about these with the PhD students. With forecast new services, we can chose to ask people whether they’d like the new service (something that the Cambridge folks did when derived the HSR forecasts) or you can try to observe what people are doing. Asking people is tough: it certainly can be through contingent valuation and carefully put together stated preference surveys. But they are problematic: without any actual experience withe service or real price information in terms of time and money, it’s hard for people to really know what they would do.

Revealed preferences can be a rum-go, too. So you go out, and look at the ridership in places where there exists a comparable service to the one you’d like to forecast. Then you go and you apply what you’ve learned there to another context. And you just don’t know: maybe people in the first context chose to live near transit despite probably higher land costs–if theory holds, it’s more expensive to live there, or we’re providing service nobody wants–and people in the second context are living where they are living because they just don’t care whether they are served by transit and it’s just not worth it to them to pay for it.

So of course the ridership in the second context isn’t going to be what it was in the first. Now over time, planners promise to change the land uses in the second context so that it becomes more like the first. Fine, dandy, but we never really know how many of those first people simply move somewhere cheaper or whether we make transit users out of them–or realistically, enough of them to justify the money we are spending.

3) I argued that I think among a certain class of people (notably young urban planners fanatical about transit) that residential self-selection happens at the regional level as well. So the people who have a great deal of job mobility and a strong strong strong preference for urbanity and transit self-select at the regional level as well.

For people like me, who aren’t perhaps as job mobile and who don’t know how to drive, the residential self-selection problem within regions is, in fact, interesting. I’ll pay more rent because I HAVE to have access to transit and taxis and food delivery when Andy travels or I will starve to death in my apartment. So I don’t even consider locations not served by transit. But for the very job-mobile–dare I say it–“creative class” types–I suspect that there is endogeneity with predicting the selection of *regions* as well as neighborhoods, with a hierarchy of cities *and* neighborhoods within them.

The problem with doing forecasts poorly isn’t just an optimum investment problem. It’s a bond rating problem if it gets too bad, and it’s good-faith problem when dealing with public-private partnerships. So it’s not solely grist for the “politics surrounding transit” mill.