Saturday, January 31, 2009

Selection into transport mode

The Reason Foundation usually gets things right, and their heart is always in the right place, but once in a while they produce a howler like this:
A policy that focuses largely on shifting travelers out of cars and into transit will reduce mobility. An examination of work trip travel times in 276 metropolitan areas found that the length of public transit trips exceeded those for private automobiles in 272 of those areas. On average, public transit riders spend about 36 minutes traveling to work while private automobile travelers commute about 21 minutes. This does not have to be the case. The innovative use of HOT Lanes, such as the networks being built in Northern Virginia and discussed in Atlanta, Houston, the San Francisco Bay Area, and Miami can finance critically needed road capacity while also providing viable bus rapid transit alternatives.

The reasoning in this paragraph, which comes from this testimony here (about which I learned from a Reason Foundation email updating me on their doings) seems to assume that individuals choose between cras and transit at random, so that the mean difference in travel times represents a causal effect, of an odd sort, of taking transit. An alternative model would suggest that higher income individuals, who are likely to have higher values of time, drive, while lower income individuals, who tend to have lower values of time, take transit. In that case, the difference is partly or even entirely selection.

On a more rarified note, one might also consider the variance of travel times, which is likely to be lower for transit in some contexts. When I was at Maryland, I quickly learned that one accident could turn a 30 minute beltway journal into a 90 minute one; as a result I took back rounds from Chevy Chase to College Park. This raised the mean a bit but lowered the variance. When you need to get to class at a specific time, variance matters.