I’m on the editorial board of Transportation Research Part A, and it’s an excellent journal by any measure. But one article this morning seemed so promising, and then rather failed to deliver:

Graham-Rowe, E., Skippon, S., Gardner, B. & Abraham, C., 2011, Can we reduce car use and, if so, how? A review of available evidence, Transportation Research Part A: Policy and Practice.

Great, right? Another review, and we probably needed another review after this inappropriately optimistic one appeared in JAPA last year:

Ewing, R. & Cervero, R., 2010, Travel and the built environment, Journal of the American Planning Association, 76(3).

The latter review was problematic because it summarized the evidence and then concluded “Yes, well, all the empirical evidence shows small effect and insignificant effect sizes, but we still think our interventions work under the right conditions.”

There comes a point where you have to wonder if those right conditions are feasible if the research can’t find them time and time again.

However, Graham-Rowe sort studies according to quality, stating what’s obvious to everybody: there aren’t enough randomized trials in applied social science research.

Gee, ya think?

There’s a reason why the high quality studies are looking at program evaluations and why the cross-sectional studies look at before and after projects. Unlike medical and psychological research, researchers in my world don’t get to randomly select controls for anything other than programs, and often not even then because there are practical problems with employers or city governments allowing some employees or residents–but not others–to participate in a program that carries a benefit, like being paid not to drive.

So undeniably, we’d have better research if I could select random samples and controls for selected interventions, use our godlike hands to pick drivers up by their heads, place them in case-control groups according to intervention versus non-intervention environments or programs, and make them live there/participate as long as we wanted them to. Unfortunately, doing that sort of thing in societies where human beings have freedom of movement and self determination tends to be frowned on.

The takeaway–AGAIN–is that self-selection and endogeniety go hand and hand. Gargh.

I don’t see a path out of this cycle of research-critique. We’ve hit a stalemate. People who are advocates of particular position–that mixed land uses and transit supply reduce auto use–are like Fox Mulder: “I want to believe.”

Social scientists can try to tinker on the margins of what we have, with instrumental variables and various econometric contraptions strapped on to different datasets, but there’s no way around the residential self-selection problems here.

We can publish critique after critique, and perhaps that’s useful, but I don’t see how. We know where we are with this research–and we also know that planning, policy, and forecasts are thundering ahead with the “I want to believe” attitude. The alternatives to believing aren’t particularly attractive, either.