Should single-equation dynamic gasoline demand models include moving average terms?

We argue that single-equation dynamic demand models applied to estimating gasoline demand should capture the slow evolution of unobservable habits that in part determine vehicle and gasoline usage. Inclusion of unobservable habits implies that single-equation models should include moving-average terms. Ordinary least squares estimation is thus inappropriate. Using examples from Australia and the US, it is shown that estimates of long-run price and income elasticities become much less precise.