The mixed logit model provides a framework for studying discrete choice that avoids the Independence of Irrelevant Alternatives problem of standard logit, permits coefficients to vary over a population, and allows for the incorporation of unobserved heterogeneity into preferences. Recent computational developments have made it feasible to estimate this model even though its choice probabilities cannot be obtained in closed form, and one can anticipate that it will rapidly replace standard logit as the tool of choice. In this paper the author studies the sensitivity of one set of policy conclusions - the feasibility of private bus transit provision in urban corridors - to the specification of the choice model. He concludes that both models largely agree on the conditions permitting private transit; however, at the detailed management level, profit-maximising route structures and peak head ways differ significantly as between the two models, though fares do not. (Author/publisher).
Abstract