Chandra R. Bhat
University of Massachusetts, Amherst, MA 01003

The nested logit model relaxes the "independence of irrelevant alternatives" (IIA) property of the multinomial logit model by grouping alternatives based on their degree of substitution. Alternatives in a nest exhibit an identical degree of increased sensitivity relative to alternatives not in the nest. An assumption maintained in the nested logit is that the degree of sensitivity among nested alternatives is invariant across agents making the choice. This assumption might be untenable in many situations. In this paper, we propose an extension of the nested logit model to allow heterogeneity (across agents) in the covariance among nested alternatives based on observed agent characteristics. We label this model as the COVNL model. The multinomial logit, the nested logit, and the COVNL model are estimated to examine the impact of improved rail service on weekday, business travel in the Toronto-Montreal corridor. The empirical results show that not accounting for covariance heterogeneity in the nested logit formulation leads to a statistically inferior data fit and also to biased model estimates of the effects of level-of-service variables.