COVARIANCE HETEROGENEITY IN NESTED LOGIT-MODELS - ECONOMETRIC STRUCTURE AND APPLICATION TO INTERCITY TRAVEL

The nested legit model relaxes the 'independence of irrelevant alterna tives' (IIA) property of the multinomial logit model by grouping alter natives based on their degree of substitution. Alternatives in a nest exhibit an identical degree of increased sensitivity relative to alter natives not in the nest. An assumption maintained in the nested legit is that the degree of sensitivity among nested alternatives is invaria nt across agents making the choice. This assumption might be untenable in many situations. In this;paper, we propose an extension of the nes ted legit model to allow heterogeneity (across agents) in the covarian ce among nested alternatives based on observed agent characteristics. We label this model as the COVNL model. The multinomial legit, the nes ted legit, and the COVNL model are estimated to examine the impact of improved rail service on weekday, business travel in the Toronto-Montr eal corridor. The empirical results show that not accounting for covar iance heterogeneity in the nested legit formulation leads to a statist ically inferior data fit and also to biased model estimates of the eff ects of level-of-service variables. Copyright (C) 1996 Elsevier Scienc e Ltd