ROBUST ANALYSIS OF DISCRETE CHOICE IN TRANSPORT WITH AN APPLICATION TO ALBERTA CYCLISTS

In this paper we apply recently-developed nonparametric conditional kernel density estimation to model discrete choice in a transportation context. Our empirical application, for the purpose of demonstrating the technique in a transport context, is to the route choices of cyclists in the Canadian province of Alberta. The empirical analysis employs the nonparametric multivariate kernel with mixed data types (continuous and discrete). This approach permits the inclusion of continuous and discrete variables where the rate of convergence only depends on the number of continuous variables; it allows for interactions between the covariates, which may be important determinants in discrete transport decisions; it allows the estimation of marginal effects without assuming a constant response regardless of the level of the explanatory variable as it would be in a linear model; and the non-parametric estimation does not impose assumptions in the underlying economic behavior such as IIA. Using a nonparametric index-free approach, the data are permitted to model the full range of relationships among variables. In our simple application of cyclist route decisions, the kernel-density-based estimator correctly classifies 76.24 % of decisions in contrast to a correct classification rate of 61.28 % for the probit model and 61.44 % for the logi