A hazard-based duration model of shopping activity with nonparametric baseline specification and nonparametric control for unobserved heterogeneity

Abstract

Activity duration is an important component of the activity participation behavior of individuals, and therefore, an important determinant of individual travel behavior. In this paper, we examine the factors affecting shopping activity duration during the return home from work and develop a comprehensive methodological framework to estimate a stochastic hazard-based duration model from grouped (interval-level) failure data. The framework accommodates a nonparametric baseline hazard distribution and allows for nonparametric control of unobserved heterogeneity, while incorporating the effects of covariates. The framework also facilitates statistical testing of alternative parametric assumptions on the baseline hazard distribution and on the unobserved heterogeneity distribution. Our empirical results indicate significant effects of unobserved heterogeneity on shopping activity duration of individuals. Further, we find that parametric forms for the baseline hazard and unobserved heterogeneity distributions are inadequate, and are likely to lead to substantial biases in covariate effects and hazard dynamics. The empirical results also provide insights into the determinants of shopping activity duration during the commute trip.