Exploring Variation Properties of Time Use Behavior on the Basis of a Multilevel Multiple Discrete–Continuous Extreme Value Model

Abstract

This study attempts to examine the variation properties of time use behavior on the basis of a multilevel multiple discrete–continuous extreme value model, which describes both activity participation and time allocation behavior by incorporating various variance components. Five major variation components are dealt with, including interindividual, interhousehold, temporal, spatial, and intraindividual variations. The MobiDrive data, continuous 6-week daily travel data, allow identification of these variations simultaneously. Two types of models are empirically examined: the model without consideration of the influence of explanatory variables (null model) and the model that introduces explanatory variables (full model). Estimation results of the null model confirm that intraindividual variation still accounts for more than 50% of the total variation (except for mandatory activities) even after incorporation of the aforementioned four other types of variations jointly. In contrast, the results from the full model reveal that most types of unobserved variations (especially the intraindividual variation) still dominate in the total variation even after the relevant observed information is introduced. These findings provide useful insights into both model development and data collection methods as well as to the understanding of the mechanisms of time use decisions.