Affiliation:
1. Key Laboratory of Road and Traffic Engineering of Ministry of Education, College of Transportation Engineering, Tongji University, Shanghai, China
Abstract
This paper further explores multiple discrete-continuous extreme value (MDCEV) models. MDCEV models are widely used for time-use analysis because they can accommodate the diminishing marginal utility with respect to time consumption and benefit from simple closed-form likelihood expressions for estimation, while considering the overall constraints of time resource. However, all of the existing models adopt a monotonically increasing form of utility function, which does not accommodate possible non-monotonic utilities with respect to time-use. This paper further develops the MDCEV model by firstly introducing utility functions with a form of parabola to capture non-monotonic preferences in time allocation behaviors. A corresponding simple and computationally efficient forecasting algorithm to predict the time-use of individuals is also developed. The proposed MDCEV model with parabolic utilities is applied to analyze time-use decisions among a variety of outdoor non-mandatory activities on weekends using data from the 2019 American Time Use Survey. The proposed model is compared with traditional MDCEV models to show its ability to describe non-monotonic preferences in time-use in association with various explanatory variables. This paper makes an initial attempt to specify non-monotonic utilities for time-use analysis, but further studies are required to develop more sophisticated non-monotonic utility formulation to improve the goodness-of-fit of the model.