Exploring Behavioral Heterogeneities of Elementary School Students’ Commute Mode Choices Through the Urban Travel Big Data of Beijing, China

Abstract

Students' commute mode choices have been recognized as an important factor affecting the physical and psychological health levels of children and urban traffic performance in peak hours. The influential patterns between most of the factors and students' commute mode vary depending on the characteristics of the city. This paper seeks to reveal such patterns specifically for elementary schools students in Beijing, China, as those students' commute behaviors have attracted considerable attention from society. The data from the Beijing School Commute Survey conducted in December 2014 and January 2015 were adopted. To account for the unobserved heterogeneity, a finite mixture multinomial logit (FMMNL) model was developed. Compared with the conventional MNL model, the FMMNL is superior due to the smaller AIC and BIC values. More importantly, the FMMNL model is flexible and able to detect some complicated mode choice behaviors. For example, the results of the FMMNL model indicate that there are two types of students, those who tend to use a car and those who tend to use a bicycle, as their grade increases. Such a heterogeneous pattern is difficult to be detected by conventional models. The finer results produced by the FMMNL model would be the references for policymakers to design more targeted policies. Findings in this paper could be the references to other cities in China and the world.