Co-dependent workplace, residence and commuting mode choice: Results of a multi-dimensional mixed logit model with panel effects

Abstract

An effective instrument of urban planning to alleviate congestion is to create a job-housing balance, which depends on the co-localisation of housing and labour. This study is based on the contention that in response to dramatic change in their decision individuals and households maximize the combined utility of housing, job and commuting trips as opposed to maximizing their utility of one of these domains, treating the other dimensions as given, as many models assume. To support this contention, a stated choice experiment was designed to mimic the multidimensional choice behaviour of interest. Instead of creating an experimental design that is the same for all respondents, we generated an efficient design in which attribute values were pivoted using the real-world data of individual respondents. Based on data collected in Shenyang, China, a Mixed Multinomial Logit (MMNL) model with panel effects which allows for unobserved heterogeneity in individual preferences, was estimated to capture the effects of different residential, job and commuting attributes on multidimensional choice, accounting for the panel nature of the data. Our findings indicate that, 1) Housing tenure, size, price, distance to the bus stop, and housing location are important housing characteristics that help explaining the residential mobility choice process. Salary, job type, co-worker relationships and job environment are significant factors in the job mobility choice process. 2) Time-related factors influence commute mode choice. Choice of public transportation modes is sensitive to commuting costs while car mode choice is not. 3) People are relatively satisfied with their current situation and do not frequently make changes. Furthermore, people are less inclined to move house relative to changing job. 4) Both unobserved heterogeneity and demographic characteristics affect the multiple dimensions of choice.