Developing a residential location choice model to capture the heterogeneity in commute vehicle miles travelled (VMT)

Abstract

Residential location choice plays a prime role regarding all types of travel decisions and has a direct influence on average trip lengths, frequencies and modes of all household members. Total Vehicle Miles Travelled (VMT) increases with the increase in trip lengths and modes as well as with the frequency of trips. Modeling choice of home location is a direct indicator of VMT and an important issue in modeling transport demand. In this research, residential location choice models have been developed to capture the heterogeneity in commute vehicle miles travelled. The models have been specified using detailed survey data collected from faculty members of two public universities (Dhaka University of Engineering and Technology, Gazipur and Shahjalal University of Science and Technology, Sylhet) of Bangladesh. Both of the universities have residential facilities for their faculty members but many of them are not currently using these. It is observed that many faculty members are living in Dhaka (and/or Sylhet in case of SUST) and commuting long distances to go to universities. In some cases, faculty members are living at or, near their workplaces and the rest of their family are staying in the major cities resulting in a ‘split’ family. Therefore, it is necessary to identify the potential variables that attract them to the capital and the major cities and thereby lead to increase in their commute VMT. The survey includes the current Revealed Preference (RP) data regarding choice of residential locations, as well as Stated Preference (SP) data where the faculty members are given some hypothetical future scenarios which include some improved facilities at or near university campus and are asked to choose a location among alternative residential locations. The SP scenarios include multiple levels of five attributes (better school facilities, reduced rent of university residence, spouse’s job opportunity, professional work scope and some additional facilities including better health care, big shopping mall etc.). Data analysis shows that in the presented SP scenarios, 61 percent commute trips are likely to be reduced in case of DUET and 78 percent commute trips are likely to be reduced in case of SUST. The analysis also shows that 68 percent commute VMT are likely to be reduced with a reduction of two-way daily commute VMT of 690 for DUET and 78 percent commute VMT are likely to be reduced with a reduction of two-way daily commute VMT of 891 for SUST in the presented SP scenarios. Discrete choice models have been developed using the SP data and the coefficients of the utility functions have been estimated using the maximum likelihood technique. The observed taste heterogeneity of the respondents has been taken into account by the introduction of socio-economic variables like age, gender, income, car ownership etc. into the model. Survey reveals significant distinction in the choice process of residential location between the two universities and therefore separate models have been developed for them. A Nested Logit Model (NL) and a Multinomial Logit Model (MNL) have been found as the best models for DUET and SUST respectively. Estimation results show that better school facilities with Bengali as well as English medium, reduced house rent, professional work scope and spouse’s job opportunity are the potential variables of choosing on-campus housing facility for DUET faculty members. Faculty members of average age 45 years have less likelihood of choosing on-campus facility whereas female faculty members are more likely to choose on-campus housing facility in DUET. On the other hand, higher standard Bengali medium schools and reduced rent are the most influential variables in the choice of residential location for SUST faculty members. Female faculty members of SUST have higher likelihood of choosing off-campus housing facility which is quite a different scenario from DUET. The faculty members who own car are more likely to choose off-campus housing facility. The unmarried faculty members of SUST have higher likelihood of living split from their family. The estimated equations can be used to predict the probabilities of shifting to on-campus facilities and calculating the corresponding change in VMT in response to a certain policy change. Therefore, the findings of this research work can help transport policy makers and university authorities in formulating policy guidelines to promote on-campus housing. Further, the methodology used in this research work can be used in future researches on residential location choice modeling of other segments of population.