Preference stability in household location choice: Using cross-sectional data from three censuses

Household location choice has become an important topic to both researchers and practitioners as more disaggregate transportation and land-use models have been developed. Modeling household location choice is complicated by the many influencing factors involved. This modeling, as far as we know, until recently, has been mostly static in the sense that models have been estimated on data from one point in time. This paper takes advantage of microdata on movers from three Canadian censuses in Montreal to estimate household location choice models in a dynamic context, with the aim of understanding if, and how, the influencing factors are evolving through time. That is, the parameters are assumed to be functions of time by testing different functional forms for expressing parameter changes. Furthermore, spatial autocorrelation across neighborhoods is included in the model. The results show a significant improvement when accounting for spatial autocorrelation. Furthermore, the appropriate distance threshold to incorporate surrounding tracts in spatial autocorrelation calculations is examined and also it is found that there is no significant autocorrelation between neighborhoods located farther than 800 m from each other. Finally, over time, household sensitivity toward a number influencing variables is observed to have changed.