A new model for residential location choice using residential trajectory data

Traditional residential location choice (RLC) models are based on the characteristics of location and demographics, revealing important patterns of RLC, but no RLC models have yet incorporated individual preferences. This study fills this gap by integrating the pattern of home-based travel into the RLC model. Firstly, by analysing residential trajectory data collected from Beijing and Shenzhen, we find that both residents' commuting time, that is, time spent commuting to work, and home-based non-commuting (HBNC) time, that is, time spent on the consumption of amenities when departing from homes, follow an extreme value distribution (EVT). This indicates that, based on time budget and financial constraints, residents strive to minimise commuting time and maximise HBNC time. Subsequently, by integrating these findings into individual-level RLC analysis, we obtain an RLC model that aligns with the gravity model. Throughout the model training process, we demonstrate that the RLC model exhibits strong robustness by incorporating control variables, changing the spatial scale of the observation unit, testing for endogeneity, and considering historical RLC. Moreover, the model performs well in applications including assessing dynamic changes in RLC behaviours and making predictions based on previous travel behaviours. The RLC model in this study advances our understanding of human habitat selection behaviour and can be utilised by policymakers to develop and implement effective urban planning and epidemic management policies.