Integrated modeling of residential and work-related mobilities: A large-scale dynamic Bayesian network approach

Residence and work-related choices are commonly assumed interdependent and associated with some other individuals’ and households’ life events as well as people’s socio-demographics. This paper proposes an integrated dynamic model to capture the determinants of residential and work-related mobilities, taking into account their associations with other life events. The change of employment is modeled with its multi-dimensional features, i.e., change of working hours and change of work place location, while residential mobility is modeled through the change of residential location as well as with whom within the household the move out take place. The Dutch micro-population data from 2015 to 2019 provided by the Netherlands CBS (Census Bureau of Statistics) are used to train and validate a dynamic Bayesian network (DBN) model. The integrated model has a high prediction precision with an average of 89.77% across life events. The results suggest that other life events, such as getting married or divorced, significantly affect the relocation of residence, while changes in working hours strongly influence the work place relocation. However, the relocations of residence and work place in the Netherlands appeared not to be directly intertwined.