Modelling residential relocation behaviour combining passive revealed preference data and stated preference survey data

Understanding how various factors shape residential relocation is crucial for effective infrastructure planning and policy. Yet, existing revealed preference (RP) datasets often lack essential demographic or dwelling details, while stated preference (SP) surveys are prone to hypothetical bias and behavioural incongruence. To fill in this gap, this study presents a residential relocation choice model that combines residential location data derived from passively generated public transport smart cards of 82,720,872 users and SP data from 971 respondents (8739 observations) in Beijing, China. Both types of data were generated or collected in the backdrop of the COVID-19 pandemic, which led to higher-than-usual residential relocations in Beijing. The integrated approach, which accounts for the scale difference between the two datasets, reveals a strong preference for city-centre locations. But higher infection risks increase the likelihood of moving away from crowded areas, whereas flexible work-from-home policies lower the inclination to relocate to the centre. These findings quantify how different pandemic-related factors alter traditional relocation drivers. The results can guide policymakers in designing more resilient housing and transport policies, especially under future disruptions like pandemics. Moreover, the data-fusion framework offers a replicable strategy for researchers and planners seeking to capture both real-world behaviours and hypothetical scenarios in residential location studies.