Unravelling intra-urban spatiotemporal disparities in housing price dynamics: An infrastructural approach using interpretable machine-learning methods in Shenzhen, China

Source abstract

Housing inequalities have long been a pressing global issue, particularly in the context of rapidly transforming urban housing markets. However, intra-urban spatiotemporal disparities in housing price dynamics remain insufficiently understood. By developing an infrastructural approach operationalized through interpretable machine-learning methods that explicitly incorporate spatiotemporal heterogeneity, this study examines the complex capitalization of key infrastructures into housing prices, drawing on 288,570 second-hand housing transactions and multi-source geographic data from 2010 to 2023 in Shenzhen, one of the largest and most dynamic urban housing markets in China. The results reveal increasingly intensified housing price disparities, which are particularly pronounced within school catchment areas and 15-minute infrastructure-accessible neighborhoods. Compared with physical infrastructures, as well as structural and locational attributes, social infrastructures—especially public schools—are more decisively and consistently associated with housing prices. Their spatiotemporally heterogenous capitalization and increasingly salient threshold effects highlight the agency of social infrastructures in leveraging urban development and (re)shaping social inequalities, thereby advancing the understanding of housing price disparities. Finally, this study provides dynamic, fine-grained evidence to support locally calibrated and stage-sensitive urban planning and housing policies.