TITLE：Unravelling the dual displacement risks of urban renewal: A causal inference framework integrating double machine learning and XAI
ABSTRACT：With urban renewal advancing globally, intensified gentrification and rising living costs are exacerbating displacement risk. However, existing research has rarely examined how renewal-induced cost pressures vary across fine spatial units, where their spatial spillover boundaries emerge, and how housing and consumption costs jointly shape displacement risk. We construct a novel analytical framework integrating Double Machine Learning (DML) for causal inference, sliding-window threshold detection for boundary analysis, and eXplainable Artificial Intelligence (XAI). Our analysis in Shenzhen reveals that urban renewal significantly increases displacement risks by raising the relative market standing of housing rents and catering consumption costs. Specifically, the model estimates that renewal causally elevates the inflation-adjusted relative market standing of housing rents by 5.46% and catering consumption costs by 9.69% within the city-wide distribution. Crucially, the spatial analysis uncovers asymmetric patterns: while rental increases exhibit a clear spillover range of approximately 1103.6 m, no significant spatial boundary was detected for consumption cost changes. Further examination identifies three displacement-risk pathways shaped by renewal's differentiated impacts, highlighting how functional replacement and the relative dominance of the “housing-cost crowding-out effect” versus the “consumption-demand substitution effect” shape displacement risk. These findings not only deepen the understanding of the fundamental processes driving direct and indirect displacement but also bridge a critical gap in conventional theory regarding spatial responsiveness to renewal, providing policymakers with spatiallygrounded quantitative evidence to advance the complementary objectives of social equity and inclusive development.
Keywords: Urban renewal;Displacement risk;Gentrification; Double machine learning ;XAI; Shenzhen
