Abstract
<title>Abstract</title> <p>Shenzhen's urban villages are former rural settlements absorbed by rapid urban expansion. They house many migrant and lower-income residents but raise questions about local service provision, environmental quality, and housing valuation. Seventeen spatial layers are integrated on a 500 m grid over Shenzhen's built-up area. Available official village locations and rural collective residential land identify 318 village cells and 1,214 formal-residential cells. At 4.98 percent prevalence, a held-out building-based indicator yields receiver operating characteristic area 0.911 and precision--recall area 0.222, providing convergent rather than boundary validation. Under spatial-block bootstrap inference, village cells retain lower floor area ratio and height, higher ground coverage, and weaker blue--green provision. In the principal parsimonious model, the buffer-union distance coefficient is 0.070 with HC3 inference and remains positive in maximum-likelihood spatial-lag and spatial-error models. A spatial-lag distance-band model estimates a direct impact corresponding to approximately 12.1 percent lower listing prices within 250 m than beyond 2 km. Theil decompositions retain zero-service cells and show that the village-versus-formal divide explains little inequality in local service provision among residential cells. Adding other built land changes the citywide spatial-structure result, but does not directly measure resident welfare. Urban villages are a distinct urban form, yet low local service provision is a broader neighborhood-wide issue rather than a simple village-versus-formal-city divide.</p>