Highly Disaggregated Land Unavailability

Standard empirical proxies contradict the canonical prediction that supply constraints amplify house price cycles. We resolve this puzzle by constructing the Land Unavailability--Machine Learning (LU-ML) Indices using high-resolution satellite imagery. Unlike existing measures, our indices capture the nonlinear and heterogeneous impacts of physical geography. We document two main facts: (1) physical constraints are a key determinant of cross-sectional price dynamics, driven largely by the intensive margin (steep slopes); and (2) looser supply constraints significantly mitigate the price effects of demand growth, overturning the "null result" in recent literature.