A1669
Title: Synthetic control method with block kriging for point-referenced data
Authors: Naoki Hiramoto - The Graduate University for Advanced Studies (SOKENDAI) (Japan) [presenting]
Daisuke Murakami - The Institute of Statistical Mathematics (Japan)
Abstract: A change-of-support-aware synthetic control method (SCM) for settings where outcomes are observed at spatially irregular, time-varying point locations is developed. Standard SCM requires a fixed spatial panel, which is unavailable when data such as housing transactions are collected at locations varying across periods. Moreover, coarse spatial aggregation can mask heterogeneous treatment effects that differ in sign across subregions. The approach combines block kriging (BK) with SCM. BK aggregates point-level outcomes into areal means for user-specified spatial units, producing a panel compatible with SCM estimation. By explicitly modeling the latent spatial process, BK corrects for imbalanced sampling and provides kriging standard errors for uncertainty quantification. Areal averages satisfy exact linear identities across nested regions, enabling algebraic reconstruction of subregion effects from a limited number of SCM fits. Monte Carlo experiments confirm that the method recovers spatially heterogeneous and sign-reversing effects hidden under coarse aggregation. An application to the Toei Oedo Line opening in Tokyo reveals positive effects near western inner-city stations and negative effects in eastern and waterfront areas, consistent with spatial reallocation of residential demand rather than a uniform metropolitan-wide price increase.
