Multilevel Modelling with Spatial Interaction Effects with Application to an Emerging Land Market in Beijing, China

• DOI: 10.1371/journal.pone.0130761
• Publication Date: 2015-06-18T18:16:10Z
• Authors: Guanpeng Dong; Richard Harris; Kelvyn Jones; Jianhui Yu
• Keywords: Economics and Econometrics; China; 330; Science; Social Sciences; Multilevel model; 01 natural sciences; Spatial Dynamics; Spatial Econometrics and Spatial Data Analysis; Spatial Structure; FOS: Mathematics; Computer Simulation; 0101 mathematics; Spatial Analysis; Global and Planetary Change; Models, Statistical; Global Analysis of Ecosystem Services and Land Use; Spatial Economics and Agglomeration Theory; Spatial Econometrics; Geography; Economic geography; Q; Statistics; R; Commerce; Economics, Econometrics and Finance; Models, Economic; Spatial Dependence; Archaeology; Spatial Modeling; Beijing; Environmental Science; Physical Sciences; Geographic Information Systems; Multilevel Analysis; Medicine; Monte Carlo Method; Mathematics; Research Article

This paper develops a methodology for extending multilevel modelling to incorporate spatial interaction effects. The motivation is that classic models are not specifically spatial. Lower level units may be nested into higher ones based on geographical hierarchy (or membership structure—for example, census zones regions) but the actual locations of and distances between them directly considered: what matters groupings how close together any two within those groupings. As consequence, effects neither modelled nor measured, confounding group (understood as some sort contextual effect acts ‘top down’ upon members group) with proximity (some joint dependency emerges neighbours). To deal this, we simultaneous autoregressive processes both outcome variable residuals. assess performance proposed method model, series Monte Carlo simulations conducted. results show performs well in retrieving true model parameters whereas provides biased inefficient parameter estimation presence interactions. An important implication study cautious an apparent neighbourhood terms its magnitude statistical significance if at lower suspected. Applying new approach two-level land price data set Beijing, China, find significant interactions parcel district levels.