ACCOUNTING FOR LOCAL SPATIAL HETEROGENEITIES IN HOUSING MARKET STUDIES
采用高斯条件自回归模型处理特征价格分析中遗漏的空间异质性，通过区域特定随机效应与邻近区域的空间交互，得到更可靠的空间相关系数估计，并发现社会经济邻里特征对房价空间变异影响有限。
ABSTRACT We adopt a novel method to deal with omitted spatial heterogeneities in hedonic house price analysis. A Gaussian variant of the conditional autoregressive (CAR) model is used to study the impact of spatial effects. In a general linear modeling framework, we include zone‐specific random effects that are allowed to interact spatially with neighboring zones. The results demonstrate that this estimator accounts for missing spatial information, producing more reliable results on estimated spatially related coefficients. The CAR model is benchmarked against a fixed effects model. Socioeconomic neighborhood characteristics are found to have only modest impact on spatial variation in housing prices.