基于地理空间数据的Hedonic房价指数构建

Constructing Hedonic Housing Price Index Based on Geospatial Data

在基于OLS的传统Hedonic房价指数构建中,房价的地理位置坐标没有被纳入模型,使观测个体之间的空间相关问题被忽略,也导致了模型中估计的参数和构建的Hedonic房价指数的偏误。文章将空间计量经济模型(SEM和SLM)和空间广义可加模型(GAM)应用于分析天津二手住宅交易数据,能够提高传统Hedonic模型的拟合与预测能力,进而提高房价指数估计精度。结果表明,传统Hedonic房价指数在房价较平稳时期有效,但在快速上涨时期因不能对交易住宅的地理位置质量变动进行完全地调整会低估房价指数。

In the construction of the traditional Hedonic housing price index based on OLS(Ordinary Least Squares),the geographic coordinates of housing price are not included in the model,which ignores the spatial correlation between the observed individuals and results in biased error of estimated parameters and the constructed Hedonic housing price index in the model.This paper utilizes two basic spatial econometric models(spatial error model and spatial lag model)and spatial generalized additive model(GAM)to analyze Tianjin second-hand housing transaction data,which improves the fitting and prediction ability of traditional Hedonic model,thus improving the prediction accuracy of housing price index.The results show that traditional Hedonic housing price index is valid when the housing prices are stable,but that the index is underestimated when the housing prices rise dramatically due to the failure to fully adjust the geospatial coordinates’quality changes of transacted residences.