Construction And Application Of A Spatio-temporal Housing Price Index

Housing price and housing price index is one of the most important indicators of the housing market.To reflect the trends and patterns of housing price accurately and promptly is the basis for the monitoring,analysis and research of housing market.As housing is a typical heterogeneous commodity,it should eliminate the impact of housing quality differences when compiling housing price index,which means homogenization.Traditional homogenized housing price index is in general calculated from parametric Hedonic pricing model using OLS estimation,which may cause missing variables errors and function setting errors,especially concerning the spatial attributes of housing.In addition,the parametric Hedonic model can not effectively capture the complex spatial distribution of housing prices and its dynamic changes over time,which causes that the traditional housing price index lacks consideration of the spatial dimensions,and can not reflect the heterogeneity of price changes in different locations within the city..Based on hedonic pricing theory,a semi-parametric hedonic model is built in this paper,in which the location characteristics are specified as a nonlinear function of the geographic coordinates in a nonparametric form,while other characteristics are specified in a parametric form.Semi-parametric hedonic model relax the assumptions of the function form of geographic coordinates,which helps mitigate the errors of traditional Hedonic model.Based on this,the spatio-temporal housing price index could be compiled and could be used to measure and compare homogenized housing prices simultaneously over time and space.The empirical models are estimated by 119,745 transaction data from 2011 to 2015 from the second housing market in Beijing,and the results show that semi-parametric hedonic model has noticeable advantages in in-sample fitting performance,out-of-sample predictive power,the consistency and efficiency of parameter estimation,and the capture ability of the spatial dependence of housing price,when compared with the traditional models.The homogenized spatial housing price index effectively captures the spatial distribution of housing prices and intuitively shows the spatial heterogeneity in housing prices.The homogenized spatio-temporal housing price index further shows the dynamic change of the distribution over time,revealing the spatial heterogeneity of housing price growth.Based on this index,the whole and local spatial structure as well as its temporal variation patterns could be analyzed,and the differences of local housing market performances could be compared.This index provides an essential tool for researchers to conduct housing price study on a more micro level.Research on the spatio-temporal index in Beijing reveals that the distribution of homogenized housing price presents a pattern of a major center plus a secondary center,surrounded by a number of local highs.After conducting the cluster anylysis,it is also found that Beijing still exhibits a centripetal spatial structure with serious imbalances between northwest and southeast regions.In addition,the spatio-temporal housing price index can also be applied to the mass appraisal techniques.It provides a correction factor at a more microscopic level to the time and location attributes of housing transaction,and improves the fitting and prediction accuracy of appraisal results.