The Research Of House Price Index Based On Matching Model And Longitudinal Unbalanced Panel Model

It has important theoretical and practical significance to study the compilation theory and method of housing price index.Accurate house price index can reflect the changes of real estate,provide basis for the government to formulate regulatory policies,and provide reference for real estate investors and real estate buyers.However,the compilation method of house price index issued in China is still the first or second generation method,the backwardness of the theoretical approach led to a deviation between the house price index and the real house price trend.The new ordinary residence is the major part of commercial housing sales in Chinese cities,this paper takes it as the research object,and deeply studies the compilation theory and method of its price index.Firstly,a matching model for the new ordinary residence price index is established in this paper.In the first step,a semi-linear hedonic price model of house price index is constructed by analyzing the influencing factors of house price.By differencing the semi-linear hedonic price model of two real estaties in the same set of upstairs and downstairs,the compilation model of house price index with the set of upstairs and downstairs as the matching space is established.Then the specific form of the error covariance matrix in the matching model is deduced,and the corresponding three-stage least squares method is designed to estimate the parameters in the matching model.To some extent,the influence of residual autocorrelation and heteroscedasticity on the parameter estimation in the matching model is eliminated.In order to prevent excessive error caused by data errors or matching errors,an iterative algorithm for eliminating outliers is designed to gradually delete the abnormal matching pairs in the data set.The empirical results show that the significance level of parameters is improved,the residual becomes smaller and the distribution of residual becomes more reasonable after the abnormal matching pair is eliminated.Secondly,a longitudinal unbalanced panel model of the house price index is constructed in this paper.To avoid splitting the sample data into matching pairs,and comprehensive utilization of all real estate transaction information in the set of upstairs and downstairs,all the trading houses in the same set of upstairs and downstairs are considered as a whole in this paper,and a longitudinal unbalanced panel model with the random effects of upstairs and downstairs set and the interaction between the type of building and the standard height is constructed to compile the price index of the new ordinary residential in China.Similarly,in the process of parameter estimation,the iterative algorithm of outlier elimination is implemented.Empirical results show that after eliminating the outlier data,the calculated housing price index can better highlight the change of house price.Thirdly,in order to ensure the consistency of the published indexes and calculate the house price index dynamically,a chain updating model of house price index is designed in this paper.The chain update model is applied to the matching model and the longitudinal unbalanced panel model.Using the dataset of this paper to check the chain update model,the results show that the price index of the latter period calculated by the chain update model is basically consistent with the result of the original model,which shows that the chain updating model is effective.Finally,the two models established in this paper are compared and analyzed in aspect of the calculation results and its statistical significance and the relative root mean square error.The results show that the two models have different advantages in these three aspects.In practical application,we could select one of the methods to compile the housing price index of new ordinary residence in China according to the distribution of real estate transaction data.