Research On Spatial Autoregressive Panel Data Model And Its Applicationonthe Dependence Of Commodity Housing Price

The acceleration of the urbanization process has led to further expansion of the public housing demend.As a result,the commodity housing prices in some cities are rising all the way.The increasingly prominent “high housing prices” has triggered a close attention of the whole society.Discussing the temporal and spatial variation is of great significance to understand the trend of commodity housing prices.In this paper,we consider the change in economic in the time dimension and construct a dynamic spatial weights matrix,which incoporates the economy-community relations and the space-distance relations.Then,under the time varying spatial weights matrix,we set up a random effects spatial autoregressive panel data(i.e.SAPD)model where the disturbance has a time correlation.Finally,we use the established model to explain the time-space dependence of Shanghai commodity housing prices.The main contents are as follows:First,the spatial weights matrix that incoporates the economy-community relations and the space-distance relations is constructed.A variety of classic spatial weights matrix construction methods are summarized.Moreover,the differences and similarities of different spatial weights matrix are analyzed,and the advantages and disadvantages are compared.Based on the existing threshold distance weights matrix,the grey relational analysis is used to design an economy-community relations matrix between spatial units.And then a time varying spatial weights matrix that incorporates information from the economy-community relations matrix and the space-distance relations matrix is developed.Then,based on the dynamic spatial weights matrix,a random effects SAPD model with a time autocorrelation in the disturbance is established.Simultaneously,the Bayesian Markov Chain Monte Carlo estimation is introduced for parameter estimation,which deduces the conditional posterior distribution systematically for each parameter and gives the sampling method that mixes the M-H algorithm and Gibbs sampling.The parameters to characterize the spatial lag and time correlation are sampled by the M-H method.The Monte Carlo experiments with different cycles are obtained with high accuracy and the validity of the MCMC estimation for themodel is verified.Finally,the established SAPD model is used to explain the time-space dependence of Shanghai commodity housing prices.The spatial autocorrelation of the commodity housing price across the Shanghai administrative regions is tested using Moran'I index.Then,the SAPD model of commodity housing price is established.The model parameter estimation and solution results are given,and the results show that the estimation results of the dynamic spatial weights matrix are superior to the estimation results of the fixed spatial weights matrix,there is a positive effect on the commodity housing prices between neighboring administrative district in Shanghai,and the SAPD model can better interpret the spatial effects of commodity housing prices and has better applicability.