Spatial Structure Of Urban Housing Land Prices Based On GWR Model

Land price is a "weatherglass" of the land market and conditions of supply anddemand, and also is governments' main instrument for macro-economic regulation.Studies on land price and its spatial structure could provide a reference to visualizethe property of urban land, formulate and implement land supply plans, and toregulate the supply structure of land. It is also an important basis for land market trade,real estate development, decision-making of construct projects and reform in realestate tax system.The former researches on land price, usually treating spatial structure change as astatic one, often use Hedonic Model to depict the spatial variety of the whole studyarea. This research method hypothesized that the error items were dividedindependently and equally. Spatial information got from this way was often infrequentand unpractical, as it neglected the local jump and individual variety of land price.Among researches on base land price, homogeneous areas were divided subjectivelyby dividing regions. Also, influence factors were selected subjectively and theinfluence was deemed to zero when it out of the homogeneous areas. This dividingaffected the express of land value badly. The existing literatures on land price inChina, almost limited to data survey and data compute, scarcely using maps to interpret why this spatial structure of land price was formed. It seems that peopledidn't know the relationships between the prices of land sites seated aside very well.In a same way, planners scarcely realized the effect of the plan on land price. Underthis condition, researches on land price need technical innovation to advanceharmonious development of cities and help governments to make decisions.Geographical Weighted Regression (GWR) model was adopted in thisdissertation to deal with technical problems related with land price. By taking housingprices in Hangzhou as an example, this dissertation used open land market data(including calling for bids, auction and listed on board, between 1998 and 2005) andother data collected from investigate and statistics. The data was quite complete andclose to market reality.This dissertation emphasized the importance of the data pretreatments at first.Treatments such as complete disposal, outlier exclusion and standardization weredone according tO Geography theory and Statistics theory. Especially during theprocess of standalization, a buffer analysis using "range" was done to make the studyarea smaller and improve the precision of the research, and it helped to selectinfluence factors in a certain distance range in the meantime. Exploratory spatial dataanalysis (ESDA) was adopted to do spatial autocorrelation analysis. The resultsshowed that the structure was not single-trend distributed, but have different spatialautocorrelation relationship at different location.In order to confirm spatial non-stationarity of land price spatial structure,Geostatistics tools and GIS tools were introduced to display the spatial structure easily.The existence of spatial non-stationarity of land price need specialized spatial analysismethod. By comparing with other models, this dissertation explored the mechanism ofGWR model and proved its feasibility and value in land price research.In applying GWR model to housing land price, the research depicted factorschoosing and their quantization in detail. It compared the difference between fixedspatial kernels model and adaptive spatial kernels model, after taking the followingsteps: computing optimal bandwidth and regression coefficients, doing ANOVA,casewise diagnostics, Monto Carlo significance test and so on. Adaptive spatial kernels model was selected. Ten influence factors were selected in interpreting thespatial structure of housing land price. These ten variables were: CBD, West Lake,Qiantang River, quick transportation system, elementary school, colleges, hospitals,main household goods markets, land area and FAR. A detail spatial structure map onland price was used to make the analysis more intuitionistic and real. Hedonic modelwas used with the same data in order to do comparison. The results showed that theGWR model was more advanced in technique.At the end of the dissertation, the results of the GWR model were applied in caseland valuation and the amendment of base land price to show its practicability andgreat social value. The former made the revise of land valuation factors morecompletely. The latter solved the technique problems existed in area amendment ofparcel land for long time and provided an effective and credible method forgovernments' decision-making.