Essays on examining the impacts of forest cover on housing prices using Bayesian model averging and geographically weighted regression.
详细信息
文摘
This dissertation presents four essays that address two general research questions. First,what is the proper role of model selection in hedonic price modeling of residential real estate? While there is usually uncertainty about the correct model to use for pricing models,this model specification uncertainty is not usually considered consistently in the final results. The second general research question involves specific characteristics of the housing good: how does the amount vegetative land cover affect the willingness of homebuyers to pay for a housing unit in the market,all else being equal? Although there are recognized benefits to having urban forest cover and other types of natural vegetation within a local neighborhood,are these features valued by local residents? Each of these four essays is used to address these questions,using single-family residential property sales in King County,Washington. The first chapter details a multidisciplinary analysis comparing measures of the social,ecological and economic functions of forest cover along an "urban forest gradient". This analysis contains a hedonic price model for sales from 1998 to 2001 including the urban forest gradient as a semiparametric term. The remaining chapters assess the correlations between price and measures of the amount of distinct types of classified land cover within the local neighborhood of properties sold in 2002 in King County and Seattle,Washington. This incorporates the use of "Bayesian model averaging",or BMA,a technique to evaluate all possible likely models and combine the results into a single estimate to account for model uncertainty. Models are developed using standard and BMA models,for the entire area and for localized models calculated with geographically weighted regression. The results show that there is promise for the use of BMA with hedonic price modeling. Averaging can select appropriate variables for use in modeling,and improve inference when drawing conclusions from output. Increases in model fit are negligible,however,due to the larger data sets used,and management of collinearity can be an issue with BMA in some cases. Modeled price effects of urban forest cover are complex and site-specific,and larger patches of urban forest may not be valued in some locations. In urban areas,there are clearer price effects from increasing vegetative cover in heavily urbanized areas over creating or maintaining large patches of urban forest in many urban neighborhoods. Future work should focus on describing these location-specific effects.