| The primary objective of this dissertation is to introduce alternative nonparametric procedures to identifying housing submarkets. A two-stage hierarchical hedonic housing price parametric model, prompted by the assumption that quality of public education is capitalized by the size of housing is considered as a base model for comparisons against kernel based nonparametric techniques. The nonparametric method will encompass three specific techniques, the Nadaraya-Watson regression estimator, a purely nonparametric based estimator, the alternating conditional expectations estimator and the average derivative estimation approach, which are semiparametric methods. Based on the housing submarkets implied by the 2-stage hierarchical hedonic housing price estimation, comparisons are made against the nonparametric techniques to validate the authenticity of claimed housing submarkets and provide as a by-product market evaluations of housing and neighborhood characteristics. The models are corrected for potential spatiotemporal autocorrelation that may exist in the residual estimates. Repeat sales coefficient estimates are also obtained and compared across submarkets under a multiplicative parametric heteroskedastic error structure, a nonparametric and semiparametric heteroskedastic estimation approach. An application of a unit root test on the residual estimates of the hedonic housing price models, is also conducted using repeat sales information to test whether housing prices have a random walk component. |
