Spatial Econometrics Revisited: A Case Study of Land Values in Roanoke County

Abstract

An increasing volume of empirical literature demonstrates the possibility of spatial autocorrelation in land value models. A number of objections regarding the methodology followed in those empirical studies have been raised. This thesis examines three propositions. The first proposition states that there is spatial dependence in the land value model in Roanoke County. The second proposition is that mechanical construction of neighborhood effects, or grouping nearby land parcels into neighborhoods, is not always the best way to capture spatial effects. Finally, the third and most important proposition states that by implementing a comprehensive set of individual and joint misspecification tests, one can better identify misspecification error sources and establish a more statistically sound and reliable model than models based on existing spatial econometric practices. The findings of this dissertation basically confirm the validity of those three propositions. In addition, we conclude that based on their development status prices of land parcels in Roanoke County may follow different stochastic processes. Changes in the values of hedonic variables have different implications for different groups of land parcels.