Date of Award
Ph.D. in Economics
Walter J. Mayer
Matthew D. Hill
This dissertation develops and estimates a spatial autoregressive with autoregressive errors model of housing prices that accounts for both the endogeneity of spatially-lagged housing prices and local school quality measured by performance on state standardized tests. Two datasets are used from Boyertown, PA and Minneapolis, MN. Homes are spatially weighted against each other using a k nearest-neighbor approach. School quality is thought to be endogenous because unobserved neighborhood amenities in the error term of a hedonic regression are very likely positively correlated with local elementary, middle, and high school quality. Following previous literature, the optimal instrument matrix is constructed as the Cochrane-Orcutt tranformed conditional means of the spatially-lagged housing prices and quality measures. As school quality is observed on a much lower frequency than housing prices, it is not possible to estimate the conditional mean of school quality using non-parametric methods as proposed previously in the literature. So in order to instrument the school quality variables, a parametric model in which school quality is a function of average home prices within its attendance zone and average home prices outside its attendance zone but still within the same school district is used. Three different methods are presented for estimating the conditional mean of the spatially-lagged housing prices, one of which is new to the literature. I find that parametrically estimating school quality can cause issues when the number of observations on quality are low as in the PA dataset. Also results are not robust to different specifications of W as small changes in k can affect the estimates by a large amount.
Smith, Jonathan Taylor, "School Quality Capitalization Into Housing Prices In Minnesota And Pennsylvania" (2013). Electronic Theses and Dissertations. 484.