Biased Prediction of Housing Values

Publisher Website
Google Scholar
Abstract
This paper introduces the use of non‐sample, prior information to the problem of predicting prices of heterogeneous products. Using data from the 1983 American Housing Survey, the predictive performance of three Stein‐like empirical Bayes estimation rules are compared to the least squares estimator and the traditional biased estimation technique, ridge regression. The biased estimators improve upon the least squares mean square error of prediction risk under certain design‐related conditions. We provide evidence of this for the housing market in this paper.