Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Finance (Business Administration)

First Advisor

C. F. Sirmans


This dissertation compares eight biased estimators as alternatives to Ordinary Least Squares estimation in the context of predicting residential real estate prices. It considers ridge rule estimation and principal components regressions, techniques that have previously been proposed for this application. It also introduces the use of Stein-like rules for predicting housing prices. The study examines relative performance of these estimators in three data settings and under four separate assumptions regarding loss criteria. The first test of the estimators uses Multiple Listing Service (MLS) data for Baton Rouge, Louisiana between 1984 and 1989 to examine relative predictive effectiveness in a time series framework with highly descriptive data. Next, American Housing Survey (AHS) data for six metropolitan areas is employed to compare the estimators in a cross-sectional context with the type of data typically used to create housing price indexes. Finally, the AHS data is used as the basis for a Monte Carlo experiment that compares estimator performance in numerically simulated repeated samples. The partitioned Stein-like estimators do well in all three data environments. Two of them provide especially impressive performance. Under quadratic loss, in the Monte Carlo experiment, these estimators outperform all compared alternatives across the entire range of generated samples.