Hedonic-based price indexes for housing: theory, estimation, and index construction

Housing price indexes should not confound the effect of changes in quality with the effects of changing house prices. A recent nonparametric regression technique, loess, allows flexible estimation of the hedonic price function and centers the estimation at fixed points, such as the beginning or ending period housing characteristics. Indexes using these estimates are consistent with the requirements of Laspeyres and Paasche price indexes. The technique is used to obtain indexes for fifteen municipalities in Alameda County from 1970:Q1 through 1995:Q1. The nonparametric hedonic-based indexes provide better controls for the effect of quality evolution on price movements than alternative methods.