A spatial model of housing returns and neighborhood substitutability

This article provides a method for estimating housing indices at the local level. It develops a ''distance-weighted repeat-sales'' procedure to exploit the factor structure of the error-covariance matrix in the repeat-sales model. A distance function defined in characteristic and geographical space provides weights for the generalized least-squares model, and allows the use of all of the repeated sales in a metropolitan area to measure returns for the specific neighborhood of interest. We use distance-weighted repeat sales to estimate return indices for all zip codes in the San Francisco Bay area over the period 1980-1994. When distance is defined in terms of socioeconomic characteristics, we find that median household income is the salient variable explaining covariance of neighborhood housing returns. Racial composition and educational attainment, while significant, are much less influential. Zip-code level indices often deviate dramatically from the citywide index, depending upon income levels. This has implications for investors and lenders. Our results indicate that rates of return may vary considerably within a metropolitan area. Thus, simply using broad metropolitan area indices as a proxy for capital appreciation within a specific neighborhood may not be justified.