Journal of Property Tax Assessment & Administration


In this paper, the author has shown that when data subject to heteroscedasticity are regressed without correction, the resulting error term is biased. Real estate data are typically prone to this condition. Therefore a means of modifying this type of data that corrects for heteroscedasticity during estimation is superior to all other estimation routines when there is significant within- and cross-priced-level variation. The author developed a means for reducing bias and improving estimation efficiency in nonrandom samples. In the pre-regression analysis of sample data, differences between a sample’s characteristic distributions and the corresponding population’s distribution of property characteristics are calculated. And the differences are used as the weighting factor in the WLS model. When no adjustment is made to correct for the differences between the population and sample parameter distributions, prediction errors with the OLS model are greater than with the WLS technique developed in this paper. The proof is complex; however, the implementation is straightforward, and the results are interpreted in the same manner as in the OSL model.

First Page


Last Page



Mass appraisal techniques