New empirical evidence on heteroscedasticity in hedonic housing models

This paper re-examines the issue of heteroscedasticity in hedonic house price models. The paper uses data for Boston, which has a high average age of dwelling. The results largely support previous findings with evidence of heteroscedasticity with respect to the age of dwelling. The iterative GLS correction, specified in terms of age, eliminates all heteroscedasticity at both aggregate and disaggregate levels. However, tests for heteroscedasticity with respect to living area show that the GLS models report significant findings. In addition, a more general EGLS approach does not even eradicate heteroscedasticity with respect to age. Evidence is presented that would support the estimation of hedonic models at a disaggregate level. It is hypothesised that this is due a greater level of homogeneity in the sample at a submarket level, leading to a reduction in reported heteroscedasticity with respect to both age and living area. (C) 2004 Elsevier Inc. All rights reserved.