Consistency of Hedonic Price Indexes with Unobserved Characteristics

Abstract

Hedonic regressions are prone to omitted variable bias. The estimation of price relatives for new and disappearing goods using hedonic imputation methods involves taking ratios of hedonic models. This may lead to a situation where the omitted variable bias in each of the hedonic regressions offset each other. This study finds that the single imputation hedonic method estimates inconsistent price relatives, while the double imputation method may produce consistent price relatives depending on the behavior of unobserved characteristics in the comparison periods. The study outlines a methodology to test whether double imputation price relatives are consistent. The results of this study have implications with regard to the construction of quality adjusted indexes.

Keywords: Hedonic imputation method, omitted variable bias, model selection, quality adjusted price indexes, new and disappearing goods

JEL Classification: C43, C52, E31