Disaggregated House Price Indices

Abstract

This paper estimates and compares methods of constructing disaggregated house price indices from existing house price models using individual sales data for Sydney. Nine alternative house price models are selected to cover the most frequently used methods in the literature: the mean model, median models (standard and stratified), hedonic models (restricted and unrestricted hedonic), repeat-sales models (age-adjusted and Case-Shiller weighted), and a hybrid of the hedonic and repeat-sales model. The unrestricted hedonic model and the hybrid model have an advantage over the other seven models in that they do not require stratification of the data for estimating disaggregated indices. Both models employ the whole sample to estimate implicit prices of house characteristics that are used to construct disaggregated house price indices. These two models eliminate variability arising from small sample sizes and provide more efficient estimates of house price heterogeneity. In addition, house characteristics that are important drivers of the variability of individual house prices are identified in the two models. Disaggregated indices constructed from these two models provide more accurate comparisons with an aggregate house price index. We quantify the extent to which disaggregated house prices indices have significantly more variability than, and differing trends from, the aggregate index.