Bias and Precision of Estimates of Housing Investment Risk Based on Repeat-Sales Indices:A Simulation Analysis

A simulation analysis is reported which examines the bias and precision of estimates of housing investment risk based on small sample indices of housing returns. The trade-off between smoothing bias (due to temporal aggregation in the index) and noise bias (induced by random estimation error) is examined in the housing return total volatility, beta, and autocorrelation statistics of the index returns. The study compares the performance of three different specifications of the repeat-sales index, under assumptions of either an informationally efficient or inefficient housing market, and at two levels of estimation data availability. Findings suggest that regression-based repeated-measures indices may be useful at a more micro-level (e.g., at the neighborhood level or for specific housing types) than has hitherto been employed.