A Real Estate Price Index for Thin Markets

This article examines a time-series-based method for estimating real estate price indexes for markets that have few transactions. The proposed method is more parsimonious than the conventional repeat sale or hedonic methods. Also, it is potentially more accurate and less prone to outliers. It achieves this by linking current transactions to preceding transactions, thereby increasing the set of comparable transactions on which to base the index. My experiments confirm that the time-series price index fares much better in thin markets than a benchmark hedonic index. It remains close to the true index when there are few transactions and it does not have the volatility of the benchmark index. While the time-series-based index developed in this article does better than the benchmark hedonic index, one surprise result is that the hedonic index is itself quite robust in small samples.