Author(s): Andrew Leventis, Principal Economist
Abstract: Under the repeat-transactions framework for constructing house price indexes, the paper analyzes the technical challenges associated with producing unbiased price indexes for homes in distinct price tiers. The basic problem is that the “tier” to which a given home truly belongs is unobservable and can vary over time. Various approaches to forming tiered indexes, including the methodology used in the formation of S&P/Case-Shiller tiered indexes, are analyzed. Some are shown to be susceptible to mean-reversion-related biases (under- or over-statement of price changes) when home prices reflect “true” home values plus transactions-related noise. Empirical comparisons of price trends in California as a whole and two California metropolitan areas—San Francisco and San Diego—are shown where indexes are constructed under various tier-classification schemes.