Measuring Home Price Trends Is Hard

Interesting note from the New Zealand Reserve Bank, “Evaluating alternative monthly house price measures for New Zealand” which highlights that whilst there are various methods which can be applied to measuring home prices, none is perfect. The data-intensive “Hedonic” approach as advocated by some in Australia, did not come out on top.

They also highlight the “quality-mix problem, which refers to the fact that the composition of houses sold will differ from period to period, making it difficult to discern whether observed price changes reflect genuine movements in underlying house prices or simply changes in the composition of houses sold. For example, prices may increase from one month to the next simply because of an increase in the average size of
houses sold. Larger homes tend to sell for higher prices, so it’s not clear whether the observed increases in prices represent genuine market movements or simply changes in sales composition. This quality-mix problem is of particular concern in the property market since
housing quality varies significantly along multiple dimensions.

This paper outlines the production of three monthly house price indices (HPIs) for New Zealand produced using data from the Real Estate Institute of New Zealand (REINZ) using three alternative methodologies. REINZ approached the Reserve Bank of New Zealand at the end of 2015 for technical guidance on possible improvements to their house price index methodology, in light of significant improvements to their dataset in recent years. The paper documents the guidance, providing an overview of the alternative methodologies and an empirical evaluation of the resulting indices.

The database provided by REINZ is a rich unit-record sales dataset with information on price, location, valuation, and property characteristics (such as the number of bedrooms and the floor area). We use the database to produce HPIs based on three well-established and widely adopted methodologies: 1) sales-price to appraisal ratio (SPAR); 2) hedonic regression; and 3) repeat sales. All three methods are found to produce credible-looking indices, which match the turning points and well-established cyclical properties of New Zealand’s existing house price statistics.

As a benchmarking exercise, the three candidate indices are evaluated alongside a simple median and a stratified median index (similar to the methodology currently used by REINZ). Applying a range of criteria to assess index performance, we find that all three alternative candidate methodologies out-perform the simple median and the stratified median methodologies.

The SPAR method is found to perform the best, due to lower month-to-month noise (especially for more disaggregated regional indices), greater stability as more data are added, robustness to sample changes, and higher accuracy in predicting sales prices.

Author: Martin North

Martin North is the Principal of Digital Finance Analytics

Leave a Reply