Despite their different methodologies, the seven major house price indices tell fairly consistent stories nationally over the long horizon. Even over shorter-term horizons (year-over-year and quarter-over-quarter), the indices show generally consistent results.
But an inspection at the metro level reveals larger differences across the indices, which in turn reveals the prudence in consulting multiple data sources when evaluating local trends over a shorter period of time, Goldman Sachs Global Economics says in a new research report.
House price indices are available from multiple sources and vary in terms of data coverage and methodology: S&P Case-Shiller, CoreLogic [stock CLGX][/stock], Federal Housing Finance Agency, Freddie Mac Home Price Index, National Association of Realtors, Radar Logic and Zillow [stock Z][/stock].
Indices that are based on the broad U.S. housing market — and not just homes purchased with mortgages backed by Fannie Mae and Freddie Mac — all peak from 2006 to 2007 and experience 30% (see chart below) price declines through 2009, and then decline another 5% to today.
They also all agree on the ranking of long-run trends across metro areas, with all of them depicting Las Vegas as the worst performing metro in terms of peak-to-trough decline, and Dallas and Denver among the best.
At the metro level over short horizons, however, there exist larger differences across the house price indices. For example, the dramatic 18% house price drop seen in the Atlanta S&P Case-Shiller index over a recent 12 month period is not reproduced in other indices, which suggest year-over-year house price growth declines closer to 8%.
Goldman gives clues as to why the deviation exists. One is the number of transaction pairs used to estimate the Case-Shiller indices: the transaction count in Atlanta in March 2012 was 2.5 times higher than the count in March 2011, while the transaction count was largely flat for the 20-city composite index, as well as for the other individual constituent cities, over the same time period.
“The findings suggest that while the choice of index may not be crucial for understanding broad long-run trends, it may be desirable to consult multiple data sources when evaluating local trends over a shorter period of time,” Goldman says in the report.
House price indices vary according to features such as:
Weighting – does the index construction algorithm treat high priced and low priced homes equally, or does it weight transaction data in proportion to market price?
Geographic coverage – does the index cover all metros within the U.S., or just a limited subset of the largest metros?
Collateral data coverage – are all home sales potentially incorporated in the index estimation, or are just a subset of homes (such as those financed by Fannie Mae/Freddie Mac mortgages) included?
Most of the differences in cumulative peak-to-trough house price depreciation rates seen in the chart below are due to differences in the respective methodologies.
The FHFA index has a relatively shallow peak-to-trough decline of 20% because it is constructed from data on sales of homes backed by Fannie Mae and Freddie Mac mortgages only and excludes homes backed by subprime and other non-conventional mortgages. Other indices that are based on transactions of homes backed by conventional mortgages and homes backed by non-conventional mortgages reflect greater home depreciation.
The indices differ more in terms of their measures of shorter horizon local growth rates. So, Goldman shows us, the more narrow the question of interest, with respect to either time-horizon or geography, the more there is to gain from consulting multiple house price data sources.
