Behind the numbers: House prices

House prices have never fallen since the government started keeping records in 1950, according to this NY Times piece. Actually, they have never fallen since the Great Depression, according to the National Association of Realtors. But they have, in the early 1990’s and since June 2006, according to the Case-Shiller composite index. Confused?

I am (and so is the Times, as they confuse the NAR and OFHEO measures, see below). This is not surprising. The various home price measures differ greatly in coverage and methodology. Here’s a handy guide to the main ones:

OFHEO House price indexes

Calculated by the Office of Federal Housing Enterprise Oversight (the folks who regulate Freddie Mac and Fannie Mae), this index starts in 1975 and has national coverage. It geometrically weighs changes in the price of single family homes on which at least two mortgages have been taken out and bought by Freddie Mac and Fannie Mae. These This helps ensure that the houses included have comparable characteristics over time, avoiding biases resulting from changes in the composition of sales.

It does have two main limitations. VA and FHA mortgages are not included, as are those that exceed the federal loan limit of $417,000. More info can be cound here.

Case-Shiller indexes

Methodologically, they're very similar to the OFHEO indexes (weighted repeat transactions on single family homes). However, they're based on transactions as recorded in county assesor and recorder offices. The geographic coverage is much more limited, as only the largest 20 metropolitan areas are included. It does have the advantage that it covers high-end homes. More info can be found here.

New Home Prices

Published by the Census Bureau, these prices are by definition not comparable to the previous two indexes, as they only cover new houses (a small part of the total real estate market). Information can be found here.

NAR Indexes

The National Association of Realtors publishes information on median and mean prices of existing homes sold (for both single-family houses and condos/co-ops), based on a sample of national transactions. It is subject to biases resulting from changes in the type of units sold and does not adjust for quality (comparable characteristics, as the C-S and OFHEO do by using repeat transactions). The methodology is presented here.

Summary

The Case-Shiller indexes seem to offer the most accurate measure of trends in house prices. However, it does have the drawback of having limited geographical coverage. OFHEO indexes don't have this problem, but the exclusion of high-value housing is a very big flaw. Lack of quality adjustments clearly makes the NAR median home price indicator a much inferior alternative. In the end, they're fairly complementary and should all be analyzed. The following graph shows that the NAR and OFHEO measures produce similar results that are very different from the Case-Shiller index.



This Fed paper (PDF) provides much more information on this topic.