In a report that manages to be a geeky advertisement for its own preferred measure of housing prices, the OFHEO last week released a report covering differences between its own HPI index and the popular S&P/Case-Shiller indices. The report claims that the HPI index values are a) geographically broader, b) include more states, c) include more complete data. Regarding HPI's use of more states:
According to the methodology materials, the S&P/Case-Shiller index does not include house price data from thirteen states. Market conditions in those thirteen states have, on average, been stronger than in the rest of the nation. OFHEO's estimates indicate, for example, that three of the five fastest appreciating states in the nation (Idaho, Montana, and Wyoming) do not have representation in the S&P/Case-Shiller index. This missing information has likely caused some of the divergence between the trends shown in the two national indexes.
What's interesting is that the first part of the paper spends great detail pointing out perceived flaws in the S&P version of the index, but the very last paragraph of the report bring up a biggie that probably should have been discussed earlier:
The impact of the conforming loan limit on OFHEO's data sample likely explains part of the difference; with less extensive data representation at the upper end of the price spectrum, OFHEO's index estimates will be less influenced by price trends for expensive homes.
That's one way to spin it. The other is that the markets with the most dramatic price movements long ago outstripped the conforming loan limit (Boston, Los Angeles, New York -- should I go on?) -- these are also the markets driving the majority of loan volume. Which means, in a nutshell, that it's entirely possible that the HFI is underrepresenting market trends because it's using the wrong sample. But that analysis, I suspect, will be forthcoming in a rebuttal from Standard & Poor's.