Which is the Most Accurate Home Price Index?

There has been much ado about which home price index is a best indicator of the health of the U.S. housing market. There are really only two worth considering since the National Association of Realtors median prices tell an unreliable tale at best.

Median prices are subject to the many vagaries of changes in the market. For example, when a market begins rising, median prices tend to be skewed lower from the sales of a greater number of lower priced starter homes as first time buyers enter the market. Conversely at the tail end of the bull market and as it begins to deteriorate, the luxury home market is generally the last to roll over skewing median prices higher. As the market has broken down, median prices have rapidly declined as the sales of the more expensive homes began to fall in earnest.

But the pivotal question remains. Which index gave us the earliest warning that home prices were in trouble?

As we see from the charts below, the regular Office of Housing Enterprise Oversight U.S. quarterly home price index registered a peak in Q2-2007 (Figure 1). This was a full year after a peak showed up on the Case-Shiller home price index in figure 2 (see Case-Shiller ). The OFHEO HPI also shows home prices were still appreciating as of September 2007.

The regular OFHEO index is a composite of both actual paired sales prices and refinancing data based on appraisals. But it only tracks homes with conforming mortgages so does not include sales in which jumbo and subprime mortgages were used. Another problem with it is that in both rising and falling markets, appraisals tend to lag since they are based on historic sales data. (For a comparison between OFHEO and Case-Shiller, see http://www.ofheo.gov/media/research/notediff2.pdf (pdf file))

click to enlarge

Figure 1 – Chart of quarterly OFHEO regular home price index (top chart) data and year-over-year rate of change (lower chart). This index did not peak until Q2-2007 and the year-over-year chart shows prices still rising. Data - OFHEO

Figure 2 – Monthly S&P Case-Shiller home price index data showing a peak in July 2006, a full year before the regular OFHEO HPI. As we see here, the CS HPI shows prices beginning to drop in January 2007 while the OFHEO HPI would have us believe that prices are still rising. Data- S&P Case-Shiller.

Figure 3 – Here is the quarterly OFHEO Purchase Only Final Adjusted HPI that is much more similar to the Case-Shiller chart above. The problem is that the latest data on the OFHEO website is a year old (March 2007) so of little use in showing recent changes in home prices. Data – OFHEO.

As a trader, I put more stock in the S&P Case-Shiller Home Price Index for two reasons. First, it forms the standard upon which the home price futures are traded (and priced) on the Chicago Mercantile Exchange that allows traders and home builders to hedge housing risk. The May 2008 Composite contract (symbol CUSK) has fallen from 211.82 in July 2007 to 189.60 for the week of February 18, 2008.

However, second and more importantly, the Case-Shiller HPI provided the first real warning that home prices were falling while the economic and fundamental investment world (including the Fed) believed they were still rising.

Here are two more points to consider. From the peak in February 2007 to December, existing home sales fell 27% and new home sales dropped 41% from their December 2006 peak. As of the latest data, new home permits and housing starts have fallen 53% and 56% respectively from their peaks (in January 2006) while the National Association of Home Builders housing market index, a monthly survey of home builder sentiment, has dropped from its peak in June 2005 of 72 to an all-time low of 18 in December (07). It challenges the bounds of credulity to believe that home prices will not continue to be negatively impacted in the face of sales and demand falling out of bed.

Although there is no guarantee that the Case-Shiller Home Price Index will be continue to be the best leading indicator of where the housing market is headed, I prefer to go with what works best and will continue to do so until there is convincing evidence to the contrary no matter what industry cheerleaders and financial analysts may say.

Comments

[locke]

Great article, Matt. It's well-researched and empirical. I hope you post an update when you see things turning around!

[MattBlackman]

Thanks Locke. I track various real estate metrics including National Association of Realtor median price and inventory data weekly as well as the Case-Shiller and OFHEO home price indexes. For the latest weekly info and charts please go to tradesystemguru.com/co.../

[Louis610]

C-S only measures the sale of previously-owned homes.
It does not measure the pricing of newly-constructed homes, nor does it measure the pricing of condominiums. Each of those belong in the category of 'housing', and each have had price reductions more substantial than pre-owned homes. Inventories of each hand over the housing market.
So perhaps C-S is better than OFHEO, but that is a flawed comparison. And C-S does not overstate the housing decline, as one Columbia professor tried to spin. The C-S methodology actually sidesteps the two housing categories with even larger declines.

[SavCD]

I still like OFHEO better. The CS model has a lot of flipper towns, and dying towns. I think if these covered all MSA's, I might feel differently.

[It Figures]

I think it might be useful to approach the "which is best" question from the point of view of what you want to use it for. If you want it to help with trading, C-S probably does the trick, as it is what most others look at. But if you are looking for a read on housing's impact on the consumer, OFHEO could be more informative, as it probably eliminates non-owner occupied spec related transactions (overwhelmingly non-conforming loans), where a great deal of the "sell at any price" and/or walkaway behavior has been concentrated.

Houses are not fungible, and neither are housing markets. To gauge the impact on the creditworthy owner occupied market (the bulk of which are going to look like the OFHEO data), I think it is useful to judge the degree of speculation in the market before coming to conclusions. Looked at this way, most markets are not declining Y/Y - just increasing very slowly. This is why the consumer spending numbers ( and WMT) keep "surprising" on the upside.

[Stanley]

also remember that when sales slow down and inventory rises, the market price (the price which would result in a given amount of housing inventory selling) is effectively falling, even though asking and selling prices might not yet be. all inventory (housing and otherwise) sells if priced correctly. increased inventory results where there is a difference between the market price and the asking price

[alemboy]

Interesting try by Mr. Blackman, but unpersuasive. Here's why. In some markets, prices did in fact begin to fall in mid 2006, but in many markets, that did not occur (nor has the drop been as extreme) until early 2007. Case-Shiller is fundamentally structurally biased towards markets that were overly hot, became excessively expensive more quickly, the market turned, and now will be overly cold and vulnerable. As such, it is not surprising that it does a better job describing those markets - for this one particular market downturn. The question is, how do you weigh amongst the entire body of US housing markets? That matter isn't proved here, it is still entirely open for debate. And any of the indices are probably going to have varying accuracy in different types of market conditions going forward...we are a long way from establishing a "one size fits all" index.

More minorly, the fact that Case-Shiller matters to a futures market is completely and utterly irrelevant in trying to assess it's historical accuracy. Futures markets are nothing more than good educated guesses about the future, do not always turn out to predict accurately (if they did, life would sure be easy!), and as such are totally inappropriate for this kind of validation test.

[tradersystemguru]

Alemboy, interesting opinion. (You wouldn't be in the real estate industry by any chance would you?)

Too bad you didn't provide more proof. OK, so not every market dropped in 2006 and in fact as of the latest Case-Shiller index, three markets (Charlotte, Portland and Seattle) have not dropped in price. That leaves another 17 that have for an average drop of nearly 10% in the last year.

Would you rather rely on median prices even though they are subject to the vagaries of changes in market dynamics (eg. the highest priced homes tend to be affected last since it is the last segment of the market to be negatively impacted in a correction thereby skewing the median price higher at the beginning of a correction)?

Chart 4 at tradesystemguru.com/co.../ shows how median prices have changed over time. As the chart shows, from their latest peak just 8 months ago, median prices have fallen 12% which is a greater drop than shown by the Case-Shiller home price index. The problem is that median prices didn't show that the market had peaked until July 2007, a full 9 months after the Case-Shiller HPI showed prices declining.

Don't know about you, but don't need an indicator that tells me what's happening long after the fact. You are obviously not a trader...

Matt Blackman

[IAS360]

Case-Shiller’s home price index could be considered noteworthy given that you only want to look at a limited number of MSA’s or the entire nation, but for practicality’s sake, this data is fundamentally incomplete. Essentially, C-S is irrelevant.

[MattBlackman]

Given that it provided the earliest warning of the property price meltdown, I find it difficult to buy your argument that the CS index is irrelevant. It may not be perfect but from from what I have observed is the most reliable index we as investors have. However, this fact seems to be ignored by those in the real estate business who have done their level best to try to discount it. You do not explain your criticism of the CS index nor do you provide an alternative. So what index do you think has done a better job?

[IAS360]

IAS360 House Price Index Provides First Monthly View of Housing Price Trends Based on Neighborhood Level Data.

Integrated Asset Services (IAS), a leader in default management and residential collateral valuation, just launched its monthly-reported IAS360 House Price Index www.iasreo.com/ias360....

The new Index represents the industry’s first clear representation of U.S. housing market trends at a county level. IAS360 House Price Index is a comprehensive housing index tracking monthly change in the median sales price of detached single-family residences in more than 15,000 “neighborhoods” across the U.S. This data is then rolled up to report on the changes in 360 counties, nine census divisions, four regions, and the nation overall. The timeliness of the data, which is based on all arms-length transactions occurring in underlying neighborhoods, makes the IAS360 the leading indicator for housing price trends in the U.S. April Index: www.iasreo.com/ias3600...

[alexcrystal]

Are there empirical evidences that SPCS would also forecast the rebound from the housing bottom better? Thanks.