Monday, April 07, 2008

The Almost Daily 2¢ - Attack The Index?

In the wake of the National Association of Realtors (NAR) absurd attack on Professor Robert Shiller and possibly even more ridiculous attempt to discredit the validity of the S&P/Case-Shiller Indices (CSI), it appears that there is a subtle, yet growing, effort on the part of local Realtors and other interested parties to publicly discount the accuracy of the CSI.

Spurious accusations abound but several center on the CSI being misleading, either because of an assumed complete bias towards the local metro markets they primarily track or because of the exclusion of condo and new construction properties from their calculations or even that the indices are simply used (with purposely dishonest intent) to portray a gloomier picture than reality.

Although these allegations are mostly transparent and desperate attempts to besmirch reputations and otherwise muddy the water, I thought it might make sense to compare the various modern home price indices to see how well they conform to each other and, in turn, see if any of the above mentioned allegations could be easily dispatched with.

There are three main modern home price index sources, two of which employ essentially the same “repeat sale” methodology and one that focuses on gauging the daily “price per square foot” in order to construct their respective indices and all vary notably in both region of coverage and property inclusion.

The Office of Federal Housing Enterprise Oversight’s (OFHEO) Home Price Indices (HPI) and the S&P/Case-Shiller Indices (CSI) both exclusively use single family home sale transactions, HPI’s coming from Fannie Freddie source data and CSI’s coming from county deed records, to construct their indices using two slightly different versions (HPI vs CSI methodologies) of the “arms-length repeat sale” methodology jointly proposed by Professor Karl Case of Wellesley College and Yale’s Professor Robert Shiller.

Probably the greatest difference between the HPI and the CSI is that the HPI source data is limited to “conforming” Fannie Freddie transactions and thus exclude all transactions made possible due to the use of “Jumbo” loans.

Another major difference is coverage area but although there are many hundreds more HPI indices than CSI, only a few of the HPIs are published as “purchase only” indices that include only data resulting from home “purchase” transactions rather than combining refinance (tends to skew the results) and purchase transactions.

Also, even though the CSI’s are strongly associated with designated metro markets they offer some fairly widespread coverage with inclusion of data from far reaching counties and even crossing state borders when appropriate (e.g. the Boston CSI index includes the counties of Essex, Middlesex, Norfolk, Plymouth, Suffolk, Rockingham NH and Strafford NH).

Radar Logic, on the other hand, constructs price indices (RPX) for 35 metro area markets derived from both new and existing single family and condo sale transactions using a very interesting and inventive approach that seeks to essentially establish the daily spot “price per square foot” for the entire residential property market.

All of the mentioned technologies are both highly analytical and logical and above all else offer a substantially more accurate view of real market price movement especially when compared to the Realtor’s more favored median selling price method (or sometimes average selling price…whichever way the wind is blowing of course).

But how well do these indices correlate with each other?

First, take a look at the following chart that shows the RPX for Boston (averaged monthly) and the CSI for Boston both normalized to a base of 100 and plotted since January of 2000 along with the month-to-month percent change to the CSI supplied for seasonal comparative purposes.

Notice that although the RPX is clearly capturing a more substantial seasonal variation, both indices show a well correlated (Pearson’s of 0.982, r-squared of 0.965) trend and comparable seasonality even in light of the fact that the RPX source data includes both condo and single family property transactions and CSI uses only uses single family.

Notice also that given this correlation the RPX, although not complete for February transactions, is quite possibly giving us a bit of a preview of the February results of the CSI which will be released at the end of April.

Next, study the following chart which adds the “purchase only” HPI for Massachusetts and forgive the “stepped” appearance of the series as it is published quarterly so I simply repeated each reported value as the interim monthly data.

Notice again that the HPI’s underlying trend is fairly well correlated to both the CSI (Pearson’s of 0.996, r-squared of 0.993) and the RPX (Pearson’s of 0.978, r-squared of 0.956) even though the HPI series broadly captures price changes from across all Massachusetts counties and excludes homes that were purchased using Jumbo mortgages.

So the key take away is merely this (and I suppose non-wishful thinkers and non-Realtors would have assumed as much) all of the most well regarded and highly analytical home price indices are essentially saying the same thing.

Whether you keep account by exclusively using single family homes or mix in condos, use whole property price or price per square foot, include metro-regional or look more broadly, when it comes to residential real estate… prices are trending down.