The Standard & Poor's/Case Shiller Index was released Tuesday, January 29th, showing an 8.4% drop in real estate prices in November, 2007 from November 2006. Case Shiller is a highly reported index that is alleged to be more accurate than the statistics provided by the Realtor associations.
The argument is that the Realtors report the average or median price of all sales from one year to the next, which is a different group of homes each year. Case Shiller only uses cases where the same home has sold twice, using the apples vs. apples argument. There is logic to this argument, but there are significant problems.
The first problem is the small statistical sampling. In any given year there may be at most about 5-7% of all homes selling. It's a much smaller group that sells twice in a short period, which is what Case Shiller is looking for. To "keep sample sizes large enough to create meaningful price change averages", they use a "three month moving average algorithm". I would argue that this is still far too small a sampling.
The index construction methodology uses a weighting of sales pairs with less weight applied to pairs with longer intervals between sales. Excluded from the calculation are sales less than six months apart, as they are not likely to be arms-length transactions. The most weight is applied to sales about a year apart. Now let's think about this: Who sells their home after just one year? It's often a foreclosure or some other distressed situation. No wonder the prices are so much lower. In fact, per Case "Subsequent sales by mortgage lenders of foreclosed properties are included if repeat sales pairs, because they are arms-length transactions". Right...Ok
Imagine, someone buys a house for $200,000, loses it in foreclosure and the bank sells it for $160,000. An investor buys it, paints it and sells it for $200,000. Case Shiller index includes the foreclosure sale because it happened after one year and excludes the investor sale because it took place in less than six months. According to the index, the property value dropped $40,000. The point here is that most weight is afforded to sales that may be distress situations.
There are ten main statistical areas in the index plus ten other areas. Averages from the ten large metropolitan areas are then used to determine the U.S average. Again, this is another example of a small statistical sampling.
In addition, you have the same problem that exists with all methods of coming up with average sales increases/declines. What really matters in real estate is the status of a specific property rather than a whole area. According to Case Shiller, my area, Minneapolis dropped in value 6.6% last year. The actual drop in average sale price was 1.3%. But there were great variances within the area. On area within the metropolitan area dropped 33%, while another increased by 12%. What's most important to homeowners, sellers and buyers is the status of their neighborhood rather than national or even city averages.
Ultimately, all statistical measurements have their flaws, including Case Shiller. I give them credit for their creativity and the work involved in constructing their complex algorithm. But in the end, nothing replaces a good Realtor in understanding the pulse of the market.