The Case-Shiller Index

Last week’s NY Times piece on home price metrics (login required) prompted some forum discussion on the accuracy of the Case-Shiller home price indices, which I use for all my long-term price charts. It ends up that I looked into their methodology a while back and have been meaning to write about it — so now seems like as good a time as any!

To start with, let me quote from an old voiceofsandiego.org piece as a means of reviewing the problems with the median price:

…while the median price is a decent measure of how much the typical buyer paid, it does not tell us what that typical buyer got for the money. When it comes to measuring actual changes in home market values, the median price can be thrown off by the following:

1. Incentives. San Diego developers routinely offer buyer incentives — televisions, cars, homeowner fee waivers, and more — of up to $30,000. These giveaways overstate the actual price being paid for the homes themselves.
2. Changing purchase patterns. Although I have not verified the numbers myself, I have heard several realtors claim that the so-called "low end" has slowed more drastically than the rest of the market. If true, this would shift the distribution of sold homes prices upward and cause the median price to rise more (or fall less) than it would have otherwise.
3. Bang for the buck. It may simply be the case that a given amount of money will buy a larger or nicer house this year than it would have last year. If buyers are spending the same amount of money but getting nicer houses, this represents a real-world price decline that is not reflected in the median price.

Shiller et al’s methodology is to string together same-home sales in an attempt to model overall price change patterns. This approach addresses item #2 above, because overall purchase patterns will not have much effect on the change to a given unit’s market price.

The Case-Shiller index also addresses #3 –sometimes. If the change in bang for the buck comes from buying a different house, we are ok, because the index will compare that same house’s prior sale price. But if the bang for the buck changes because of modifications to the condition of a given house, things start to get weird.

For example, if someone put $100,000 worth of improvements into a house, a $100,000 increase in the home price would only represent the increased quality of that house. It would not indicate any change in overall housing market prices. But Case-Shiller would note that as an increase in the price of a given home.

Shiller does take steps to address this problem. If the change in a given home’s price is out of line with price changes of surrounding homes, it is assumed that the first home’s price changed because of some non-market factor such as home improvements. So that first home is given a lower weighting in the index.

This helps a bit, but if for instance everyone is making improvements*, then the resulting price increases will seem like the norm and will be ascribed by Shiller’s methodology to an increase in market value.

* This actually seemed to be the case for a while when San Diego had collectively convinced itself that a dollar spent on kitchen fixtures was two dollars earned at the home sale.

The approach of giving lower weightings to outliers should also partially address the first median problem mentioned, incentives. If the sale price of a house is more than it "should" be based on local non-incentived houses, then it will be given a lower weighting. But again, if everyone is granting incentives, incentive-based prices will seem like the norm, and people who aren’t granting incentives will seem like the outliers and get a lower index weighting.

This may all help explain the Shiller index’s apparent overstatement of home prices in recent times. Many people are improving their homes to stay competitive in a glutted market, and more importantly incentives are all the rage. Both factors would tend to overstate quality-adjusted home prices in a way that probably isn’t fully captured by the Case-Shiller index.

So the index isn’t perfect — nothing is — but it’s pretty good. I still think it’s valuable when analyzed in the context of other available data, and that it offers the best available metric of long-term home price trends.