Tuesday, March 18, 2008

It's not just the mean, it's the distribution

David Brooks this morning says that house prices have fallen by 10 percent from peak and that "experts" think it will fall another 20. Actually, the biggest national peak-to-trough number I have seen is negative 8.9 percent (Case-Shiller), and I, for one, guess house prices nationally will fall but another 5-10 percent in nominal terms.

But that doesn't mean I am much more sanquine than Brooks. Around those mean changes are distributions, and I am wondering whether the distribution of house price outcome is getting larger--i.e., whether the standard deviations of the house price indexes is getting larger. If they are, from the standpoint of mortgage performance, the increased number of houses with large price declines will not be offset by an increased number of houses with rising prices.

Much to my surprise, when I look up my wife's and my house on Zillow, its value continues to increase; not by much, but nevertheless it is increasing. And it doesn't matter from a mortgage performance standpoint: we have a lot of equity in the house, so more won't do anything to improve our contribution to default performance. But of course if our house is holding its own in value, there is another house in the DC area that is doing much worse than a declining average, and therefore is very likely to default.

To really characterize this, it would be nice for Case-Shiller and Ofheo and NAR to not just give means and medians, but also quintiles or even deciles of house price performance. Then we would know what we are up against.

2 comments:

Anonymous said...

I agree with you. There are many factors that need to be taken into account before coming to a conclusion.

Anonymous said...

ofheo does have volatility estimates on their website. not as user friendly as percentiles, but percentiles don't naturally fall out of the regression process used to estimate a house price index.