Interpreting Housing Indicators
Finding the right economic indicators is a challenge for investors. Often the same data are presented in several different ways. How does one make the right choice?
Today's data on home prices from S&P Case-Shiller provides a useful example. As everyone knows, prices are down significantly from peak values and the annual data have a strong seasonal component. There are three quite different approaches.
Month-over month changes. The 20-city home price index for April fell by 0.6% from March. This decline was reported by some media sources, but ignores the seasonality in the data. When the seasonal effect is strong, it can be quite misleading.
Year ago changes. Most solve the seasonality problem by comparing the prices in April, 2009, to those in April of 2008. Sources using this approach cited a price decline of 18.1%. This ran as a headline on some stories and as a subtitle on CNBC.
The problem with these year-over-year changes is that it is difficult to see improvement fast enough to be helpful for investment decisions. Let us illustrate this with an unlikely and extreme example. Suppose that the index went up 10% from April to May. The year-over-year value would still be a decline of 9.2%.
To avoid this problem, those using the year-over-year method compare the annual change in one month to that of another. The conclusion often reached is that prices are declining at a lower rate. This is not correct. In the example given, prices would be increasing, not declining at a lower rate. It is not easy to get real insight from a string of year-over-year numbers.
Seasonally adjusted data. S&P also puts out a seasonally adjusted version of the series. This allows the user to focus on the month to month change, the real time movement of greatest interest, while removing the regular seasonal pattern. Using this approach, prices declined by 0.9%, worse than suggested by the other two methods.
Conclusion
Using seasonally adjusted data is frequently the best solution for this sort of problem. Many of our fellow data consumers are suspicious of any adjustments to raw data. They are then forced to make their own seat of the pants guesstimates about how important the changes are.
Calculated Risk, a favorite and featured source, also focuses on the seasonally adjusted data. You can check out the latest update to this series, comparing it to the bank stress test assumptions, in this article.
Excellent info. I find it amazing how often the wrong numbers become the focus of the equity and bond markets based on however bloomberg or reuters intially report it. In the short term that’s the interpretation that matters but when the next related data point comes out the mistaken emphasis often shows up.
On the latest Case-Shiller data the most interesting thing I found was that for the five “non boom” cities the change in prices was basically 0 over the last 5 months (with a range of -2% to +1%) while in the boom cities the average was -3.8% (range -1% to -7%). Clearly a signficant difference, both statistically and practically speaking. So the recent trend is that in the non boom cities prices are almost stable despite the rising unemployment while in the cities which had booms the price falls are slowing but still there.
I’ll blog some more at http://reflexivityfinance.blogspot.com/ about what are likely to be longer term dynamics in housing markets/foreclosures.
Sorry typing error above – was talking about the last 2 months not the last 5 months.
“Calculated Risk, a favorite and featured source, also focuses on the seasonally adjusted data.” Risk vs Reward has always been the mantra of investors, unfortunately both of those values are in such flux it is impossible to know which way to turn