The April 2012 S&P Case Shiller home price index shows a -1.9% decline from a year ago for over 20 metropolitan housing markets and a -2.2% decline for the top 10 housing markets from April 2011. Home prices are now comparable to May 2003 levels for the composite-20 and March 2003 for the composite-10, but up for the month. Both composites are now down -34% from their 2006 home price bubble peaks. March 2012 home prices were record lows. Below is the yearly percent change in the composite-10 and composite-20 Case-Shiller Indices. These are not seasonally adjusted, but comparing from April 2011.
Below are all of the composite-20 index cities yearly price percentage change, using the seasonally adjusted data. We can see the burning of Atlanta, year, down 17% from a year ago While we have a Phoenix rising, up 8.6%, bear in mind Phoenix is down 51% from its peak.
S&P reports the not seasonally adjusted monthly data for their headlines. Housing is highly cyclical. Spring and early Summer are when most sales occur. See the bottom of this article for their reasoning.
This month the not seasonally adjusted March to April change for the composite-20 was +1.3%. The seasonally adjusted composite-20 monthly change was +0.7% (maroon). The composite-10 not seasonally adjusted monthly change was also +1.3% and the seasonally adjusted composite-10 monthly change was +0.7% (blue). The below graph shows both composites, using the seasonally adjusted series.
The above graph shows the composite-10 and composite-20 city home prices indexes, seasonally adjusted. Prices are normalized to the year 2000. The index value of 150 means single family housing prices have appreciated, or increased 50% since 2000 in that particular region. These indices are not adjusted for inflation.
News headlines on the S&P Case-Shiller Housing Index often differ. Some in the press use the seasonally adjusted data, and others do not. S&P themselves use the not seasonally adjusted housing price data. To make matters worse, some in the press compare just the monthly not seasonally adjusted data and others compare the housing price index to one year ago, without telling you what number they are quoting from S&P. Below are the seasonally adjusted monthly home price percentage changes for each City reported by S&P.
Below are the seasonally adjusted indices for this month. Folks spin these indexes with percentages while the index itself tells you what has happened to home prices, per city, from the year 2000. Remember the below is seasonally adjusted.
Either way you look at this month's data, most metropolitan area housing prices came off from their bottoms, except for Detroit. Both Boston and New York had little change. From the S&P press release, using the not seasonally adjusted home price indices:
Detroit and New York stand out this month as the only two MSAs that saw their annual rates of return deteriorate compared to March. While Detroit posted a positive annual rate of 1.2%, it was still well below March’s +3.9%; New York was -3.8% in April down from -3.0% in March. Detroit was also the only city to show a monthly decline, down 3.6%.
To Season or Not to Season, That is the Question:
The S&P/Case-Shiller Home Price Indices are calculated monthly using a three-month moving average and published with a two month lag. Their seasonal adjustment calculation is the standard used for all seasonal adjustments, the X-12 ARIMA, maintained by the Census.
So, why would S&P report the not seasonally adjusted data? According to their paper on seasonal adjustments, they claim the not seasonally adjusted indices are more accurate. It appears the housing bubble burst screwed up the cyclical seasonal pattern. What a surprise, although those steep cliff dives are now going back to 2009, one would think the seasonally adjusted data would now start to converge back to it's cyclical, seasonal pattern.
The turmoil in the housing market in the last few years has generated unusual movements that are easily mistaken for shifts in the normal seasonal patterns, resulting in larger seasonal adjustments and misleading results.
To see S&P's argument in action, look at the below graph. The maroon line is the seasonally adjusted national index, reported quarterly. The blue line is the not seasonally adjusted national index. As we can see before the housing bubble burst, we see a typical cyclical pattern difference between the seasonally adjusted and not seasonally adjusted data points. Yet after the bubble burst we see large swings, which would throw off a seasonal adjustment adaptive algorithm. This is going to become a major question among statisticians, how does one adjust for seasonality in the face of tsunami like economic events?
Not seasonally adjusted data can create more headline buzz on a month by month basis due to the seasonality of the housing market. S&P does make it clear that data should be compared to a year ago, to remove seasonal patterns, yet claims monthly percentage changes should use not seasonally adjusted indices and data. This seems more invalid than dealing with the statistical anomalies the massive housing bubble burst caused. Below is the seasonally adjusted and not seasonally adjusted Composite-20 Case-Shiller monthly index.
For more Information:
S&P does a great job of making the Case-Shiller data and details available for further information and analysis on their website.
Here is our overview from last month.