I've been meaning to go over some of the details in last week's Housing data. I was surprised to hear several commentators discuss the imminent turnaround in Housing.
Mind you, these were the people who missed the entire significance of Housing before the collapse, who insisted it was contained, and would not infect anything else. "Why are you so obsessed with Housing?" I was asked repeatedly by this crowd from 2005 thru '07. Perhaps I shouldn't be all that surprised at what the usual crowd of shills and Pollyannas states, given their horrific track record.
Recall that back in March, we noted in passing that both Bloomberg and the WSJ had misinterpreted the Existing Home Sales data. I said bluntly they both "got it wrong." The spin doctors at the NAR scored an unquestionable propaganda victory, with a front page WSJ article that omitted the year over year data.
That led to some back and forth between myself, the reporters, and a senior editor. Our email conversations led me to the inescapable conclusion that many finance journalists simply don't understand statistics, especially seasonality. (See Existing Home Sales, Non Seasonally Adjusted, Explained).
Let's (once again) look at how seasonality affects home sales, using some charts that may help to explain this.
First, note that New home Sales are reported based on purchase contracts (subject to cancellation), while Existing Home Sales (single-family, townhomes, condominiums and co-ops) are recorded at closing (when the actual transaction takes place). This causes a somewhat different form of seasonality between the two.
Why? A few ideas: New Homes are a mix of already finished units, some partially built, some mostly finished but waiting for certificates of occupancy -- and some have yet to even see ground broken. No one knows when a unit will actually be completed. Everyone I know who bought a new home tells stories of delays and closing postponements waiting to finally move in.
Existing homes sales (about 85% of all house purchases) behave with a more distinct seasonality. A few major and minor factors play into to the pattern, but the big ones are 1) school year, 2) weather, and 3) holidays.
What would you expect from seasonal patterns? Well, Winter weather is a less than ideal time to shop for a home, especially the further North you travel. Snow covered houses are not ideal for buyers to view or inspect -- the landscaping and foliage is not visible, flaws may be hidden, roof, vents, A/C not accessible. Then there's the holiday season. From Thanksgiving to Christmas to New Year's, people tend to be busy with other things -- year-end work, holiday shopping, travel, parties, etc.
But the really big one is the school year. As any family with school age children will tell you, its greatly preferred not to disrupt the kids' school year, and smooth the transition to a new neighborhood in a different school district. The ideal time to start shopping is the Spring. That gives you a few months to look around, find a house, negotiate a transaction. The ideal closing date is between May and August -- with the actual move in date at the end of the school year, but before classes start in September.
Looking at the raw data -- Non-Seasonally Adjusted -- that is precisely the pattern we see each year.
Existing Sales (closings) bottom in January, as people ware mostly otherwise occupied in the 4th quarter. Closings pick up steadily from there, rising each subsequent month until transactions reaching a high in the month of June, followed closely by August at 2nd, and and July coming in 3rd for total monthly existing home closings. From August, they fall off the rest of the year, plateauing in November/December (perhaps for tax reasons?) and then bottoming in January.
One would imagine that the seasonal adjustments would remove the effects of this, but there are many different ways to apply these adjustments (using four key factors: Seasonal, Trend, Cyclical, and Error). For example, in June, the non-seasonally adjusted data for New Home sales showed a 3.9% decline, so while the headline was positive, the seasonal adjustment factor was the source of the gains. (New Homes show much less seasonality than existing homes -- they tend to peak in March, then drift somewhat lower thru December). I suspect the adjustments are applied in a neutral numerical way that balances for the annual total, but not the odd pattern that accompanies existing home sales (See BLS and Census methodologies for Seasonal Adjustments at bottom).
Regardless, we can look at the pattern that exists in the non seasonally adjusted data, below:
Existing Home Sales, Non-Seasonally Adjusted
Chart courtesy of Calculated Risk
New Home Sales and Recessions
Chart courtesy of Calculated Risk
Existing Home Sales, Non Seasonally Adjusted, Explained (March 25, 2008)
How Counter-Productive is Realtor Association Spin? (March 25, 2008
Fact Sheet on Seasonal Adjustment in the CPI
Bureau of Labor Statistics
FAQs on Seasonal Adjustments
US Census Department
Graphs: Existing Home Sales
Calculated Risk, July 26, 2008