by Calculated Risk on 10/26/2006 04:09:00 PM

UPDATE: Dirk van Dijk says it was one more Saturday: see Zacks Equity Research Analyst Blog

In discussing the NAR Existing Home Sales report, I noted that the SA factor changed as compared to 2005. Several people have asked me about this - so let me explain. (SAAR: Seasonally Adjusted Annual Rate, NSA: Not Seasonally Adjusted)

First, here is the NAR data (NAR excel file).

Sept 2005:
NSA 630,000
SAAR 7,200,000
Factor: 11.43 (divide 7,200,000 by 630,000)

Sept 2006:
NSA 527,000
SAAR 6,180,000
Factor: 11.73

SAAR decline: 14.2%
NSA decline: 16.3%

I'm pretty sure NAR just uses a computer program to calculate the SAAR number. They plug in the raw data and it gives them the answer. It isn't anything nefarious. So why did the factor change?

First, the idea behind presenting the SAAR number is to make it is easy to compare between different periods. For example, sales are seasonally slow in January and February, so the adjustment to the NSA data is much larger than say for July and August. So the factor could be significantly different month to month, but all else being equal, the factor would be the same for the same month each year.

However some things are different. A common method to calculate SAAR from monthly data uses three components: seasonal, trend, and irregular (like weather, holidays or number of weekends in a month). If the trend stays the same, and there are no irregular component adjustments, then the seasonal adjustment would be the same.

I don't know if NAR uses any irregular adjustments. I don't think they adjust for weather, but they might be adjusting for the number of weekends during the sales period. But one thing is clear: the trend changed, and that impacts the SAAR number.

I have no idea what the NAR program is using for the trend, but during periods when the trend clearly changes, I prefer to use the NSA data to compare to previous years.