Understanding Seasonal Adjustments

Following my rant about the putzes at the NAR, a few people asked me to better explain the Seasonality Adjustment issues.

Here goes nuthin:

I certainly understand that we have to do seasonal adjustments. One cannot report that Retail Sales fell 80% in January (for obvious reasons) but most of all, because to do so would be misleading. The sources of data report information to inform the public, media repeats what is said, and we pass along interpretations to make things clearer, to get at an objective truth.

The NAR does the opposite.

Let’s look at the specifics of the adjustments this year and see where they went awry.

Whenever we have an outlier year — like Sept 2009 — then we know that seasonally adjusted results will be utterly misleading. That is an issue when we seasonally adjust, as every statistician, economist and number cruncher is well aware.  An honest broker of information recognizes that, and reports it the data in a way that is not misleading.

The NAR is no such honest broker (pun intended).

Most people are unfamiliar with what goes into the methodology of Seasonal adjustments, and how they are performed. When people misunderstand statistical methods, it allows folks like the NAR to make major misrepresentations, and get away with their misrepresentations. It is incumbent on the people who are “numerate” — who understand mathematics — to explain it.

There is a mathematical assumption in SA that the annual seasonal changes will occur around the same time each year. There is also a presumption that the month-to-month changes will be approximately equal, or at least of similar magnitude, from season to season. This forms the baseline for the seasonal adjustment.

Hence, when we are discussing EHS, the prior years’ monthly August-to-September drops are the basis for making the newest seasonality adjustment.

As Rex pointed out, the past decade of August to September EHS changes were:

1999: -19%
2000 -17.7%
2001 -26%
2002 -17.1%
2003 -12.5%
2004 -15.5%
2005 -15.2%
2006 -19.2
2007 -28.9%
2008 -10%

This range was 10% to 28.9%. That averages to 17.2% in the typical September. This is the key element in impacting any subsequent seasonal adjustment (different SA methodologies may use differing time periods).

This year, the fall was 5.3%. Hmmm, that was highly aberrational — I wonder why?  We (and the NAR) know the reason: Due to ZIRP and the soon to be expiring 1st time home buyers $8,000 Tax credit, the drop was minor – much smaller than it usually is when we go from August to September in EHS.

The tax credit very likely extended the selling season by at least a month. It pulled some sales forward, and perhaps created other sales where there might not have been.

But the seasonal adjustment does not know that; The math PRESUMES THE AUGUST/SEPTEMBER DECLINE IS OF TYPICAL MAGNITUDE OF THE PRIOR 10 YEARS.

That creates a misleading — lets even say false — appearance when the seasonally adjustments are performed.

Again, someone trying NOT to mislead will inform the reader of that directly. But calling it a SURGE?  Only if you are innumerate — or a liar. Any honest statistician who worked on these numbers KNOWS that the seasonal adjustment was going to create a big bump, a misleading number, based on the historical data.

And thats the whole point. The NAR knows that calling this a surge will mislead readers, but they report the data — DOWN 5.3% — as a  “SURGE.”  What else might their goal be BUT TO MISLEAD THE PUBLIC?

I refuse to facilitate that. And I will call anyone an unprofessional liar, a distorter of the data who claims this was surge. THIS MEANS YOU, NAR !

The folks who are unfamiliar with seasonal adjustments will get caught in the scam. This was not an ordinary seasonal adjustment — it was highly misrepresentative.

I know better. And now, you know better. Unfortunately, most folks do not.