Unemployment Claims: Big Drop to 352K SA (15% Year-Over-Year) Predominantly Due to Seasonal Factor Change; NSA 521 K (Only a 5% Drop); See Updates
SEASONALLY ADJUSTED DATA
In the week ending January 14, the advance figure for seasonally adjusted initial claims was 352,000, a decrease of 50,000 from the previous week’s revised figure of 402,000. The 4-week moving average was 379,000, a decrease of 3,500 from the previous week’s revised average of 382,500. …
The advance number of actual initial claims under state programs, unadjusted, totaled 521,613 in the week ending January 14, a decrease of 124,606 from the previous week. There were 549,688 initial claims in the comparable week in 2011.
While it would be really easy to say, “Yeah, that’s great,” someone is going to have to explain to me how a 5% drop in actual claims from the same week a year ago (522K vs. 550K) generates a 15% drop after seasonal adjustment (352K vs. 415K).
The seasonal adjustment factor for the same week a year ago was 132.5 (i.e., 550K divided by 132.5 equals 415K seasonally adjusted).
This year’s factor was 148.2 (i.e., 521,613 divided by the seasonally adjusted result of 352,000). Readers can find the factor used by going to the interactive tool here.
If last year’s factor had been used on this year’s raw number, seasonally adjusted claims would have been 394,000 (521,613 divided by 132.5), which is not much better than last week’s (revised upward, as usual) 402,000.
Once again, we see a lesson in the value of looking at the raw numbers — and we see that things are not better to anywhere hear the extent that the seasonally adjusted number would make us believe.
It will be interesting to see how the press plays this.
UPDATE: I have sent DOL an email requesting an explanation for the large seasonal factor change.
UPDATE 2: Thanks to Scott Gibbons at DOL for the quick response, which is as follows –
You are focusing on a single week, but this can be misleading, as holidays can fall in different weeks in different years. In addition, some years have their high point in the first week of January and some years have the high point in the second week in January. It’s more useful to look at that part of the annual clams cycle where we expect the peak in claims.
I get what Scott is saying, but thinking in terms of real-world meaning, both weeks involved in the 2010 and 2011 comparison represented the five days before the Martin Luther King holiday, and both represented the first full week of work (except for Alabama and LSU fans :–>) following the Christmas season. The stat guys may be able to justify the calculations of the seasonal factors, but the real-world drop of only 5% in raw claims from 2010 to 2011 indicates that this was a “lucky” week for the seasonally adjusted unemployment claims stat. Though I’d like to be wrong, today’s number of 352,000 doesn’t look like a level that will last.