>I wasn't clear. I believe that they were alluding to the revisions
>that occur in the series: "Today's data should remind observers that
>seasonal distortions and errors in measurement can drive the bulk of
>the month-to-month variation in employment." They think that Easter
>messed up some earlier seasonally-adjusted indicators, including
>jobless claims. Many trends are looking better - for now.
Ah. The seasonals on the jobless claims series are amazingly volatile. That doesn't mean, as idiots like John Crudele like to say, that the gov is cooking the books with them - seasonal adjustment is essential to strip out weird calendar and other timing effects. But it's really hard to adjust a weekly series accurately. Here's what the adjustment factors for the last few months look like <http://www.workforcesecurity.doleta.gov/unemploy/claims.asp>. To test their effect, imagine a constant 300,000 unadjusted figure, apply the seasonal adjustment factor, and see what changes in the adjusted series result. As the table below shows, they're big - swings of 40,000-50,000 from week to week:
unadj factor adj change
1/29/05 300,000 113.0 265,487
2/5/05 300,000 117.4 255,537 -9,950
2/12/05 300,000 100.3 299,103 43,566
2/19/05 300,000 98.3 305,188 6,086
2/26/05 300,000 91.9 326,442 21,254
3/5/05 300,000 99.1 302,725 -23,717
3/12/05 300,000 93.4 321,199 18,475
3/19/05 300,000 88.1 340,522 19,323
3/26/05 300,000 82.8 362,319 21,797
4/2/05 300,000 86.6 346,420 -15,899
4/9/05 300,000 102.3 293,255 -53,165
4/16/05 300,000 95.6 313,808 20,552
4/23/05 300,000 93.5 320,856 7,048
4/30/05 300,000 86.6 346,420 25,565
Doug