Something curious happened earlier today when the DOL revealed its latest initial claims number [3]: while the seasonally adjusted print declined by 10,000 to an expectations beating 316K (a change that identically matched what happened to the Seasonally Adjusted print a year ago), the unadjusted number rose by 37,229 to 363K. That's ok: after all that's what "seasonal adjustments" are for - to take a volatile number which historically posts an abnormal jump or drop in any given week and smooth it out, right? Wrong. Because as the DOL also reported a year ago, the supposedly same "seasonal adjustment" applied to the same week in 2012, when the claims number was 390K adjusted and 359K unadjusted, should have been adjusted in the same direction. And while the 390K claims print in 2012 was indeed a 10,000 drop from the prior week's 400K, the unadjusted number instead of being an increase, was actually a drop, one of 44,768 jobs. How does this same "recurring" seasonal adjustment look further back - after all it is seasonal, so there should be some recurring logic for a specific time of the year? The answer is shown on the chart below.
[4]
In other words, the "same" adjustment that in the past was applied to an NSA weekly change that was a greater drop than the seasonally adjusted print, somehow in 2013 ended up having its sign flipped, and made the weekly spike in claims look much better than it actually was. For the exactly same week.
One wonders just what other goalseeking intentions (and directive) the BLS had when it ordered the Arima "adjustment" software to make such a radical departure in this specific week?