Courtesy of Mish
Weekly unemployment claims have been all over the map recently. Here are the seasonally-adjusted Weekly Unemployment Claims totals for the last 5 weeks.
Jan 27, 454,000
Jan 20, 403,000
Jan 13, 447,000
Jan 06, 411,000
Dec 30, 388,000
The first three numbers above are from the current report. I calculated the January 6, number. The December 30 number is from the archives.
The reported seasonally-adjusted number on January 6 reporting was 409,000. It was revised up but no one saw that revision.
The reason no one can easily spot revisions is the weekly report only gives the latest 3 weeks. I calculated January 6th number from the 4-week moving average, now reported as 428,750.
A similar calculation looking at the January 20 Weekly Claims Report shows that December 30, was revised up from 388,000 to 391,000. These are small revisions but even large ones would be hard to spot if you do did not do the math or go to the archives.
Computing the Missing Number and Hidden Revisions
The 4-week moving average is constructed from the current 4 weeks. However the report only shows 3 weeks. To compute the week not shown, take the 4-week moving average (SA) and multiply by 4. Subtract the last three weeks shown on the report. What remains is the hidden 4th week used to compute the 4-week moving average.
Moreover, the difference between that number and was was originally reported for that number is a hidden revision.
Gaming the 4-Week Moving Average
If you want to pace a bet on whether the 4-week moving average will rise or fall, you need to know the number to beat and how to calculate it.
The number to beat is the missing number (as described above), about to roll off. In this case, 411,000.
Assuming no revisions, a number higher than 411,000 will cause next week’s 4-week moving average to rise. A number below 411,000 will cause next week’s 4-week moving average to drop.
My guess is the 4-week moving average will rise next week and fall the following week when the January 13 of 447,000 rolls off the report.
Clearly, if you are attempting to predict such numbers, it is critical to look at the number about to roll off.
What’s With The Ping-Pong?
Revisions and hidden numbers aside, inquiring minds are asking about the ping-pong.
What’s happening is most analysts are misinterpreting seasonal data. It is clear that seasonal hiring picked up in October for the Christmas season. This hiring pattern happened one month sooner than normal. Moreover, some of those people were no doubt kept on (not let go in the traditional after-Christmas employee dump).
In my opinion this is a one-time effect, not a mad rush by retailers (or anyone else) to hire people. With that in mind, one should have expected, in advance, to see large drops in seasonal claims.
Previously, I had commented on the possibility that we might see a couple of hot jobs numbers at the beginning of the year. Certainly the February BLS Jobs Report (for January data) might be impacted by atypical Christmas skew that started early. Also the January data itself is subject to BLS twice-annual revisions. Finally, the BLS is changing the way it does revisions next month, going to quarterly revisions.
If you can game all that, good luck.
While the trend in unemployment claims is likely getting better. I rather doubt the numbers are as good as most think.
Certainly yesterday’s reported number of 454,000 may have been impacted by snow, but what about the January 13th number of 447,000?
If this was not messy enough already, next week may again have skew because of snow. Regardless, I do not think we can successfully interpret data for a few more weeks until it is clear that skews caused by changes in Christmas season hiring have ended.
When it does, I believe we will see improvements in weekly claims, but not enough to get overly excited about or enough to cause anyone to believe a big source of jobs is just around the corner.
