You Almost Have to be Unemployed to Have the Time to Understand the Unemployment Numbers

Although the data is showing the percentage unemployed dropping, a trend I have been expecting, when the number came out today, I also warned against over reading too much into any one  number. I wrote:

Keep this drop in unemployment in perspective, part of the drop was a result of of a 504K decline in the labor force. This could be as a result of the storms, retirees, discouraged. I'm still expecting a gradual improvement in unemployment but these numbers can swing pretty wildly. This [report] is going in the direction I expect, but the drop does appear pretty dramatic.

It is much more meaningful to understand a trend over time, than it is to put too much weight on a stand alone number. 

The Slate's Annie Lowery does a great job of explaining what goes in to (and doesn't go in to) the numbers on a monthly basis as they are calculated:

First and foremost, economists never recommend reading too much into any one dispatch. The numbers tend to jump around quite a bit—in the preferred economic parlance, the data are "noisy." That is because the jobs survey is just that—a survey, not an actual report on the status of every American in the labor force. In the winter, the reports tend to be especially noisy, as factors like the weather and seasonal hiring influence the numbers. And especially strange things happen in January.

To understand why, it helps to know that the unemployment rate and the jobs growth figure are determined by two different surveys. To get the unemployment rate, federal workers visit and interview members of about 60,000 households, asking about careers and job searches and decisions not to work. This is called the household survey, or the current population survey. To get the sector-by-sector and job growth numbers, the BLS surveys businesses in something called the establishment survey.

The week businesses reported to the BLS happened to be the week of a massive snowstorm that battered the Midwest and a number of Southern states. About 886,000 Americans stopped heading into work—twice as many as in a normal January. That disturbed the data. The construction and transportation industries slowed during the storm—meaning the data for that week might not be representative of the data for the month. And the stormy weather perhaps also changed job search and hiring patterns. Economists believe this means the job growth number may actually be higher than reported or might tick up significantly next month.
The decline in the unemployment rate, on the other hand, might be exaggerated—and, one way or another, is not easy to parse in the context of December's numbers. For one, the BLS revised its estimates of the unemployment situation in December and November. It now believes that employers added more jobs in those months than previously thought: 121,000 positions in December, up from 103,000; and 93,000 jobs in November, up from 71,000. That perhaps accounts for some of the very steep drop.

Other one-off changes need to be accounted for as well. In January, the Labor Department adjusts its estimates of population size, meaning December and January can't be fairly compared in the same way November and December can. According to the BLS, between December and January, the number of employed workers increased by 117,000, but the number of unemployed workers fell by a massive 622,000. That implies that discouraged jobless workers simply fell out of the labor force—but the BLS says the size of the labor force did not change. In short, the numbers do not add up terribly well, at least not this month.

We'll know more in February and March, when there may be further adjustments to the numbers released today and when month-to-month data become easier to compare.