The US unemployment rate dropped by a quite large 0.4% points in November, from the previous 9.0% to 8.6%, according to the December 2 release of the US Bureau of Labor Statistics.   But the drop did not generate much excitement among analysts.  Why?

While at first glance, such a drop would generally be viewed as quite positive, an analysis of the underlying numbers suggests caution.  The estimated number of unemployed fell by 594,000, to a total of 13,303,000.  A fall of 594,000 is major.  However, the estimated increase in the number of employed was only 278,000.  Why the difference?  The reason was an estimated reduction in the size of the civilian labor force of 315,000.  Hence only a bit less than half (47%) of the fall in unemployment came from more workers being employed.  A bit more than half (53%) came from workers leaving the labor force.

Workers may leave the labor force for many reasons, including retirement, but also because they may be unemployed and see little prospect in getting a reasonable job.  Furthermore, the labor force will normally grow (rather than contract) over time as the population grows, but month to month it can vary.  In recent years it has been unusually flat or even falling in the face of high unemployment and hence poor job prospects for many.  Older baby boomers may also be dropping out of the labor force and taking early retirement.

In addition, it is not clear that the estimated increase in the number of employed, of 278,000 in the month, is accurate.  In the same December 2 release, the BLS stated that the number of employed in the month was only 120,000.  Why the difference?  The reason is that both are estimates obtained from surveys, and the figures come from two different surveys.  The 278,000 figure that accompanied the unemployment estimates comes from a survey of households.  The 120,000 figure comes from a survey of establishments (generally firms, but including governmental entities and other employers).

Over time the figures will on average be close to each other, but they can differ in any given month.  Coming from surveys, they are both subject to sampling errors, where it should be kept in mind that both of these figures are the relatively small differences between two estimates (for October and November) of the much larger number of total workers employed in that period.  Furthermore, the timing of the surveys within the month will differ.  Finally, there are some definitional differences, although such differences probably do not affect too much the estimates of the change in employment levels between the two periods.  The household survey asks whether the person has been employed in any occupation during the period.  The establishment survey surveys employers of non-farm workers.  Hence the household survey will include (and the establishment survey will not) those employed on farms, the self-employed, unpaid family workers, private household workers, and those on unpaid leave from work.  The establishment survey, on the other hand, will count each worker, and hence a worker with two jobs will be counted twice.  There are some other small differences as well.

The difference between the two estimates, of 278,000 more employed vs. 120,000 more employed, is large.  The estimates are also preliminary and hence subject to change.  Observers will therefore be looking at the trends over the next few months to see which estimate was likely the more accurate.