In the Nation's Interest

A Puzzle in the Employment Report

By Joseph Minarik

This is just a brief thought about this morning’s employment report.  As you may have noted, the headlines lean negative, but the numbers are not conclusive.  We don’t have a decline in the main employment number, but the increase appears lethargic.  Is the economy lumbering along just above stall speed, in danger of losing momentum?

With somewhat greater sophistication than decades ago, observers are focusing more of their attention on the survey of employers – which is the one that yields the headline “jobs created” number – and less on the survey of households, which yields the formerly front-and-center unemployment rate.  The employer survey is larger and generally more reliable than the survey of households.  Employers answer their survey seriously; many households don’t, or fail truly to understand the concepts involved.  So the employer survey’s number of only 113 thousand new jobs – barely enough to accommodate all anticipated new entrants into the labor force – suggests that the economic recovery could be stumbling.  And there are no particular spitballs in the industry mix that would suggest that the bottom line of new jobs would be misleading.

But there is dissonance from the household survey.  It shows a robust 638 thousand new jobs, and a substantial 523 thousand new entrants to the labor force.  (Recall that this survey has been showing the labor force as shrinking in recent months because discouraged workers have ceased looking for jobs; this reversal suggests that workers might perceive better prospects.)  How can the two surveys show such different pictures?

Well, there is always sampling error – either a rogue sample, or poor responses.  Any survey is subject to error, though the household survey is generally more vulnerable, and reflex would be to reject its findings.

But there is also the possibility of one type of systematic error pushing in the other direction.  The employer survey cannot keep up with new business formation in real time.  As a result, the BLS surveys a fixed population of businesses, and estimates the number of new jobs being created in new businesses.  The numbers are trued up later in a benchmark revision.

In contrast, the household survey (assuming an accurate sample and consistent accuracy of responses) would pick up employees of new businesses every month.  So despite its inherent weaknesses, the household survey can provide an early notice of one channel for employment growth.

I don’t know that this explanation will prove to be true.  It is more common than not that one month’s employment report leaves economists scratching their heads in wait for the next one.  But it is a possible explanation of the differences between the two surveys, and a suggestion that the instant headlines may be a bit too pessimistic.

STAY UP TO DATE