How did we add 117,000 jobs when employment fell by 1.2 million?

I dunno. Ask the people at the Bureau of Labor Statistics and the folks in the media that rip and read BLS press releases.

Following up on Jon Ham’s latest Media Mangle, reporting on monthly jobs reports often varies from abysmal to downright fraudulent. Part of that has to do with the complicated way BLS collects and churns data, and how little they do to help journalists (who often have trouble with math anyway) report that information.

The advantage Jon, I, and our colleagues at JLF share is that we work with Don Carrington, who was the deputy director of the Labor Market Information Division of North Carolina’s Employment Security Commission for five years prior to joining JLF in 1995. So he’s been deciphering these numbers for two decades. Don recently has been sharing his expertise with Carolina Journal readers. And I’ll do my best to share some more with you.

If you go to the latest jobs report (you can do so here), you’ll see from the release:

Total nonfarm payroll employment rose by 117,000 in July, and the unemployment 
rate was little changed at 9.1 percent, the U.S. Bureau of Labor Statistics 
reported today. Job gains occurred in health care, retail trade, manufacturing, 
and mining. Government employment continued to trend down.

According to the seasonally adjusted data, there were 131,190,000 nonfarm employees on the date in July the BLS establishment survey — a survey asking business enterprises and government agencies how many payroll checks they issued — was taken. That’s up from a seasonally adjusted figure of 131,073,000 nonfarm employees on the date in June the BLS establishment survey was taken.

I’m using such mind-numbingly precise language to emphasize the randomness of those figures. “Seasonal adjustment” is an estimate of the number of jobs BLS’ computer model bumps up or knocks down employment data based on the date on the calendar. (Or as Don called it when we talked today, “a guess of a guess.”) For instance, if you assume that employment in  coastal counties will go up in the summer, you make a seasonal adjustment in the expected number of jobs. Likewise, if you anticipate that school teachers will not be in their classrooms when the academic year has ended, you make a separate seasonal adjustment.

Seasonal adjustments are useful to make comparisons from year to year, to look at larger trends in employment, and to calculate unemployment rates, but they are not a precise measure at any time of the number of jobs or the number of workers (which are different, because one person can hold several jobs). Nowhere in that paragraph does the release say that these numbers are based on seasonally adjusted data, and that can be misleading, because the first paragraph is often as far as journalists read when they’re relaying that information.

What do the actual numbers say? Looking at Table B-1 of the report, you can see that the number of not seasonally adjusted nonfarm employees overall fell from 132,151,000 in June to 130,920,000 in July — a drop of 1,231,000 employees. That’s right. 1.2 million fewer people were at their job sites on the date of the survey in July compared with the date of the survey in June. As Bob Stein notes at The Corner, most of that can be explained by the fact that 1.26 million fewer state and local educators were working in July than in June. That’s understandable, as K-12 schools are on summer break and most colleges rely on smaller staffs during summer sessions.

Stein says it’s “misleading” to focus on the decline in payrolls. But the fact is, those 1.26 million people were not at their jobs when the survey was taken.

Meantime, the 154,000 jobs “added” by the private sector is a myth as well. The actual unadjusted count showed a drop, from 110,066,000 in June to 110,062,000 in July — or a loss of 4,000 private sector jobs.

Now, let’s look at the BLS Household Survey, a sample of 60,000 households nationally (1,500 of them in N.C.), during which BLS determines the number of adults in each household and finds out how many were working, how many were unemployed and available for work and seeking jobs, and how many were unemployed and available for work and not seeking jobs.

According to the latest Household Survey (Table A-1), the number of people who were employed went up last month. The civilian work force (not seasonally adjusted — in other words, the actual survey count) went from in June 140,129,000 to 140,384,000 in July. 255,000 more people were working and yet the number of jobs fell by 1.2 million. (By contrast, the seasonally adjusted figures estimated that the work force shrank by some 38,000.)

How do you have more people working but fewer jobs? As Don says, it’s as if he had three jobs, was laid off one and quit a second one that his unemployed neighbor then took … multiplied by 255,000. It’s not credible. But then again, these numbers are no more than estimates. Sophisticated estimates, but  nothing to serve as a basis for policy based on what we can learn from this report. Besides, the BLS report is an estimate based on a survey, and the figures are revised up or down all the time.

What does the current jobs report tell us about the state of the work force or the broader economy? For one thing, relying on seasonally adjusted data would suggest that more people worked last month but fewer jobs were available. Which means that the seasonally adjusted info is useful for analysts examining broader trends, but nothing more than background noise if you’re trying to make a policy statement.

The temptation will be strong to try to distort the estimates for political purposes. So know where to look for the unadjusted (I’d call it real) data, read carefully, and beware.

Rick Henderson / Editor-in-chief, Carolina Journal

Rick Henderson became editor-in-chief of Carolina Journal in August 2016 after being managing editor of Carolina Journal since April 2009. Before that, he had worked more than...

Reader Comments