The U.S. is adding workers at a strong pace over the past three months. It is also losing workers.
The conflicting employment data come from two different surveys—one of the employers and one of the households—used to calculate employment, unemployment, and other key figures in the Labor Department’s monthly jobs report. The divergence raises questions about the labor market’s overall strength as more signs point to a slowing economy.
The survey of employers shows nonfarm payrolls growing by an average of 375,000 jobs a month over the past three months. The household survey shows the economy losing an average of 116,000 jobs a month during the same span.
Many economists consider the establishment survey more reliable—in part because the household survey has a smaller sample size and a larger margin of error. The two series tend to converge over time. But some are watching the deteriorating household data closely for early signs the labor market is at a turning point.
“While the household survey is much noisier than the establishment survey on a month-to-month basis, it picks up changes in net new firm creation in real-time and therefore often outperforms the establishment survey at cyclical turning points, provided both measures are averaged over several months,” Goldman Sachs economist Jan Hatzius said in a July 11 research note. “This suggests that the still-robust nonfarm payroll prints of recent months probably overstate true job growth.”
Taken alone, a few poor months for household data alongside solid nonfarm payroll gains might not merit such concern. But the recent losing streak corresponds with other data suggesting slowing growth and some cracks in the labor market.
Gross domestic product, a broad measure of economic output, contracted at a 1.6% annual pace in the first quarter of the year and could decline again in the second quarter. Economists surveyed by The Wall Street Journal this month on average estimate that GDP advanced at a mere 0.37% annual rate in the second quarter, down from 2.98% in an April survey.
Job openings remain robust but have come off recent records. Initial claims for unemployment benefits, a proxy for layoffs, hit their lowest in more than a half-century in March but have since trended upward, and hit their highest level since last November in the week ended July 9. The Institute for Supply Management’s survey of purchasing managers at U.S. factories showed employment contract in May and June; ISM’s service-sector index showed employment contracting in three of the past five months, though businesses sometimes complained they weren’t hiring because they couldn’t find qualified applicants.
Anecdotally, more companies are trimming staff, slowing hiring, or sometimes rescinding job offers. Alphabet Inc.’s Google last Tuesday became one of the latest tech companies to brace for a weakening economy when it announced a slowdown in hiring.
Google's new Bay View campus in Mountain View, Calif. The tech giant recently announced a slowdown in hiring.
PHOTO: DAVID PAUL MORRIS/BLOOMBERG NEWSGuy Berger, the principal economist at LinkedIn, said none of the labor-market data is “screaming recession.” But the household survey, jobless claims, job openings, LinkedIn’s in-house data—which shows hiring via the LinkedIn platform in June was down almost 12% from a year earlier—and other indicators all suggest the labor market is slowing down.
“To my mind, I’m a little nervous when people run around saying look at these really good nonfarm payroll numbers. Maybe they overstate the strength a little bit,” Mr. Berger said.
The Labor Department also revises its nonfarm payroll numbers several times, but not the household data. Research on previous recessions and recoveries indicates that initial payroll estimates tend to understate job creation in the early stages of a recovery and overstate it in the early stages of a downturn.
The divergence is partly because the surveys define employment differently. The household survey, for example, includes workers in private households such as nannies or housekeepers, farmworkers, and the self-employed, while the establishment survey doesn’t. Someone with two jobs is counted twice in the payroll survey, and once in the household survey.
When the Labor Department adjusts its household survey to the payroll definition of employment, the picture isn’t quite as bad—employment rose over the past three months, though by a still-meager average of 37,000 a month.
Tracked side by side, though, both sets of household numbers show a sharp drop-off in hiring since March.
“The conclusion I would make is that I wouldn’t have confidence that either measure is telling the full truth. The truth could be somewhere in the middle, though I lean toward the household survey,” said Derek Holt, head of capital markets economics at Scotiabank.