Labor data is flashing warning signs

• Initial claims for unemployment benefits ticked up last week, while continuing claims for benefits jumped to their highest level since February — suggesting that people who are out of work are taking longer to find a job.

Why it matters: It could be a sign that beneath the surface of a job market that appears exceptionally strong, things might be getting bumpier.

The details: The monthly jobs numbers are based on a survey of about 131,000 employers — asking how many people are on their payrolls, how much they were paid, and so forth.

• The unemployment rate and related data are calculated based on a survey of about 60,000 households, asking ordinary Americans whether they are working, or looking for work.
• Both sets of numbers bounce around from month to month, but over time you would expect to see the same trends from both. If employers say they are hiring a lot more, you also expect more people to report that they are employed.

By the numbers: That isn't what has happened since March. Employers report having added 2.7 million jobs in that span, but only 12,000 more people in the household survey report being employed.

• Indeed, that category fell each of the last two months, and four of the last eight.

Context: There are many potential reasons for this divergence. One is simply sampling error. But there are also important differences in who counts in the two surveys.

• A person who works multiple jobs would count as employed only once in the household survey, but multiple times in the establishment survey, so the gap could be explained in part by more people taking on a second or third job.

• People who are self-employed, work in agriculture, are on unpaid leave, or have informal or off-the-books employment arrangements don't count in the establishment survey — but do in the household survey.

Between the lines: The gloomier picture of job growth shown by the household survey may be an early warning sign about a softening labor market, argues Kathryn Anne Edwards, an economist at the RAND Corp.

• "The people not captured evenly between the two surveys tend to be marginal workers," Edwards tells Axios. "They are the people on the fringe of employment who we would expect to be affected by a downturn first."
• There was a similar divergence in 2007. The household survey showed 300,000 jobs added that year, versus 1.1 million reported in the establishment survey. A recession began that December.

The bottom line: "It's one of many potential warning signs of a recession," Edwards said.