Two May Jobs Surveys Show a Rare Coincidence

Traditionally, the government's seasonally adjusted payrolls estimate — based on its "establishment survey" of 400,000 work sites — requires further seasoning with more than a few grains of salt and comparison with the separate household survey of 60,000 "responsible persons," which is used to calculate the unemployment rate.

The two surveys are often at odds about employment trends. Earlier in the decade, the household survey implied the creation of hundreds of thousands more jobs than showed up in the establishment survey of payrolls for a given month, inviting speculation that the latter was overlooking an army of newly self-employed consultants. And indeed, annual benchmark revisions to the payrolls data subsequently suggested this might have been the case.

In contrast, April's miserly payroll gain (downgraded today to just 80,000) looked downright perky next to a loss of 468,000 jobs reported by all those responsible householders. This had sparked concern that payrolls were underestimating the extent of the springtime economic slowdown, by overestimating the number of jobs created by new businesses not yet reporting to the government.

Fast forward to today, when the establishment survey showed a net gain of 157,000 jobs in May, some 25,000 above expectations. And how many additional jobs did the separate household survey pick up? That's right: 157,000.

Statisticians calculate confidence intervals and sampling errors to gauge the odds that their stats don't match reality. But you don't need a slide rule to figure out that two separate surveys arriving at exactly the same figure improve the chances that the figure is in the right ballpark.

One could quibble, of course, by noting that, without a seasonal adjustment, nonfarm payrolls would actually have increased by a whopping 880,000 in May, counting the 203,000-job "net birth/death adjustment" for all the new businesses the government guesses have recently cropped up and old ones it surmises have gone out of business. Or one could be a killjoy and note, as the Bureau of Labor Statistics did, that monthly payroll gains are averaging 133,000 this year, down from 189,000 per month in 2006.

"That's very consistent with a 1% GDP," says Barry Ritholtz, skeptic-in-chief at Ritholtz Research. "You need a 150,000 jobs a month just to keep up with population growth. In other words, this is essentially no job creation, relative to the population increase.

"With something as big and complicated as the U.S. economy and 143 million employees, you never want to take a single data point. What you want is the trend, the average of the past three months, the past six months. And what we've seen so far this year is that job creation is slower than it was last year."

Ritholtz also takes umbrage at the government's assertion that new construction outfits it doesn't yet know about added 40,000 jobs in May, net of those lost when builders went out of business. "There is a planet somewhere in the Star Wars/Star Trek solar system that added 40,000 construction jobs this month," he says. "But it's not the United States on Planet Earth. Look at what the home builders are saying. Look at what the commercial construction companies are saying. They're not hiring — they're laying off left and right. It's basically a statistical fabrication." The BLS, it should be noted, reported an overall gain of just 7,777 construction payrolls for May, indicating job losses at known construction firms.

The truth is out there somewhere. And shards of it are no doubt embedded in the latest jobs report. But the truth is we probably don't know much more about the future than we did before the data came out. At best, we can be somewhat reassured about the recent past. On that score, mission accomplished.