When the Referee Writes His Own Review: BLS, JOLTS, and the Fed

The Bureau of Labor Statistics just released a paper that can be summarized as follows:

we investigated our own data and, after a lengthy internal inquiry, have concluded that we are largely correct.

It is a rare spectacle when a government agency publishes a formal defense brief for itself, complete with charts, footnotes, and the gentle implication that everyone else has been overreacting.

What JOLTS Actually Counts, Beneath the Hype

For two years, the Job Openings and Labor Turnover Survey was the celebrity of US labor statistics. JOLTS, as the survey is charmingly named, is a monthly questionnaire sent to about 21,000 nonfarm establishments that asks employers how many jobs they have open, how many people they hired, how many quit, and how many were shown the door. The official FAQ from BLS spells that out in dry, methodical prose, the way only a federal document can.

During the pandemic recovery, those job openings numbers exploded. From April 2020 through March 2022, the JOLTS series for openings shot upward in a way that thrilled central bankers and terrified anyone who dislikes interest rate hikes. The story was simple: far more vacancies than unemployed workers, which implied a blazing hot labor market that could keep inflation high.

Then the backlash arrived. Journalists wrote about ghost job postings that never seemed to result in actual hiring, including analysis in the business press that accused fake and duplicate online ads of distorting the picture. Policy researchers and market analysts asked whether the JOLTS openings figures were inflated by overcounting, especially at large firms with many locations and corporate level requisitions. If one requisition is open at 200 stores, is that one opening or 200 openings.

Into this mini scandal walked three BLS researchers, Larry Akinyooye, Mark Crankshaw, and Sean McIllece, with their paper in the Monthly Labor Review titled, without irony, “Debunking common misconceptions about the Job Openings and Labor Turnover Survey: BLS research disproves overcounting and other myths in JOLTS data.”

What the BLS Found When It Checked Its Own Homework

The authors directly confront the overcounting narrative. They argue that the steep climb in openings from 2020 to early 2022 was the result of real economic forces: a surge in labor demand as the economy reopened, fewer available jobseekers, and the rise of remote hiring that broadened recruiting pools. In plain English, employers were genuinely scrambling for workers, and the openings data reflects that scramble rather than some methodological hallucination.

They then present what amounts to an internal audit. At the micro level, they flag establishment reports that look suspiciously large and examine them in detail. The punchline is a statistic that BLS understandably repeats with some pride: only about 0.27 percent of reports were identified as potential overcounts, and those reports accounted for roughly 0.31 percent of weighted employment in the sample. Their conclusion is that even if every single one of those were wrong, the effect on national openings estimates would be minimal.

At the macro level, they compare trends in openings reported by single unit establishments to those reported by multi unit firms. If inflated corporate requisitions were massively skewing the data, you would expect multi unit employers to show very different patterns. Instead, the two groups track each other closely, which the authors present as further evidence against a systemic overcounting problem.

They also remind readers of a detail that got lost in the ghost jobs fever: JOLTS does not scrape data from online job boards. It collects information directly from surveyed employers. Whatever problems exist in online listings, they are not mechanically imported into the published JOLTS statistics.

So far, so reassuring. Then, in a section not blessed with a headline, the authors acknowledge that JOLTS response rates have been falling and that this is a concern that deserves further research. In a paper devoted to proving that everything is mostly fine, this is the line that should linger.

When JOLTS Met the Federal Reserve

The BLS paper would be an internal cleanup operation if JOLTS had remained a niche series for labor economists and trivia buffs. It did not. The Federal Reserve elevated job openings to starring status in its story about inflation and “labor market tightness.” The June 2023 Monetary Policy Report, for example, points explicitly to JOLTS openings as evidence that the labor market remained very tight relative to available workers.

That tightness narrative helped justify higher interest rates and a willingness to keep them elevated while inflation cooled. In that framework, high vacancy rates implied firms were still bidding aggressively for scarce workers, which in turn threatened to keep wage growth and prices uncomfortably high.

Not everyone bought the vacancy story. Economists at the Federal Reserve Bank of Minneapolis took a hard look at vacancy measures and hiring outcomes and found something more sobering: fewer openings, and a hiring process that seemed harder for workers, led them to conclude that the US labor market was likely softer than headline indicators suggested. In other words, if you relied on the shiny openings numbers alone, you risked overestimating how much heat was left in the labor market.

Outside the Federal Reserve System, Congress and oversight agencies circled the same issue from different angles. The Congressional Research Service produced explainers walking through what JOLTS does and does not measure and why lawmakers suddenly cared about a survey that had been running quietly since 2000. The broader debate also drew on Indeed and other online job-posting data as economists explored whether private vacancy measures could supplement or cross-check the official survey. That question entered the congressional record when Reuters reporting on the divergence between JOLTS and private measures was submitted for inclusion in a House Agriculture Committee hearing by Committee Chairman Rep. Glenn Thompson of Pennsylvania.

The Government Accountability Office, asked to take a broader look at federal economic statistics, concluded that core measures generally meet user needs but also highlighted opportunities to improve data quality and coordination. Translation: reliable enough to steer policy, but do not confuse the dashboard with reality.

Ghosts, Divergences, and the Comfort of Clean Narratives

The public fixation on ghost jobs speaks to a broader discomfort. When Forbes publishes pieces arguing that fake or stale postings might distort perceptions of demand, it is channeling the suspicion that the labor market story was too neat. A flood of online listings, corporate performative hiring, and pro forma posts created the impression of abundance even as job seekers told very different stories.

Experimental indicators have not always backed up the official narrative either. The Conference Board, for instance, has maintained experimental vacancy nowcasts that at times diverge from JOLTS readings. That does not prove JOLTS is wrong, but it does remind us that different measurement tools can paint different pictures of the same labor market.

Even within the federal data machine, the message is more nuanced than the BLS headline suggests. The GAO calls for quality improvements. Regional Fed researchers warn that vacancy data may exaggerate tightness. CRS patiently reminds lawmakers that every economic indicator has limits. Then BLS publishes a paper that tells the world: relax, the overcount is about a quarter of a percent and hardly moves the needle.

Who Watches the Watchers of the Labor Market

The 0.27 percent figure is not trivial. It is genuinely useful to know that a microscopic slice of employer reports appears to drive very little of the national openings total. If the critique was that JOLTS was riddled with phantom vacancies invented by clerical error, the BLS researchers have assembled a solid rebuttal.

But the most consequential sentence in their paper is the one about declining response rates and the need for more research. That issue goes straight to the heart of statistical legitimacy. If fewer employers answer the phone or the email, and if the ones who do differ in meaningful ways from those who ignore the survey, then the problem is not inflated corporate counts. It is that the sample may drift away from the real economy while everyone keeps treating the top line numbers as gospel.

Meanwhile, the Federal Reserve has already used JOLTS openings as a key input in deciding how hard to lean on the economy. Congress has already held hearings that treat these statistics as largely authoritative. Job seekers have already lived through a cycle where advertised abundance coexisted with stubbornly difficult searches.

So yes, it matters that BLS finds only a 0.27 percent overcount. It also matters that, beneath the triumphant title, the same agency quietly flags a basic vulnerability in the survey on which a great deal of policy now rests. The data guardians have graded their own exam and awarded a reassuring score, while noting in the margin that the testing room is getting emptier.

The open question is not whether JOLTS is fraudulent. The open question is who, exactly, will keep interrogating the numbers the next time a single jobs series becomes the hinge for trillions of dollars in economic decisions. When the referee writes his own review, the least the rest of us can do is read the footnotes.

---