The August Openings Figure Rests on One Monday, the Jobless Figure on Mid-August

A vacancy total measured on a single Monday

On Tuesday, September 29, 2026, the U.S. Bureau of Labor Statistics releases the Job Openings and Labor Turnover Survey for August. Within minutes the August openings level will be divided by the number of unemployed people, and the quotient will be described as the tightness of the American labor market.

Both figures are real. They are not measured at the same moment, by the same survey, or over the same population. The August openings number will describe positions that stood open on Monday, August 31. The August unemployment number, already published on September 4, described people interviewed about the week running Sunday, August 9 through Saturday, August 15. That is a 16-day gap between the close of the household survey's window and the day the establishment survey takes its photograph.

That gap is not a flaw anyone is hiding. It is written into both methodologies. The Current Population Survey uses what the BLS calls "the survey reference week (usually the week that includes the 12th of the month)," and classifies people as unemployed if they "do not have a job, have actively looked for work in the prior 4 weeks, and are currently available for work." JOLTS, by contrast, records openings that exist on the last business day of the reference month, while its hires and separations figures cover the whole month. One release therefore carries a point-in-time stock and a full-month flow side by side, and the ratio built from it borrows a stock from a different survey taken two weeks earlier.

Three conditions have to hold at the same instant

The looser reading of a job opening is that some employer somewhere would like to hire. The BLS definition is narrower, and it is the narrowness that decides what the printed number contains. The Handbook of Methods requires that "A specific position exists and there is work available for that position," that "The job could start within 30 days, whether or not the employer can find a suitable candidate," and that "The employer is actively recruiting workers from outside the establishment to fill the position."

Read the third condition closely. Recruiting from outside is the test, so a position that an employer intends to fill by internal transfer, promotion, demotion, or recall does not enter the count, even when the work is real and the budget is approved. Neither does a position the employer expects to cover with a staffing agency, a contractor, or a consultant. Neither does a role with a start date more than thirty days out, which quietly removes much of the seasonal and project-based hiring that firms plan a quarter in advance.

The second condition cuts in the opposite direction. Because the job "could start within 30 days, whether or not the employer can find a suitable candidate," an opening that has gone unfilled for eight months still counts every month it sits there. A stale, hard-to-fill requisition contributes exactly what a fresh, easily filled one contributes. The openings level therefore mixes genuine near-term demand with a residue of chronic vacancies, and nothing in the headline distinguishes them.

Active recruiting is defined generously. The BLS lists "advertising in newspapers, on television, or on the radio; posting internet notices, posting 'help wanted' signs, networking, or making word-of-mouth announcements; accepting applications; interviewing candidates; contacting employment agencies; or soliciting employees at job fairs." Merely accepting applications qualifies. That breadth is one reason JOLTS openings and commercial job-board postings can drift apart. The Federal Reserve's own July 2026 Monetary Policy Report notes that "Job openings, as measured in the Job Openings and Labor Turnover Survey (JOLTS), have moved a little higher since late last year," while "an alternative measure of vacancies that uses job postings data produced by Indeed, a large online job board, has flattened out this year following declines in previous years." The two series answer different questions.

The openings rate does not share a denominator with the hires rate

Readers who move past the level to the rates run into a second trap. The three headline rates in the release look like members of one family, and they are not. The BLS computes the hires and separations rates by "dividing the number of separations by employment and multiplying that quotient by 100." The openings rate is computed by "dividing the number of job openings by the sum of employment and job openings and multiplying that quotient by 100."

The openings rate places openings in its own denominator; the hires rate does not. The two cannot be subtracted, stacked, or compared as shares of one base. In July 2026 the openings rate printed at 4.4 percent and the hires rate at 3.2 percent. That 1.2-point spread is not a shortfall of hires against vacancies, because the 4.4 is a share of jobs plus vacancies while the 3.2 is a share of jobs alone. The same two series were read as a gap rather than as rates in Fault Lines Weekly, The Gap Between Job Openings and Actual Hires.

The convention is defensible: it puts openings and filled jobs on a common footing as total labor demand. The Federal Reserve uses the same construction when it discusses "the gap between the total number of available jobs (measured by employed workers plus job openings) and the number of available workers (measured by the size of the labor force)." The problem is only that the release prints the three rates in one column and leaves the reader to notice that one of them is built differently.

June was published at 7,359,000 and now reads 7,182,000

The third structural feature is revision. JOLTS publishes a preliminary estimate, then revises it in the following month's release as late respondents report. May 2026 stands at 7,537,000. On August 4, 2026 the BLS put June job openings at 7,359,000, a rate of 4.4 percent. On September 1, with July data, June was restated at 7,182,000 and a rate of 4.3 percent. The revision removed 177,000 openings from a month that had already been reported, discussed, and traded.

Consider what that did to the widely quoted ratio. The BLS itself published a note reading: "There were 7.1 million unemployed people in June 2026 and, on the last business day, 7.4 million job openings. This yielded a ratio of 1.0 unemployed person per job opening." With the revised openings figure of 7,182,000 against 7,094,000 unemployed, the ratio is 1.01 openings per unemployed person rather than 1.04. Rounded to one decimal, the published ratio is unchanged at 1.0. A 177,000-opening revision left no visible trace in the ratio that circulated most widely.

For July 2026 the openings level stands at 7,271,000 against 6,916,000 unemployed, or 1.05 openings per unemployed person. The month-over-month move from June to July was an increase of 89,000 openings, which is roughly half the size of the revision applied to June. Anyone who built a narrative on the June-to-July direction was reading a difference smaller than the correction that had just been applied to one of its endpoints.

A sample of 21,000 units, benchmarked to a survey that is itself revised

JOLTS draws on roughly 21,000 sampled units, with the Quarterly Census of Employment and Wages as its frame. Establishments stay in the panel for a long stretch: the Handbook of Methods states that "Most sampled establishments remain in the survey for 36 months and, after completing time in sample, are not sampled again for at least 3 years." One panel enters and one leaves each month, so the composition of the sample turns over continuously beneath a series that looks smooth.

The weighted results are then anchored to another survey. The methodology is explicit: "JOLTS estimation cells are benchmarked monthly to the current employment level from the U.S. Bureau of Labor Statistics (BLS) Current Employment Statistics (CES) program." Because CES employment is itself revised, and re-benchmarked annually to the QCEW, a revision to payroll employment propagates into JOLTS levels without any respondent changing an answer, a mechanism traced in Fault Lines Weekly, Benchmark Revisions Land on One Month. On top of that, each annual update recalculates seasonal factors so that "The seasonally adjusted estimates are recalculated for the most-recent 5 years." Five years of history can move on one release day: the January 2026 data, which carried that year's recalculation, went out on March 13, 2026.

Scale is worth keeping in view. The Federal Register notice of September 22, 2025 covering this collection under OMB control number 1220-0170 estimates 6,446 respondents, 77,352 annual responses, and 12,892 burden hours. A series that moves bond yields is assembled from a monthly form collected across a sample of about 21,000 establishments.

Concrete Framework

A practical reading order for the September 29 release, and for the ones after it.

First, separate the stock from the flows. Openings are a single-day count; hires, quits, layoffs and total separations cover the entire month. A month with a strong hires total and a weak openings total is not contradictory. It can describe a month in which vacancies were filled and not replaced.

Second, read the revision line before the headline. Each release restates the prior month. Compare the size of the revision with the size of the month-over-month change you are about to interpret. When the revision is larger, the change carries little information on its own.

Third, refuse to mix the two rates. The openings rate has employment plus openings in its denominator; the hires and separations rates have employment alone. Compare openings rates with openings rates across months, and hires rates with hires rates.

Fourth, date both halves of any tightness ratio. For August, the openings term refers to Monday, August 31 and the unemployment term to the week of August 9 through 15. Write both dates next to the ratio. If a policy argument depends on the ratio crossing 1.0, check whether it survives the rounding that hid June's 177,000-opening revision.

Fifth, treat the exclusions as a known bias with a known direction. Openings filled internally, through agencies, or with start dates beyond thirty days are outside the count. In industries that hire heavily through staffing firms or fill senior roles by promotion, the measured vacancy level understates the positions being staffed.

Where This Doesn't Apply

None of this argues that JOLTS is uninformative. For the questions it was designed to answer, it is the cleanest series available, and several of the cautions above matter less than they appear.

Over long horizons the reference-period mismatch largely washes out. A sixteen-day offset matters when you are reading a single month against a threshold; it matters much less across a multi-year move, because that offset stays the same size while the change being measured does not. Analysts tracking the Beveridge curve over a cycle are on firm ground.

The revision concern is also bounded. The 177,000 restatement of June 2026 is large against a monthly change but small against a level above seven million, and its sign is not fixed by construction: revisions reflect late respondents and updated benchmark employment. Check the revision each month rather than assume a direction for it.

Quits deserve separate handling. The quits rate is a full-month flow with no reference-day problem and a denominator shared with the hires rate, though it carries an exclusion of its own: retirements and transfers to other locations are classified as other separations, not quits. When the Federal Reserve calls quits "an indicator of the availability of attractive job prospects," it is leaning on the cleanest measure in the release. Much of the caution in this piece is specific to the openings line.

What to Watch Next Week

The week beginning Monday, September 28 puts two labor releases within four days of each other, and they are measured on different clocks.

On Tuesday, September 29 at 10:00 a.m. Eastern, JOLTS for August arrives. Three things are worth reading in order. The revision to July's 7,271,000 opens the release; compare its magnitude against the June-to-July change of 89,000 before drawing any trend. The August openings level refers to Monday, August 31 alone. And the quits level, a full-month flow, will say more about worker confidence than the openings level says about employer demand.

On Friday, October 2, the September Employment Situation follows, with a household survey reference week of Sunday, September 6 through Saturday, September 12. August payrolls rose 162,000 against a prior twelve-month average of 31,000, with unemployment at 4.1 percent and 7,031,000 people unemployed. If August openings land near July's 7,271,000, the August ratio against the 7,031,000 already published for that month is about 1.03 openings per unemployed person, below July's 1.05. The September report will not settle that ratio, because its household reference week belongs with September openings, which publish on November 3. That is a number assembled from two surveys, two reference periods, and one estimate that has not yet been revised.

Primary documents for the week: the U.S. Bureau of Labor Statistics Handbook of Methods, Job Openings and Labor Turnover Survey, Concepts; the U.S. Bureau of Labor Statistics Handbook of Methods, JOLTS Calculation; the U.S. Bureau of Labor Statistics Handbook of Methods, JOLTS Design; the U.S. Bureau of Labor Statistics Job Openings and Labor Turnover Summary for July 2026; the U.S. Bureau of Labor Statistics How the Government Measures Unemployment; the Office of the Federal Register notice on Agency Information Collection Activities, Job Openings and Labor Turnover Survey, 90 FR 45413; and the Board of Governors of the Federal Reserve System Monetary Policy Report, July 2026.