Job Openings Are Back at 2019 Levels. The Hiring Rate Is Not.

Two lines in the same monthly survey, published the same morning, have pointed in different directions for most of the past year. In June 2026 the U.S. job openings rate stood at 4.4 percent, indistinguishable from where it sat through most of 2019. The hires rate stood at 3.4 percent, roughly half a point below its 2019 average of about 3.9 percent. Posted demand looks pre-pandemic normal. Realized movement does not.

That divergence is the most informative thing in the Job Openings and Labor Turnover Survey right now, and the part most easily overread. The gap between openings and hires is not a clean measurement of anything: it sets a stock against a flow, it rests on a survey whose response rate has fallen by roughly half since early 2020, and its level has drifted so far across the life of the series that cross-decade comparison is close to meaningless. What follows is the data, the interpretation it can carry, and the conditions under which that interpretation stops working.

What JOLTS Counts, and On Which Day

Most confusion around the gap dissolves once the reference periods are visible. BLS uses three measurement windows inside one release.

• Job openings are counted on the last business day of the month. A position qualifies only if three conditions hold at once: a specific position exists with work available for it; the job could start within 30 days, whether or not a suitable candidate is found in that window; and there is active recruiting from outside the establishment.
• Hires are all additions to the payroll during the entire calendar month, including rehires and recalls after a formal layoff of more than seven days.
• Separations cover everyone leaving the payroll that month, split into quits (voluntary, excluding retirements and transfers), layoffs and discharges (employer-initiated, including suspensions beyond seven days), and other separations (retirement, transfer, death, disability).
• Total employment is measured for the pay period containing the 12th.

The denominators differ too. The job openings rate divides openings by the sum of employment and openings, that is, by all filled and unfilled jobs. The hires, separations, quits and layoffs rates divide by employment alone. The openings rate and the hires rate are therefore not two slices of the same pie, and their difference in percentage points has no clean economic meaning. What is meaningful is how each has moved against its own history.

Timing is the other constraint. JOLTS publishes roughly five weeks after the reference month ends: June 2026 data arrived August 4, 2026, and July 2026 is scheduled for September 1, 2026. By then the payroll survey for that same month is about four weeks old and the following month's payroll report has usually already printed. JOLTS is a texture reading on a period faster data has already described, not news.

The Data

The June 2026 report, the most recent available, reads as follows.

Series (June 2026)	Level	Rate	BLS characterization
Job openings	7.4 million	4.4%	Little changed
Hires	5.3 million	3.4%	Unchanged
Total separations	5.4 million	3.4%	Little changed
Quits	3.2 million	2.0%	Unchanged
Layoffs and discharges	1.8 million	1.1%	Unchanged

Four consecutive "unchanged" or "little changed" verdicts is itself the story. The useful comparison is not month over month but against the last stretch of conditions anyone would call normal.

Rate	2019 monthly average	February 2020	Cycle extreme	June 2026
Job openings	4.5%	4.4%	7.5% (March 2022)	4.4%
Hires	3.9%	3.9%	4.6% (November 2021)	3.4%
Quits	2.3%	2.3%	3.0% (November 2021)	2.0%
Layoffs and discharges	1.2%	1.2%	7.0% (April 2020)	1.1%
Total separations	3.8%	-	-	3.4%

Read down the last two columns. Openings made a full round trip: 4.4 percent pre-pandemic, a record 7.5 percent in March 2022, 4.4 percent now. Hires did not. They run about 0.5 points below the 2019 average and have held a narrow 3.1 to 3.5 percent band across 2025 and the first half of 2026. Quits are about 0.3 points below 2019. Layoffs sit at the low end of the 2019 range, not above it.

What the Gap Can and Cannot Support

As a ratio rather than a difference, the picture is easier to hold. Dividing the hires rate by the openings rate gives roughly 0.85 in 2019 and 0.77 in June 2026 — about nine percent less completed hiring per unit of posted demand. The construct is crude, since the denominators differ, but its direction within one regime is informative.

The supported claim is narrow: employers are posting at a normal rate, converting those postings into payroll additions more slowly than in 2019, and separating from existing staff at a rate that is normal or slightly below. That is a low-churn labor market, not a deteriorating one. The two look alike in a headline and behave very differently.

The quits rate is where the distinction becomes tractable. Quits are the only line in the survey measuring a decision made by the worker rather than the employer, and they read as a confidence signal because leaving voluntarily usually implies an offer in hand or the belief one is obtainable. Research from the Federal Reserve Bank of New York supports treating quits as more than coincident: its work on labor market tightness finds the quits rate, and a composite index built partly on quits, among the best available predictors of wage growth in the following quarter, while the traditional vacancy-to-unemployment ratio "began to falter" as a tightness measure around 2015.

That is a substantive reason to weight the quits line above the openings line. It is not a claim that quits predict recessions. It is a claim about which series carries information on wage pressure over a one-quarter horizon.

The vacancy-to-unemployment ratio frames the aggregate. With 7.4 million openings against 7.09 million unemployed, both June 2026, the ratio sits just above 1.0 — near parity between posted positions and people counted as looking. By July 2026 the unemployment rate was 4.1 percent, unemployed persons 6.92 million, participation 61.4 percent, and nonfarm payrolls -23,000. A ratio near one is neither shortage nor glut. It is where composition, not aggregate scarcity, decides outcomes.

A Prior Episode Where Postings Led and People Did Not Move

The JOLTS series begins in December 2000, so it contains exactly one recovery widely described at the time as jobless. The 2003 to 2004 period is the closest structural analogue inside the survey's own history.

Annual average rate	2003	2004	Change
Job openings	2.4%	2.7%	+0.3 pp (about +12%)
Hires	3.7%	3.8%	+0.2 pp (about +5%)
Quits	1.8%	1.9%	+0.1 pp (about +7%)

Openings recovered faster than either realized flow, and the December 2004 openings rate of 3.0 percent ran well ahead of where worker behavior had moved. The same asymmetry appears in 2009 to 2010: the openings rate rose from roughly 1.9 percent in mid-2009 to about 2.3 percent by late 2010 while the quits rate crawled from 1.3 to 1.5 percent. In both episodes posted intention led realized movement, and the lag was measured in quarters.

One caution attaches to those tables. The level of the 2003 openings rate is not comparable to 2026. The hires-to-openings ratio was roughly 1.5 in 2003 against 0.85 in 2019 and 0.77 now. A series that moves from 1.5 to 0.8 over two decades has a trend running through it, plausibly tied to how much cheaper posting a vacancy became. Direction inside a period is usable; the level across periods is not.

What Would Make This Reading Wrong

Five conditions would invalidate or weaken the reading above. At least two are live.

1. The stock-versus-flow problem is real, not pedantic. Openings are a snapshot on one day; hires accumulate over roughly 21 business days. If average vacancy duration lengthens for reasons unrelated to hiring appetite — slower internal approvals, more interview rounds, requisitions left posted to build a pipeline — the stock inflates while the flow is unchanged, and the gap widens with no change in employer intent. Nothing in the published tables separates these cases.

2. Survey quality has deteriorated materially. The JOLTS collection rate ran near 58 percent in January 2020, fell roughly ten percentage points by April 2020, and stood near 33 percent as of September 2024 with no recovery since. The sample is about 21,000 establishments. A response rate near a third raises the risk that nonresponse correlates with what is being measured: a firm in a hiring freeze has less reason to complete a hiring survey. Single-month moves of a tenth of a point are noise.

3. Revisions are large relative to the signal. Every release restates the prior month. In the June 2026 report, May figures moved -57,000 for openings, +82,000 for hires, +159,000 for total separations. Separately, five years of data are revised each January for updated employment benchmarks and recalculated seasonal factors. A narrative built on two or three months can be revised out of existence.

4. Composition can produce the same aggregate from a different economy. A 3.4 percent hires rate is consistent with broad-based slow hiring and with a few large, high-turnover sectors slowing sharply while the rest is unchanged. In June 2026 openings rose 97,000 in transportation, warehousing and utilities and 39,000 in federal government, while falling 74,000 in wholesale trade and 55,000 in nondurable goods manufacturing — offsetting sector moves inside a headline BLS called little changed.

5. The frame fails immediately if layoffs move. The whole "low churn, not deterioration" reading rests on the layoffs and discharges rate holding near 1.1 percent, a level it has not exceeded by more than a tenth in any month of 2025 or the first half of 2026. Should it hold at the top of the 2019 range for two or three consecutive months while hires stay at 3.4 percent, the correct description shifts from low mobility to a market shedding workers, and this framing becomes the wrong lens.

Risk Factors Worth Pricing

Three asymmetries deserve weight rather than a single point estimate.

The re-entry problem. Where hires run half a point below normal and layoffs run normal, the burden falls on anyone entering or re-entering, since incumbents are protected by the same inertia holding the layoffs rate down. Aggregate rates do not register this; unemployment duration and the unemployment-to-employment flow would.

The fragility of low churn. A market held together by employers declining to fire, rather than by employers hiring, gives a demand shock a direct path to the layoffs line, because the buffer of attrition has already been spent. Conditionally stable, not robust.

Measurement risk runs both ways. At a collection rate near a third, openings could be overstated by stale postings or understated if nonresponse concentrates among firms actively hiring. The published tables cannot settle it.

What to Watch Next Week

The week ahead is dense with second-tier labor releases, one of which matters more than its billing suggests.

• Weekly unemployment insurance claims, Thursdays at 8:30 a.m. ET (U.S. Department of Labor). Initial claims were 209,000 for the week ending August 8, 2026, four-week average 199,000, insured unemployment 1.777 million at a 1.2 percent insured rate. The fastest read on the layoffs line this frame depends on.
• State Employment and Unemployment, July 2026 — August 21, 10:00 a.m. ET. Tests whether the low-hire condition is broad or concentrated geographically.
• Productivity by Industry — August 26, 10:00 a.m. ET. Bears on whether output is sustained on flat headcount, one coherent explanation for normal postings alongside subdued hiring.
• Employment Projections and Worker Displacement — August 27, 10:00 a.m. ET. Displacement data speaks to re-entry outcomes for separated workers.
• Current Employment Statistics preliminary benchmark — August 28, 10:00 a.m. ET. The one to read closely. JOLTS employment is benchmarked monthly to the CES employment level, and every JOLTS rate except the openings rate uses employment as its denominator. A material benchmark revision moves the denominator underneath every series discussed above.
• JOLTS, July 2026 — September 1, 10:00 a.m. ET. The next data point on the gap, roughly five weeks after the reference month closed.

Concrete Framework

A repeatable procedure, in order. Each step checks against a specific number rather than a judgment call.

1. Open the revision paragraph before the headline. If any single restatement exceeds roughly 100,000, treat the prior month as unsettled. May 2026 separations moved 159,000.
2. Record the hires rate first. Compare it to the 2019 average of 3.9 percent and to the trailing three-month average, not to the prior month. Anything inside 3.1 to 3.5 percent is continuation, not news.
3. Split total separations. In June 2026 quits were about 59 percent of separations and layoffs about 33 percent. A flat total with a five-point shift in that mix is a change in kind.
4. Set a layoffs threshold in advance. The 2019 range was 1.1 to 1.3 percent. Two consecutive months above 1.3 percent, with hires at or below 3.4 percent, retires the low-churn interpretation.
5. Compute the hires-to-openings ratio, and compare only within the regime. Roughly 0.85 in 2019, roughly 0.77 now. Track it over four quarters. Do not compare it to the 2003 level of about 1.5.
6. Match reference months when computing the vacancy-to-unemployment ratio. Openings of 7.4 million against 7.09 million unemployed, both June 2026, gives roughly 1.0. Pairing a JOLTS month with a different payroll month is the error the five-week lag invites.
7. Cross-check a weekly series and two industry tables. Claims cover a period JOLTS will not describe for over a month, and stability built from large offsetting sector moves is a different economy from broad-based stability.
8. Discount any single month by construction. At a 33 percent collection rate, a three-month average is the shortest window that should carry an argument.

The honest summary: posted demand has normalized while realized movement has not, the configuration holds so long as the layoffs line stays where it is, and the survey producing the evidence has weakened enough that any conclusion drawn from one month deserves less confidence than the precision of the numbers implies.

Disclaimer: Macroeconomic analysis for general information. Not investment, financial, or employment advice, and no prediction is made about future data releases or market outcomes.