The Bureau of Labor Statistics publishes the monthly Employment Situation release at 8:30 ET on the first Friday of the month, and the desk reads the release as three prints under a single timestamp: the initial nonfarm payroll headline for the reference month, the second-close revision for the month two prior, and the third-close revision for the month three prior. The initial print lands with a first-close establishment-survey (Current Employment Statistics, CES) response rate that has held near 65 percent through 2025 and 2026, well below the roughly 75 percent range it ran at through the 2015 to 2019 window. The two-month revision incorporates response rates closer to 93 percent. That gap is the entire reason the strip has spent the last twelve months pricing the back-revision as the higher-signal line inside the release.

The response-rate arithmetic

The CES first-close cutoff for the initial monthly release runs roughly two weeks after the reference-period pay period ends. Establishments that report inside that window generate the initial headline. Establishments that report inside the following two months (the second and third closes) generate the revised figures. BLS publishes the first-, second-, and third-close CES collection rates monthly on its response-rate table, and the pattern through the last twelve monthly cycles reads consistently: first-close rates in the 60 to 68 percent band, second-close rates in the 90 to 93 percent band, third-close rates in the 93 to 95 percent band.

The gap has widened since the pandemic. Pre-2020, the first-close rate held in the 70 to 78 percent band and the second-close rate in the 92 to 94 percent band, meaning the initial-to-second-close response-rate gain ran roughly 15 to 20 percentage points. The 2025 to 2026 window has run the gain closer to 25 to 30 percentage points. The larger the gain, the more information the second close adds relative to the first, and the more the revised figure differs from the initial in expectation.

The direction of the average revision has run negative through the entire post-2022 window. Across the last twelve monthly releases, the two-month revision has averaged roughly negative 45 thousand per month across the second- and third-close rounds combined. Roughly 60 percent of releases have carried negative combined revisions, roughly 25 percent have carried positive revisions, and the remainder have landed inside plus or minus 10 thousand combined. The distribution has a left skew: the largest negative combined revisions have run past negative 100 thousand, while the largest positive combined revisions have topped out closer to plus 60 thousand.

Why the systematic downward tilt

Two structural drivers carry the downward tilt in the revision distribution.

The first is the birth-death model. BLS estimates the net contribution of establishment births and deaths (openings and closings) to monthly nonfarm employment change through a statistical model, because new firms cannot be surveyed on the initial release timeline and closed firms cannot report a zero. The birth-death adjustment feeds directly into the initial headline. When the actual pace of firm openings runs behind the model’s central estimate, the initial headline overstates the true change and the second-close revision walks it down. Through 2025 and 2026, small-business formation data from the Census Bureau’s Business Formation Statistics has run below the pace the birth-death model has priced, which has pulled the systematic revision component negative.

The second driver is response-rate composition. Larger establishments respond first at higher rates than smaller establishments. When smaller establishments come in at the second and third closes, they carry more of the small-business labor-market signal than the initial sample does. Through periods when small-business hiring is cooling faster than large-firm hiring (the pattern the June 2026 NFIB employment component at negative 3 net percent flagged), the second-close incorporation of the smaller-establishment responses pulls the revised figure below the initial figure.

What the desk carries out of the arithmetic

The desk carries three read-throughs into the 8:30 ET stamp.

The first: the initial headline is a lower-signal noisy point estimate against the true underlying labor-market state, and the desk applies a wider two-sided band around the initial print than the historical band the coverage flow prices against. A 30 thousand miss on the initial headline against consensus carries less repricing weight than an equivalent miss on the second-close revision, because the revision is closer to the true state.

The second: the combined two-month revision at the 8:30 ET stamp carries as much or more front-end repricing weight as the initial headline itself. A May two-month revision that walks a May initial 154 thousand print down through 100 thousand carries the labor-side softening trajectory the September cut probability at 99 percent is pricing, independent of whatever the July initial prints. A June one-month revision that walks the June initial 139 thousand print down through 90 thousand walks the December second cut probability from its Thursday-close 68 percent line back through 75 percent, again independent of the July print.

The third: the year-ahead SR3 strip carries an implicit trend correction against the initial-headline series. If the market simply took the initial headlines at face value, the strip would systematically over-price the labor-market state relative to the revised series, because the revisions have run negative. The desk correction runs the trailing three-month average of the revised headline (currently at roughly 147 thousand through June) as the primary trend read, not the sequence of initial prints.

What to watch inside today’s stamp

Today’s 8:30 ET stamp publishes the July initial print, the May two-month revision (from the initial 154 thousand), and the June one-month revision (from the initial 139 thousand). The strip has priced a base case of a modest negative combined revision consistent with the trailing twelve-month pattern, roughly negative 30 to negative 60 thousand combined. A combined revision inside that band reads as quiet news for the strip regardless of what the initial July headline prints. A combined revision more negative than negative 80 thousand walks the December second cut probability inside the 75 percent band without the initial July headline needing to move. A combined revision positive on both months carries more repricing weight in the hawkish direction than a firm July initial headline would carry on its own, because it breaks the twelve-month pattern the strip has priced into the back-revision as the trend read.

The initial July headline lands under the same timestamp and reads against the plus 100 to plus 150 thousand band and the 4.1 to 4.2 percent unemployment-rate band and the plus 0.2 to plus 0.3 percent average-hourly-earnings band. But the higher-signal number inside the 8:30 ET stamp is not the July headline. It is the May two-month revision.

Sources