Nonfarm Payrolls vs Unemployment Rate

Nonfarm Payrolls is published from the Current Employment Statistics (CES) program, a BLS survey of roughly 122,000 establishments covering 666,000 individual worksites. It counts payroll jobs at participating businesses and is a flow measure that captures hiring net of separations during the reference pay period containing the 12th of the month. A worker holding two jobs is counted twice. The Unemployment Rate is published from the Current Population Survey (CPS), a monthly Census Bureau survey of roughly 60,000 households. It counts people, not jobs, and a worker holding two jobs is counted once.

The two surveys can produce sharply different pictures. March 2026 is the cleanest recent example: CES showed payrolls up 178,000 against the Dow Jones consensus of 59,000, while CPS showed civilian employment down 64,000 and the labor force down 396,000. The unemployment rate fell from 4.4% to 4.3% mechanically because the labor-force denominator shrank faster than the unemployed numerator. Labor force participation hit 61.9% in March 2026, the lowest since November 2021, and the employment-to-population ratio held at 59.2%, both signaling that the apparent improvement in the unemployment rate did not reflect job creation in the household survey at all.

The breakeven payroll rate, defined as the pace of job creation needed to keep the unemployment rate flat given labor-force growth, has fallen sharply across the 2024-2026 window. In the 2010s the breakeven sat at roughly 100 to 120 thousand per month, reflecting trend population growth and stable participation. In 2022-2023 it rose to roughly 150 thousand because immigration boosted labor-force growth. By early 2026 the St. Louis Fed estimated breakeven had fallen to 15 to 50 thousand per month, reflecting tighter immigration policy under the second Trump administration, demographic aging, and falling participation among prime-age men.

The collapsed breakeven changes the interpretation of every payroll print. The March 2026 reading of 178,000 is unambiguously above breakeven and is consistent with falling unemployment, which is what happened. The February 2026 reading of -133,000 was the first negative print since December 2020, but with breakeven near 25,000 it represents a much larger underrun than the same number would have meant in 2019. Three-month average payroll growth at roughly 60,000 in March 2026 is therefore mid-range relative to the new breakeven, not weak in absolute historical terms but flatter than the 2022-2024 expansion pace.

On August 21, 2024 the BLS released its preliminary annual benchmark revision showing payroll employment overstated by 818,000 jobs through March 2024. The revision was based on the Quarterly Census of Employment and Wages (QCEW), which uses state unemployment insurance records covering more than 95% of US workers, against the CES sample of just 122,000 establishments. Over the April 2023 to March 2024 window, CES showed 1.9% employment growth while QCEW showed 1.3%, a 60-basis-point gap that translated into the 818,000-job overstatement.

The final benchmark issued in February 2025 reduced the revision to roughly 598,000, still the largest downward correction since 2009. The episode mattered for two reasons. First, it established that the strong 2023-2024 payroll prints were partially overstated, which affected retrospective interpretation of the Fed's 2024 easing decisions. Second, it introduced a wider error band on monthly CES prints: the implicit revision uncertainty per month over that window was roughly 50,000 to 70,000, larger than typical. The 2025 preliminary benchmark released August 21, 2025 was smaller at 130,000 jobs, suggesting the divergence between CES and QCEW had narrowed. Practitioners now check the QCEW release each quarter as a more reliable cross-check on CES, with a 6-month lag.

The Sahm Rule, formalized by economist Claudia Sahm in 2019, triggers a recession signal when the 3-month moving average of the unemployment rate rises 0.5 percentage points or more above its trailing 12-month minimum. From 1970 through the COVID era, the rule had identified every US recession with one false positive. The trailing 12-month minimum unemployment rate ahead of the current cycle was 3.6% in mid-2023; the 3-month moving average reached 4.1% by July 2024, triggering the rule. As of April 2026 the rate has run between 4.1% and 4.4%, persistently above the trigger threshold, with no NBER-declared recession.

The 21-plus months since trigger without recession is the longest false-positive in the rule's 54-year history. Sahm herself has flagged in public commentary that the 2024-2026 trigger is likely a structural false positive because the unemployment rise was driven by labor-force expansion (immigration plus re-entry, roughly 3 to 4 million workers added 2022-2024) rather than job destruction. The mechanism: when the denominator (labor force) expands faster than employment can absorb new entrants, unemployment rises without layoffs. This is materially different from the historical pattern where the rule triggered. The April 2026 question, with payrolls returning to growth and the labor force now contracting, is whether the rule has been redeemed by avoidance or whether a delayed recession is still pending.

March 2026 payroll growth of 178,000 was concentrated heavily: healthcare and social assistance added 42,000 jobs (24% of the total), construction added 15,000, and transportation and warehousing added roughly 11,000. Across 2024-2026, healthcare has consistently accounted for 30 to 40% of monthly job gains, reflecting demographic-aging-driven demand and the sector's near-immunity to cyclical pressure. Government employment has been the second-largest contributor, with state and local hiring covering education and public safety running ahead of federal headcount cuts under the second Trump administration's reorganization.

The concentration matters for interpretation. Strip out healthcare and social assistance, and the March 2026 print falls to roughly 136,000, which puts payroll growth ex-healthcare close to the new breakeven. Cyclical sectors (manufacturing, retail trade, professional and business services) have been flat to modestly negative through 2024-2026. The implication is that the headline job market is being held up by a structural-demand sector while the cyclical sectors have stalled. If healthcare hiring slows, even modestly, the broader labor-market read could deteriorate quickly. Watching the BLS healthcare-ex print monthly is therefore more diagnostic than the headline number for cyclical positioning.

The Fed's reaction function weights both series but treats them differently. Payrolls is the high-frequency activity check; unemployment is the structural check on the dual mandate. The May 7, 2026 FOMC meeting is expected to hold at 3.75-4.00% with implied probability above 90%, in part because the March payrolls print and stable 4.3% unemployment do not require an easing response, and the energy-driven inflation surge takes precedence in the near term. The April 2026 employment report due May 8 will be the first post-Iran-Hormuz snapshot and is the swing print for the June 17-18 FOMC.

For portfolios, the operating rule is that 3-month average payrolls below 100,000 sustained, combined with unemployment rising above 4.5%, would intensify recession positioning toward Treasuries and away from cyclicals. The current setup with payrolls at roughly 60,000 three-month average and unemployment at 4.3% sits right at the tipping point. Bond markets have priced this ambiguity: 10-year Treasuries at 4.31% reflect a fiscal premium and persistent inflation, not a clean recession trade. The Convex Recession Probability Index synthesizes these labor signals with claims data and the yield curve, and its current reading sits in the elevated-but-not-confirmed zone consistent with a Sahm-Rule-triggered cycle that has not yet broken into recession.