US payrolls grew 316,000 in the year to July — one sector added 550,000, and without it employment fell

The July 2026 jobs report was reported as a one-line miss: nonfarm payrolls fell 23,000 on the month, the figure the day's coverage led with (NBC, CNBC in Notable). Read the same seasonally adjusted series across a full year and the composition is the story. From July 2025 to July 2026, total nonfarm employment rose 316,000, to 158.858 million (BLS series CES0000000001). Over the same twelve months, one supersector — private education and health services — added 550,000 jobs, to 27.962 million (CES6500000001). That single sector's gain is 174% of the entire net increase in payrolls. Subtract it, and the other ten supersectors together shed 234,000 jobs. We read all eleven mutually exclusive supersectors that sum to total nonfarm. Five grew; six contracted. Besides education and health, the gainers were professional and business services (+115,000), leisure and hospitality (+83,000), construction (+82,000) and other services (+49,000) — 329,000 between them. The six that shrank cut 563,000: government (−315,000), financial activities (−114,000), information (−81,000), trade/transportation/utilities (−36,000), manufacturing (−14,000) and mining and logging (−3,000). Government's 315,000 decline was the largest of any supersector and nearly equals the whole year's net gain. The private sector added 631,000 over the same twelve months; government subtracted 315,000; the two net to +316,000. Inside the shrinking trade/transportation/utilities group, retail trade actually added 11,000 while transportation and warehousing lost 63,000.

A single vantage — the month's headline, or a glance at the health line — misses what the whole record shows: 2026's payroll growth is not broad, it is one sector deep. Private education and health services (which excludes public schools and public hospitals, counted under government) did more than carry the year; it offset almost the entire 563,000 lost across the six shrinking supersectors. The arithmetic is exact and re-pullable, but it is only a delta, not a diagnosis. These figures say what changed, not why. We do not assert a cause for government's 315,000 decline or for education and health's rise — the establishment survey counts jobs, not motives (and jobs, not people — a worker with two payroll jobs is counted twice). And the split of that government drop across federal, state and local is a separate question the top-line supersector series does not resolve: BLS carries those breakouts in distinct CES series (federal, state, and local government), not in the single government total used here. What the eleven-sector read does establish is concentration: on the current preliminary data, strip out one line and the US labor market added no net jobs over the year — it lost them.

The strongest counter is vintage. These are preliminary, seasonally adjusted estimates: BLS revised May and June down by a combined 103,000 in this same release, and the establishment survey is rebenchmarked each year to unemployment-insurance tax records. The +316,000 net could shrink materially when the benchmark lands — but because that annual revision also restates the individual supersector figures, not just the top line, whether it sharpens or softens the "one sector carries it" reading depends on how education and health is itself revised, and cannot be assumed. A second, fair objection: concentration is not automatically weakness. Demand for health and elder care is widely argued to be demographic and durable, and on that view a services-led economy can grow fastest where an aging population spends — an interpretation these payroll counts are consistent with but do not themselves establish ^sspeculative. A third: government is not monolithic — a federal decline reads very differently from a state or local one, and the top-line supersector figure blends them.

Read one way, this is a labor market quietly stalling: six of eleven supersectors down, factory and mining payrolls among them (−17,000 combined), and a single still-growing sector papering over the rest. Read the other way, payrolls still grew, construction added more jobs than any other goods-producing sector (+82,000), and health-sector strength is, on the aging-economy reading, roughly what a services-heavy economy might be expected to show ^sspeculative. Both readings sit on the same eleven numbers; the disagreement is about what a one-sector expansion means, not about the count.