Employment Revisions Turn Positive

Over the last few years, reactions to the monthly employment data have often taken a cynical or skeptical air: “Wait until these numbers get revised down.” That skepticism has lingered into 2026, despite a run of relatively good job reports. But as we’ll discuss in the rest of this piece, this negativity is outdated. In fact, it’s possible that labor market watchers should be skeptical in the opposite direction: recent numbers are likely to be revised up. The rest of this essay will discuss these revisions.

The monthly employment report sources data from two surveys: the Current Population Survey (the CPS or “household survey”) and the Current Employment Statistics (the CES or “establishment survey”). And every year, historical data from the CES undergo a substantial annual revision that incorporates more complete data from a 3rd, more comprehensive data source: the Quarterly Census of Employment & Wages (QCEW).1 The QCEW is published with a substantial lag (about five months after the CES) and is much noisier than the CES, so the QCEW-derived “benchmark” level of nonfarm employment is applied to CES data only once a year.2 For instance, in January of 2026, a benchmark revision using primarily QCEW data was applied to the level of March 2025 nonfarm payroll employment.34

Over the last few decades, benchmark revisions have been both positive and negative. But the past three rounds have all been negative: March 2023 (-266K; -0.2% of nonfarm payroll employment), March 2024 (-589K; -0.4% of nonfarm payroll employment), and most dramatically, March 2025 (-898K; -0.6% of nonfarm payroll employment). A lot of folks, myself and the Fed included, assumed we were headed for another negative benchmark revision early next year when the data is revised from (to the level of employment as of March 2026).

But that assumption seems increasingly incorrect. The QCEW is released quarterly, and last week we received the Q4 2025 data. That means we now have three-quarters of the data that will be incorporated into the eventual benchmark revision. And that “three-quarters complete” QCEW data indicates that, between December 2024 and December 2025, employment grew substantially faster than the payrolls data currently indicate (by about 230K). At the same point in 2023, 2024, and 2025, the QCEW was showing slower growth than NFP by 694K, 735K, and 907K, respectively. In each of those years, we had negative revisions (though not quite as negative as QCEW implied).