Firms took the first claim on the post-2019 productivity revival, and it shows up at a different point on the income statement than the headline numbers imply. A new reading of 10-K filings by New York Fed economist Jeremy Pearce finds that public-firm real revenue per worker accelerated from 0.7 percent a year over 2010 to 2019 to 2.2 percent a year over 2019 to 2025. Over the same years, sales-weighted gross margins rose from about 34 percent of sales to 37 percent, and operating margins rose from 16 percent to 18 percent.
The growth rates suggest the operating line did the heavy lifting: 2.4 percent a year against 1.3 percent for gross margin. The arithmetic in percentage points says otherwise. Gross margin widened by roughly three points and operating margin by roughly two. That means the overhead layer between them got slightly larger as a share of sales, not smaller. The productivity gain was captured where goods and services are actually produced. Firms then reinvested part of it in the overhead lines, where research and development sits.
Two Independent Ledgers Record the Same Turn
The revival is not an artifact of any single dataset, and that matters for everything that follows. The Bureau of Labor Statistics measure of nonfarm-business output per hour grew 1.0 percent a year from 2010 to 2019 and 2.2 percent a year from 2019 to 2025. That roughly doubles the 2010s pace, although it remains short of the near 3 percent run of the late 1990s.
The firm-level series is built differently. Pearce takes Compustat sales, deflates them by industry output prices, and divides by employment. He also drops agriculture, utilities, finance, real estate, and public administration, where output is notoriously hard to measure. The two series start from different raw material and arrive at the same inflection. In the firm data the acceleration is actually larger, a tripling from 0.7 to 2.2 percent.
The deflation step deserves attention because it closes off the easiest objection. Real revenue per worker is not simply nominal sales inflated by the 2021 to 2023 price surge. Industry output prices strip out that effect. Whatever drove revenue per worker higher, it was more real output per employee, at least as industry price indexes measure it.
That settles the productivity question. It leaves open the question the title asks, which is a different one: who keeps the extra output?
Growth Rates Flatter the Operating Line
A profit margin splits each sales dollar into what the firm keeps and what it pays out. Gross margin subtracts the direct cost of production, the cost of goods sold. Operating margin also subtracts selling, general, and administrative expense, which in Compustat typically includes R&D.
Through the 2010s neither margin moved much. Gross margin grew about 0.4 percent a year and operating margin was roughly flat. After 2019, both turned up.
The mechanism becomes clear once the two margins are translated into cost shares:
- Cost of goods sold fell from roughly 66 percent of sales to roughly 63 percent. This is the line that carries direct production labor, materials, and the cost of delivering the service.
- Overhead, measured as the gap between gross and operating margin, rose from roughly 18 percent of sales to roughly 19 percent.
- Operating profit absorbed the net difference, about two points.
These are averages of sales-weighted ratios with rounding at each step, so the decomposition is approximate rather than exact. The direction is still unambiguous. The faster operating growth rate comes from a smaller starting base. It does not mean overhead was squeezed. Firms produced more per worker, paid proportionally less to produce it, and spent some of the savings on SG&A and R&D.
This is where intuition tends to go wrong. The textbook expectation is that productivity gains reach workers through wages as firms compete for labor, or reach customers through lower prices as firms compete for sales. Either channel would hold margins roughly flat while output per worker rose. Margins did not stay flat. The cost of production fell relative to the price received, and the difference stayed with the firm. That pattern is consistent with the continuing decline in labor's share of output, and the author frames the result that way. The extra revenue per worker did not fully pass through to the people producing it.
Why R&D Intensity Picks the Winners
The cross-section explains why the gain landed at the production line. Firms that entered 2019 with more R&D relative to sales saw faster margin growth afterward. On Pearce's estimates, 10 percent more R&D expenditure relative to sales in 2019 corresponds to about 0.7 percent higher operating-markup growth over 2019 to 2025, and 0.5 percent over 2022 to 2025. Capital intensity shows no comparable pattern.
The difference between those two forms of investment is the mechanism. Physical capital is rival and scales with output: a second factory has to be built to produce a second factory's worth of goods. Intangibles such as software, designs, process knowledge, and data are built once and reused. When demand rises, a firm whose product rests on a codebase or a chip architecture serves the extra sales at a low marginal cost. Its cost of goods sold grows more slowly than revenue, and gross margin widens. That is the same signature visible in the aggregate decomposition, now traced to the firms most able to produce it.
Accounting makes the result stronger, not weaker. R&D is expensed through the overhead lines, so R&D-intensive firms report operating margins after paying for their current research. The fact that their operating margins still widened faster suggests the returns came from an accumulated stock of intangibles that kept earning, rather than from cutting current spending.
The 2022 to 2025 coefficient matters for durability. That window runs from the peak of the inflation surge into disinflation. If the margin advantage had been purely a reopening-era pricing windfall, it would likely have faded as price pressure eased. A smaller but still positive effect in the later window points to something more persistent. The effect is modest in size, though. It shifts margin growth at the edge; it does not account for the entire aggregate move.
The contrast is easy to see among familiar large caps, with one caution. The companies below are an illustration chosen for this article and are not drawn from the study's sample. A software and cloud business and a chip designer turn sales into operating profit at a far higher rate than an integrated oil major or an automaker, whose cost structures are built on plants, rigs, and production lines.
That gap is old news. What the 10-K evidence adds is that since 2019 the gap has been widening along the intangible dimension specifically.
The Countercase: Windfall, Headcount, and Concentration
Three objections deserve a fair hearing, and the third is the strongest.
Pricing rather than productivity. The 2019 to 2025 window contains the sharpest inflation episode in four decades. Many firms raised prices ahead of their input costs in 2021 and 2022, and nominal margin ratios would record that as a wider gross margin even with unchanged efficiency. This objection is partly met by the deflated revenue-per-worker series and by the persistence of the R&D effect after 2022. It is not fully met, because margins are nominal ratios and part of the three-point gross widening could be a pricing gain that later reverses.
Headcount accounting. Compustat employment counts the firm's own employees. A company that moves work to contractors, offshore suppliers, or platform workers raises its revenue per worker without any change in how efficiently the work is done. Outsourced labor usually reappears as purchased services in cost of goods sold, which limits the effect on margins, but it can still inflate the per-worker productivity figure. The close match with the BLS output-per-hour series, which counts hours across the business sector rather than within individual firms, works against this explanation as the main driver.
Concentration dressed as investment. R&D intensity is not randomly distributed. It clusters in software, semiconductors, and pharmaceuticals, sectors where scale effects often produce dominant firms. A firm with a large R&D budget in 2019 may also have been a firm with market power in 2019. In that case, the coefficient measures the returns to market position, and R&D is simply the variable that identifies who held it. Pearce leaves this question explicitly open, asking whether rising margins reflect returns to intangible investment or a windfall. The distinction is not academic. Returns to investment tend to attract imitators and get competed down over time. Returns to market position can last as long as the position does.
What Would Settle the Split
The decision variable is whether the roughly three points of gross margin hold once labor markets reprice the productivity gain. Two observable outcomes separate the readings.
If gross margins stay near 37 percent while wage growth catches up to the faster productivity trend, the gain is structural. Output per worker would be rising fast enough to fund both higher pay and wider margins. If labor's share of output stops falling and gross margins drift back toward the 2019 level of about 34 percent, the post-2019 widening was largely a pricing windfall from the inflation years that labor eventually recovered.
The cross-section offers a second check. If the R&D advantage reflects genuine returns to intangibles, the tools behind it should spread, and the margin gap between R&D-heavy and capital-heavy firms should narrow as competitors adopt them. If the gap keeps widening while R&D-intensive firms report rising margins year after year, the windfall interpretation gains ground, and the answer to the title's question shifts from a cyclical yes to a durable one.
