生产率：亮眼的标题数据，疲软的现实

Productivity: Strong headlines,softerreality Recent productivity data have fueled optimism that the US isentering a structurally faster growth phase, driven by AI andsurging tech investment. But much of the recent strengthappears to reflect cyclical normalization and measurementquirks rather than a clear productivity regimeshift. Jonathan Millar+1 212 526 4876jonathan.millar@barclays.comBCI, US Marc Giannoni+1 212 526 9373marc.giannoni@barclays.comBCI, US •Headline productivity looks strong—but gains overstate the underlying trend.Nonfarmbusiness productivity rose 2.5% y/y in 2025 and 3.5% in H2, prompting optimism about astructuralshiftand some upward revisions to trend growth at the March FOMC. But thesegains are highly sensitive to measurement choices and cyclicaleffectsthat can inflateunderlying strength. Pooja Sriram+1 212 526 0713pooja.sriram@barclays.comBCI, US Colin Johanson+1 212 526 8536colin.johanson@barclays.comBCI, US •Measurement divergences weaken signals.Since 2022, expenditure-and income-sideoutput measures have meaningfully diverged, with the income side pointing to slower,steadier growth. Smoothing puts productivity closer to ~2.5%, tempering the signal fromsome of the more eye-catching recent prints. •Cyclical utilization is doing heavylifting.Much of the recent pickup reflects post-pandemicnormalization—with broad swings from labor shortages to overhiring, then to labor hoarding—leaving lasting imprints on the estimates. Adjusting for utilization suggests a sizabletailwind to measured productivity in 2025 that is unlikely to be durable. •Structural acceleration is real, but not transformational.Afteraccounting formeasurement and utilizationeffects,structural productivity appears to be running around1.5–2.0% y/y. That is better than pre-pandemic norms, but still short of a true regimeshift. •AI capex is booming, but capital deepening is not.Hyperscaler capex is eye-catching, butaggregate capital growth remains modest relative to the existing stock and depreciation.Without a broad and sustained investment boom or an unprecedented standalone pickup inmultifactor productivity, AI alone is unlikely to materially raise potential growth in the nearterm. An inflection point for productivity? With the latest updates, nonfarm business productivity is estimated to have increased 2.5%Q4/Q4 in 2025, and 3.5% during H2 2025. This will likely reinforce hopes that the US is entering aperiod of structurally faster productivity growth, supported by technological advancements inAI and surging AI capex. Headline measures of hourly output have strengthened, on balance,over the past two years, and this improvement isoftencited as evidence that a sustainedacceleration has taken root. But a closer look reveals that the evidence in favor of sustained acceleration is still in dispute.Media accounts frequently ignore the sensitivity of measured productivity growth to cyclicalinfluences and accounting choices.Differentoutput measurement conventions (such asexpenditure- versus income-based accounting) can yield materiallydifferentconclusions, andcyclical swings in the utilization of capital and labor can be easily mistaken for structuralchanges. In this piece, we look behind the headline measures to illustrate the many measurement andconceptual issues at play. The upshot of our analysis is that underlying productivity growth islikely running in a range of 1.5-2.0% y/yafteraccounting for measurement issues and adjustingfor utilization. While this is at the upper end of assessments prior to the pandemic, it remainswell short of what one might reasonably label a transformational change. This pace also may beinsufficientto overcome the aggregate growth drag from agingdemographics and slowed immigration. Standard growth accounting indicates that potentialGDP growth relates directly to the increases in labor services, capital services, and structuralacceleration in multifactor productivity (MFP). Contrary to the popular perception, the scope ofthe AI capex boom is still much too narrow to fuel a meaningful strengthening ofaggregatecapital growth, with overall levels of fixed investment growing at a moderate pace andremaining relatively low in relation to the overall capital stock. Absent this, the most plausiblechannel for a sustained productivity acceleration is MFP, but this tends to react slowly totechnological advances and has never carried a productivity boom without support from strongcapital deepening. Mind the output measure A key source of uncertainty in recent productivity data is how output is measured. The standardmetrics published by the BLS rely upon concepts of output that are reckoned by expenditureaccounting, as opposed to income accounting. In principle, the two approaches areconceptually identical. In practice, estimates have diverged meaningfully. The BLS's workhorse productivity metrics are for the nonfarm business (NFB) sector, whichstrips from GDP the output of farms, governments,