Goldman Sachs: Tariffs and AI Statistical Bias Overstate U.S. Inflation by ~3.4 Percentage Points
nashnova research
Goldman Sachs said in a September 22 report that all of the excess in U.S. core goods inflation versus other advanced economies is explained by two temporary, U.S.-specific factors — tariffs and an AI-related statistical bias — which together add roughly 3.4 percentage points of phantom inflation, suggesting markets may be overstating the case for further Fed tightening.
Where do those 3.4 phantom percentage points come from?
Goldman attributes the entire gap between U.S. core PCE goods inflation and other advanced economies to two temporary, U.S.-specific factors.
Factor one: tariffs, contributing roughly 2.4 percentage points — current duties directly push up imported-goods prices.
Factor two: AI-related statistical bias, contributing roughly 1 percentage point — the PCE basket's "software and accessories" category carries outsized weight and has channeled the memory-chip price surge into consumer inflation.
This means → strip away these two layers, and U.S. goods inflation looks broadly similar to the rest of the developed world.
How does AI distort the inflation data?
The PCE's "software and accessories" category — which includes memory chips and carries a large weight — partly captures enterprise purchases through retail channels, causing the sharp rise in memory prices earlier this year to be over-represented in consumer inflation.
The category also applies no quality adjustment, so real performance gains from AI tools are not netted out.
In plain terms = memory getting more expensive is being counted as "consumers spending more," but much of it is really enterprise procurement and a measurement quirk — it has little to do with everyday household spending.
The contrast is stark: memory price increases contribute less than 0.1 pp to core goods inflation in other advanced economies, versus 1 pp in the U.S. — a tenfold gap.
When will these two biases fade?
Tariff effect: Goldman expects the boost to diminish sharply in H2 next year as the tariff impact is gradually absorbed.
AI statistical bias: Goldman's equity strategists see limited further upside for memory prices; in addition, the Bureau of Economic Analysis (BEA) will adjust the weight of "software and accessories" in the PCE basket later this month, and the bias is expected to unwind through 2027.
This means → both distortions come with a clear "expiry schedule" — they are noise that will dissipate on its own, not structural problems.
After adjusting for measurement gaps, is U.S. services inflation actually lower?
Official data show U.S. non-housing services inflation (PCE basis) running above most advanced economies, but Goldman argues this comparison is distorted by methodological differences.
Specifically: the PCE treatment of financial services prices differs markedly from other countries and is highly sensitive to equity-market swings; health insurance measurement is also inconsistent.
After harmonizing methodologies across countries, Goldman finds U.S. core services inflation is actually lower than in other major advanced economies.
This reflects a broader pattern: much of the apparent "U.S. inflation is higher" narrative is an illusion created by measurement methods.
What does the forward-looking labor-cost indicator tell us?
Goldman highlights unit labor cost growth — the change in how much it costs a firm in wages to produce one unit of output — as the core driver of medium-term non-housing services inflation.
U.S. unit labor cost growth has now fallen back close to its pre-pandemic average, while other economies remain notably elevated.
This means → economies outside the U.S. face greater persistent inflation pressure — their wage-price spiral risk is higher than America's.
Housing inflation: the U.S. has normalized — have others?
U.S. rent inflation has largely returned to normal, but other advanced economies remain well above their long-run trends.
Goldman attributes part of this divergence to immigration policy differences — economies where net immigration inflows have pulled back the most (such as the U.S., Canada, and New Zealand) have seen clearer improvement in rent inflation.
In plain terms = fewer immigrants → less rental demand → slower rent growth — and the U.S. is furthest along in this adjustment.
The report's overarching conclusion: after stripping out tariffs and the AI statistical bias, underlying U.S. inflation pressure is actually weaker than in other major advanced economies. If these distortions fade as expected, markets may be overestimating how much current inflation data constrain the Fed's policy path.
市场有风险，内容仅供研究参考，不构成投资建议。