US August Core PCE Falls Short of Expectations: A Cooling Report Manufactured by a "Definition Change"Will the Fed Buy It?

U.S. consumer spending in August grew at the fastest pace in more than a year, while the year-over-year reading of the Fed's most closely watched inflation gauge fell back markedly. Data released Wednesday by the U.S. Bureau of Economic Analysis (BEA) showed that inflation-adjusted personal consumption expenditures rose 0.6% month over month in August, the largest single-month gain since March 2025; nominal personal consumption expenditures rose 0.9%, above the market expectation of 0.8%. The personal consumption expenditures price index (PCE), the Fed's preferred inflation gauge, rose 0.3% month over month, in line with expectations; core PCE, which excludes food and energy, rose 0.2% month over month, below the expected 0.3%. The pullback was more pronounced on a year-over-year basis: headline PCE was up 3.4% year over year (expected 3.7%), and core PCE was up 3.0% year over year (expected 3.3%). Data Panorama: Income Trails Spending, Consumption in Full Swing By component, nominal PCE rose by $190.8 billion in August, with goods spending up $114.1 billion and services spending up $76.7 billionautos, furniture and clothing were the main drivers, as consumers maintained a strong willingness to spend against a backdrop of high gasoline prices and rising overall costs. The other side of the ledgerincomelooked thin: personal income rose only 0.2% month over month in August (expected 0.4%), disposable personal income rose 0.3%, while inflation-adjusted disposable income was flat month over month (0.0%). In other words, the August consumption expansion was not primarily driven by income growth. Total personal saving was $990.2 billion, and the saving rate was 4.1%. Bank of America credit card data show that U.S. consumer spending rose 6.9% year over year in the week ended September 19, with gasoline spending up 26.5% and still up 5.7% excluding gasolineconsumer resilience has so far shown no sign of loosening. The Water in the "Cooling": How Much Did Three Methodological Adjustments Contribute? The most important thing to be careful about in this report is that the "decline" in the year-over-year reading and the "rebound" in the month-over-month reading occurred at the same time. On the year-over-year side, headline PCE was up 3.4%, far below the market expectation of 3.7%, and the July prior reading was also revised down sharply from 3.7% to 3.4%; core PCE was up 3.0% year over year, below the expected 3.3%, with the July prior reading revised down from 3.3% to 3.0%according to compilation by FX678, this was the lowest reading for core PCE year over year since February. The month-over-month side, however, was accelerating: headline PCE rose to 0.3% from a revised 0.1% in July, and core PCE rose to 0.2% from 0.1%. The former was in line with expectations, while the latter was below the expected 0.3%, but both were moving upward. What created this divergence of "cooling year over year, heating up month over month" was two sets of revisions implemented simultaneously on Wednesday. First, this release included the annual revision of the national accounts covering the past five years (retroactive to January 2021); second, the BEA adjusted the calculation methods for three components of the PCE price indexportfolio management and investment advice services, computer software and accessories, and legal services. The largest impact came from the portfolio management services component. The old method directly deflated nominal spending using an industry producer price index, so asset price gains were recorded as "higher services prices"; this component's year-over-year increase over the past 12 months was once as high as 21.6%, making it the second-largest contributor to core PCE inflation. The new method instead uses total hours growth from employment surveys to measure "real services volume." UBS economists including Alan Detmeister estimate that this alone lowered core PCE year over year by about 0.21 percentage points; Goldman Sachs's Manuel Abecasis and others estimate that the adjustment to the software and accessories component will reduce core PCE year over year by 0.05-0.1 percentage points in May and by 0.1-0.2 percentage points in December; the legal services component, meanwhile, slightly pushes it up by about 0.04 percentage points, partially offsetting the first two. Taken together, this round of methodological adjustments lowered core PCE year over year by roughly 0.2-0.3 percentage pointsroughly equivalent to the drop in the year-over-year reading from 3.3% to 3.0%. In other words, after stripping out the definitional effect, the inflation reading was roughly in line with expectations, and a considerable portion of the "surprise cooling" the market saw was at the statistical level. MarketWatch columnist Jeffry Bartash put it more bluntly: the new method "appears to have shaved a few tenths of a percentage point off the prior inflation rate, but not enough to suggest a major change in trend"; under either the old or new definitions, inflation is too high. A background worth recording is this: former Fed governor Miran pointed out last December that the current method "treats what should be recorded as an increase in the quantity of services consumed as a price increase"; in May of this year, he co-authored a paper with Fed economists Barbarino and Diercks that systematically laid out measurement flaws in categories such as portable storage devices and video games. But the timing of the adjustment came during a sensitive periodTrump had previously fired the head of the Bureau of Labor Statistics over weak employment data, and the president has continued to call for rate cuts, prompting Vikas Patel of the liberal think tank Employ America to urge the BEA to "increase transparency regarding the timing of adjustments, weights, and historical revisions." Whether the technical basis holds up and whether the public trusts statistical independence are becoming two separate questions. The revisions on the GDP side also cannot be ignored. The final estimate of second-quarter real GDP growth was revised up by 0.7 percentage points to 2.2% quarter over quarter at an annualized rate (expected 1.5%), and first-quarter growth was revised up to 2.5%; the second-quarter PCE price index was revised down to 5.0% (previously estimated 5.3%), and core PCE was revised down to 3.3% (previously estimated 3.6%). Growth being revised up and inflation being revised down at the same time substantially weakened the "stagflation" narrativeand this came just as the Fed had restarted rate hikes and needed to judge "whether the economy can withstand more tightening." Market Reaction: Yields Fall, Gold Spikes After the data was released, the bond market reaction was more direct: the two-year Treasury yield fell 4.57 basis points to 4.843%, and the ten-year yield dropped about 5 basis points to 5.205%. The starting point of this pullback mattersjust the day before (Tuesday), the 30-year Treasury yield rose intraday to 5.619%, the highest since June 2002; the 10-year yield briefly approached 5.3%, the highest since 2007, and it was already the 8th record high within September. The surge in yields dragged the three major U.S. stock indexes to close lower on Tuesday (Dow 0.26%, S&P 0.17%, Nasdaq 0.09%). Jose Torres, senior economist at Interactive Brokers, observed that stocks "are struggling to hold current levels, but tightening financial conditions are strengthening bearish confidence while raising investor interest in downside hedges." Gold briefly surged more than $10 to $4,205/oz, with spot gold up 0.62% at $4,207.75; the dollar index had risen 0.17% on Tuesday to 101.366 and was up 1.5% against a basket of currencies in September, with the euro falling to a 16-month low against the dollar at one point, before the dollar weakened somewhat after the data. In oil, WTI fell 3.48% on Tuesday to $89.38 and Brent fell 2.56% to $102.59 (as Middle East crude outflows resumed), before rebounding during Asian hours on Wednesday after Trump denied he would ease sanctions on Iran. There is also a background data point that cannot be ignored: the U.S. Conference Board consumer confidence index for September, released Tuesday, was only 81.9, below expectations and the lowest since 2014a jarring contrast with the booming consumption data, and one that plants the seeds for the question of "how much longer consumption can stay hot." Policy Outlook: October Rate-Hike Expectations Cool CME FedWatch pricing after the data showed: the probability that the Fed leaves rates unchanged at its October 28 meeting was 65%, and the probability of a cumulative 25 basis point hike was 35%, down somewhat from before the data release; for the final meeting of the year on December 9, the probability of no change was only 12%, the probability of a cumulative 25 basis point hike was 59.5%, and the probability of a cumulative 50 basis point hike was 28.4%. Traders have reduced bets on an October hike. But it is more accurate to place this pricing in the context of the previous day: on Tuesday, New York Fed President Williams had already said "there is no need to rush to act, and there is time to gather more information before the October meeting," and the probability of an October hike had already fallen below 50% before that. He also stressed at the same time that 3.7% inflation was "unquestionably too high" and that one more hike this year "may be appropriate"with only the October 28 and December 9 meetings remaining. Evercore analyst Krishna Guha interpreted this as leaning toward skipping October and implementing a hike in December. The hawkish side has not softened. Fed Governor Barr said core PCE had met the 2% target in only 2 of the past 20 months, that "further policy adjustments are likely necessary," and attributed part of the price pressure to AI spending driving up chip costs; Governor Cook also said on September 28 that in the short term AI appears to be adding inflationary pressure to the economy and delaying inflation's return to 2%. Cleveland Fed President Hammack, the Philadelphia Fed president and others have also expressed concerns that policy "may need to become more restrictive." After the September 16 FOMC voted 12-0 to raise rates by 25 basis points to 3.75%-4.00%, the dot plot showed a median year-end federal funds rate of 3.6%, with 16 of 18 policymakers expecting at least one more hike before year-end. The market's pricing and officials' disagreement together constitute the entire suspense before October 28. What Institutions Think JPMorgan (Michael Feroli): expects tightening to end with just one additional hike (in December), saying "inflation continues to look supply-shock driven, so we do not foresee the hiking cycle extending into next year"; the case for a hike is that "core PCE inflation has been above 3% every month this year, with little recent progress toward the 2% target," while Chair Warsh "has repeatedly and sternly warned of intolerance for inflation, and without some action to back that up, institutional credibility could be at risk." He and Abiel Reinhart also noted: the saving rate has been falling steadily this year, partly due to the wealth effect from rising stocks, but the pace of decline has accelerated over the past six months, which may mean some consumers are stretching to spend amid high living costsonce the wealth effect fades, consumption resilience will be tested. Citi: had previously expected August core PCE at 0.29% month over month and 3.14% year over year, and expects core PCE year over year at about 3.1% in Q4 2026, below the Fed's September median projection of 3.4%. Nomura: had previously expected 0.278% month over month and 3.30% year over year, and maintained its view of one more hike in December and no hikes in 2027. Goldman Sachs: inflation data may show some unfavorable changes in the coming months, then return to a moderate trend. Allianz Trade senior economist Dan North: core inflation has yet to show a convincing pullback, and the Fed still finds this level hard to ignore. RBC Wealth Management: a 30-year Treasury yield of 6% is "not entirely impossible." One common point: institutions had already warned before the data release that "methodological revisions would push down the year-over-year reading," so most treated the August month-over-month figure rather than the year-over-year figure as the real testthe 0.2% core month-over-month reading landed exactly in a spot that was "not hot enough, nor cold enough," insufficient to make hawks back off, and also insufficient to make an October hike consensus again. What to Watch Next In chronological order: the September nonfarm payrolls report due Friday (October 2), with the market expecting about 100,000 new jobs and the unemployment rate rising to 4.2%; inflation and consumption data in the following weeks; the October 28 FOMC decision; and the advance third-quarter GDP reading on October 29 (one day after the policy meeting, with greater implications for the December meeting). In addition, Micron Technology will report earnings after the close on Wednesdayat a time when AI stocks are propping up the Nasdaq while market breadth has narrowed to historic levels (the index near a 52-week high while the average stock is far below its high), this earnings report itself is a stress test for the AI trade. To sum up in one sentence: the message from this data is not "inflation has cooled," but "inflation is not hotter than expected, and the economy is sturdier than expected"for the Fed, the former provides a reason to wait, and the latter provides the wherewithal to hike again; the October coin is still in the air.