The market is pricing three more Fed hikes off an inflation gap that is mostly an accounting artifact. One of its largest pieces is the stock market feeding into its own discount rate. Core PCE inflation at roughly 3.2% against core CPI at 2.4% looks like a hawkish warning. Broken into its parts, the gap rests on four line items. Two of them, financial-services fees and computer software, measure asset prices and an AI pricing spike, not the persistent consumer inflation a tightening cycle exists to stop. If that reading holds, the policy debate matters less than what happens across assets. A front end priced for three hikes is carrying premium against a Fed whose own dot plot shows one.
A Gap Built From Four Line Items
Start with how unusual the reading is. Over the past four decades, core PCE has run about 0.43 percentage points below core CPI on average. By July 2026, year-over-year core PCE was 0.88 points above core CPI. That is the widest positive gap in more than 40 years. Eighteen months earlier, in January 2025, the sign was reversed: core PCE at 2.8%, core CPI at 3.3%. The gap swung by roughly 1.4 points in a year and a half.
The structural reason PCE normally runs cooler has also faded. The CPI uses a fixed basket, and fixed baskets overweight categories whose prices are rising fastest because they ignore substitution. The Bureau of Labor Statistics shortened its reweighting interval from once a decade before 2002, to every two years, to annually from 2023. The average PCE discount shrank in step: 0.7 points from 1986 to 2001, 0.3 points from 2002 to 2022, and 0.2 points since. That formula effect was stable over the last 18 months, so it did not cause the swing.
The Bureau of Economic Analysis publishes a monthly reconciliation of the two measures. Using that framework, the swing breaks down as follows.
- Shelter. Rent and owners' equivalent rent make up about 42% of core CPI but only 17% of core PCE. In January 2025 the official shelter measures were up 4.2% and 4.6%, even as new leases showed inflation between -0.4% and +2.8%. Official shelter has since cooled to about 3.0%. Because the CPI weight is so much heavier, that cooling cut core CPI by roughly 0.35 points relative to core PCE.
- Computer software and accessories. The category weighs 0.031% of the CPI basket against 1.1% of PCE. Annual inflation in the category went from 0.4% to 21% in both indices. Because the weights differ, the identical price move added about 0.2 points to the gap.
- Financial services. Fees and commissions carry a 2.83% weight in the PCE deflator and about 0.2% in the CPI. They are currently estimated in a way that tracks a booming equity market. Inflation in the category topped 14% in the July accounts and added more than 0.4 points to core PCE, with almost no effect on CPI.
- Auto insurance. CPI weight 2.6%, PCE weight 0.5%. CPI's measure went from +11.8% to -4.5%; PCE's went from +6.8% to +0.4%. Combined with the weighting difference, this added about 0.5 points to the gap.
These four total roughly 1.45 points against a swing of about 1.4. In other words, they explain essentially the whole move. Health care, the category often blamed for scope differences between the indices, netted to zero. The gap is not a broad divergence in how American consumers experience prices. It comes from a few categories whose weights differ sharply between the two indices, each moved by its own unusual shock.
When the Stock Market Shows Up in the Price Index
The financial-services component deserves separate attention, because this is where a rates story becomes a cross-asset loop.
Part of what the PCE deflator counts as financial services is priced off asset values. When equity markets rally, measured fees and commissions rise, and core PCE goes up with them. The Fed has said it has little tolerance for persistently elevated inflation and backed that up with last week's hike. So it is now reacting to a print that is partly inflated by its own target market's strength. Higher equity prices lift measured PCE, which strengthens the case for hikes, which tightens financial conditions. That chain runs through an accounting convention, not through wages, rents, or goods prices.
That is a fragile basis for a multi-hike path. A fragile basis tends to break suddenly, and here there is a scheduled date. The BEA's annual re-benchmarking at the end of this month introduces a new methodology for financial services, which is expected to shrink this effect going forward. The size of the revision is not yet known, and revised history may not match current-vintage estimates point for point. The direction, though, is toward a smaller financial-services contribution, and that change lands within weeks.
Software points the same way, with less certainty. A 21% annual rise in software and accessories prices looks like an AI-era pricing surge, not a steady state. If the category merely slows, the 0.2-point contribution starts to unwind. If AI capital spending keeps product pricing hot, it fades more slowly than the base case assumes.
What the Front End Is Actually Pricing
This is where the rates lens matters. Futures price three more hikes over the next year. The Fed's dot plot projects one more, in December, which would leave fed funds at 4.00% to 4.25% through 2027. As of October 1, the two-year Treasury yielded 4.78%. Three more quarter-point hikes from the current 3.75% to 4.00% range would take the policy rate to 4.50% to 4.75%. The two-year sits on top of that path, not on top of the dot plot. This is an inference from the levels rather than a decomposition of the curve, but the arithmetic is hard to read any other way.
The gap between those two paths is roughly 50 basis points of terminal rate. The four-item breakdown is the best evidence for which path wins. Suppose the gap closes because PCE falls: financial-services fees get revised down, software cools, and shelter's drag on CPI ends. Then the Fed's own projection becomes the likely outcome, and the front end has about two hikes of premium to give back.
The long end tells a different story. The 10-year at 5.24% sits 46 basis points above the two-year. That is a positively sloped curve during an active tightening cycle, which is not the usual pattern; hiking cycles tend to flatten or invert the curve. The long end is not pricing a Fed that wins. It is demanding compensation for something else, plausibly term premium tied to fiscal supply and the energy shock. The source material does not establish that cause, so it should be treated as unproven. Even so, it has a practical implication. A PCE-led closing of the gap would pull the two-year down more than the 10-year, so the curve would steepen from the front end. Investors who expect lower inflation readings to also bring 10-year yields down sharply may be disappointed.
The dollar is the next link, and the least certain one. A front end priced for three hikes supports the dollar on rate differentials. Removing two of them takes away part of that support. How much depends on what other central banks are doing, which is outside this evidence. The direction is clear; the size is not.
The Case for CPI Doing the Converging
The opposite case is credible and should be stated fully. The gap is highly mean-reverting, but it can close from either side, and some of its parts point toward CPI rising.
Auto insurance is the clearest example. A CPI reading of -4.5% annual inflation reflects the clumsy way the CPI tracks insurance, and most observers expect it to partly reverse. Because the category carries 2.6% of the CPI basket, a rebound pushes core CPI up directly. Shelter has little room left to fall. Official measures near 3.0% already sit close to market rents, which removes one of CPI's sources of disinflation. Add the energy shock, an AI capex boom that keeps software and computing costs firm, and a Fed that has just shown its willingness to act. That combination supports the futures path. In that scenario the gap closes with CPI rising toward 3% while PCE holds near 3.2%, and three hikes look reasonable.
The base case has its own vulnerable assumptions. It assumes Middle East energy flows normalize by year-end and that no further fiscal stimulus arrives. Tariff rates below last year's levels, slow rent growth from weak demographics, and soft wages all support gradual disinflation. Each of those is a forecast, not an observation. If energy does not normalize, headline inflation will keep pressure on the Fed's reaction regardless of how core PCE is reconciled.
Overall, though, the balance still favors PCE coming down. Of the four drivers, financial services is set for a methodology change that should reduce it. Software is likely to fade. Shelter's remaining drag on CPI is small. Only auto insurance clearly argues for CPI rising, and it accounts for about a third of the swing. Three of the four pieces push toward PCE falling. That is the core of the argument: the convergence is more likely to come from the PCE side.
The Print That Settles It
Two releases will decide this. The first is the BEA's end-of-month re-benchmark. If revised history shows financial-services inflation running well below 14%, a large part of the 0.88-point gap is removed by revision rather than by economic change. The 3.2% core PCE figure that worried Warsh would then become a lower number in the revised record. The second is the monthly path of CPI auto insurance. A fast rebound toward positive annual inflation would be the strongest evidence for the hawkish path and would make the futures curve look correct.
The thesis fails if core CPI moves toward 3% by year-end while revised core PCE stays above 3%. That would mean the gap closed the hawkish way and the three-hike path was justified. Short of that, the front end is overpricing tightening on the strength of a fee index tied to equity prices. The clearest signal is a two-year yield falling faster than the 10-year once the revised PCE data are published.
