Notes on the Week Ahead Sep 21, 2026
With David Kelly, Chief Global Strategist at J.P. Morgan Asset Management
Core PCE inflation ran 0.88% hotter than core CPI in July, the biggest positive gap in more than 40 years. Over those 40 years the average has run the other way, with core PCE 0.43% cooler.
The Federal Reserve raised rates last week and Chairman Kevin Warsh cited two numbers at the press conference: core PCE at about 3.2% and core CPI at about 2.4%. David Kelly does not think the Fed should have raised at all, and he does not think the higher of those two numbers is measuring inflation.
"On balance, I don't believe that the Federal Reserve should have raised interest rates last week. However, I must confess that there was one reference in Chairman Warsh's press conference that left me distinctly uneasy."
Kelly, Chief Global Strategist at J.P. Morgan Asset Management, on Notes on the Week Ahead, works the gap apart using the Bureau of Economic Analysis's own four-part methodology and the monthly numbers the BEA publishes against it, and ends with a rate forecast that is one hike short of what futures markets have priced.
The full episode is covered here so you can skip it.
Here are the 6 arguments that matter.
Key Takeaways
Core PCE running 0.88% above core CPI is the biggest positive gap in more than 40 years, and for those 40 years the average ran 0.43% the other way
The direction the gap closes decides the rate path — three more hikes are priced if CPI rises to meet PCE, one if PCE falls to meet CPI
The formula effect is not the culprit, because the CPI basket now reprices annually instead of once a decade
Core PCE ran 0.7% cooler than core CPI from 1986 to 2001, 0.3% from 2002 to 2022 and 0.2% since the start of 2023
Four categories opened the gap, and Kelly expects all four to reverse: shelter, computer software, financial services and auto insurance
Shelter is 42% of core CPI and 17% of core PCE, so falling rent measures cut CPI by roughly 0.35% relative to PCE
Financial services carry a 0.2% CPI weight and a 2.83% PCE weight, and inflated core PCE by over 0.4% on a booming stock market
The BEA is changing the financial services methodology this month, which should shrink that contribution on its own
Kelly's call is one more hike in December and then nothing, with the funds rate at 4.00%-4.25% through 2027
1. Warsh's Two Numbers
Kelly opened by separating two disagreements. He does not think the hike was warranted. Separately, he was unsettled by which numbers the chairman leaned on.
The rate decision and the reasoning are two different objections
On balance, I don't believe that the Federal Reserve should have raised interest rates last week. However, I must confess that there was one reference in Chairman Warsh's press conference that left me distinctly uneasy.
David Kelly
The reference was a pair of year-over-year rates: core PCE at about 3.2% and core CPI at about 2.4%. On the lower of the two, Kelly said there is nothing to be alarmed about. Energy shocks like the current one are short-lived, core inflation predicts future inflation better than headline does, and the Fed's target is set on headline PCE.
A 2.4% core CPI print is on target, not above it
Over the past 40 years, core PCE inflation has run an average of 0.43% cooler than core CPI inflation. Since the Fed's long-term target is 2.0% headline PCE inflation, a core CPI inflation rate of 2.4% would seem to be right on track.
David Kelly
2. A 40-Year Record Gap
The problem is that the usual relationship has inverted. PCE is now running hotter than CPI rather than cooler, and by an unprecedented margin.
The July reading is the widest positive gap on record
In fact, in July, the year-over-year increase in the core PCE deflator was 0.88% higher than core CPI inflation, the biggest positive gap in more than 40 years.
David Kelly
Eighteen months earlier the relationship looked normal: in January 2025 core PCE was at 2.8% against 3.3% for core CPI. The gap is highly mean-reverting, Kelly said, so it will close. What matters for markets is which side moves.
Which way it closes is the whole question for rates
However, if the PCE-CPI gap closes by CPI inflation rising, the Fed may well follow through with the three additional rate hikes that futures markets have priced in for the next year.
David Kelly
If PCE drifts down to close it instead, the Fed may be satisfied with one more hike at the end of this year, which is what its own dot plot projects. Kelly's stated method for choosing between the two is to ask how the gap opened in the first place.
3. The Formula Effect
The BEA published a methodology in 2007 for decomposing the difference between the two measures and updates the numbers monthly, most recently for July 2026. It buckets the difference four ways: a formula effect, a weight effect, a scope effect and "other".
The formula effect is the one most people reach for first. CPI uses a fixed-weight basket; the PCE deflator uses a chain-weight basket that moves with actual spending. Because people substitute away from what is getting expensive, a fixed-weight index overweights the high-inflation categories.
Kelly's point is that this effect has been shrinking for two decades, because the Bureau of Labor Statistics keeps rebasing the CPI basket more often — once a decade before 2002, every two years from 2002 to 2022, and annually since the start of 2023.
The formula effect has been fading for 20 years
This is a key reason why year-over-year core PCE inflation ran an average of 0.7% cooler than core CPI inflation between August 1986 and December 2001, but just 0.3% cooler between January 2002 and December 2022 and just 0.2% cooler between January 2023 and July 2026.
David Kelly
And it did not move over the past 18 months, which is the window in which the gap opened. So it is not the explanation.
4. Shelter And Software
The weight effect is where Kelly finds the first real cause. The two indices apply very different weights to the same goods, and shelter is the extreme case.
Shelter is 42% of one index and 17% of the other
Rent and owners' equivalent rent account for 42% of core CPI but just 17% of core PCE.
David Kelly
Both measures smooth housing costs and both lag the actual rental market. In January 2025 new leases were running between -0.4% and +2.8% year-over-year on industry data, while the government's rent and owners' equivalent rent series were up 4.2% and 4.6%. Because CPI weights shelter so much more heavily, that gap was pushing CPI above PCE. The government series have since come down to roughly 3.0%.
That one adjustment cut core CPI by 0.35% relative to core PCE
This change alone has cut core CPI inflation by roughly 0.35% relative to core PCE.
David Kelly
The second weight mismatch is smaller and stranger. Computer software and accessories is 0.031% of the CPI basket and 1.1% of the PCE basket — a 35-fold difference in weight on a category that was irrelevant when its inflation rate was 0.4% in both indices in January 2025.
Software inflation went to 21%, and only one index noticed
However, by July 2026, the year-over-year inflation rate had jumped to 21% in both indices, adding another 0.2% to the PCE-CPI gap.
David Kelly
5. Fees And Auto Insurance
The scope effect covers categories that appear in one basket and not the other. Health care is the obvious candidate: CPI counts only what consumers pay out of pocket, while PCE includes what employers and others pay on their behalf. Kelly checked it and found nothing — both measures have seen health care inflation fall since the start of 2025, and PCE's heavier weight is offset by a larger decline in the CPI measure, so the net effect on the gap is zero.
Financial services is the small category doing large damage. The BLS gives fees and commissions a 0.2% weight, treating them as closer to saving than to consumption. The BEA gives them 2.83%, and calculates them in a way that tracks the stock market.
A booming market is being counted as inflation
This category saw year-over-year inflation of over 14% in the July PCE accounts, adding over 0.4% to core PCE inflation although having a negligible impact on CPI inflation and thus providing a considerable boost to the PCE-CPI gap.
David Kelly
That one is scheduled to shrink by administrative fiat: in the annual re-benchmarking at the end of this month, the BEA is introducing a new methodology for financial services that Kelly expects will reduce the effect going forward.
The last contributor is auto insurance, which carries a 2.6% weight in CPI against 0.5% in PCE, and which swung violently. In January 2025 CPI had rates up 11.8% against 6.8% in PCE. By July of this year the CPI measure was down 4.5% year-over-year while PCE was up 0.4%.
The auto insurance swing alone added half a point
A little arithmetic shows that this sharp swing in auto insurance rates, combined with their different measurement and weight in the two indices, added 0.5% to the core PCE-CPI gap over the past 18 months.
David Kelly
6. One More Hike, Then Stop
Kelly's reconstruction leaves January 2025 looking like the anomaly rather than today. Back then core CPI was 0.5% above core PCE, which he attributes to the heavier CPI weights on auto insurance and owners' equivalent rent at a moment when both were running hot. In the long run he expects core CPI to sit just 0.1% or 0.2% above core PCE.
The four things that opened the gap since then all reverse, on his reading: CPI shelter inflation has a little further to fall, the software surge fades, the financial services contribution gets downgraded by the methodology change, and the collapse in CPI auto insurance partly reverses.
His forecast assumptions are explicit: some normalization in Middle East energy flows by year end, no further fiscal stimulus, lower tariff rates than a year ago, sluggish rent growth from weak demographics, soft wage growth, and a continuing AI capital spending boom.
Both measures drift down, and PCE falls faster
Moreover, because most of the forces that opened up a positive PCE-CPI gap are likely to reverse, we expect PCE inflation to fall more than CPI inflation.
David Kelly
That is the comfort he thinks the Fed needs to stop.
One more hike in December, then a funds rate held through 2027
This should come as some comfort to the Federal Reserve allowing them to stop an abbreviated tighten cycle with just one more rate hike in December, leaving the federal funds rate in a range of 4.00% to 4.25% throughout 2027.
David Kelly
That is less tightening than the market has priced, which he says should generally support risk assets. His closing caveat is a portfolio one rather than a macro one: it does not reduce the case for rebalancing away from concentrated portfolios, which carry plenty of risks that have nothing to do with inflation or interest rates.
Bonus Insights
The source he is working from
The whole analysis runs off a single BEA framework, set out in a 2007 Survey of Current Business article by Clinton P. McCully, Brian C. Moyer and Kenneth J. Stewart, which Kelly footnotes. The BEA updates the decomposition monthly, and the latest data he uses are for July 2026.
Kelly's bottom line is that the number making the Fed uneasy is largely an artifact of how two agencies weight rent, software, brokerage fees and car insurance, and that when those four reverse it is PCE inflation that falls to meet CPI rather than CPI that rises to meet PCE — which is why he expects one more hike rather than the three the market is paying for.
Books & Resources Mentioned
Comparing the Consumer Price Index and the Personal Consumption Expenditures Price Index – Clinton P. McCully, Brian C. Moyer and Kenneth J. Stewart, Survey of Current Business, November 2007 (The BEA methodology Kelly's entire analysis runs on: it splits the CPI-PCE difference into formula, weight, scope and "other" effects, and the BEA updates the numbers against it every month)
Listen to the full episode
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: