The Macro Investing Tool

Warsh is holding out on inflation. The data is about to come his way.

RAOUL'S TAKE · THE COMPRESSION · 31 AUGUST 2026

Kevin Warsh stood up at Jackson Hole on Friday and told us he isn't convinced on inflation yet. He named PCE as the gauge he will act on, said financial conditions are not restrictive, and left a rate hike on the table for the coming months.

The front end took him at his word, so the two-year is sitting at 4.36% against a funds rate of 3.63%, i.e., the market is still pricing the best part of 70 basis points of hikes. So this week the job was to go through every inflation dashboard we have and answer the only question that matters right now: is inflation actually still trending lower? Because that is Warsh's game, as I see it.

He is waiting for the data to confirm the trend, whether he does a rate hike in the meantime or not, and if the data confirms it, the front end falls, and the curve steepens. On his own gauge, he has a point, which is worth being honest about. Core PCE has been flat at 3.3% for four months, so when he says the underlying trends haven't improved, that is what he is looking at. But the year-on-year rate is the rearview mirror, and it is currently carrying the whole March-to-May oil spike inside it.

The place to look is the three-month and six-month annualised rates, because they catch the turn two to three quarters before the annual numbers do, and we have now built them into every inflation series on the dashboard.

They all say the same thing. Headline CPI is 3.3% year on year, but the six-month annualised rate is 3.8%, and the three-month rate is 0.5%. That is not a typo.

The momentum of US inflation over the last quarter was half a percent annualised, down from over 8% in May, because the oil shock came in fast and it is leaving just as fast. Strip out shelter, which lags the real world by about sixteen months, and the three-month rate is minus 0.5%, i.e. prices ex-shelter actually fell over the last quarter.

Core is 2.5% year on year with a three-month rate of 1.7%, and supercore, which is core ex shelter and energy and is the closest thing to the domestically generated part of inflation, is 1.9% year on year with a three-month rate of 1.2%.

Every one of those fans is closing from below, with the three-month under the six-month under the year-on-year, and that's usually what a spike looks like on the way out, not what a new inflation regime looks like on the way in.

The rest of the inflation tab backs it up...

Core goods are at 0.8% year on year, and the three-month rate is zero, so there is still no tariff spiral, and goods lead services down the chain, with services at 3.0% and grinding lower.

Energy tells you where the whole thing came from, as WTI was up 64% year on year in May and is up 18% now, and if oil just sits where it is, those base effects keep falling and eventually flip negative.

The breadth of inflation, i.e. the share of the CPI basket rising faster than 5%, spiked to 34% in May and is already back at 20%, right around its long-run average of 18.6%, so the shock never spread into the basket.

And wages suggest it keeps deflating...

The forward-looking stuff agrees...

The regional Fed price surveys, which lead CPI by around nine months, eased again in August...

and the market never believed it in the first place, with the 10-year breakeven at 2.31% and never above roughly 2.5% even at the peak of the panic.

The one genuinely sticky series is core PPI at 3.6% year on year, though even there the six-month rate has rolled over.

Add the shelter pipeline, where home prices at around 1% year on year feed into shelter CPI with a sixteen-month lag and mechanically drag a 37% chunk of the index towards 2% into 2027, and the picture is clear. This was a supply shock that never propagated into the rest of the basket, and once you look past it, the underlying trend in inflation is still down.

Into the autumn this plays out fairly mechanically, i.e. the August CPI lands on the 11th of September and then the spike months start dropping out of the annual numbers one by one, which means headline falls through the autumn while the momentum data keeps printing soft.

That is the confirmation Warsh says he is waiting for. He may still hike into it, and honestly a hike into falling inflation momentum just steepens the curve by another route, because the market would price the mistake and the eventual reversal at the same time. Either way, the direction of travel for the front end is lower over time, and I still think the hikes priced into the market go to zero.

The entire market's focus now sits on the bond market itself...

For a week in which the Fed Chair floated rate hikes, the long end behaved remarkably well. The 10-year finished at 4.72% versus 4.75% last week, and the 30-year came in from 5.33% to 5.21%, so the long end rallied while the front end sold off, which is the market quietly telling you where it thinks the pressure really is.

Bessent's enlarged buyback operations actually begin on the 9th of September, running through early November, so we are about to find out whether the Treasury intervenes more aggressively, and everything he said in the last two weeks suggests he will.

Our trend Monitor flipped on bonds in August to yields lower, the first constructive bond signal in a while. The curve at 47 basis points has plenty of room to steepen, and the steepener is the whole ballgame, because bank lending is already running at 7.5% year on year and it is the steeper curve that lets the banks turn the deficit into money in the real economy. That remains the mechanism that extends this cycle.

Financial conditions are where the argument gets decided... We are testing two versions of the GMI financial conditions index side by side. The version built off the 10-year yield plus the dollar has tightened dramatically, harder over six months than in all but roughly 8% of its history, and on its usual nine-month lead that tightening would arrive in the economy between November and April.

The version built off three-month rates plus the dollar, which is Julian's original construction, barely tightened at all and has been easing since March.

What settles it for me is the Nasdaq. With its usual 90-day lead, it has tracked the three-month version very well, holding 20% to 30% year-on-year gains straight through an episode that the 10-year version says should have rolled it over by now.

The market is trading the three-month version, i.e., it is treating this as a narrow long-end and oil squeeze rather than a genuine tightening of conditions, and given that, it makes more sense to give this more room and wait rather than treat the dramatic version as gospel. It helps that even the 10-year version has been turning back up since the end of July as the dollar leg faded.

The dollar is the other key measure here, because it's what decides whether any of this tightening actually bites. The tightening we have had came mostly from the long end and from oil, while the dollar has been falling since March and has been the offset, and only two of the four pressure channels are squeezing at once, which is the kind of narrow squeeze the economy has historically absorbed.

If the dollar were to turn and join the squeeze, that changes the whole equation, which is exactly why we would rather see it keep weakening. This week it bounced to 99.4, the bounce we talked about last week, and it held under 100, which is what mattered. Gold at around $4,440, down on the week, is just that same dollar bounce read through the conditions lens, and I would treat both moves as noise inside the trend rather than a change to it.

Meanwhile everybody is in Asheville, where Bessent and Warsh are hosting the G20 finance ministers and central bank governors on Monday and Tuesday, with Iran sanctions, tariffs and global imbalances on the agenda. The output from that lands later in the week and should set the tone, because it is the first time we get the whole cast in one room since the Treasury started managing the long end.

Iran is still the loose end. Oil at $85 is the other channel doing the squeezing; it is the reason the energy base effects haven't finished falling, and the situation needs resolving soon so oil can fall and the dollar can continue its move lower. Nothing about my expectation there has changed, but the clock matters more now, because the midterm electioneering window opens in a few weeks and everything above works better with oil at $70 than at $85.

And none of this is happening in a weak economy. The August ISM lands tomorrow, with the last print at 55.6 and new orders ahead of it at 56.7; temp hiring is still positive year on year, lending standards aren't tightening, and high-yield spreads at 2.6% say there is no credit stress anywhere in the system. Strong growth with inflation momentum fading underneath it is the combination that extends this cycle, and it's easy to forget that when the whole market is arguing about one man's reaction function.

And then there's crypto... Bitcoin finally flipped the trend Monitor to rising in August, which was the confirmation we said we were waiting for last week, and it sits at around $78,500. It still comes down to excess liquidity, which needs the curve to steepen, rates to start moving, and the dollar to keep falling. They're all the same trade, and the inflation data above is the thing that unlocks it.

Into next week I'm watching the G20 output from Asheville, the ISM on Monday, the start of the bigger buyback operations on the 9th, the August CPI on the 11th, the dollar staying under 100, and any movement at all on Iran. The data is walking towards Warsh now, whatever gets said in Asheville, and I don't think he can hold out for long.

Raoul