Steno Signals

Happy Monday, and welcome back to my editorial on everything macroeconomic!

This is a week with one data point that matters more than the rest, so I will build the editorial around it. The September CPI lands on Wednesday. Consensus is for headline at 0.6% on the month and 3.6% year-on-year, and core at 0.2% and 2.5%, with both annual rates RISING.

The Fed has told us, through Warsh and Williams, that this is the number it hikes on: the headline, the breadth, and whether the monthly core crosses the 0.25% milestone. Growth and labour have been told to wait. So Wednesday decides whether one more hike becomes two (for the rest of the year), and whether the curve inverts in the autumn or in the spring.

Today I will address three questions:

What does our nowcast say about Wednesday’s CPI, and how much do I believe it?

What is Trump actually doing with all these ceasefires, and why do they matter for the inflation outlook?

Is this inflation high enough to kill the cycle yet?

Follow along below.

Theme 1: The CPI Preview: Mildly Below Consensus, and Not a Hill I Will Die on

The Nowcast IQ projection for September has headline CPI at 0.55% on the month and core at 0.14%. Consensus is 0.6% and 0.2% on the median, 0.55% and 0.24% on the average. So we are mildly below on both, and the difference sits almost entirely in the core.

The headline is energy and nothing else: Our model has energy commodities up 12% on the month, gasoline up 12.1%, fuel oil up 11.2%, which contributes about half a point on its own. Food is quiet at 0.13%. Strip the fuel out and the picture is the one we have described for months. Core goods are flat to negative at -0.02%, with used cars down 0.85% and medical commodities down 0.2%. Core services run at 0.20%, with shelter at 0.23%, rents at 0.24% and owners’ equivalent rent at 0.22%, which are soft numbers by any post-Covid standard.

The only warm lines in the core are transportation services at 0.34%, which is jet fuel showing up in airfares, and alcohol, which I will not pretend to explain. That is a 0.14% core.

横向滑动查看完整图表

Now the conviction. It is not high, and I want to be explicit about why, because regular readers know I have been a little too low on core twice this year, after otherwise having had the best forecasting track record of basically all economists tracking the U.S. economy. Two things argue against our number. The first is seasonality in the surprises themselves. U.S. CPI has a calendar pattern that the sell side does not seem to have noticed: In the spring, prints come in LOWER than consensus far more often than not; from September onwards they come in HIGHER. Residual seasonality, methodology, whatever the cause, the autumn is when the surprise skews against a soft call.

And September is the worst month of all for a soft call. Look at the bar for the September print in the chart below: Since 1998, it has come in ABOVE consensus 44% of the time, in line 39%, and below just 18%. That is the most lopsided month on the calendar, worse than October, and the exact opposite of March and May, where downside surprises run above 50%.

Which brings me to the chart I added on Sunday, and it cuts both ways. Our daily core nowcast sat around, or just below, 2.4% year-on-year during the September reference week, the days the BLS actually collects the prices, which is why our projection has core at 0.14% on the month and 2.4% on the year against a 2.5% consensus. So the live data backs a soft print for THIS report. But look at what the same series has done since: It has re-accelerated from about 2.15% in late August to roughly 2.55% today, a move of four tenths in six weeks, on the prices-paid components and the pass-through from the crack. Momentum in our own data is up, not down. That is the second reason the conviction is low. We may well get the soft core on Wednesday, and then spend the October print, out in November, giving it back.

The recent history is even less kind. The September 2022 print was the one that blew up the autumn: Core at 0.6% against 0.4% expected, and the curve inverted within the week. September 2023 beat on the headline. September 2024 beat on both headline and core by a tenth. Only last year, September 2025, delivered the downside surprise we are looking for, and that was a print delayed into late October by the shutdown, which nobody should build a model on.

So four of the last five Septembers have surprised to the upside, and the one that did not was a statistical orphan. The mechanics are partly known: The autumn apparel and education resets, airfares and lodging reversing the summer pattern, and a BLS seasonal-adjustment process that has under-corrected for post-Covid price-setting since 2021. Whatever the cause, when my own chart tells me the month I am calling soft is the month that disappoints the soft callers most often, I listen to the chart.

The second is that core is already running above the model in the official data: The PCE decomposition two weeks ago and the September CPI consensus both imply a core that is firmer in the books than it is in our live readings. The live readings have been right on direction all year, and wrong on timing twice. So, I take the nowcast as a reason to lean, not to bet. A 0.1% core would be a dovish surprise that pulls the second hike completely out of the question. A 0.3% core would be a hawkish one that puts the second hike in and the curve inversion on the autumn calendar. The base case is in between, and the asymmetry is not strong enough for a big position either way.

What I do have conviction on is the direction of the core over the next two quarters, and the reason is the chart we have banged the drum on since September. Core PCE now sits about 0.6 points above core CPI, close to a record. That only happens when core is trending down, because PCE, with its stickier imputed components, lags the CPI on the way down; 1983 is the parallel. Whatever Wednesday prints, the PCE-CPI spread tells you the direction. As long as Warsh is anchored to the PCE, his life is difficult, but the CPI leads and the PCE follows.

Theme 2: Trump’s Numerous Ceasefires... All of Them Aimed at the Crack Spread

Count them. In April, a two-week ceasefire with Iran, conditional on Hormuz reopening, brokered by Pakistan. In May, a three-day Russia-Ukraine truce for Victory Day. In June, a renewed Israel-Hezbollah ceasefire. In July, the interim truce with Iran collapsed. And this weekend two more: an immediate “energy ceasefire” between Russia and Ukraine, under which both sides halt attacks on each other’s energy infrastructure, announced on Truth Social with the instruction “Do not break it!” (so far, not followed), and a deal with Putin to allow Russian diesel exports to reach American and global markets, which Kyiv is furious about. Add the statement on Thursday that the U.S. will not attack Iran before the midterms, and the expectation that talks with Tehran resume this week after Trump rejected the seven-day proposal. That is a lot of ceasefires for a president who says a deal is not a priority.

There was one small complication with Saturday’s energy ceasefire: neither side appears to have known about it. Kyiv’s first reaction was to ask what, precisely, had been agreed, since nobody in the Ukrainian government had agreed to it. Moscow said it had received no proposal and would study one if it arrived. Which leaves a ceasefire between two parties, declared by a third, with the instruction “Do not break it!” addressed to people who had not been told it existed. I have been in enough negotiations to know that this is, technically, a method: You announce the deal, and then you negotiate it. It is how Donald Trump got the Iranians to the table in April, and it is how he got the Israelis to Hezbollah in June. The market, to its credit, has learned to trade the announcement and read the fine print later. Diesel cracks fell on the headline before anyone in Kyiv had finished the sentence “we were not consulted”. The ceasefire may yet turn out to be real; a lot of Trump’s do, eventually, once the people on the ground have caught up with the Truth Social post. But for now the most accurate description is that Russia and Ukraine are observing an energy ceasefire in the same way my children observe bedtime: It has been announced, in capital letters, by someone who is not in the room.

Every one of these moves targets the same variable: The price of refined products at the American pump, four weeks before the midterms. The Russia-Ukraine energy ceasefire stops Ukraine hitting Russian refineries, which is the single largest source of lost refining capacity in the world, around 3 mb/d at the peak of the strike campaign. The diesel deal with Putin puts Russian product back into the market that the sanctions regime took out. The promise not to attack Iran keeps Gulf crude flowing, which it already is at 93% of last year’s level. None of this is about crude. Crude is fine; we have said so for months. It is about the crack spread, the margin between crude and the gasoline and diesel it becomes, which is above the 2022 peak and is the reason the headline CPI prints with a 12% monthly gasoline number while the core prints 0.14%. Trump has worked out that he cannot nowcast his way to a lower CPI through Hormuz. So he is going after the refineries, Russia’s by ceasefire and ours by export ban.

Two consequences. First, if the energy ceasefire holds for even a month, the crack spread comes in, because 3 mb/d of Russian refining capacity stops being hit and starts being repaired, and because Russian diesel resumes flowing to the Atlantic basin. That is the single most disinflationary thing that could happen to the headline CPI between now and the midterms, and it is exactly what we have said a Hormuz products deal would do; Trump has found a cheaper way to the same outcome.

Second, it tells you something about the Iran track too. A president who hands out ceasefires at this rate is a president who wants the energy problem gone before November 3 and does not much mind who signs. Tehran is broke and pushing for the phased reopening; Trump has told them he will not attack but has not told them he will deal. The talks resume this week. We are past peak escalation, and the ceasefire habit is the evidence.

Theme 3: Why Inflation Is Not Yet High Enough To Kill the Cycle

Which brings me to the question underneath all of this: is the inflation we have, headline above 3%, energy up 12% on the month, a central bank hiking into it, enough to end the cycle? Two charts say not yet, and both are about proportion rather than direction.

The first is the necessity impulse against the ISM. Our basket of input costs for food, energy and chemicals producers, the things an economy cannot do without, is up more than 25% on the year, and the impulse leads the ISM by about fourteen months. That sounds bad until you put 2022 next to it: the same basket was up 100% or more then, two to three times as severe as now, and the ISM duly fell to the low 40s. Today the two are “worlds apart”, as the chart says: a 25% necessity shock is the kind of thing an economy absorbs through margins and substitution, not the kind that forces a recession. The impulse has also peaked and started to roll, which the 2022 version did not do until the Fed had hiked 400 basis points.

The second is the ratio of inflation surprises to economic surprises, which I use as the sell signal for equities.

The logic is simple: When inflation keeps surprising to the upside faster than the economy does, the central bank is forced to tighten into weakness, multiples compress and the earnings do not arrive to rescue them. The ratio leads the MSCI World by about four months, and when it spiked in 2008, 2015 and 2021-22, equities followed it down within the quarter. It is not spiking now. Inflation is surprising, but the economy is surprising alongside it; growth is holding up in Europe and Japan and the ISM is still above 50 in the U.S. That is not the setup for a bear market. It is the setup for a sideways, narrow, frustrating market, which is what we have, until one of two things happens: the inflation surprises fade, which Wednesday and the ceasefires decide, or the economic surprises collapse, which the ten-month lag from the hikes decides, and which is a 2027 question.And for now, the model, which is by far my best forward-looking indicator for equities, actually suggests that we trade more than 25% higher year-on-year by early 2027, which leaves a few very decent months ahead. That is also my base case. We will see surprisingly strong returns through the earnings season that starts in earnest this week, and the market is not ready for it.

And the regime engine agrees. The U.S. rising-inflation probability has crept back to 44% on the prices-paid components and the crack, but it remains below neutral, while the growth probability has fallen to 19% and liquidity holds at 74%. That is still QE-like: inflation risk below neutral, growth fading, liquidity intact.

Lame Duck QE, the regime where inflation wins and the dollar with it, needs the inflation probability above 50%, and that is what a hot core on Wednesday could potentially deliver. So the regime, like the trade, hangs on the core.

The bottom line: Our nowcast has the September CPI mildly below consensus, 0.55% headline and 0.14% core, and I hold it loosely, because the autumn calendar skews surprises to the upside and the official core has run warmer than our live data twice this year. Lean, don’t bet.

Trump’s ceasefires, the energy truce in Ukraine, the diesel deal with Putin and the no-attack promise on Iran are all aimed at the crack spread, which is the real driver of the headline, and if they hold, they are the most disinflationary thing on the calendar. And the inflation we have is not yet high enough to kill the cycle: The necessity shock is a third of 2022’s and already rolling, inflation surprises are not outpacing economic surprises, and the regime is still QE-like. A hot core on Wednesday potentially changes the regime. A soft one changes the Fed. The base case is neither, and the pain trade stays up.

Have a great week ahead, and good luck out there!