On Wednesday afternoon, the market did two things that are not supposed to happen at the same time.
It lowered the odds of a near-term rate hike, and it drove the 30-year Treasury yield to its highest level since 2007.
The two-year yield fell. The long end sold off. Odds that the Fed leaves rates unchanged in September jumped 20 percentage points to 45% on the CME.
That is not a market pricing a direction. That is a market pricing a machine that can no longer produce one.
That is not a market pricing a direction. That is a market pricing a machine that can no longer produce one.
Here is the thing most readers get wrong about Kevin Warsh's decision to abolish forward guidance. It was sold as flexibility — a Fed freed from its own promises, able to react to data instead of defending stale projections.
What it actually did was remove the only instrument that ever translated internal disagreement into a single readable path.
Last week, in the "hike-or-hold machine" thread, the point was that the committee had become a mechanical device: soft data in, hold out; hot data in, hike out. This week the machine got fed an input it cannot resolve, and the absence of guidance means it has no way to tell you what it did with it.
Start with the mechanical fact. Rates have sat at 3.5%-3.75% for five straight meetings. The July vote was 9-3.
Three named voters — Beth Hammack of Cleveland, Neel Kashkari of Minneapolis, and a third — reportedly dissented, with Hammack reportedly dissenting in favor of a hike. That is a hard, countable minority.
But a 9-3 vote used to be the beginning of a sentence, not the end of one. The chair would take the dissent and metabolize it into guidance: we hold now, but the balance of risks tilts toward tightening, and here is roughly when. Dissent became a vector.
Warsh removed the second half of that sentence on purpose. "I understand the desire for rolling forecasts and commentary from this committee," he said in his opening remarks, "but for our part, we need to observe market reaction to developments, direct and unfiltered."
Direct and unfiltered. The problem is that the committee itself is neither.
Because the split isn't binary — it's a spectrum, and that spectrum no longer maps to any signal.
At one end you have Hammack, a sitting voter who dissented in July for an immediate hike, worried that policy isn't actually restrictive and that businesses will keep borrowing and pushing prices up.
At the other end you have Austan Goolsbee and Tom Barkin, who think the recent inflation — the tariff pass-through, the oil spike from the Iran war — is a set of one-time shocks that will pass, making further tightening unnecessary. Goolsbee said as much on Fox: he expected these to be one-time disruptions, and he hopes they're soon behind us.
Under the old regime, the median dot and the chair's language would have adjudicated between Hammack's urgency and Goolsbee's patience. You'd get a number, a lean, a rough date.
Now you get a "family fight" — Warsh's own phrase — and, as CNN put it, he's leaving it to Wall Street to choose between dueling narratives.
This is where the oil shock becomes uniquely corrosive, and it's the new input since last issue.
Most macro shocks push the committee one way. A demand collapse is dovish. An overheating labor market is hawkish. The machine can process those.
An oil shock feeds both wings at once. It hands Hammack and Kashkari a live inflation threat — energy prices spiking, headline running hot, exactly the environment where waiting looks reckless.
And it hands Goolsbee and Barkin their cleanest possible case, because an oil spike from a specific war is the textbook definition of a one-time, fading shock that the Fed should look through rather than chase.
The same barrel of oil is simultaneously the hawks' reason to move and the doves' reason to wait. It is a genuinely two-sided input.
Forward guidance was precisely the tool built to adjudicate a two-sided shock — to say, in effect, here is how we're weighing the inflation impulse against the fade, and here is where that nets out.
Warsh threw away the adjudication tool at the exact moment the committee got the one input that most requires adjudicating. That is the whole story.
Now watch how the outside world tries to price a machine with no output signal.
Goldman's Jan Hatzius calls a September hike "very unlikely." His argument isn't one data point — it's a run of misses, soft payrolls, weak retail sales, decelerating inflation, and his read that none of the committee's doves has any reason to flip toward a hike.
Goldman has hike odds around 30% and sees the next move pushed from December to January.
But the market as a whole is still closer to a coin flip — two-year yields holding above 4%, unchanged-odds at only 45% even after Wednesday's 20-point jump.
That gap between Goldman's 30% and the market's near-even stance is not sloppiness. It is the market holding two mutually exclusive committees in its head because the chair refuses to say which one is real.
Half the distribution believes the doves win and the fade materializes. The other half believes the hawks are right and the oil impulse forces a hike later, harder.
With no median to collapse those two worlds into one, the front end prices both. That's a bimodal path — a stalled rate today sitting on top of an unresolved fork tomorrow — and it shows up as a market that can move the near-term hike odds down and the long-end yield up in the same afternoon.
The long-end move belongs to the sibling piece on the 30-year and term premium. Here it matters only as the symptom that proves the disease: when the front end can't resolve the committee's next step, the indeterminacy has to go somewhere, and it went straight into a contradictory two-directional reaction.
Which is why the press conference made things worse, not better.
Warsh had an off-ramp. June CPI printed down 0.4%. He could have seized it. Instead he said it was "not much" of a consideration and called inflation still "elevated."
He then declined to say what would actually make him raise rates, and floated that one of his task forces might downplay PCE as the official target after January. "We are sticking with PCE for now," he said. "But, who knows, come after next January, what we might say about strategy."
So: won't name the reaction function, won't commit to the target the reaction function is measured against, won't translate a 9-3 vote into a path.
Jon Hilsenrath, writing to clients, put it plainly: Warsh "needed to articulate what it would take for him to want to eventually raise interest rates." He didn't. "Warsh didn't convey the message clearly or explicitly, and the bond market puked on him."
AllianceBernstein's Eric Winograd called the presser "confusing and often internally contradictory." JPMorgan's Michael Feroli said the refusals raise "questions about the new chair's credibility in delivering lower inflation."
The irony is exact. Warsh built his critique of Powell around credibility, argued that a credible chair keeps long rates in line, and then delivered a performance where long rates blew out to a 2007 high on his watch.
The deeper point is not that Warsh was unclear on a bad day. It's structural.
He abolished the mechanism whose entire purpose was to convert a divided committee into a legible signal, and he did it right before a two-sided shock made the division unresolvable.
The hike-or-hold machine still runs. It just no longer prints a receipt — and September is now the coin flip that nobody, including the twelve people voting, can call.