The June Dot Plot Already Said a Hike Was Coming. The Market Took Ten Weeks to Believe It.

On 17 June 2026 the Federal Open Market Committee published a median projection of 3.8% for the federal funds rate at the end of this year. Because the SEP expresses that median as the midpoint of a target range, 3.8% points to 3.75%-4.00% - one quarter-point above the 3.50%-3.75% range the Committee held that day and still holds now. In plain terms, the Fed's own central projection said a hike was coming, and said it in a document anyone could download.

Ten weeks later, on 28 August, CNBC described the September decision as a coin flip. By mid-September, CME futures were pricing a quarter-point increase at somewhere between 80% and 90%. The projection did not change across those ten weeks. Nothing was revised, withdrawn or amended. What changed was how much the market was willing to pay for information the Fed had already given it.

Wednesday's decision is close to fully priced, which means the decision is not the story. The Summary of Economic Projections published alongside it is. And reading that document usefully requires knowing something about how its headline number is manufactured.

Eighteen dots, not nineteen

The FOMC has nineteen members. The June 2026 SEP contains projections from eighteen of them. The missing one is Kevin Warsh, the new chair, at his first meeting - he did not submit projections at all. The St. Louis Fed's own FRED Blog notes this plainly in its June write-up: the release 'includes projections from 18 of the 19 members of the FOMC.'

That is not a footnote. The dot plot is read as a guide to where policy is heading, and the single person whose views move markets furthest - the chair - has no dot in it. Whatever the June distribution showed, it showed it without the one participant who would spend August moving the market more than any data release did.

It also changes the arithmetic. With an even number of submissions, the published median is not any participant's actual view; it is the midpoint between the ninth and tenth values. Nobody necessarily holds it.

A median is a summary statistic, and summary statistics throw information away

Three things about the dot plot are routinely misread, and all three follow from the same property.

• It is not a vote. Participants submit their individual assessment of appropriate policy, each conditional on that participant's own economic forecast. Those forecasts differ from one another, so the dots are not even answers to quite the same question.
• It is not a committee forecast. No body ever deliberated over 3.8% and adopted it. The number is computed after the fact from eighteen independently submitted values.
• It is not a commitment. The SEP is explicitly a projection of appropriate policy under each participant's own outlook, and outlooks change.

The consequence that matters for reading Wednesday's release is that the median tells you where the middle sits and nothing whatsoever about how the rest is arranged. A 3.8% median is equally consistent with a committee split down the middle between hawks and doves, and with eighteen people who independently arrived at almost the same number. Those are radically different committees with radically different odds of moving again in December, and they publish an identical headline figure.

June was closer to the first case than the second. Reporting on the June SEP indicates that nine of the eighteen submitting participants projected at least one hike in 2026 - three penciling a single quarter-point and five a half-point - while one projected a cut. The median moved up to 3.8% from 3.4% in March. Both facts are true, and only one of them made headlines.

The ten weeks

Set the published projection against the market's price for the same outcome.

• 17 June: the Committee holds at 3.50%-3.75% and publishes a 2026 median of 3.8%, implying one hike. The document is public within minutes.
• 28 August: after Chair Warsh's Jackson Hole remarks, CNBC characterises the September decision as a coin flip. Roughly ten weeks after the Fed published a projection implying a hike, the market rates it a 50-50 proposition.
• Early-to-mid September: hotter August data - headline CPI at 3.4% year over year, core up 0.3% month over month - lands, and pricing moves decisively. CME futures reach the 80%-90% region.

It is worth being precise about the phrase 'the market', because it is not one number. In a single snapshot reported by Yahoo Finance as CME FedWatch crossed 60%, Kalshi traders read 57% and Polymarket traders 49%. Different venues, different participants, different collateral and fee structures, eleven percentage points of disagreement about the same binary event. One aggregator, FedRateCalc, has published readings as high as 98%. Any single figure quoted for 'the odds' is a reading from one venue at one moment, and the spread between venues is itself information.

The case that the market was right to discount it

There is a serious argument that none of this is a failure, and it deserves stating at full strength before it is tested.

First, a projection is conditional and the SEP says so. A participant projecting 3.875% is saying that would be appropriate if their forecast holds. Discounting it is not ignoring it; it is applying the document's own terms.

Second, the June SEP came from a committee whose brand-new chair abstained. A dot plot missing the chair's dot is a genuinely weaker guide to what the chair will do, and pricing it at less than certainty was reasonable rather than obtuse.

Third, and most forcefully: the June dots predate the August inflation prints. A forecast that turns out correct for reasons that arrive afterwards was not, at the time it was made, better information than the market had. On this reading the sequence is ordinary Bayesian updating and there is nothing to explain.

What that argument does not account for is the shape of the move. The largest single repricing followed a speech on 28 August, not a data release. If the market had simply been waiting for the data, the August CPI print would be the pivot. It is not.

What Wednesday can and cannot tell you

The 2026 dot is close to spent. The number worth reading is 2027, where the June median sat at 3.6% - below the 3.8% projected for 2026. That encodes a hike-then-cut path. If Wednesday's 2027 median flattens toward the 2026 figure, the Committee is signalling that the coming increase is not a one-off correction to be unwound, and that is a materially different world from the one currently priced. The 2028 median, 3.4% in June, sets the far end of the same curve.

Three things the median still will not tell you. It will not show dispersion, so a unanimous committee and a bitterly split one look identical. It will not show intensity, since a participant one basis point from the line counts the same as one a full point away. And it will not tell you who holds which view - the dots are anonymous, and the set of participants who submit projections is not the set who cast votes.

None of that is a defect. The SEP publishes the full distribution table and the central tendency alongside the median precisely because the median alone is insufficient. The document is more honest about its own limits than most of the coverage that follows it.

So the practical instruction for 2:00pm ET Wednesday is narrow: skip the headline dot and open the distribution table on the following page. Count the dots above and below the median for 2027. That count, not the median, is what determines whether the Committee has one more move in it or several - and unlike the rate decision, it is not priced.