The Fed's 3.8% Median Is a Coin Toss, Not a Dot

At 2 p.m. today the Federal Reserve released the minutes of its 16-17 June meeting, and for once the minutes matter more than usual. They are the only on-record account of a committee that has quietly split down the middle on whether to raise interest rates this year, and whose most famous output, the 'dot plotdot plotA chart the Fed publishes four times a year in which each policymaker marks, anonymously, where they think the interest rate should be at year-end, shown as dots stacked on a rate scale.' medianmedianThe middle value in a list; when there is an even number of values it is the average of the two in the middle, which is exactly what happens here with 18 dots., was built in a way that hides that split rather than revealing it.

The number every trading desk fixated on after the June meeting was 3.8 percent: the median projection for where the federal funds ratefederal funds rateThe short-term interest rate the Fed steers; it ripples out to the cost of mortgages, credit cards, business loans and bonds. ends 2026. In March that median was 3.4 percent. The jump to 3.8 was read widely as the Fed leaning toward a hike, and in one important sense that read is fair: in March not a single policymaker penciled in a 2026 hike, and in June nine did, while the core inflation forecast was revised up six tenths of a point. That is a real hawkish shift, and this piece does not dispute it.

What we wanted to test is narrower and more specific: does the 3.8 percent median itself, the single figure markets anchor to, represent a committee whose center of gravity has moved toward hiking? Open the primary document, do the median by hand, and the answer is no. The level is not a lean; it is the exact midpoint of a committee split nine to nine on whether to hike at all. The direction moved. The median just cannot tell you that half the room still sees no hike.

Four times a year, each Fed policymaker writes down where they think the policy interest rate should sit at the end of the coming years. Those anonymous guesses are plotted as dots stacked on a rate scale, the 'dot plot.' Markets do not really trade the individual dots; they trade the median, the middle dot, because it is treated as the committee's center of gravity. Rate expectations across mortgages, corporate loans and bond prices take their cue from where that one figure lands.

One mechanical detail matters here. Dots can only sit at the midpoints of the Fed's quarter-point bands: 3.375, 3.625, 3.875, 4.125, and so on. There is no 3.750 option on the grid, it is a band boundary, not a place a policymaker can mark. Keep that in mind.

The claim only stands if the published dots actually straddle two buckets rather than piling onto one. So step one was to open the June 17 Summary of Economic Projections and count the dots ourselves. Here is the year-end 2026 distribution, taken directly from the Fed's projection table.

Count the camps by the only question that moves markets, hike or not. Nine dots sit at 3.875 percent or above; that is at least one hike from today's setting. The other nine sit at 3.625 percent or below, eight at 3.625, exactly the midpoint of the current 3.50-3.75 percent target rangetarget rangeThe Fed sets a band rather than one exact rate (currently 3.50-3.75%); the midpoint of that band is 3.625%, which is why a 'hold' vote shows up there. (a vote for no change), plus one lone cut at 3.375. Nine hikes, nine not-hikes. The wires reported the split as nine hikes, eight holds and one cut; that is right too, it just bundles the eight holds and the one cut into a single 'no hike this year' block of nine.

Now do the median. With 18 dots, the middle is the average of the 9th and 10th when you line them up lowest to highest. The 9th dot is the top of the not-hike block, 3.625 percent. The 10th is the bottom of the hike block, 3.875 percent. Their average is (3.625 + 3.875) / 2 = 3.75 percent, which the Fed rounds to 3.8.

3.75 percent is not a shelf anyone declined to stand on, it is not even a valid grid point. It is the arithmetic seam between the highest 'no hike' vote and the lowest 'hike' vote, and it falls there precisely because the committee is split exactly nine to nine. That is real information. It just is not the information markets took from it.

Why is the count 18 and not the usual 19? Because the chair sat it out. Kevin Warsh, confirmed on 13 May by a 54-45 Senate vote, unusually narrow for a Fed chair, and described in the coverage as the most divisive such vote in the Fed's history, declined to submit a dot of his own. In his own words, from the June 17 press conference: 'I, however, have refrained from offering any projections of my own, consistent with my long-held views on the SEP, at least as currently structured.' By the Fed's record he is the first chair to withhold a projection since the dot plot began in January 2012.

Even-numbered dot plots are not exotic. The committee has run with 17 or 18 participants whenever Board seats sit empty, and in those years the median is always an average of two dots. Nobody calls that an artifact. What is unusual here is the specific geometry: the missing dot is the chair's, and its absence leaves the remaining 18 straddling the hike line nine to nine. Put a 19th dot back in, the chair's, and the median stops being a seam and snaps onto a rate someone actually chose.

The two scenarios in the table are not Warsh's only arithmetic options, but they are the plausible ones: 3.625 (hold) and 3.875 (one hike) are the grid values immediately on either side of the empty median, the natural marks for a centrist chair. Either one moves the published 'signal' a full quarter-point in a committed direction. As it stands, 3.8 percent is the one value the plot could show that commits to nothing. The number markets treated as the Fed's center of gravity is balanced on an absence.

None of this denies that the committee turned hawkish between March and June. It plainly did. The cleanest way to see it is to count, in each meeting, how many policymakers projected at least one 2026 hike, using the full committee as the denominator each time.

Zero to nine is a genuine shift, and it did not come from nowhere. The same June projections marked the core inflation forecast up sharply while trimming growth, the shape of a committee worried that prices are proving stickier than output is holding up.

So the argument here is not that the Fed is secretly dovish. It is that the direction (more hawkish) and the level (3.8 percent) are telling you different things, and only the direction is solid. The level is a rounding of a tie. Treat it as a measurement of consensus and it will mislead you, because there is no consensus at 3.8, there is a nine-nine standoff with an empty middle.

There are two honest readings, and they are worth keeping apart. The uncharitable one: Warsh engineered a median that tells the market nothing while the committee privately tilts hawkish, keeping the chair's hands free for September without owning a direction. The charitable one, which we think is at least as well supported: Warsh has spent years arguing against 'forward guidanceforward guidanceWhen a central bank signals in advance what it plans to do with rates, to shape market expectations; Warsh is a longtime critic of the practice.', the practice of pre-committing markets to a rate path, and is simply declining to feed a number he thinks the Fed should not be publishing in its current form. At his confirmation he said flatly, 'I don't believe in forward guidance,' and at the June meeting he launched a review of Fed communications and floated phasing the exercise out entirely.

Both readings share the same practical consequence, and that is the point for a reader trying to trade or plan around this: the headline median has been hollowed out. Whether by design or by principle, 3.8 percent no longer carries the information a normal SEP median would. The only way to recover the committee's actual balance is to do what this piece did, go past the median to the underlying dots, where the real signal is a coin-flip nine to nine, not a hawkish tilt.

And the coin is still spinning. Since June the data have cut the other way: the June payroll report showed just 57,000 jobs added, with unemployment at 4.2 percent, and futures markets pared the odds of another 2026 hike to roughly four in ten. That is exactly the kind of incoming data Warsh says a pre-committed dot would have fought against. It also means the September meeting is genuinely open, which is what a nine-to-nine committee looked like all along, and what a single rounded median at 3.8 percent was never going to tell you.