← Research
Research · Fed watching, scored

Are Markets Underconfident in Predicting the Fed?

Every FOMC meeting since February 1994 · data frozen 14 September 2026

Before each FOMC decision the fed funds futures market puts a number on what the Committee will do. This is the record of those numbers against what actually happened — 264 meetings, measured at the close two business days before each decision.

Who wrote this. The research, the code, the charts and almost all of the writing here (with one minor exception, noted by its attribution to me) are Claude’s (Anthropic’s model), working under my direction: I asked the questions, chose the framing and checked the numbers, and it did the rest. I am posting it because I found it useful, not because I wrote it. Errors are mine for publishing them.

The short version: Market pricing of Fed moves has been systematically underconfident. Every one of the 63 times from 2000 through July 2026 that the market has put the probability of a move at 60 percent or higher, the Fed has moved — including all 12 times the pre-meeting probability was between 68 and 90 percent. There has never been a near miss either: the highest reading going into a meeting where the Fed held is 59 percent, and the next highest is just 39 percent. And conversely, all 141 times the probability of a move was below 50 percent the Fed held — including all 23 times it was between 10 and 50 percent. (Jason)

The full write-up explains how the series was built, how far back it can go and why, and the one thing worth knowing before treating these as probabilities at all.

Read the analysis →

The record, meeting by meeting

Click any column heading to sort. Filters stack within a row — pick as many bands, outcomes or eras as you like. The four figures recompute with whatever is showing.

—meetings shown
—changed rates
—mean implied probability of a move
—mean absolute surprise
Probability band
Outcome
Era
Meetings

Defaults show the 251 scheduled meetings with a usable reading.

Meeting↕Era↕Priced as of↕P(move)↕Expected↕Outcome↕Actual↕Surprise↕P(actual outcome)↕Called it?↕Source↕

Reading the columns. P(move) is one minus the probability the market put on no change. Expected is the probability-weighted expected change in basis points; Surprise is actual minus expected. P(actual outcome) is the probability on the exact outcome that happened — it is zero whenever the Fed moved more than one quarter-point step from the expected rate, which is a limit of the method rather than a market view. A ! marks a meeting that defied its reading.

Sources. 2015 onward is CME’s own FedWatch end-of-day data. Before that the same calculation is rebuilt from daily settlement prices for individual 30-day fed funds futures; over the 67 meetings where both can run they agree to a mean of 1.0 basis point. Reliable readings hides five meetings in 2007–08 where the effective funds rate had detached from the target far enough that futures no longer identify it. Unscheduled meetings have no probability and are hidden by default. The data is frozen at 14 September 2026 and runs through the meeting of 29 July 2026; it is a snapshot, not a live feed, and is not updated as new meetings happen.