Built for teams that track policy shifts and need an early, structured view.
Each meeting is reconstructed from records on individual members, rotating seats, prior debates, statements, speeches, and minutes.
The model produces a logged committee-level call, with a decision trace that can be reviewed after the official statement is released.
The FOMC is the first committee modeled.
The FOMC met July 29, 2026.
The simulation locked a HIKE call on July 20, before the statement.
The Committee's decision differed from the locked call. The comparison below shows the hits and the remaining gaps.
We simulated the July 29 FOMC meeting 19 times using the same public inputs. Nineteen runs is the minimum threshold we use to identify patterns that persist across repeated simulations.
The outcomes were divided: 6 runs ended in a hike and 13 in a hold. The variation was concentrated among a small group of members. Five supported a hike in every run, two consistently supported a hold, and the remaining five shifted across runs. Their positions accounted for most of the difference in the committee outcome.
Members’ simulated policy positions remained closely aligned with their predefined profiles, with a Spearman rank correlation of ρ = 0.95. Although the committee outcome varied, the member-level pattern remained consistent.
As of July 15, CME FedWatch assigned a roughly 15% probability to a hike. A hike occurred in 6 of our 19 simulations, or about 32%. We incorporated this difference, along with the pattern of member-level shifts, into the final call of a hike.
Once the actual decision is released, we will publish where the simulation aligned with the outcome and where it differed.
Calls are locked before each meeting and scored only after the decision is released.
Each component can be traced through the deliberation transcript.
Request access to inspect the process turn by turn.
Each member is modeled from statements, votes, and speeches. Membership and voting seats are updated for every meeting. Rotations and new appointments are applied from their effective meeting onward.
Every source document carries a release date. Each simulation uses only information that was available at the time. Sources include the Beige Book, minutes, statements, and transcripts.
The process runs in four phases: briefing, debate, statement drafting, and vote. The draft is grounded in the policy regime of that period.
Three independent runs across the same 17 historical meetings.
The figures below report mean results.
Per-run breakdowns are available on request.