Skip to main content

Macro Forecast and Review Ledger

This page is an append-only forecast contract. Published event definitions, probabilities, and deadlines are never overwritten. Resolution status, scores, revisions, and reviews are appended. The reasoning remains in the corresponding time slice.

Status Rules

  • open: the resolution date has not arrived;
  • true: the event occurred under the pre-specified source;
  • false: the event did not occur;
  • void: the series ceased, its definition changed materially, or the original rule cannot resolve it;
  • revised: first-release and revised data imply different outcomes, with both scores retained.

By default, resolution uses the first official release available by the resolution date, followed by a revision check 90 days later. Policy rates use the official target effective at the close of the specified date, not market expectations.

Open Forecasts

IDT0Resolution/check dateEvent and definitionpSourceStatusyBrier
M-202608-012026-08-092026-10-31NBS first release of Q3 2026 real GDP growth ≥ 4.0% year over year0.70NBS GDP releaseopen
M-202608-022026-08-092026-11-01At least two NBS manufacturing PMIs for August, September, and October 2026 < 50.00.65NBS PMI releasesopen
M-202608-032026-08-092026-12-31Effective lower bound of the federal-funds target ≥ 3.50%0.70Federal Reserveopen
M-202608-042026-08-092026-12-31Effective Bank of Canada overnight target ≤ 2.25%0.70Bank of Canadaopen
M-202608-052026-08-092026-12-31Fed target midpoint minus Bank of Canada overnight target ≥ 125bp0.65Fed + Bank of Canadaopen

After resolution, a binary forecast receives (py)2(p-y)^2. If a 70% event fails, its Brier score is (0.700)2=0.49(0.70-0)^2=0.49; if it occurs, the score is (0.701)2=0.09(0.70-1)^2=0.09.

Review Log

2026-08-10 — Ledger Opened

  • Registered five forecasts from the 2026-08-09 baseline.
  • All remain open; no unreleased data were used to settle them early.
  • The first leading check is scheduled for 2026-09-09. China's Q3 GDP is expected to be the first resolvable event.
  • There are zero settled observations, so no mean score or calibration conclusion is reported.

Every later review appends new data, due status, outcome, score, error class, and any prospective method change. Even a clean sweep requires checking whether the probabilities were merely too conservative.

Calibration Dashboard

Through dateSettled NMean Brier50–60% hit rate65–75% hit rateLargest issue
2026-08-100No sample yet

Wait for at least 20 resolved forecasts before judging calibration. Before then, describe individual mistakes rather than presenting short-run luck as a methodological edge.

Resolution Template

### YYYY-MM-DD — <review window>

- `<ID>` → true / false / void / revised
- Resolution source and release date:
- First-release value:
- Revised value, if any:
- Brier score:
- Error type: direction / probability / definition / timing / revision / omitted variable
- What changes in the next slice (never retroactively):