China–US–Canada Macro Scenario Lab

This folder is neither a news digest nor a place to rewrite old calls after seeing the outcome. Each analysis is frozen as an auditable time slice: record what was knowable, state priors and evidence, assign probabilities, and return on a scheduled date to score the result.

Every probability is a research hypothesis, not investment advice. Probability expresses uncertainty; it does not turn an opinion into a precise fact.

File Structure

File	Responsibility	Editing rule
Bayesian inference method	Defines hypotheses, updating, dependence controls, and scoring	Update when the method changes and disclose the change
Forecast and review ledger	Stores resolvable forecasts, deadlines, outcomes, and scores	Append outcomes; never overwrite original probabilities
2026-08-09 baseline	Stores the first China–US–Canada observations, scenarios, and inference	Freeze it; record any correction explicitly
snapshots/YYYY-MM-DD.md	Future time slices	Create one page for every information cutoff

Time Slices

Every slice uses five clocks:

1. T0 / information cutoff: freeze available information, release dates, and definitions.
2. T+30 / leading check: inspect PMIs, labor, inflation, and policy communication without settling the main call early.
3. T+90 / quarterly update: create a new slice and explain how the posterior moved.
4. T+180 / main resolution: mark due forecasts true, false, or void.
5. T+365 / calibration review: summarize Brier scores, overconfidence, omitted variables, and revisions.

Publication lag is explicit: the month in which activity occurred, the statistical release date, and the date recorded here differ. Data released after T0 must not leak into an older slice.

Operating Loop

official observations
  → compare with previous forecasts
  → define exclusive scenarios and priors
  → update likelihoods with evidence
  → normalize the posterior
  → write dated, resolvable forecasts
  → settle and review after expiry

Every call has three layers:

• Observation: what an official source directly reported;
• Interpretation: the mechanism through which it may matter;
• Forecast: what observable event may occur, by which date, and at what probability.

Country and Cross-Border Views

Each slice checks:

• China: domestic demand, property, manufacturing, exports, and fiscal/monetary transmission;
• United States: consumption, labor, inflation, productivity, fiscal impulse, and the Fed reaction function;
• Canada: household rate sensitivity, housing, population, energy, Canada–US trade, and Bank of Canada policy;
• Cross-border: US–Canada rate differentials, the dollar cycle, commodities, tariffs, and supply chains.

Country analyses cannot simply be added. The same oil shock may raise inflation in the United States and Canada while supporting Canadian nominal income through its terms of trade. One US demand release may affect both Canadian exports and the Canadian dollar. Correlated evidence receives one effective vote.

Reading Path

Read the Bayesian inference method, inspect the current baseline, then visit the forecast and review ledger to see whether the calls survive contact with time. Quarterly updates add snapshots rather than deleting old ones; otherwise review quietly turns into hindsight.