Methodology
A summary of how BCR research is produced, graded, and bounded. A full methodology paper documenting the validation record is in preparation.
The transmission chain
Every judgment runs from event to mechanism to downstream consequence to observable indicators to a falsification condition. The mechanism carrying a claim is named, the observations that would confirm it are listed, and the observation that would kill it is written down before the outcome is known.
Estimative language
Confidence words map to fixed numeric ranges and are held constant across every publication, so a phrase like likely means the same thing in issue seven that it meant in issue three. A conclusion states what the evidence supports rather than more than it carries.
Falsification and grading
Research calls are registered with dates, resolution windows, and where applicable pre-committed probabilities. When a window closes, the call is graded by hand and published on the Forecast Scorecard whichever way it went. Nothing is backfilled: where a date or a confidence level was never recorded, the record says so rather than estimating one.
The data pipeline
An automated pipeline pulls 290+ economic and market series directly from primary public sources, including the Federal Reserve, the U.S. Treasury, the CFTC, the FDIC, and the BIS, refreshed daily and versioned. No proprietary or non-public data is used.
What is published and what stays private
The regime model's outputs, the trigger levels, and the graded record are public. The full voting method inside the regime classifier is proprietary. Where a surface is public it shows real readings or honest gaps; nothing is estimated to fill a hole.
Analytical boundaries
BCR owns the scenario architecture, consequence analysis, and decision framing. Specialized technical claims are validated with appropriate domain specialists.