Research methodology

1. Sources and missing data

US macro series come from FRED, which republishes BLS, BEA, Federal Reserve, Treasury and other providers. Each indicator links to its series and displays its reference period. Missing responses show “Unavailable”; no constant replaces a failed observation. A delayed observation is flagged after 10 days for daily, 21 days for weekly, 75 days for monthly or 150 days for quarterly series. These are operational thresholds, not official release deadlines.

Market quotes come from Yahoo Finance and may be delayed. A snapshot collection time does not prove the last trade time. The calendar is an estimated schedule and must be verified against official release calendars. CME probabilities remain unavailable until a verified current feed is connected.

2. Transformations and liquidity

• Year-over-year change: (index ÷ index from the same month one year earlier − 1) × 100.
• Month-over-month change: (index ÷ prior calendar month’s index − 1) × 100. CPI and PPI series used here are seasonally adjusted; official headline annual figures may use unadjusted series.
• Real GDP annualized quarterly growth: ((quarter ÷ prior quarter)⁴ − 1) × 100.
• Payroll change: current payroll level − prior month’s level, in thousands of jobs.
• A change in an inflation rate or yield is expressed in percentage points, not percent growth. One percentage point equals 100 basis points.
• Liquidity proxy: WALCL/1,000 − WTREGEN/1,000 − RRPONTSYD, all in USD billions. Weekly asset dates are aligned to the latest component observation at or before that date, allowing at most a 10-day gap. M2 is separate. This is a proxy, not a measure of all market liquidity.

3. Regime classification

Growth uses industrial-production year-over-year growth; inflation uses CPI year-over-year inflation. Each is standardized as z = (current − historical mean) ÷ historical standard deviation over up to 33 years of available monthly observations. The four quadrants reflect above/below historical norms, not whether the economy is literally growing or contracting: positive growth z and negative inflation z = Goldilocks; both positive = Reflation; both negative = Deflation; negative growth z and positive inflation z = Stagflation.

The existing risk-off override takes precedence if Sahm ≥0.5 pp, high-yield OAS ≥5%, or the recession risk score ≥65. Missing CPI or industrial-production observations suppress classification. Asset-allocation commentary is a qualitative research framework; no measured strategy performance is claimed.

4. Recession risk score

The 5–95 score is a bounded, rule-based sum of a 10Y–3M yield-curve base and step adjustments for the Sahm indicator, the NFCI z-score and the three-month change in unemployment. NFCI is standardized over 52 weekly observations. Thresholds and contributions are configuration rules, not an estimated event frequency. “33/100” does not mean a validated 33% chance of recession. Missing inputs suppress the score. The Sahm input is a published recession indicator; the combined score is an unvalidated custom heuristic.

5. Histories, revisions and daily changes

The explorer contains up to ten years of the selected series. Histories use the currently published, revised series and are not point-in-time vintages. They are unsuitable for claiming a backtest that knew those revisions in advance. Snapshot comparisons use the last saved New York daily snapshot on or before the requested date, display the exact baseline date and omit missing values. Before a baseline exists, no historical snapshot is invented. Latest-observation comparisons use each series’ own cadence, so a daily Treasury change and a monthly CPI change cover different intervals.

6. Home-page animation

The fan graphic is an illustrative, deterministic animation generated from seeded paths. It is not fitted to economic data, a Monte Carlo forecast or a confidence interval. Its colors and paths encode no estimated probability.