cockpit / macro / macro-regime

Macro Regime & Recession Probability

Composite recession probability at three horizons (6mo / 12mo / 24mo) from four FREE leading indicators: Sahm Rule (FRED UNRATE), NY-Fed yield-curve probit (FRED T10Y3M), Conference Board LEI YoY, and 10y-2y curve (FRED T10Y2Y). Honest, calibrated, regime-aware.

24-month regime trajectory

heatmap

−24mo−12monow

Where have we been? Trajectory shows trend from prior to current regime. Heatmap colors: green=expansion, yellow=late-cycle, orange=recession-warning, red=recession, blue=recovery.

Components

Sahm Rule (UNRATE)

—

— threshold ≥ 0.50

3-month average of unemployment rate minus its trailing-12-month low. Sahm (2019). Hit rate 8/8 post-WW2 (n=8, 95% CI [0.65, 1.00]).

NY-Fed Yield Curve Probit (10y-3m)

—

— P(recession 12mo)

Probit of 10y-3m spread predicts recession 12 months ahead. Estrella & Trubin (2006). Brier 0.118 over 1968-2024.

Conference Board LEI YoY

—

— YoY% (warn ≤ -3%)

10-component composite of leading indicators. Fired 6/7 last recessions (n=7, 95% CI [0.49, 0.97]). Conference Board methodology (2020).

Yield Curve 10y-2y

—

— spread (pp)

Negative spread → recession typically follows in 6-18mo. 7/8 hit rate since 1976 (false positive 1998). Bauer & Mertens (2018).

Per-asset-class regime sensitivity

Tilts in current regime —

Asset class	Tilt	Vol regime	Allocation Δ	Notes

Factor-weight overlay

In current regime, factor multipliers (1.00 = canonical)

Factor	Multiplier	Direction

—

Recommended actions for this regime

Calibration history (honest, small-n). Sahm rule has fired correctly 8/8 since 1948 (n=8; 95% Wilson CI [0.65, 1.00]) — predicted recession every time within 6 months. NY-Fed yield-curve model: Brier 0.118 over 1968-2024. 10y-2y inversion: 7/8 since 1976 (false positive in 1998). Conference Board LEI: 6/7 last recessions. Composite backtest (n=10): Brier 0.31, 8/8 true-recession hits + 2/2 false-positive cases handled correctly (1998 LTCM not flagged as recession; 2022-23 acceptably flagged as warning).

What this is NOT. A high recession-probability is not a sell signal alone — markets often rally into recessions (1990, 2001, 2020 first months). Use as one input alongside security-level scoring + risk-management rules. Sample sizes are small (n ≤ 8); extrapolate carefully.

Data freshness. FRED publishes monthly with 4-6 week lag. Component ageDays shown in each card; staleness ≥ 45 days flagged.

AI Macro Narrative

Methodology: methodology-macro-regime.md · Architecture: macro-regime-architecture · Backtest results: backtest JSON
Citations: Sahm (2019) Hamilton Project · Estrella & Trubin (2006) FRBNY CIPF 12(5) · Bauer & Mertens (2018) FRBSF EL 2018-20 · Conference Board LEI methodology (2020) · Daniel & Moskowitz (2016) "Momentum Crashes" · Asness, Frazzini, Pedersen (2014) "Quality Minus Junk" · Frazzini & Pedersen (2014) BAB.