Yield-curve shape & inversion

Definition

The yield curve plots Treasury yields across maturities; its SHAPE — steep, flat, inverted — is among the most-studied macro signals in finance. Inversion (short yields above long yields, classically the 10-year minus 3-month or 10-year minus 2-year spread going negative) has preceded every US recession since the 1960s, with lead times of roughly 6-24 months. Estrella-Mishkin formalized it: the 10y-3m spread outperforms other financial variables at predicting recessions 2-6 quarters ahead. The KB treats curve shape as a REGIME variable — slow-moving, probabilistic, and with a documented recent complication (2022-24’s longest-ever inversion without an NBER-dated recession inside the classic window).

How it works / structure

• What the shape encodes: long yields ≈ expected path of short rates + term premium; inversion therefore prices EXPECTED CUTS — i.e., the market’s own forecast that policy is tight enough to require reversal; the recession signal is partly the market forecasting, partly tight policy actually causing the slowdown (both channels documented).
• The standard readings: STEEPENING from inversion (“bull steepening” via short-end collapse) has historically been the recession-imminent phase, not the all-clear — the inversion warns, the RE-steepening times; “bear steepening” (long end rising) prices term premium/supply, a different regime (episode-rates-shock-2022 autumn 2023).
• Transmission to equities and credit: inversion compresses bank net-interest margins (fa-sector-banks borrow-short-lend-long arithmetic runs backwards), tightens lending standards with a lag, and historically precedes the indicator-credit-spreads widening rather than coinciding with it — the curve leads, spreads confirm.
• Parameterization (engine-relevant): spread definition (10y-3m per Estrella-Mishkin; 10y-2y common in practice — they disagree at the margin and the choice must be labeled), inversion depth and DURATION (one-day touches vs sustained quarters), and the probit-model recession probability as the academic output form.

When it applies

Regime classification input (alongside regime-rate-environments — curve shape is the rates regime’s leading edge); recession-probability framing for cyclical exposure decisions (macro-business-cycle); bank-sector analysis; bond positioning (steepeners/flatteners are the direct expression — ext-bonds-rates).

Risk profile & failure modes

• Lead-time uselessness for timing (the core limit): 6-24 month leads mean inversion-triggered equity exits have historically forfeited large late-cycle rallies (2006-07, 2022-23) — the signal is regime-grade, not entry-grade.
• The 2022-24 counterexample: the deepest, longest modern inversion was followed by no classic-window recession — post-QE term-premium distortion is the documented suspect; the signal’s base rate now carries this exception, and the KB requires citing it whenever the indicator is invoked.
• Definition shopping: 10y-3m and 10y-2y invert at different times — selecting the spread that fits the thesis is a researcher degree of freedom (quant-backtest-hygiene).
• Small-n honesty: eight-ish US recessions of usable data — the confidence interval on “always precedes” is wide, and international evidence is weaker (documented in the follow-on literature).

Evidence & limits

Estrella-Mishkin (1998) is the canonical probit study; the NY Fed publishes a live version of the model. The mechanism (expected cuts + causal tightening) is well-understood; the base rate is strong but n is small, and the post-QE era adds a live structural caveat the KB carries explicitly.

Falsifiable-thesis examples

Illustrations only, not signals:

• “A 10y-3m inversion sustained >3 months is followed by an NBER recession within 24 months (classic Estrella-Mishkin window)” — falsified by a no-recession case (one now exists; the thesis’s base rate must count it).
• “Bank index net-interest margins compress within 4 quarters of sustained inversion (transmission check)” — falsified by the margin series.

Cross-references

• The regime frame: regime-rate-environments, macro-business-cycle
• The confirming credit signal: indicator-credit-spreads
• The instrument mechanics: ext-bonds-rates, event-fomc
• The recent counterexample: episode-rates-shock-2022