How the index is built

What the number means

The Recession Conditions Index is a resemblance gauge. It asks one question: how closely do today's leading indicators read like they did in the year before past US recessions? A score of 0 means conditions look nothing like prior pre-recession periods; 100 means they strongly resemble them.

It is not a forecast and never outputs a probability. It describes the present in historical terms — it does not predict the future. This distinction is the whole design: stating "conditions resemble 2006" is honest; stating "a recession is X% likely" is a claim this gauge deliberately refuses to make.

The five inputs

Each indicator is scored 0–100 by where today's value falls between a calm-expansion anchor and a pre-recession anchor, then blended by weight.

The yield curve carries the most weight because it is the most-studied pre-recession signal in the literature. But it has a known flaw — it inverts early and often un-inverts in the months just before a recession actually begins. That's why faster-moving credit and labor data are weighted alongside it rather than relying on the curve alone.

How each indicator is scored

Calibrated indicators (yield curve, claims, CFNAI)

Anchors are derived from the historical distribution of each series since 1976. The pre-recession anchor is the level typically reached in the 12 months before recessions began. The calm anchor is the level during genuine expansion — and here a subtlety matters that the audit surfaced.

Anchors use robust quantiles rather than means, so single extreme outliers (COVID drove claims to ~6 million and CFNAI to −18 in weeks) don't distort the scale.

Fixed-anchor indicators (HY spread, sentiment)

Two inputs can't be distribution-calibrated and use documented fixed thresholds instead:

• HY credit spread: the FRED API serves only ~3 years of this ICE BofA series, too short to calibrate against 2001/2008/2020. Fixed anchors: ~350 bps (calm expansion) to ~750 bps (pre-recession stress) — levels well-established in credit literature.
• Market sentiment: the live FearGreedChart index has no deep recession history. Fixed anchors treat greedy/complacent readings (~70) as calm and fearful readings (~30) as recession-like. Sentiment is treated as coincident here, not contrarian — falling sentiment accompanies deteriorating conditions. It carries the lightest weight (10%) precisely because it is the noisiest recession signal.

Honest limitations

It leans on three fully-calibrated inputs. Of the five, three (yield curve, claims, CFNAI) are calibrated against the recession distribution since 1976; two (credit spread, sentiment) use fixed thresholds. The "calibrated against every recession since 1976" framing applies fully to three of five inputs — stated plainly here rather than overclaimed.

When conditions are calm, it behaves like a single-indicator gauge. If four of five inputs sit near their calm anchors, the composite is driven by whichever indicator is moving (often the yield curve). That's correct behavior — it means little is deteriorating — but it's worth understanding that a low reading reflects broad calm, not five independent confirmations of safety.

Leading indicators lead imperfectly. The yield curve can invert 6–24 months before a recession or give false signals; claims can spike on one-off events; sentiment whipsaws. The blend smooths this, but no leading gauge has clean timing. Treat the number as one input to your own thinking, not an answer.

Data & updates

Macro series come from MacroRead and FRED; sentiment from FearGreedChart. The index recomputes every weekday evening after the underlying data updates. Calibration anchors refresh monthly (they barely move). Nothing here is proprietary — it's public data, presented as one readable number.

← Back to the index