Business Cycle Index

Two complementary reads of the US business cycle. The Cycle Clock identifies cycle phase from direction-of-change, financial stress, and trend position. The Structural Health read measures broad indicators against a 20-year baseline. They sometimes disagree — and that disagreement is itself information. Updated .

Cycle Position

Cycle Clock — last 10 years

Three sub-components of the Cycle Clock plotted over time. Sign convention is the same across all three: positive = good (above-trend / improving / calm), negative = bad. The composite "level" used for phase classification is a 50/50 weighted average of Trend Position and Stress; Direction is the orthogonal momentum axis. Watch the components individually — when they diverge (e.g., trend positive while direction negative), that's a regime transition.

Cycle Position score — last 5 years

Cycle Position score is the percentile rank of the level component within its trailing 10-year history. 50 = median. Above 50 = above-recent-average activity + financial conditions; below 50 = below-recent-average. The score is a regime-stable summary; the phase label gives the qualitative read.

Structural Health (BCI factor)

The BCI factor measures broad indicator levels against a 20-year baseline (2005+). When it sits below zero — as it does today — it is saying the panel is currently below long-run norm, regardless of the cycle phase. The two reads can tell different stories: today the BCI factor reads negative (below long-run norm) while the Cycle Clock reads MID CYCLE (cycle in healthy expansion). Both can be true: the post-2020 regime sits structurally lower than the pre-2020 baseline (rates higher, sentiment slower to recover, manufacturing soft) and the cycle is in expansion. That divergence — "feels recessionary, isn't recessionary" — is itself a useful read.

Cycle supplement

Global BCI

Country-level business / consumer confidence panel, GDP-weighted into a global composite. Cross-check on the US cycle — divergence is itself the signal.

Cycle supplement

Recession Ensemble

Three-model recession nowcast (yield-curve probit · multi-indicator logit · Sahm-rule sigmoid) weighted into one ensemble probability with published risk bands.

Methodology

This page surfaces two complementary cycle metrics built from the same underlying indicator panel.

Cycle Clock (Fidelity-style phase classifier)

The Cycle Clock combines three orthogonal sub-factors into a 4-quadrant phase rule mirroring the Fidelity / BlackRock / PIMCO business-cycle framework:

Trend Position — weighted average of real-economy indicators (activity, labor, consumer, housing, leading composites) z-scored against a 10-year rolling window. Positive = above 10y trend, negative = below. Excludes credit / markets / inflation, which belong to the other axes.
Direction — common DFM factor over 6-month differences of the full -indicator panel. Captures whether the broad set of indicators is improving or deteriorating in unison. Smoothed with a 9-month moving average.
Financial Conditions — composite of cross-asset stress signals (NFCI, KCFSI, VIX, BAA spread, term premium), rolling-10y z-scored, sign-aligned so positive = calm/healthy.

Cycle level = 0.5 × Trend + 0.5 × Stress. Cycle position score is the percentile rank of the level over a trailing 10-year window. Phase is the 4-quadrant of (level, direction):

	Direction up	Direction down
Level > 0	MID CYCLE	LATE CYCLE
Level < 0	EARLY CYCLE	LATE CYCLE

RECESSION fires only under a stricter compound rule: either level < −0.5σ deep with negative direction, or both Trend AND Stress simultaneously below −0.10σ. This prevents stress-only blips (financial stress without real-economy weakness) from triggering RECESSION calls.

Hysteresis: ±0.10σ on level, ±0.05σ on direction. When either axis is in its deadband, prior phase is held — preventing single-month flips from noise.

Structural Health (BCI factor)

The original BCI is a single latent factor extracted via a Dynamic Factor Model with AR(2) transition over the full 55-indicator panel — same econometric backbone as the NY Fed Staff Nowcast, Goldman Sachs' Current Activity Indicator, and the Fidelity / BlackRock business-cycle framework.

Measurement: $X_t = \Lambda F_t + \varepsilon_t$ — each indicator loads on the common factor plus idiosyncratic noise.
Transition: $F_t = A_1 F_{t-1} + A_2 F_{t-2} + \eta_t$ — factor follows an AR(2), capturing both level and momentum persistence.

Standardization is expanding-window — the z-score at date $T$ uses only data through $T$. Full-sample z-scores would contaminate every historical observation with future distribution information; we don't use them.

Phase classification combines factor level (deadband ±0.10σ) with 6-month momentum (applied to a 3-month trailing-smoothed factor) to produce four regimes: Early Cycle, Expansion, Late Cycle, Recession.

Why the two reads can disagree

The Cycle Clock answers "where are we in the cycle?" — direction-aware, regime-adaptive (rolling 10y baseline), 4-quadrant Fidelity-style.

The BCI factor answers "is the broad economy at, above, or below long-run norm?" — level-anchored, expanding-window baseline, structural health.

When subscribers feel recessionary but the cycle clock disagrees, both are correct: structural levels sit below pre-2020 norms (BCI) and the cycle is healthy (Clock). Two complementary reads of the same panel.

Validation

Cycle Clock: catches 2008 GFC and 2020 COVID as RECESSION (both compound triggers fire); correctly does NOT call RECESSION for the 2024-25 manufacturing slowdown (real-economy soft patch without credit stress); 2022 inflation panic correctly fires RECESSION via the both-components-negative rule.

BCI factor: ρ ≈ +0.54 against CFNAI on the matched monthly sample; flagged 6 of 6 NBER-dated US recessions since 2006, zero false positives.

Data & limitations

Built on Some sub-indicators (HY/IG OAS via FRED) are limited to 2023+ in current cache due to ICE BofA licensing changes. The Cycle Clock's Direction component excludes those indicators; Stress uses BAA10Y as the long-history credit-spread proxy. Paid data sources would extend history (BAML/ICE long-history credit indices), add granularity (sub-asset spreads, earnings momentum, Conference Board LEI which is now paid), and improve real-time signal — that's a future track.

Canon

Stock & Watson (2002, JBES); Giannone, Reichlin & Small (2008, JME); Bańbura & Modugno (2014, JAE); Chauvet & Piger (2008, JBES); Croushore & Stark (2001, J. Econometrics). Fidelity Asset Allocation Research Team — quarterly Business Cycle Update.

Forward-looking complements

The BCI factor and Cycle Clock both read the current state of the cycle from realized data. Three calibrated nowcasts read the forward cycle from leading indicators, each with a published track record on the Calibration Ledger:

Inflation Nowcast — forward 3m / 6m CPI YoY from a 2-input pipeline composite (PPI + core PCE 6m annualized) via 60-month rolling-origin OLS. OOS corr +0.88 at h=3m. 6-regime categorical sidecar.
Wage Growth Nowcast — forward 6m / 12m Atlanta Fed Wage Growth Tracker from a 7-input composite (JOLTS quits + hires, unemployment gap, prime-age employment, WEI, breakevens, manufacturing payrolls).
Industrial Production Nowcast — forward 3m / 6m INDPRO YoY from an 8-input composite (manufacturing payrolls, capacity utilization, manufacturing hours, new orders, truck tonnage, WEI, initial claims, inventory-to-sales).

The Recession Watch ensemble combines yield-curve probit + multi-indicator logit + Sahm rule sigmoid into a single cycle-end probability. The Credit-Cycle Melt-Up Monitor watches the financial-cycle complement to the business cycle.

Inputs: FRED, ALFRED (first-release vintages where available), Yahoo Finance. Nightly rebuild. See the methodology index for the full indicator manifest.