Can Leading Indicators Really Anticipate the Business Cycle?

Leading indicators flag changes in conditions, not turning-point dates. Their reliability erodes when the monetary regime that originally calibrated them shifts. Yield curve inversion preceded each of the seven US recessions since 1970. In 2022 it inverted again. By mid-2024, it had been negative for 22 consecutive months — the longest stretch on record — and the recession it was supposed to announce had not arrived. This single episode reframes the entire debate on leading indicators: not as predictive tools whose accuracy can be measured in calendar terms, but as regime-conditional signals whose meaning depends on what is driving them. The framework matters for anyone using these series to read where the cycle stands. Leading indicators capture changes in financing conditions, ordering pipelines and construction permitting — variables that historically led GDP by six to twelve months. What they do not specify is the magnitude, the timing or the depth of the inflection they announce. Treating them as oracles is a category error. This is why the framework of real business cycle phases reads them as part of a constellation rather than as standalone verdicts.

The yield curve, a regime-conditional signal

The 10Y-2Y Treasury spread is the textbook leading indicator. Its track record before 1990 was nearly mechanical: inversion, then recession within 12 to 18 months. The 2022-2024 episode breaks that template. The Cleveland Fed reported in December 2025 that the spread had finally normalised to +35 basis points, a configuration that has historically coincided with the onset of recession rather than its avoidance — a footnote that reverses the usual reading direction. The reason the signal weakened is mechanical, not anecdotal. Curve inversion reflects two distinct forces: monetary policy expectations and growth expectations. When the Fed maintains policy rates above neutral to compress inflation while activity holds up, the curve inverts because short rates are anchored high — not because long-end growth is collapsing. The same shape, two different macroeconomic stories. Reading inversion as an unconditional recession signal ignores which of those two stories is dominant in any given cycle. Investment and productivity dynamics shape the real cycle at horizons the bond market cannot price by itself.

Composite indicators face the same regime sensitivity

The Conference Board’s Leading Economic Index aggregates ten components: building permits, ISM new orders, the term spread, average weekly hours in manufacturing, jobless claims, consumer expectations and four others. The LEI declined for 24 consecutive months between 2022 and 2024 — its longest contraction outside an actual recession. The recession never materialised. In 2025 the index stabilised and rebounded marginally, without the Conference Board issuing a structural reassessment of the diagnostic. Two factors degraded the signal. The first is composition: components weighted on manufacturing and housing pick up the duration shock of a rate-hiking cycle, while services — now over 70% of US GDP — barely register. The second is the indicator’s calibration period, which spans cycles dominated by demand shocks, not the supply-and-policy shocks of the post-pandemic regime. Confidence surveys feed into the LEI with the same regime fragility: their correlation with subsequent activity has weakened markedly since 2020, a degradation also visible in the recurrent errors of economic forecasting models over the past five years. The takeaway is not that these tools are obsolete. Their value lies in directional convergence, not in any single component’s level. When the yield curve, ISM new orders, building permits and weekly hours move together for several quarters, the probability of a turning point rises materially. When one or two move in isolation, the diagnostic is fragile. Structural frameworks for cycle analysis use leading indicators as the first line of a triangulation that runs through real-activity data, financial conditions and policy stance — never as a standalone read. This is decomposed carefully in how reliable leading indicators are.

Last updated — 12 July 2026