What Confidence Surveys Actually Measure About the Cycle

Confidence surveys measure a climate of opinion, not activity. The correlation between sentiment and subsequent GDP has weakened sharply since 2020 — and that weakening is the signal worth reading.

The reflex is to treat household and business confidence surveys as leading indicators of the cycle. They are not. They are surveys of perception, run on self-reported questionnaires, and their correlation with what actually happens to output has always been imperfect. What changed in the post-pandemic regime is that this imperfection became structural rather than statistical noise.

European household confidence remained depressed long after the post-Covid activity rebound. In the United States, consumer sentiment slid through 2024 and 2025 against a labour market that kept printing solid payrolls. The textbook framework — confidence weakens, then spending weakens, then GDP weakens — broke. The forecasting community is still working out whether the break is cyclical or permanent. This article makes the case it is mostly the latter, anchored in the framework of real cycle indicators.

Perceptions, not transactions

Confidence surveys — the University of Michigan Consumer Sentiment Index, the European Commission’s Economic Sentiment Indicator, INSEE’s household confidence indicator, the ZEW expectations survey in Germany — rest on a stylised question: do you perceive an improvement or deterioration in your situation and in general economic conditions? Answers blend lived experience, media exposure and prospective fears. They do not count transactions, invoices or hires.

The European Commission published an ESI at 95.2 for the euro area in December 2025, below its long-term average of 100, while euro-area GDP printed positive year-on-year growth. The gap reflects what the index measures and what it does not. The real business cycle, built on investment and productivity flows, can follow its trajectory without lining up with reported sentiment. The structural lag between macroeconomic indicators and the real economy compounds the problem: surveys add a perceptual filter on top of an already lagged measurement system.

The media climate sets the floor

Academic work has documented the correlation between negative media tone and confidence indices, independently of actual activity dynamics. The University of Michigan reported in 2025 that the current-conditions component of its index had diverged from the expectations component for more than twelve months — a sign that respondents were reacting to perceived inflation and political coverage rather than to their own employment, wage or wealth situation. The divergence is not a glitch; it is the survey doing exactly what it is designed to do, which is to capture sentiment.

This is where the operational problem sits. Mainstream forecasts still incorporate confidence surveys as leading indicators of the cycle, weighted as if their predictive properties were stable across regimes. They are not. The OECD noted in its November 2025 Economic Outlook that the correlation between the business confidence index and six-month-ahead GDP growth had fallen from 0.72 to 0.41 between the 2010-2019 period and the 2020-2025 period. A correlation of 0.41 is no longer a leading signal in any operational sense — it is barely better than coin-flip noise once you account for sampling variance. The point is set out at length in how PMI surveys anticipate GDP.

The case for keeping confidence surveys in the toolkit is narrower but real. When household and business sentiment deteriorate simultaneously for several months, the convergence can signal a regime shift in spending and investment behaviour — particularly on big-ticket discretionary items. Structural cycle analysis frameworks incorporate them as one signal among others, weighted by their persistence and by their convergence with hard data. Their diagnostic value comes from corroboration, not from any single reading.

Last updated — 14 June 2026