Why Economic Forecasts Miss Cycle Turns — Systematically

Why economic forecasts miss cycle turns — and why the errors are systematic, not random. Economic forecasts occupy a central place in public debate and investment decisions. Their recurring failure at cycle turning points raises a methodological question that is rarely posed directly. They rely on continuity assumptions and linear models poorly suited to ruptures. Endogenous shocks, threshold effects, and adjustment lags are structurally undercaptured. The mechanism behind growth-driver reading is laid out in the Eco3min framework on real growth drivers. This limitation explains why forecasters rarely flag recessions in advance and revise their scenarios late, all at once, after the data has already turned. The real issue is not that forecasts are wrong — every projection carries an error margin. The issue is that they are wrong systematically in the same direction. Forecast errors are not random: they underestimate downside reversals and overestimate the speed of recoveries. That structural bias points to a methodological problem far more than to a competence gap among forecasters.

Continuity bias: extending the present instead of pricing the rupture

Forecasting models are built on historical regularities and behavioural equations estimated in normal periods. When the macroeconomic regime changes — a transition from low rates to abrupt tightening, a supply chain disruption, a geopolitical shock — these models extrapolate the past instead of capturing the discontinuity. The IMF acknowledged in its October 2025 World Economic Outlook that its growth forecasts for advanced economies had been revised by an average of 0.6 percentage points between initial estimate and realised figure over the 2020–2024 period — a gap that dwarfs the precision implied by the projections themselves. The real cycle proceeds through nonlinear sequences: credit tightening can run for several quarters without visible effect before triggering a sharp investment adjustment. Linear models capture these threshold effects poorly, which explains why every cycle, structurally singular in its mechanics, catches forecasters off guard. The error is not in the model — the error is in expecting a linear model to detect a nonlinear turn.

Consensus: stabilising mechanism or collective blind spot

Consensus forecasts — the average of major institutions and bank desks — carry an additional bias: they converge toward a central scenario that leaves little room for outcomes outside the median. The OECD released in November 2025 a global growth forecast of 3.0% for 2026, with a gap between the most optimistic and the most pessimistic scenario of only 0.4 percentage points — a strikingly narrow range given the geopolitical and monetary uncertainty of the period. That compression of scenarios reflects an anchoring bias: each institution calibrates its forecast on what others anticipate, generating an artificial convergence that masks the true breadth of risks. Forecasters do not want to be alone with a contrarian call; the career cost of being wrong with the crowd is far smaller than the cost of being wrong against it. The leading signals that flag cycle turns deliver readings that consensus integrates late — precisely because no single isolated signal is enough to dislodge the median forecast from its anchor. Forecasts retain analytical utility: they provide a reference framework and a central scenario against which to calibrate decisions. The issue is not their use, but the treatment of them as point estimates rather than as distributions with structurally biased tails. Structural frameworks for analysing the economic cycle situate forecasts within their methodological context and confront them with market signals and high-frequency data — a more robust approach than treating a central figure as a conclusion.

Last updated — 16 June 2026