Why experts are worse at forecasting their own field

A panel of political analysts is asked to predict the outcome of an election two years away. Their forecasts are confident, detailed and wrong at about the same rate as a simple extrapolation of recent polls. The audience does not notice, because the confidence was the part they were listening to.

What everyone sees

Expertise is expected to improve prediction, and for a narrow range of repeatable, well-specified problems it does. Outside that range the expectation persists, because depth of knowledge looks indistinguishable from depth of foresight. A person who can explain the past in great detail appears to be someone who should be able to explain the future in great detail too.

What is actually happening

Tetlock’s twenty-year forecasting study showed that domain experts performed no better than attentive generalists when predicting political and economic events, and that the most confident experts were often the least accurate. The mechanism is overconfidence fed by fluency: a mind stocked with causal narratives can always construct a plausible story for any outcome, which makes every outcome feel predictable after the fact and the next outcome feel predictable in advance.

Why it stays hidden

The illusion hides because experts are selected for confidence, not calibration. Media, institutions and hiring panels reward decisive answers and penalise expressed uncertainty, so the survivors of any public selection process are disproportionately the ones whose confidence exceeds their accuracy. The audience then mistakes the selection effect for a trait of expertise itself.