What a Consensus Forecast Is, and What It Is Not

Somebody ran a survey

The expected figure printed beside a release did not come from the statistical agency. A data provider asks a panel of economists at banks and research houses what they think the release will show, collects the individual answers, and publishes a central value from them — usually the median, sometimes the mean. That value is the consensus, and it is a summary of opinions rather than a property of the statistic.

The panel is the thing to ask about

Consensus figures differ because the surveys differ. Providers poll different panels, of different sizes, close their surveys at different points before the release, and treat late or updated submissions differently. A forecaster who revises after a related release lands may or may not be captured, depending on when the survey closed. None of this is hidden — providers document their process — but none of it is visible on the calendar row that prints the result either.

Why two calendars can show different expectations

The practical consequence is that the same release can carry different expected numbers on two sites, and therefore be a beat on one and a miss on the other. Neither is wrong; they are reporting different surveys. Before treating a beat as information, it is worth knowing whose consensus your source uses, because that is the entire content of the label.

The median hides the disagreement

A single central value says nothing about how much the panel agreed. A tight cluster of forecasts and a wide spread can produce the same median, and they describe very different states of knowledge. Where a provider publishes the range or the individual contributions, that spread is the more informative number: it tells you whether the outturn landed outside what anyone expected, or comfortably inside the set.

What a beat or a miss establishes

A gap between the published figure and the consensus establishes one thing: the panel's central estimate was off by that much, in that direction. It is a statement about the forecasters. It is not a statement about whether the economy is strong or weak — a weak number that was expected to be weaker is still weak — and it is not a prediction of what follows.

Consensus is not the only expectation in the room

Surveys are one measure of what is expected, and they close before the release. Prices, positioning and later commentary carry expectations too, and can move after the survey closed. That is why a figure landing exactly on consensus can still be treated as news, and why a release that beats a stale consensus need not have surprised anyone at the time.

What to do with it

Treat the consensus as context for reading the release: what informed forecasters thought was coming, and how far out they were. If you keep notes, record which provider it came from, because a consensus without its source is not comparable with anything later. And treat commentary that explains a market move purely by reference to a beat or a miss as an assertion rather than an account of causation — it is one visible number among many things happening in the same second.