The reason a prediction can cause its own failure

A credible forecast says a bridge will be overcrowded on Saturday. Drivers hear the forecast and take an alternative route. The bridge is empty on Saturday. The forecast was wrong, and it was wrong because it was believed by the very people whose behaviour it was trying to predict.

What everyone sees

The wrong forecast is filed as a forecasting error: the model overestimated demand, and next time it should be adjusted downward. The possibility that the forecast itself was the cause of the undershoot is not standard in the debrief, because forecasts are assumed to observe a system rather than to participate in it.

What is actually happening

Merton described the self-defeating prophecy (the inverse of his more famous self-fulfilling prophecy): a prediction that provokes action which prevents the predicted outcome. Grunberg and Modigliani formalised the problem: if a forecast enters the decision set of the agents it describes, the system is no longer the same system the forecast was made about. Simon showed that public predictions in economics alter the behaviour they predict, which is why unconditional forecasts in social systems are structurally unstable.

Why it stays hidden

The instability hides because each individual driver’s response is rational: they received information and acted on it. The problem is not in any single decision but in the aggregate: if every driver responds to the forecast, the forecast destroys its own basis. The debrief sees only the empty bridge and concludes the model was wrong, because the reflexive loop is not a standard input to post-mortem analysis.