When to Stop Trusting a Forecast

The most underused epistemic move: discard a forecast entirely when you discover the forecaster was willing to misrepresent it.

The intuitive correction — "they're optimistic, so shade their numbers down by 20%" — doesn't work. Once you know someone is willing to lie about a project, you don't know the direction of the distortion, the magnitude, or what else they've distorted. You can't adjust for a lie you can't characterize. The forecast becomes unusable, not just discounted.

The implication is stronger than it sounds. When evaluating any proposed initiative, the first question is not "does the model make sense?" but "are the people presenting it being honest?" If the answer is no — if you catch them misrepresenting the costs, projecting impossible timelines, or dismissing real objections with hand-waving — you can skip the model. The dishonest promotion is itself evidence against the initiative.

This is not cynicism. It's a useful heuristic because it's asymmetric: good ideas rarely require sustained dishonest advocacy to gain acceptance. When a project can only be sold with misrepresentation, something is structurally wrong with it — either the promoters know it and are hiding it, or the pressure to make the project work has distorted their judgment past the point of usefulness.

The heuristic applies symmetrically to information sources. A newsletter, analyst, or institution that has been caught misrepresenting conclusions doesn't become a slightly-discounted source — it becomes an unusable one. The value of a forecast is its correlation with reality, and a forecaster who lies destroys that correlation regardless of how sophisticated their model is.

The inverse also holds: institutions and individuals who maintain honest assessment under pressure — acknowledging when their projections were wrong, updating visibly — become more valuable as sources over time. Epistemic integrity compounds.

Source: D-squared Digest (Daniel Davies, 2004). The essay was written in the context of the Iraq War but the principle generalizes.