Learning to Interpret Forecasts: Evidence Quality and Source Familiarity

Abstract

We study how learning about a forecasting method affects a seller's choice of evidence. More evidence improves the current forecast but makes the method's persistent component harder to identify. A monopolist may therefore collect less evidence initially to help customers interpret later forecasts. Because the seller captures only part of customers' gains from costly evidence, it switches to less initial evidence sooner than a welfare-maximizing regulator. A minimum evidence requirement improves welfare only for intermediate weights on future benefits. An independent assessment removes this learning trade-off but may be privately unprofitable despite being socially worthwhile.