The reason a forecast range matters more than its midpoint

A projection is reported as one figure and the range that accompanied it is dropped in transmission. The figure is then treated as the forecast. What was discarded is the only part that indicated how much confidence the estimate deserved, and two forecasts with the same midpoint and different widths are not the same forecast at all.

What everyone sees

A point estimate is read as the forecast and the interval as a technical caveat, so reporting keeps the number and drops the range. This treats uncertainty as an admission of weakness rather than as content. The width is a quantity in its own right, and a projection without it cannot be evaluated or compared.

What is actually happening

Gigerenzer and colleagues surveyed how people understand probabilistic weather forecasts across several countries and found the reference class for a stated probability was interpreted in several incompatible ways, showing that a single number without an explicit frame communicates less than it appears to. Alpert and Raiffa’s work on training probability assessors found the intervals people state are systematically too narrow, with true values falling outside stated ranges far more often than the stated confidence implies. The two findings compound: intervals are already too tight when given, and dropping them removes the only signal about how tight they were.

Why it stays hidden

The loss hides at each retelling. The forecaster published a range, the summary quoted the midpoint, and the decision cited the summary. Nobody discarded anything deliberately; each step preferred the more usable number, and by the final step the estimate carries an implied precision that no one in the chain ever claimed.