The Nordhaus Test with Many Zeros

We reformulate the Nordhaus test as a friction model where the large number of zero revisions are treated as censored, i.e., unknown values inside a small region of “imperceptibility.” Using Blue Chip individual forecasts of U.S. real GDP growth, inflation, and unemployment over 1985-2020, we find pervasive overreaction to news at most of the monthly forecast horizons from 24 to 1, but the degree of inefficiency is very small. The updaters, i.e., those who make non-zero revisions, are not found to perform better than their “inattentive” peers do.