Evidence from Survey Data of Inflation Forecasts
We develop a novel characterization of participants’ forecasts with a mixture of normal variables arising from a Markov component. Using this characterization, we formulate five behavioral specifications, including four implied by the diagnostic expectations approach, as well as three implied by REH, and derive several new predictions for Coibion and Gorodnichenko’s regression of forecast errors on forecast revisions. Predictions of all eight specifications are inconsistent with the observed instability of individual CG regressions’ coefficients, based on inflation forecasts from 24 professionals. Our findings suggest how to build on key insights of the REH and behavioral approaches in specifying individuals’ forecasts.