Biases in Expectations: Experimental Evidence
41 Pages Posted: 3 Oct 2017 Last revised: 16 Feb 2020
Date Written: February 13, 202
Abstract
We study belief formation in a large-scale experiment where participants forecast a stable random process, across a rich set of conditions. We have three main findings. First, the rational expectations hypothesis is strongly rejected and we find little evidence of learning over time. Second, expectations display over-reaction, which is particularly pronounced for less persistent processes, consistent with evidence in macroeconomic forecasts (Bordalo et al, 2019). Third, commonly-used expectations models, estimated on our experimental data, do not easily account for the way over-reaction varies with process persistence. A simple model with "forward-looking extrapolation" performs well in accounting for this empirical regularity.
Keywords: Expectation formation
JEL Classification: D03; D84; D9; E7; G4
Suggested Citation: Suggested Citation