By Nikolay Braykov
Microeconomic models often use the Rational Expectation Hypothesis (REH) instead of including expectation data. This paper examines the validity of the REH using subjective probability questions about mortality, fertility and education outcomes from panel data. First, I ask whether expectations are accurate and homogenous at the individual level; I find substantial forecast biases that depend on the nature of the outcome and decrease with ability and elimination of focal responses. I then propose a Bayesian learning framework to explain biases and find evidence of partial learning, suggesting probabilities become more accurate over time. Finally, I find subjective probabilities have predictive power over and above objective estimates, suggesting they contain private information about anticipated events.
Advisor: Frank Sloan