Learning in Survey Expectations of Inflation




The main question of this paper is how private agents form their expectations. To answer this, we approximate survey expectations of inflation by constant gain learning algorithms. In particular we develop a Bayesian constant gain estimator, and show that this gives a much better approximation to survey expectations then simple constant gain estimators. We find that in more volatile economies private agents pay more attention to recent data when they form their inflationary expectations.Finally, when a regime change occurs private agents do not follow mechanically an adaptive rule, but understand it immediately that statistical relations between economic variables will change.