Learning, Expectations and the Business Cycle

Standard RBC models predict forecastable movements in output, consumption and hours that differ from those obtained from a VAR estimated on US data. The paper investigates whether introducing bounded rationality and learning generates business cycles properties which are empirically plausible. In particular we focus on, (i) the forecastable components of output, consumption and hours and (ii) the expected co-movement of output, consumption and hours. We set up a three-sector RBC model with structural change and bounded rationality, where the economic agents gradually learn about changes in the growth rate of productivity. We then estimate the model using indirect inference methods. We evaluate the empirical fit by comparing the model with learning with a version that imposes rational expectations. Given that the asymptotic behavior of the agents’ beliefs depends only on the deep parameters of the model, our econometric approach does not require the estimation of extra free parameters, compared with the RBC model under rational expectations.