Formation and dispersion of expectations : an empirical analysis with special focus on inflation

This thesis is a collection of papers that use survey data to analyze expectations about macroeconomic variables and the way these expectations are formed. Using a new approach for modeling forecast errors in a structural way, we show that most of the individual forecasts in the Consensus survey data set have fairly good properties with respect to unbiasedness and efficiency. We provide empirical evidence in favor of partial equilibrium models of the sticky information type that explain the development of inflation dynamics and the formation process of inflation expectations of households respectively. We show that there is a dichotomy between disagreement about real variables, which is more strongly affected by real factors, and disagreement about nominal variables, which is related to the institutional setting of monetary policy.