Adaptive learning in expectations-driven business cycle models

The dissertation focuses on expectations-driven business cycles (EDBC) models. That is, economic fluctuations are generated by news about future productivity or sunspots. The stability properties and dynamics in the EDBC models are examined under adaptive learning.;Chapter One examines two news-driven models' stability properties: 1) the standard real business cycle (RBC) model with habit persistence in consumption preference and adjustment costs to investment; 2) the RBC model with complementarity between consumption and investment and adjustment costs to investment. The news-driven rational expectations equilibria (REE) in the above models are stable under adaptive learning. Learnability does not depend on the strong informational assumption, in which econometric forecasts and outcomes are simultaneously determined.;In comparison to the rational expectations assumption, Chapter Two displays the impact of exogenous and endogenous news under infinite horizon learning in response to news about future total factor productivity (TFP) based on the standard real business cycle (RBC) model. Intertemporal substitution effects on labor caused by shifts in beliefs offset income effects produced by exogenous anticipated future productivity changes. Although there are no comovement between output and consumption in response to news under learning, the strengthened substitution effects generate a weak acyclicality. There are larger impact effects under learning. The dynamics of economic variables move in opposite directions under different expectations assumptions in the anticipatory phase.;The rational expectations equilibrium (REE) in the determinate Ramsey model is stable under finite horizon learning. However, multiple equilibria in the indeterminate Farmer-Guo model are not learnable, when forecasters have a one-period horizon. In Chapter Three, multiple equilibria are still unstable under.finite horizon learning. The speeds of convergence towards or explosion away from sunspot equilibria vary across planning horizon and learning implementation.