Essays on heterogeneity, learning dynamics, and aggregate fluctuations

This dissertation investigates the relationship between learning, heterogeneity, and aggregate fluctuations in an economic model with multiple equilibria. In chapter 2, I investigate the stability properties, under econometric learning, for a model with heterogeneous expectations. Stability properties depend on the proportion of agents using each learning mechanism. Furthermore, the equilibria "exchange" stability when the level of heterogeneity is such that the mean squared error of the inefficient learning mechanism is minimized.;In chapter 2, agents are allowed to change their learning mechanisms based on relative performance in an evolutionary game theoretic model. Only some Nash equilibria are evolutionary stable when disturbed by a "mutant" population. Local and global stability properties are shown for the equilibria for homogeneous expectations when another expectations operator is available. Also, under some conditions, heterogeneous expectations are stable under combined learning and evolutionary dynamics.;In chapter 3, a confidence level is introduced into the model that affects aggregate demand. The structure of the model is allowed to change from a two-state Markov process which changes stability properties for each equilibrium. As a result, at any state change, confidence decreases leading to a lower level of output for some time. As agents change to using the efficient expectations operator, confidence increases and output returns to its equilibrium value. Learning produces a business cycle with a short period of negative (or slow) growth and a longer period of expansion to the equilibrium level of GDP.;Chapter 4 focuses on a change to the costs of using each expectation operator instead of structural changes to the model. Due to changes in stability, agents will change their forecasting operator leading to a change in the level of heterogeneity and a decrease in confidence followed by an increase in confidence as agents learn the equilibrium parameter values. It is possible to reach an outcome of heterogeneous expectations in the long run. This state leads to an outcome where collusion would produce a superior outcome. As in chapter 3, learning accentuates the confidence affect from the state change producing asymmetry in the business cycle.