Adaptive Learning, Informational Frictions and Monetary Exchange

It is well known that both expectations and informational frictions are key determinants of macroeconomic conditions. This dissertation studies how expectations are formed and the macroeconomic outcomes when agents use an adaptive learning rule. The second chapter replaces rational expectations in a random matching model of monetary exchange to re-visit three central issues in monetary theory; the tenuousness of monetary equilibria, the non-neutrality of money, and the social cost of money growth. Chapter three studies the interaction between market structure and expectations. The fourth chapter is an empirical study on the heterogeneity of adaptive learning rules used by professional forecasters.;Chapter two finds that when agents use a simple adaptive learning rule to learn about the value of money in the search theoretic framework, monetary exchange is a robust arrangement in the sense that the monetary steady state is locally stable and agents never learn the non-monetary steady state. The learning process takes time, however, so a change in the money supply has real effects in the short-run, but is neutral in the long-run because agents eventually learn the lower steady-state value of money. The social cost of a money growth policy is lower than previous estimates because of the short-run benefits associated with the learning dynamics.;The third chapter shows the structure of the market plays a key role the local stability of a monetary steady state and the learning dynamics. When agents learn about prices and search intensities the economy can fall into a negative feedback due to the congestion effect. Also, the economy can fall into a negative feedback loop due to a large liquidity premium when buyers are sensitive to new information and have high bargaining power.;Finally, I show that among the individual professional forecasters there are three sources of heterogeneity: agents learn at different speeds, the type of learning rule used depends on the period in which the forecast is made, and the learning rule used differs depending on the macroeconomic variable being forecast. I show that the sources of heterogeneity go undetected in when forecasts are aggregated using a mean or median.