Access Type

Open Access Dissertation

Date of Award

January 2015

Degree Type


Degree Name




First Advisor

Tatsuma Wada


The first chapter of this dissertation analyzes the necessary and sufficient conditions for stability under recurring structural changes. Using a finite state Markov process to model stochastically evolving, state-dependent parameters I find that by employing the conditions unique to mean-square stability, the minimum state variable (MSV) solution, found in non-linear models of this reduced form, is also stable in the learning sense. However, the choice of parameter values limits the robustness of this result. Furthermore, to illustrate this outcome I develop empirical results for a model similar to Cagan’s 1956 work on hyperinflation for Germany and the United States. I find that during the time of active currency market intervention, monetary policy was not mean-square stable for both the U.S. and Germany.

In the second chapter, I analyze if economic agents could have learned the policy decisions of the Plaza and Louvre accords. New techniques in Markov switching Adaptive Learning models (MSAL), shows that economic agents would not have learned the rational expectations outcomes of exchange rate interventions and therefore contributed to exchange rate overshooting and excess volatility during this time. These finding help to explain why forecasts of short-term exchange rates have historically been poor while long-run forecasts do much better at matching the data.

The third chapter analyzes empirical data from the forward exchange rate premium to interpret the puzzle, made famous by Fama, using Markov Switching Adaptive Learning (MSAL) techniques. This chapter addresses the need for using Mean-Square Stability as the criterion for stability rather than traditional stability conditions. Moreover this chapter observes the possibility for a self-referential solution to occur under specific conditions similar to what is found empirically. Furthermore, this chapter is able to replicate the results typically found during the analysis using a Markov-switching constant gain model, indicating that economic agents may posses some form of bounded rationality or information asymmetry which produces the observed bias. A central tenant of this chapter is that agents facing a regime which tend to produce the forward premium bias present in most empirical applications even in the face of highly persistent fundamentals.

Included in

Economics Commons