| Chapter 1 of this dissertation uses a more flexible modeling methodology in order to test for time varying coefficients in the classic uncovered interest parity regression. By allowing the coefficients to vary over time, but also sharing information across currencies in a random effects type setup, we get clearer picture of the evolution of UIP in financial markets. We find that the once standard failure of the equilibrium condition is no longer as pronounced.;Chapter 2 extends the work done in chapter 1 by forming currency portfolios that exploit the failure of UIP in order to be profitable. We show through the use of dynamic linear models as well as a non parametric random portfolios methodology that excess returns on these strategies are no longer quite so pervasive, adding evidence to the claim that UIP holds better now than in the past.;Chapter 3 switches gears and revisits a classic paper on long swings in the exchange rate. By testing the model on more currencies over a longer time period, and gauging predictive accuracy based on the signals it gives for portfolio formation, we find that the model does not give adequate predictive performance. We conclude that when trends are defined by a markov switching Gaussian model, it is highly unlikely that exchange rates still exihbit long swings. |
