Essays on nonlinearity in exchange rates

The three essays included in the thesis examine the issue of nonlinearity in exchange rates from different perspectives. In the first essay the issue of model selection in financial markets is examined and it is demonstrated that in order to evaluate models of nonlinearity in financial asset returns the Quadratic Loss function is not the best one to use. A new criterion based on the Exponential Utility loss function is proposed and it is argued that this loss function is much more closely related to the decision rules used by financial market participants. In the second essay the new criterion developed in the first essay is used to evaluate the two state Markov Switching model for exchange rate returns measured at a monthly frequency. It is found that with the new criterion it is not possible to reject the null of no difference between the Random Walk and the Markov Switching model for the British Pound, the Japanese Yen and the Deutsche Mark. The importance of using appropriate loss functions is also demonstrated in this essay. In the third essay the dynamics of financial market volatility are examined for exchange rates, interest rates and the S&P 500 index. For all these assets it is found that the standard GARCH(1,1), EGARCH(1,1) and Stochastic Variance with AR(1) variance specification are not able to capture the rich volatility dynamics present in the data