Exchange rates, Markov switching models, technical trading rules and fundamental data: Are they related? (Germany, United States)

Instead of using fundamental models of exchange rate determination, foreign exchange traders commonly use moving average technical trading rules to forecast future exchange rate developments. These types of technical trading rules solely rely on past price observations and do not explicitly incorporate fundamental data in the forecasting process. This dissertation analyzes the behavior of the German mark/US dollar exchange rate over a nine year period to determine what type of stochastic processes are present in the exchange rate to allow the moving average rules to successfully predict future exchange rate values. With the help of a Markov switching model, the research shows that the exchange rate is segmented into periods of appreciating and depreciating currency values with low volatility, and into periods of high volatility that are not clearly associated with either appreciating and depreciating currency values. Further, this research shows that simple technical trading rules are able to detect these periods and that traders can use either a Markov switching model or a moving average trading rule to profitably forecast future exchange rate development.; To answer the question whether fundamental data can be associated with the different period of appreciating and depreciating currency values, this dissertation looks at news announcements to serve as proxies for fundamental data. This news data set is shown to be related to the periods predicted by the Markov switching model only and not to the periods predicted by the technical trading rules, implying that Markov switching models adapt more rapidly to exogenous shocks as compared to the moving average rules. Further, this dissertation suggests that even though technical trading rules are profitable, traders can improve on predicting future exchange rates by following a Markov switching model instead of the more inert moving average trading models. The decision making process for selecting a profitable trading rule is less subjective for Markov switching models as compared to moving average models, and, in addition, changes in fundamental data appear to be incorporated faster in the Markov switching models as compared to the moving average trading rules.