| Popis: |
The literature in exchange rate forecasting has met a lot of interest from the academia and practitioners, as well. Since most exchange rates entered the free floating regime, the forecasting ability of the models has been challenged. In this thesis, we explore several aspects of exchange rate forecasting. We first examine the contribution of technical indicators to exchange rate forecasting. Next, we create a new methodological approach that is a hybrid of the Iterated Model Combination and the Constrained Predictors approach, on which we also apply positivity constraints in the forecasts. Last, we focus on the realized volatility of exchange rates.\ud \ud In Chapter 2, we test the forecasting ability of several theoretically motivated models along with the forecasting ability of technical indicators, an atheoretical tool that identifies patterns and produces market signals. We use monthly data ranging from January 1974 to December 2014 for six widely traded currencies. We show that both types of predictors provide valuable information about future currency movements. To efficiently summarize the information content in candidate predictors, we extract the principal components of each group of predictors. Our findings suggest that combining information from both technical indicators and macroeconomic variables significantly improves and stabilizes exchange rate forecasts versus using either type of information alone.\ud \ud In Chapter 3, we focus on forecasting daily exchange rate returns of six widely traded currencies using financial predictors and combination and dimensionality reduction methods. We propose a hybrid Iterated Combination with Constrained Predictors (ICCP) approach. In addition, we examine the impact of positivity constraints on the forecasting ability of each method. Our results indicate that the proposed hybrid method outperforms the simple linear bivariate method and both the Iterated Combination and the Predictor Constrained approaches. Furthermore, positivity constraints significantly improve the forecasting ability of all methods. We provide several robustness tests by changing several specications of the forecasting experiment.\ud \ud Chapter 4 provides empirical evidence in forecasting realized volatility in exchange rates. Forecasting realized volatility in exchange rates is very important for practitioners and has been vividly discussed among academics. Our target is to contribute to this dialogue by providing a comprehensive analysis of forecasting realized volatility in exchange rates. For the purposes of our analysis, we use data from January 1986 to December 2012 for four widely traded currencies, GBP, CHF, YEN, EUR and a composite one (FX Aggregate). We show that macroeconomic and financial variables provide additional information to the autoregressive term and can benefit the forecasting accuracy. We apply a large set of 38 variables, supported by the literature, which shed light on different macroeconomic aspects. We answer the question of variables selection by extracting all available information with the use of several shrinkage methods, machine learning techniques, dimensionality reduction techniques and combination forecasts. In order to resolve the problem of method selection the forecaster faces, we aggregate all methods and form an amalgamation of forecasts. We test whether outliers drive the performance of this type of naive combination. We apply different specifications of naive combination by trimming the first, second and third outlier from the top and bottom. Our findings suggest that macroeconomic variables should be accounted when forecasting realized volatility. Moreover, the amalgamation of forecasts benefits the forecasting experiment significantly, irrespective of the specification under consideration. |
