Nonlinear currency modelling in the scope of self-exciting threshold autoregressive models

The numerical values of currencies as a macroeconomic variable has a great importance in outward-oriented economies especially when the mutual dependence between the economies is taken into consideration. It becomes clear when the issue is interpreted in terms of economy that the targeted level of the currencies is of vital importance in economies that has the characteristics of export-driven growth and that struggle for keeping the photograph of the macroeconomic design smooth. When the issue is considered in a global scale, the numerical values of the currency regimes and policies that are reflected in parity may even give rise to a recession effect towards the countries that apply the currency policy. In this sense, the currency policies of the dominant economies will manipulate the economies of the other countries and change the global economic projection at one instant. The currencies which are so important for macro economies are in close contact with the other macro economic variables. However; this situation does not resolve the issue of what the value of the currency in (t+1) period will be; and tries to explain a variable that has a scholastic characteristics with another scholastic variable. In this situation, the problem will grew even worse. Right at this point, we must to tell why the currency estimation is so important and nothing more. In addition, handling the issue as a mere econometric cause-effect will only mean dealing with limited number of scholastic independent variables. Therefore, in addition to the economic dimension, the time series analysis dimension must also be included in the modelling process. When the fact that each time series has its own internal dynamics (sometimes these dynamics are described as time series components) that are sensitive is considered, it is observed that these dynamics give coordinates in terms of estimation, and eliminate the obligatory dependency to some external variables at a serious level. This is exactly the thing which is dealt with in this study. Forex (foreign exchange) market is the spot currency market where the abovementioned currencies are purchased and sold or processed for short or long term periods. The Forex Market, which has the highest process volume in the world, is one of the data mines in which the time series experience can be lived in the most beautiful sense. The Forex Market ensures a very serious activity area for the account owners accompanied by great leverage rates. Right at this point, the thing which is emphasized is that the importance of currency parity in economies is not only in the activity given to external commerce but also in attracting the attention of individual and institutional owners of savings. For this reason the estimation of currency parity has the quality of a very different and serious technical analysis for investors. In this study, it is claimed that the modeling approach, which is performed in estimation, will be included in the Forex Investment Platform in a future that is not so far. The linear time series analysis that is relevant with the issue is given in detail in the thesis study, and the results that are based on this analysis and the descriptive statistics are given in the Appendix-1 of the study for all currency parities in separate time points. Then, the nonlinear time series analysis is dealt with in the scope of linearity tests, and these tests are applied for all parities and for all time points. Then the SETAR modeling, which is the title of the thesis, is applied in detail to explain a pattern that is not linear. The analyses of the SETAR Modeling and all the other descriptive statistical analyses are applied in all parities for separate time points.