Designing And Application Of A DTGARCH Exchange Rate Model Based On Theories Of Target Zones

Since 1990s, great attentions are paid to'Exchange Rate Target Zone'by various researchers, not only because the target zone (TZ) regime is one which has stability owned by the fixed exchange rate regime and flexibility owned by the float exchange rate regime, but also because many countries use exchange rate target zone as one policy to stabilize their exchange rates. Due to the difficulty of TZ models'application in empirical studies, the dissertation attempts to build a time series model of three-regime double-threshold generalized autoregressive conditional heteroskedastic (DTGARCH) model based on TZ theories, and then use the model to describe exchange rate behavior and forecast exchange rates. In the end, the model is extended to describe the behavior of exchange rates under the middle exchange rate regimes, from which characteristics of managing exchange rates are inferred.The dissertation firstly builds a three-regime DTGARCH model based on reasonable TZ theories in order to discretely approximate the model behind the TZ theory selected. The conditional mean part of the model can reasonably describe threshold behavior of exchange rates, and the conditional variance part can explain characteristics of exchange rate's volatility.Then, the three-regime DTGARCH model is used to describe the behavior of exchange rates and examine the rationality of TZ theories behind the model. Exchange rates of member countries of the European Monetary System are chosen as sample in the empirical study, because the Exchange Rate Mechanism is one typical exchange rate TZ regime. The results of the empirical study indicate that the model can describe the behavior of exchange rates quite well and the TZ theories behind the model are reasonable.Since the three-regime DTGARCH model can describe the behavior of exchange rates well, the excellent forecasting performance of the model is expected. The forecasting study is carried out with two sets of data, one set data being artificial ones obtained by Monte Carlo experiment, and the other being true data which have been used in empirical study above. The motive of using artificial data is to guarantee the similar nonlinearity appearance in forecasting periods the same with that of the past and to eliminate the possibility of fading away of nonlinearity which will affect the forecasting performance of the model. In addition, the purpose of using true data is to...