Research On Exchange Rate Volatility Forecasting Based On FIGARCH-MIDAS Model

With the rapid progress of economic globalization around the world,the role of exchange rates in the economic and trade transactions of countries around the world has become increasingly obvious,and the fluctuation of exchange rates will have a direct impact on the country’s trade in goods and capital flow situation of the country.On the other hand,floating exchange rate systems are now commonly used in countries around the world.In the case of the CNY,a managed floating exchange rate system has been in place since July 21,2005,when it stopped being singularly pegged to the US dollar.On August 11,2015,the Chinese central bank announced an adjustment to the quotation mechanism of the CNY exchange rate mid-price,which further liberalized the fluctuation range of the CNY exchange rate fluctuations.In conclusion,in the course of the continuous marketization of the CNY exchange rate,its volatility and frequency of fluctuation have increased and accelerated.Therefore,it is important to forecast exchange rate fluctuations more accurately,which is of great significance for exchange rate risk management in China.This thesis investigates how to construct an econometric model based on mixedfrequency data to develop more accurate forecasting effects on exchange rate fluctuations.First,starting from the analysis of the influencing factors of exchange rate fluctuations,we sort out that although there are many influencing factors affecting exchange rate fluctuations,it can be found from the basic monetary exchange rate model that money supply differences and output differences are still the main factors affecting exchange rate fluctuations.Therefore,this thesis takes these two factors as currency shocks to forecast exchange rate volatility.Meanwhile,the existing literature shows that exchange rate volatility has obvious characteristics of long memory,volatility aggregation,and autocorrelation,and the fractionally integrated generalized autoregressive conditional heteroskedasticity(FIGARCH)model can well describe the above characteristics.However,the traditional FIGARCH model can only use historical volatility for forecasting,ignoring other economic information on the impact of volatility.Therefore,this thesis further considers how to combine both currency shocks,which have a significant impact on exchange rate volatility,and the FIGARCH model.Considering that macro fundamental data often have mixed frequency data characteristics,and the Mixed Data Sampling(MIDAS)model can handle mixed frequency data well and effectively retain the integrity of data information.Based on this idea,this thesis constructs a FIGARCH-MIDAS model to forecast exchange rate volatility,which is a mixed-frequency data econometric model that divides exchange rate volatility into two long-term volatility components and short-term volatility components.The long-term volatility component is described by the MIDAS process that uses output variance and money supply variance as influencing factors,and the short-term volatility component is explained by the FIGARCH process that can characterize the long memory of the exchange rate.Using the exchange rate returns of the Canadian dollar,Australian dollar,Japanese yen,euro,Korean won,Chinese yuan,and British pound against the US dollar as the explanatory variables,along with the output differences and money supply differences between the home country(Canada,Australia,Japan,the euro area,Korea,China,and the United Kingdom)and the target country(the United States)as the explanatory variables,this thesis empirically analyzes the predictions of the FIGARCH-MIDAS model on the volatility of the above seven exchange rates The results are as follows.First,from the results of AIC and likelihood function,it is found that the FIGARCHMIDAS model has better volatility fitting advantage than the traditional FIGARCH model which does not consider the influence of currency factors and the GARCHMIDS model which does not consider the long memory feature of exchange rate fluctuations,for all exchange rates except the Australian dollar exchange rate.Second,the FIGARCH-MIDAS model is able to capture the long-memory nature of exchange rate fluctuations well.Moreover,currency shocks have a positive effect on long memory in the other six more developed exchange rate markets except the CNY.Third,money supply differentials have a greater effect on home exchange rate volatility than output differentials,and both have a positive effect on exchange rate volatility.Fourth,for the Canadian dollar,euro and yen exchange rates,the FIGARCH-MIDAS model is not affected by the sudden structural changes generated by the events of the 2008 global financial crisis.For the other four exchange rates,the FIGARCH-MIDAS model is more influenced by the 2008 global financial crisis or the 2015 CNY exchange rate reform,and there are sudden changes in the model structure.Finally,this thesis tests the out-of-sample forecasting ability of the FIGARCH-MIDAS model.The Va R(Value at Risk)backtesting methods(penetration rate HR,loss function LF,LR test and DQ test)are used to test the Va R forecasting effects of the underlying FIGARCH model,GARCH-MIDAS model and the FIGARCH-MIDAS model constructed in this thesis,and it is found that the FIGARCH-MIDAS model has better out-of-sample Va R forecasting effects.Based on the empirical findings,this thesis concludes that the FIGARCH-MIDAS model can forecast exchange rate volatility well,and also puts forward the following policy recommendations: continue to expand the depth and breadth of the foreign exchange market and strengthen the market-oriented reform of the foreign exchange market;actively focus on the impact of economic fundamentals factors,especially money supply,on exchange rate volatility;continue to enhance the flexibility and flexibility of the CNY exchange rate,while enhancing the transparency of monetary policy;and be alert to the exchange rate risks arising from changes in the international environment.The thesis consists of 3 figures,11 tables and 116 references.