Research On Volatility And Nonlinear Dynamic Adjustment Of RMB Exchange Rate

In recent decades,China's economy has developed rapidly,and China's position in the world has significantly improved.China plays an important role in global import and export trade,investment,Internet and many other aspects.In recent years,trade in the United States has been in deficit and has been expanding gradually.According to the "301 clause" survey,there is a 375 billion 200 million dollar trade deficit between China and the United States.The United States believes that the transfer of lower labor force,production conditions and intellectual property rights in China leads to the US trade deficit.The US Federal Reserve continues to raise interest rates.Under the weak dollar,the market tends to the two-way fluctuation trend of the RMB against the US dollar.A reasonable analysis of the trend of the RMB exchange rate and timely adoption of monetary and fiscal policies is of great significance for maintaining the sustained and healthy development of the national economy.Because of the impact of external factors,there may be dynamic adjustment process and volatility clustering in the real time series of macroeconomic.Compared with the STAR model and the GARCH model,this paper considers that the STAR-GARCH model can better describe the characteristics of the "nonlinear,peak,thick tail and heteroscedasticity" of the macroeconomic time series.In this paper,the two step method is used to estimate the coefficients of the model.Before estimating the conditional variance equation of the STAR-GARCH model,the conditional mean equation must be estimated,that is,the STAR model.Because most macroeconomic and financial time series have volatility aggregation,that is,the sequence has heteroscedasticity.Compared to the STAR model with normal distribution of error,this paper focuses on the effectiveness of the nonlinear test and the test of the symmetry test for the LM-TYPE test with heteroscedasticity.Therefore,this article mainly did these aspects:First,on the basis of the linear test of LM-TYPE test and the correlation theory of symmetry test,this paper compares the normal distribution with the error obeying the variance of 1,and generates the random number of STAR model with heteroscedasticity.Monte Carlo simulated 1000 times,and obtained the test results of LM-TYPE test's nonlinear test and symmetry test.The conclusion shows that the STAR model with heteroscedasticity based on LM-TYPE test can choose the correct delay order compared to the STAR model whose error obeys the normal distribution,and the test efficiency is higher.Next,we will make an empirical analysis of the dynamic adjustment and volatility of the RMB exchange rate.Secondly,before establishing the conditional mean equation STAR model of the STAR-GARCH model,the following 3 steps must be taken.Firstly,according to the autocorrelation function and the partial autocorrelation function graph,the lag order of the single variable sequence is selected.Second,the unit root test and ARCH-LM test are used to determine the stability and heteroscedasticity of the time series,and the next step of the nonlinear test and the symmetry test are prepared.The results show that: the five sets of exchange rate returns show non unit roots,and except for the JPY/CNY sequence,the four sets of sequences show heteroscedasticity.Third,on the basis of the second step,we get the descriptive statistics of the five sets of RMB exchange rate series and speculate the general distribution.At the same time,the nonlinear test and symmetry test are carried out,and the corresponding delay order and the selected model form of the five sets of sequences are obtained.Then,on the basis of the above analysis,the STATA14.0 software is used to estimate the estimated values of the coefficients of the five series of STAR models according to the nonlinear least square method,and the mean equation of each sequence condition is determined.Five groups of residual values are diagnosed and plotted with corresponding scatter functions.The results show that the residual values of the USD/CNY,EUR/CNY,GBP/CNY and KRW/CNY sequences are all heteroscedasticity at the 10% significant level except for the JPY/CNY sequence,and the JPY/CNY sequence has the largest conversion coefficient and the model tends to be linear.Finally,on the basis of the known results of the previous residual value diagnosis,the conditional variance equation of the five sets of sequences is estimated,that is,the GARCH model is established for the residual value,and the magnitude of the sequence impact coefficient is obtained.Compared with the STAR model and the GARCH model,the MSE statistic in the STAR-GARCH model is the smallest and has good prediction ability.The results showed that,under the 1% significant level,the sequence of the four groups showed fluctuation concentration except the JPY/CNY sequence.The USD/CNY sequence was most affected by the external information impact and the fluctuation was abnormal,and the duration of the KRW/CNY sequence was the longest.