基于门限分位数自回归模型的人民币汇率波动及预测研究  

Research on RMB exchange rate volatility and forecast based on threshold quantile autoregressive model

作　　者：康宁 刘霆 KANG Ning;LIU Ting(School of Economics,Nanjing University of Finance and Economics,Nanjing Jiangsu 210023,China)

机构地区：[1]南京财经大学经济学院,江苏南京210023

出　　处：《阜阳师范大学学报（自然科学版）》2022年第2期24-32,共9页Journal of Fuyang Normal University:Natural Science

基　　金：教育部人文社会科学研究青年基金项目(18YJC790069);全国统计科学研究一般项目(2018LY18)。

摘　　要：汇率是衡量宏观经济状况的重要指标之一,对其波动特征以及短期预测的研究受到广泛关注。已有文献多采用均值框架下的非线性模型研究人民币汇率的波动及预测问题,难以揭示汇率分布的完整特征。文章根据2015年“8.11”汇改到2021年8月6日的人民币对100美元汇率中间价数据,建立门限分位数自回归模型,深入挖掘人民币汇率波动的非线性和异质性特征,并与传统的TAR(Quantile Autoegressive model)模型、QAR(Quantile Autoegressive model)模型和ARIMA(Autoregressive Integrated Moving Average model)模型进行比较。实证结果表明:人民币汇率不仅在波动幅度上具有两阶段的非线性特征,而且呈现典型的异质性,即不同分位点处当期汇率均受到前期汇率的影响,具有一定的惯性效应,但其影响程度随分位点的变化而变化。此外在预测能力方面,TQAR模型能够较好地预测汇率波动趋势,不但提供比TAR、QAR和ARIMA模型更高的预测精度,而且通过条件密度曲线细致刻画出人民币汇率条件分布的位置、散布与形状等信息。The exchange rate is one of the important indicators to measure the macroeconomic conditions,and the research has received extensive attention on its fluctuation characteristics and short-term forecasts.The existing literature mostly uses the nonlinear model under the mean frame to study the volatility and prediction of the RMB exchange rate,and it is difficult to reveal the complete characteristics of the exchange rate distribution.The paper builds up a threshold quantile autoregressive model(TQAR)based on the data of the central parity rate of the RMB to 100 USD exchange rate from the“8.11”exchange rate reform in 2015 to August 6,2021,deeply explores the non-linear and heterogeneous characteristics of RMB exchange rate fluctuations,and compares with traditional TAR model,QAR model and ARIMA model.The results show that RMB exchange rate not only has two-stage nonlinear characteristics in the fluctuation range,but also presents typical heterogeneity.The current exchange rate at different quantiles is affected by the previous exchange rate,which has a certain inertia effect,but the influence degree varies with the change of quantiles.In addition,TQAR model can better predict the trend of exchange rate fluctuations,not only providing higher prediction accuracy than TAR,QAR and ARIMA models,but also carefully describing the location,dispersion and shape of the conditional distribution of RMB exchange rate through conditional density curve.

关 键 词：人民币汇率 分位数自回归 门限自回归 条件密度预测 

分 类 号：F224.0[经济管理—国民经济]