基于目标变量的非参数汇率波动预测

A Targeted Predictor Based Nonparametric Prediction of Exchange Rate Volatility

在有管理的浮动汇率制度下,人民币汇率波动幅度在汇改后逐渐提升,故有效预测汇率波动成为央行、外汇管理局以及外贸公司等汇率敏感部门的迫切需求。传统波动模型中的线性假设与较低的包含高阶滞后项能力限制了其预测精度。为提高汇率波动的预测水平,文中结合目标预测变量与非参数回归方法对美元、欧元及日元兑人民币汇率进行了不同预测期限的滚动窗口预测。通过损失函数与预测能力检验,本文发现基于目标预测变量的非参数方法较经典波动模型可显著提升汇率波动预测的精度,且对不同币种、预测期限及损失函数都表现出较强的稳健性。

As a result of managed floating in exchange rate,the volatility of Renminbi exchange rate keeps raising,which makes the precise prediction of its volatility important to PBOC,State Administration of Foreign Exchange,international trading sector and other exchange rate sensitive sectors.The linear hypothesis and the limitation of including high order lagged information reduce the predictive power of traditional volatility models.In order to enhance to performance,this paper employs targeted predictor along with nonparametric method to forecast the exchange rate volatility of USD,EUR and YEN in different forecasting horizon by rolling window.According to loss functions and predictive power tests,the method based on targeted predictor and nonparametric regression outperforms traditional volatility models significantly.And this improvement is robust across different currencies,forecasting period and loss function.