首页|时变风险厌恶与人民币汇率波动率—基于GARCH-MIDAS-SK模型的实证研究

时变风险厌恶与人民币汇率波动率—基于GARCH-MIDAS-SK模型的实证研究

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
经验研究表明汇率收益率分布呈现出时变高阶矩(偏度和峰度)特征,其对于汇率波动率建模和预测具有重要作用.同时众多研究表明,时变风险厌恶(RA)包含了金融波动率预测的相关信息.鉴于此,本文构建带时变高阶矩的GARCH-MIDAS-SK模型框架,进一步将RA指数引入该模型框架,实证检验RA对人民币汇率波动率的影响以及预测作用.实证结果表明:RA对人民币汇率波动率具有显著负的影响;人民币汇率收益率分布展现出明显的时变高阶矩特征;引入RA和时变高阶矩有助于提高模型的样本内拟合效果.基于损失函数和MCS检验证实了引入RA和时变高阶矩能够显著提高模型的样本外预测精度,且预测结果具有关于不同样本外预测阶段的稳健性.
Time-Varying Risk Aversion and RMB Exchange Rate Volatility:An Empirical Study Based on GARCH-MIDAS-SK Model
Empirical studies demonstrate that exchange rate return distributions exhibit properties of time-varying higher moments(skewness and kurtosis),which plays an important role in modeling and forecasting exchange rate volatility.Moreover,numerous studies show that time-varying risk aversion(RA)contains useful information for forecasting financial volatility.In light of this,this paper proposes the GARCH-MIDAS-SK framework that incorporates time-varying higher moments and RA to empirically investigate the impact and predictive role of RA on RMB exchange rate volatility.The empirical results show that RA has a significant negative impact on RMB exchange rate volatility.The RMB exchange rate return distributions exhibit obviously properties of time-varying higher moments.Incorporating RA and time-varying moments improves the in-sample fitting of the model.Furthermore,the loss functions and MCS test confirm that incorporating RA and time-varying higher moments leads to significantly more accurate out-of-sample volatility forecasts.The finding is robust to different out-of-sample forecasting periods.

time-varying risk aversiontime-varying higher momentsRMB exchange rate volatilityGARCH-MIDAS-SK modelMCS test

吴鑫育、梅学婷、周海林、尹学宝

展开 >

安徽财经大学金融学院,安徽蚌埠 233030

时变风险厌恶 时变高阶矩 人民币汇率波动率 GARCH-MIDAS-SK模型 MCS检验

国家自然科学基金项目安徽省高校自然科学研究重点项目安徽财经大学研究生科研创新基金项目安徽省自然科学基金项目安徽省高校杰出青年科研项目安徽省高校学科(专业)拔尖人才学术资助项目

71971001KJ2019A0659ACYC20201852208085Y212022AH020047gxbjZD2022019

2024

10.13860/j.cnki.sltj.20240510-002

数理统计与管理
中国现场统计研究会

数理统计与管理

CSTPCDCSSCICHSSCD北大核心
影响因子：1.114
ISSN：1002-1566
年,卷(期)：2024.43(4)
  • 6