摘要
本文以4种农产品期货的高频数据为样本,在实证考察预测因子对农产品期货已实现波动率的预测能力基础上,通过假定时变HAR模型的参数遵循独立正态-伽马自回归过程先验分布,构建了具有时变稀疏度的HAR模型(TVS-HAR),以同时考虑预测模型参数的时变性和预测模型的时变性,并采用MCS检验评价和比较该模型和其他HAR族模型的样本外预测性能.实证结果表明:TVS-HAR模型能较好地识别和拟合潜在预测因子对农产品期货市场波动率的预测的重要性和影响程度的时变性;跳跃成分对我国农产品期货市场已实现波动率具有一定的预测能力;相对于其他几类HAR模型,TVS-HAR模型的预测性能最好.
Based on investigating the forecast power of the potential predictors for realized volatility of four agricultural commodity futures by employing the high-frequency data from Chinese markets, we propose a HAR model with time-varying sparsity （TVS-HAR） to consider the time-varying regression coefficients and the time-varying forecasting models simultaneously, which is constructed by utilizing the independent normal-gamma autoregressive processes priors for the regression coefficients of a time-varying HAR model, and then we use MCS test to evaluate and compare out-of-sample forecasting performance of proposed model and other HAR type models. Our results indicate that the proposed model can adequately account for the time-varying effect of the regression coefficients and identify predictors that are most relevant over time, and the jumps have some forecast power for realized volatility forecast. Furthermore, the proposed model appears to be the most effective models for forecasting the realized volatility of agricultural commodity futures.
作者
田凤平
杨科
TIAN Fengping YANG Ke(International School of Business and Finance, Sun Yat-sen University, Guangzhou 510275, China School of Economics and Commerce, South China University of Technology, Guangzhou 510006, China)
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2016年第12期3003-3016,共14页
Systems Engineering-Theory & Practice
基金
国家自然科学基金面上项目(71673089)
国家社会科学基金(15CJY004)
教育部人文社会科学基金(14YJCZH141)
广州市哲学社会科学发展"十二五"规划课题(15Y21)~~
关键词
已实现波动率
时变稀疏度
HAR模型
样本外预测
realized volatility
time-varying sparsity
HAR model
out-of-sample forecasting
