GARCH族的模型平均估计方法  

Model Averaging Estimation for GARCH Family

作　　者：赵国庆[1] 姚青松[1] 刘庆丰 

机构地区：[1]中国人民大学经济学院 [2]日本小樽商科大学商学部

出　　处：《数量经济技术经济研究》2017年第6期104-118,共15页Journal of Quantitative & Technological Economics

摘　　要：研究目标:对模型平均方法进行理论扩展,构建GARCH族的模型平均估计量及相应权重选择准则。研究方法:蒙特卡洛模拟实验方法。研究发现:在一定条件下最小化权重准则选择的权重向量将在渐近意义上最小化真实KL偏离度;蒙特卡洛模拟结果表明,与AIC准则、BIC准则、AIC模型平均、BIC模型平均的估计结果相比较,本文提出的模型平均法具有更小的KL偏离度。研究创新:将模型平均估计方法引入条件异方差模型族中。研究价值:本文结果将为捕捉金融市场资产的时变波动性提供强有力的研究工具。Objectives： This paper makes a theoretical extension of model averaging meth- ods, proposes the model averaging estimators for GARCH family and constructs the corre- sponding weight choosing criterion. Research Methods： Monte Carlo simulation methods. Research Findings： According to our results, the weight vector selected by minimizing the criterion will asymptotically minimize the true KL distance or model approximation error. Simulation results indicate that comparing with AIC, BIC, AIC-averaging and BIC-averaging, the model averaging method leads to relatively lower KL distance. Research Innovations： This paper extends model averaging method to conditional heteroskedasticity model family. Research Value： Our results provide an effective tool for empirical analysis and theo-retical forecasting.

关 键 词：模型平均 GARCH 渐近最优性 

分 类 号：F224.7[经济管理—国民经济]