基于模型错误指定的资产定价模型选择方法

发布时间：2019-05-10 23:35

【摘要】：资产定价模型选择问题是实证金融研究领域一个基本的问题。AIC、BIC是模型选择中常用的两个信息准则。可是,在实际应用中,AIC通常需要假设备选模型均是真实模型,或者至少是真实模型的较好近似。BIC通常依概率1选择真实模型,如果备选模型中没有真实模型,BIC选出的模型不具备现实意义。文章基于Fama-French三因子模型,在模型错误指定下,首次运用TIC准则,选择出模型集合中的最优模型。实证分析表明,TIC与AIC、BIC间的结果差距是显著的,从而说明了AIC、BIC的不可靠性,同时TIC也非常显著地支持了Fama-French三因子模型,这加强了运用该模型进行实证研究的说服力。
[Abstract]:Asset pricing model selection is a basic problem in the field of empirical financial research. AIC and BIC are two common information criteria in model selection. However, in practical applications, AIC usually needs to assume that the alternative models are real models, or at least a better approximation of the real models. BIC usually selects the real models according to probability 1, if there is no real model in the alternative models. The model selected by BIC is not of practical significance. In this paper, based on the Fama-French three-factor model, the optimal model in the model set is selected by using the TIC criterion for the first time under the model error assignment. The empirical analysis shows that the result gap between TIC and AIC,BIC is significant, which shows the unreliability of AIC,BIC. At the same time, TIC also supports the Fama-French three-factor model very significantly. This strengthens the persuasion of using the model to carry on the empirical research.
【作者单位】：中国人民大学汉青经济与金融高级研究院;
【基金】：中国人民大学科研基金资助项目(34312027)
【分类号】：F224;F830.91

【共引文献】