Empirical study of the GARCH model with rational errors

24th IUPAP Conference on Computational Physics

物理学;计算机科学

Chen, Ting Ting^1 ; Takaishi, Tetsuya^2

Faculty of Integrated Arts and Sciences, Hiroshima University, Higashi-Hiroshima 739-8521, Japan^1

Hiroshima University of Economics, Hiroshima 731-0192, Japan^2

关键词: Bayesian inference;    Empirical studies;    Error distributions;    Fat-tailed distributions;    Information criterion;    Metropolis-Hastings algorithm;    Realized volatility;    Tokyo Stock Exchange;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/454/1/012040/pdf
DOI  :  10.1088/1742-6596/454/1/012040

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

We use the GARCH model with a fat-tailed error distribution described by a rational function and apply it to stock price data on the Tokyo Stock Exchange. To determine the model parameters we perform Bayesian inference to the model. Bayesian inference is implemented by the Metropolis-Hastings algorithm with an adaptive multi-dimensional Student's t-proposal density. In order to compare our model with the GARCH model with the standard normal errors, we calculate the information criteria AIC and DIC, and find that both criteria favor the GARCH model with a rational error distribution. We also calculate the accuracy of the volatility by using the realized volatility and find that a good accuracy is obtained for the GARCH model with a rational error distribution. Thus we conclude that the GARCH model with a rational error distribution is superior to the GARCH model with the normal errors and it can be used as an alternative GARCH model to those with other fat-tailed distributions.