A novel time-varying FIGARCH model for improving volatility predictions

Abstract

The FIGARCH model has received wide attention due to its ability to capture the features of volatility long-memory persistence and clustering. The classical FIGARCH model is based on the difference scheme of Grünwald–Letnikov fractional operators. This paper introduces the new class of FIGARCH processes for improving time-varying volatility predictions. Firstly, a novel FIGARCH model based on the Caputo fractional operators (FIGARCH-C model for short) is proposed. Secondly, a quasi-maximum likelihood estimation (QMLE) is used to estimate the parameters of the FIGARCH-C(1, d, 1), the FIGARCH(1, d, 1) and GARCH(1, 1) models. Finally, we apply the three models to Brent crude oil and S&P 500 returns and provide the comparison results of the three models. The results show that the FIGARCH and FIGARCH-C models outperformed the GARCH model in capturing the long memory in volatility. It is also found that the FIGARCH-C model is more sensitive to capture the change in the volatile period.