Modeling and forecasting for volatility of the natural gas price via asymmetric ARCH models

초록

We explore the modeling and forecasting of volatility in financial time series using an asymmetric ARCH(1) model. The conditional variance of the model is different from those of the conventional ARCH or TGARCH models, but it has a simple form with an asymmetric coefficient. The model framework is extended by incorporating different distributions for the innovations, including Gaussian, t-distribution, and Generalized Gaussian distribution. The Maximum Likelihood Estimation (MLE) method is applied to estimate the model parameters, ensuring robust and consistent estimates even when the true error distribution is unknown. The natural gas dataset is applied to this model and a comprehensive empirical analysis is conducted to compare the performance of various distributions in capturing the asymmetric volatility patterns observed in the data. The forecasting results show that the asymmetric ARCH model with t-distribution is the best for the natural gas dataset with the smallest prediction errors among other conventional volatility models.

키워드