A Hybrid Markov Switching Garch Model Approach For Improving Volatility Dynamics

• Main Author: Hossain, Md Jamal
• Format: Thesis
• Language: English
• Published: 2021
• Subjects: QA1 Mathematics (General)
• Online Access: http://eprints.usm.my/53225/1/MD%20JAMAL%20HOSSAIN%20-%20TESIS24.pdf; http://eprints.usm.my/53225/

Time series analysis has long attracted the attention of researchers in a variety of fields. Past two decades, time series have been analyzed using linear models, which have a number of advantages. However, the question of whether there are other methods that can help understand and predict actual data than linear models have been presented. The historical time series data show nonlinearity, evidence of structural changes, and are extremely volatile. In this case, linear models are incapable of explaining volatility and predicting future values. The GARCH family models explain volatility and forecasting very well for nonlinear time series data but collapse when structural breaks and market turbulence are present. This research aims to incorporate a new nonlinear time series model comprised of the nonlinear conditional mean model, ExpAR, and the nonlinear conditional variance model, MSGARCH. This hybrid model was developed to capture nonlinearity in both the mean and variance equations during structural changes and extreme market conditions. As a result, it can be a valuable method for fitting, illustrating, and capturing downside risk in nonlinear time series data. Moreover, it can offer some enhancement in both fitting and explaining volatility dynamics compared to the benchmark model. Since the launch of the proposed model, similar time series data has been generated from the proposed model and the benchmark model. Then the generated data are fitted into these models. Later, the real-world time series data were fitted into these models and their performance was compared.