Forecasting Volatility: A survey on the forecasting capability of Autoregressive Conditional Heteroskedasticity-models and Historical Volatility-models

Keywords
Loading...
Thumbnail Image
Authors
Issue Date
2020-08-21
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Volatility Forecasting is arguably one of the most pivotal aspects of empirical financial asset pricing and risk management. During the 2008 Financial Crisis it became evident that many standard forecasting models where inadequate to predict severe volatility. This thesis attempts to provide a thorough survey on existing volatility forecasting literature. It expands on the topic by comparing different forecasting models by means of a Root-Mean-Square analysis. Forecasts have been performed both insample and out-of-sample. It then focusses on the effects that certain market conditions might have on the forecasting capability of the models. By taking advantage of a more recent dataset it utilizes the COVID- 19 pandemic of 2020 as a turbulent sample period. This thesis has found that exogenous shocks severely impact the forecasting capability of all addressed forecasting models. The HIS-models in particular are heavily impacted by these shocks. Furthermore, HIS-models appear poor out-of-sample estimators compared to ARCH-models. The thesis concludes that the GJR-GARCH 1.1 model appears to be the best forecasting model out there. In contrast, it is concluded that the Naïve method of volatility forecasting is the worst model available.
Description
Citation
Supervisor
Faculty
Faculteit der Managementwetenschappen
Specialisation