This research study investigates three new considerations for improving the performance of volatility modeling of financial returns. Two of them are related to the intraday volatility modeling and the other one is about the use of overnight information for daily volatility modeling. / About the intraday volatility modeling, the limitations and potential problems of using Andersen & Bollerslev's approach are addressed and distinct modifications are proposed to tackle the corresponding issues. The first suggestion is about the utilization of the interaction between the intraday periodicity and the heteroskedasticity while the second is about the modified normalization for the estimation of the intraday periodicity. / The proposed modifications are tested with different ARCH structures, including GARCH(1,1), FIGARCH(1,d,1) and HYGARCH(1,d,1), by using simulated data and market data. Apart from studying the 1-step-ahead out-of-sample performance, several multiple-step-ahead forecasting results are also addressed. Under the same level of model flexibility (parameterized portions), our proposed modifications always outperform the original method in both in-sample fitness and out-of-sample performance on various forecasting horizons. / On the other hand, the third suggestion is about the inclusion of overnight information for the estimation of daily volatility. This study explores the possibility of incorporating the overnight variance indirectly through the use of linearly combined daily volatility estimators. The empirical results demonstrate that the inclusion of overnight variance can produce substantial influence when the minimum-variance constraints are relaxed. Besides, the influence is revealed to be not monotonic as an increase of the overnight proportion does not necessarily produce a larger influence. / Furthermore, it is demonstrated that the inclusion of overnight variance can improve the prediction accuracy of the Chicago Board of options Exchange (CBOE) volatility indexes (VIX and VXD) under specific weight combinations. The findings contradict the common perception that overnight return does not contain useful information for daily volatility modeling.
| Identifer | oai:union.ndltd.org:CHENGCHI/U0003491979 |
| Creators | Chu, Chun Fai Carlin. |
| Publisher | The Chinese University of Hong Kong (Hong Kong). |
| Source Sets | National Chengchi University Libraries |
| Detected Language | English |
| Type | text |
| Rights | Copyright © nccu library on behalf of the copyright holders |
Page generated in 0.0018 seconds