Forecasting stock market realized volatility: the role of global terrorist attacks

ABSTRACT

In this study, we provide the predictive linkage between global terrorist attacks and stock market volatility. We propose the predictive model by extending the prevailing heterogeneous autoregressive model for realized volatility (HAR-RV) with global terrorism and denote it as HAR-RV-GT. According to the Diebold – Mariano test and the model confidence set, we consistently find the superior forecasting performance delivered by the HAR-RV-GT model. For comprehensive empirical results, we extend the key finding to various settings including the consideration of popular jump and leverage effects, the use of more types of forecasting models, and the inclusion of long-horizon global terrorist attacks as well as the domestic terrorist attacks in the US. Additionally, the substantial economic gains based on a mean-variance investor confirm the valuable forecasting role of terrorist attacks. Our results are robust to a wide range of checks.

KEYWORDS:

• Volatility forecasting
• global terrorist attacks
• HAR model
• stock market

JEL CLASSIFICATION:

• C53
• C22
• G11
• G15

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ In the extension analyses, we consider several prevailing extension forms of the standard HAR-RV model, such as the considerations of jump and leverage.

² The website of the Global Terrorism Database is https://www.start.umd.edu/gtd/.

³ By employing the Augmented Dicky-Fuller test (ADF), we provide the stationary test for all considered time series. The ADF statistics regarding RV, RVW, RVM, and GT suggest that they have no unit root and are stationary.

⁴ Given the page limitation, the results are not reported and are available upon request.

⁵ In the following section of the robustness checks, we further provide the evaluation results on the basis of the direction-of-change.

⁶ We also assess the out-of-sample forecasts using an expanding window and alternative evaluation periods in the robustness section.

Additional information

Funding

This work is supported by the financial support from the National Natural Science Foundation of China [71901122, 72071114, 72001110, 71722015]; the Fundamental Research Funds for the Central Universities [30919013232]