Cathy W.S. Chen,Takaaki Koike,Wei-Hsuan Shau
Abstract:This research incorporates realized volatility and overnight information into risk models, wherein the overnight return often contributes significantly to the total return volatility. Extending a semi-parametric regression model based on asymmetric Laplace distribution, we propose a family of RES-CAViaR-oc models by adding overnight return and realized measures as a nowcasting technique for simultaneously forecasting Value-at-Risk (VaR) and expected shortfall (ES). We utilize Bayesian methods to estimate unknown parameters and forecast VaR and ES jointly for the proposed model family. We also conduct extensive backtests based on joint elicitability of the pair of VaR and ES during the out-of sample period. Our empirical study on four international stock indices confirms that overnight return and realized volatility are vital in tail risk forecasting.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to improve the accuracy of tail - risk prediction by combining after - hours information (i.e., overnight information) and realized volatility. Specifically, the paper proposes a new family of semi - parametric regression models - the RES - CA ViaR - oc model, aiming to simultaneously predict Value - at - Risk (VaR) and Expected Shortfall (ES), especially improving the immediate prediction ability after the market opens. The following are the core content and methods of the paper:
### Research Background and Motivation
- **VaR and ES**: VaR (Value - at - Risk) is a commonly used indicator to measure the potential loss risk, but it only provides the maximum loss at a specific confidence level. ES (Expected Shortfall) further measures the average loss exceeding the VaR threshold, so it captures tail risks more comprehensively.
- **Deficiencies in Existing Research**: Existing VaR and ES prediction models often ignore the influence of overnight information, while overnight information (such as the difference between the opening price and the previous closing price) makes a significant contribution to the total return volatility.
### Methods and Models
- **RES - CA ViaR - oc Model**: The paper proposes a new model, RES - CA ViaR - oc. Based on the traditional CA ViaR model, this model adds overnight information and realized volatility as predictor variables. The specific form of the model is as follows:
\[
Q_t=\beta_1 + \beta_2Q_{t - 1}+\beta_3RV_{t - 1}+\beta_4I(OC_t>0)|OC_t|+\beta_5I(OC_t\leq0)|OC_t|
\]
\[
ES_t = Q_t - w_t
\]
\[
w_t=\begin{cases}
\gamma_1+\gamma_2(Q_{t - 1}-r_{t - 1})+\gamma_3w_{t - 1}, & \text{if }r_{t - 1}\leq Q_{t - 1}\\
w_{t - 1}, & \text{otherwise}
\end{cases}
\]
where \(Q_t\) is the conditional VaR, \(ES_t\) is the conditional ES, \(OC_t\) is the overnight return rate, and \(RV_t\) is the realized volatility.
- **Bayesian Estimation**: Use the Bayesian method and the Markov Chain Monte Carlo (MCMC) algorithm to estimate the unknown parameters in the model and conduct joint prediction of VaR and ES.
### Empirical Analysis
- **Data Set**: The research uses data from four international stock markets, including the NASDAQ Composite Index (the United States), DAX (Germany), the Hang Seng Index (Hong Kong), and the Nikkei 225 Index (Japan). The data covers the daily opening price, closing price, and realized volatility from January 3, 2011, to June 28, 2022.
- **Model Comparison**: The paper compares the proposed RES - CA ViaR - oc model with several existing models, including ES - CA ViaR (Taylor, 2019) and RES - CA ViaR (Chen et al., 2023).
- **Evaluation Methods**: Use multiple back - testing methods to evaluate the prediction performance of the model, including unconditional coverage (UC), conditional coverage (CC), dynamic quantile (DQ) test, and comparison back - testing based on loss functions.
### Main Findings
- **Importance of Overnight Information**: Empirical results show that the RES - CA ViaR - oc model, which incorporates overnight information and realized volatility, performs better in predicting tail risks, especially in Asian markets (Japan and Hong Kong).
- **Model Advantages**: Murphy diagrams and hypothesis tests further verify the superiority of the RES - CA ViaR - oc model, especially its prediction advantages under different scoring functions.
### Conclusion
By introducing overnight information and realized volatility, the paper significantly improves the prediction accuracy of VaR and ES. This improvement is of great significance for risk management and helps financial institutions better understand and manage tail risks.