Forecasting US stock market volatility: How to use international volatility information

Abstract

This paper aims to accurately forecast US stock market volatility by using international market volatility information flows. The results show the significant ability of the combined international volatility information to predict US stock volatility. The predictability is found to be both statistically and economically significant. Furthermore, in this framework, we compare the performance of a large set of approaches dealing with multivariate information.Dynamic model averaging (DMA) and dynamic model selection (DMS) perform better than a wide variety of competing strategies, including the heterogeneous autoregressive (HAR) benchmark, kitchen sink model, popular forecast combinations, principal component analysis (PCA), partial least squares (PLS), and the ridge, lasso, and elastic net shrinkage methods. A wide range of extensions and robustness checks reduce the concern regarding data mining. DMA and DMS are also able to significantly forecast international stock market volatilities.

The predictability of carbon futures volatility: New evidence from the spillovers of fossil energy futures returns

“…In addition to extracting the common factors among variables, forecast combination is another efficient manner to integrate different information in individual variables. Following the related studies (see, e.g., Liu & Wang, 2021; Rapach et al, 2010; Zhang, Wang, et al, 2021; Zhang, Luo, et al, 2022), we combine the volatility forecasts from individual models to generate combination forecasts, which is given as $$${\hat{\text{RV}}}_{\text{comb},t+1}=\sum _{i=1}^{N}{\hat{\varphi }}_{i,t+1}{\hat{\text{RV}}}_{i,t+1},$$$where $${\hat{\text{RV}}}_{\text{comb},t+1}$$ and $${\hat{\text{RV}}}_{i,t+1}$$ are the combination and individual forecasts of volatility at month t + 1, respectively, $${\hat{\varphi }}_{i,t+1}$$ is the ex ante estimated weight on $${\hat{\text{RV}}}_{i,t+1}$$, and N is the total number of predictive models with individual fossil energy returns.…”

“…Following the convention in the studies on volatility prediction (see, e.g., Gong & Lin, 2018b; Patton & Sheppard, 2009; Wang et al, 2016; Zhang, Wang, et al, 2021), we employ the approach of MCS test proposed by Hansen et al (2011), to identify whether the model of interest has statistically different out‐of‐sample performance from others. The MCS is a subset of a few models with the best forecasting performance.…”

Forecasting Bitcoin volatility: A new insight from the threshold regression model

“…Given the availability of high‐frequency data and the considerable attention given to using high‐frequency data for exploring cryptocurrency markets, we use intraday Bitcoin prices to compute the realized volatility (RV), which is proposed by Andersen and Bollerslev (1998) and given as where denotes the i th intraday Bitcoin return on day t , , and is the sampling frequency. Here, we use a 5‐min interval as our sampling frequency, which is widely recommended by many related studies on volatility forecasting (see, e.g., Andersen et al, 2007; Corsi et al, 2010; Haugom et al, 2014; Zhang, Wang, et al, 2020). Moreover, as evidenced by the influential study of Liu et al (2015), the 5‐min RV is hard to be outperformed by any other measures among 400 volatility estimators across 31 different financial assets.…”

“…In the out‐of‐sample test above, we use the rolling window method to generate the out‐of‐sample forecasts. Alternatively, we consider an expanding window, which is usually used as an estimation window in some related studies (see, e.g., Dai, Kang, Wen, 2021a; Dai & Zhu, 2020; Dai, Zhou, et al, 2021b; Lei et al, 2020; Zhang, Ma, et al, 2019; Zhang, Wang, et al, 2020; Zhang, Wei, et al, 2019). In Table 9, we report the corresponding results and find that the HAR‐RV‐RT‐MS 3 model delivers the largest MCS p values of 1 for all the loss functions.…”

Forecasting China's crude oil futures volatility: New evidence from the MIDAS-RV model and COVID-19 pandemic

“…To ensure that our results are robust, we perform many tests in this section, including different k max , out-of-sample R 2 , alternative benchmark model, and direction-of-change test. In addition, these robustness test methods are also widely used in financial forecasting research (see, e.g., Yang et al, 2015 ; Tian et al, 2017 ; Li et al, 2020a ; Li et al, 2020b ; Wen et al, 2020 ; Zhang et al, 2020 ; Li et al, 2021 ; Lu et al, 2021 ; Zhang et al, 2021 ).…”