The information content of Chinese volatility index for volatility forecasting

ABSTRACT

In this paper, we investigate whether the model-free implied volatility index iVX officially launched by the Shanghai Stock Exchange has incremental explanatory power for future volatility in the SSE 50 ETF. In particular, we concentrate on Heterogeneous Autoregressive model of realized volatility and iVX (HAR-RV-iVX). We use both in-sample and out-of-sample predictive regressions to empirically indicate that the iVX significantly improves the forecasting performance of the realized volatility of SSE 50 ETF.

KEYWORDS:

• Volatility forecasting
• realized volatility
• Chinese volatility index
• SSE 50 ETF

JEL CLASSIFICATION:

• G10
• G17

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Although Qiao et al. (2019) had studied whether iVX can improve the forecasting ability of future volatility through the CSI 300 index and futures data from 9 February 2015 (the first trading day of the SSE 50 ETF option) to 29 March 2017, their results are different from the results of this paper. Specifically, their empirical results indicated that the multi-step ahead forecasting errors are higher than one-step ahead forecasting errors.

² Some of the literature utilizes the terms realized volatility for the realized variance. However, we use the terms realized volatility for the square root of the realized variance in this paper, because on average the forecasting ability of the square root of the realized variance is better than that of the realized variance, see, for example, Andersen, Bollerslev, and Diebold (2007), Forsberg and Ghysels (2007), Byun and Kim (2013) and among others.:

³ We chose the logarithmic rather than plain realized volatility for modelling because of the plain realized volatility is skewed, see Gonçalves and Meddahi (2011) for a more detailed discussion on this topic. In a regression context, this implies that regression coefficient estimates have large variances, and a few large volatilities can have an overly large influence on the estimates. Thereby, using the logarithmic HAR-type models results in more efficient coefficient estimates with lower variances.

⁴ Due to the problem of data access rights, we can only obtain high-frequency sample data for the past three years at the time of writing of this paper.

⁵ The construction method of the iVX is the same as the CBOE volatility index VIX. As for the detailed construction method of iVX, one can refer to the announcement issued by the Shanghai Stock Exchange, http://www.sse.com.cn/market/sseindex/diclosure/c/c_20161104_4198915.shtml. Meanwhile, for more details of the CBOE VIX formula, see https://www.cboe.com/micro/vix/vixwhite.pdf. In fact, we can get the iVX data from an option BBS in China, which is an option website jointly established by several securities and futures exchanges and is currently the most active and professional option community in mainland China, see https://www.optbbs.com/. It is important to note that we have verified that the iVX data released by the BBS are consistent with the data released by the Shanghai Stock Exchange.

⁶ To clearly report the coefficient estimate of each forecasting variable, we multiply the original value of realized volatility RV${ }_t$ by $\sqrt{252}$ to convert it into annualized realized volatility. In addition, since iVX is a percentage value, we divide it by 100 to make it consistent with the dimension of quantity of the annualized realized volatility.

⁷ Based on the S&P 500 index options, Jiang and Tian (2005) found the empirical evidence that the model-free implied volatility subsumed all information contained in the Black-Scholes implied volatility and past realized volatility and was a more efficient forecast for future realized volatility.

⁸ This procedure is similar to Corsi and Renò (2012), Byun and Kim (2013), Bugge et al. (2016), Chang et al. (2019) and among others.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (Grant Nos. 71901124, U1901223 and 71720107002) and the Natural Science Foundation of Jiangsu Province (Grant No. BK20190695).