Abstract
This paper considers CSI 300 Index futures and the underlying index from April 2010 to December 2018 based on high frequency data to test the price discovery function and spillover dynamics of the futures market given the change in futures market regulation in September 2015. The new regulation restricted the futures market intraday trading volume. This can be considered a natural experiment, offering us the opportunity to explore the factors that affect the price discovery function of the Chinese futures market. Information shares, the price lead-lag relationship, intraday returns and volatility spillovers are tested to reflect the price discovery function at both the long- and short-term intraday levels. We further compare the two subsamples before and after the regulation using static and dynamic approaches. The results suggest that shortly after the new regulation, the futures market was more sensitive to new information which dominated the price discovery process. However, the price discovery function of a futures market became much weaker after the regulation in the long run, due to a lack of liquidity. The regulation increased only the short-run price leading effect of the futures market and stabilised the market by limiting intraday arbitrage. We find that margin trading in the stock market significantly affects the price discovery ability of the futures market. Specifically, our results indicate that the Chinese stock index futures market was not the driving force of the market crash in 2015.
Acknowledgements
We would like to thank the participants at the 14th International Conference on Chinese Financial System Engineering and Risk Management in Haerbin, 2016, the 13th China Finance Annual Conference in Dalian, 2016, the Asian Economic Community Forum in Inchon, 2016 and the 2017 International Forum on Chinese Financial Systems in Tianjin, 2017 for their valuable suggestions. The usual disclaimer applies.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 RV is also calculated with 1- and 10-minute data to ensure robustness. The different frequencies for calculating RV do not affect our results.
2 Theissen (Citation2002) argues that both CFW measurements have similar qualitative results.
3 The result is robust for the forecast step and time window selection. We use 1 week and 1 trading year as examples. A smaller forecast step would make the index fluctuate more, while a larger forecast would result in a smoother line. A five-steps-ahead forecast was used to filter the noise and keep the real effect obvious. Our time window was selected to balance the sample size and ensure the accuracy of the model estimation and market condition change.