Empirical investigation of price movements of China stock index futures

Abstract

This thesis examines the price movements of the stock index futures of China. In history, forecasting price movements has been explored and attempted by financial researchers and practitioners; nevertheless, theoretical conclusions on its possibility are lacking thus far. Studies of emerging markets, such as China, India, and Brazil, are also of interest in their own right because the evolving process of the markets can address the question of whether efficiency improves as the institutional framework for the markets develops and as trading volumes increase. This Thesis consists of three studies. In the first study, the random walk hypothesis (RWH) is tested in the context of the index futures of China. Both traditional inter-day returns and new intra-day returns are tested. Compared with inter-day analysis, test results on intra-day data can describe the movements of intra-day markets more effectively because intra-day analysis eliminates overnight news propagation, thus generating more precise conclusions on the intra-day market for intra-day traders. Five statistical tests (i.e., Jarque-Bera test, runs test, augmented Dickey- Fuller test, variance ratio test and its variant) are applied to test the RWH. The idea of the new variant is based on the fact that the expectations of volatilities estimated from high-frequency and inter-day prices should be equal if the normality holds. For high-frequency data, closing prices are reported every 500 ms. However, the correlation of two consecutive closing prices is not zero, which invalidates closing prices when analyzing the RWH. Therefore, bid1 prices are used in the new variant test. Results of the five tests suggest that the futures price is lacking in randomness. In the second study, text mining and statistical models are deployed to explore the relationship between the Shanghai Stock Exchange Composite Index and the collective emotions of individual investors. The emotions of individual investors are quantified by analyzing and aggregating investor online posts that contain finance-related keywords. To identify a set of finance-related keywords, three years of blogs from a famous financial blog site are segmented by an automatic text segmentation method; meanwhile, in the literature, people typically select keywords manually. Posts that discuss the keywords are extracted out of all types of topics from Sina Weibo, the largest microblog platform in China. Statistical results reveal the relationship between daily posts and daily opening prices with a one-day lag, which indicates the existence of information (news) propagation lag and provides the possibility of forecasting the market. In the third study, the movement patterns of the CSI 300 stock index futures are examined. The daily price patterns of index futures from April 2010 to April 2013 are investigated using the cluster methodology of data mining. Results indicate the presence of trends in the price patterns. Four types of trend-following trading strategies are devised, namely, fixed-time entry strategy, opening-range breakout strategy, dual thrust strategy, and quadruple thrust strategy. The fixed-time entry strategy is improved by testing the stability of the trend. The dual thrust strategy includes two variants, namely, strategies with one trade and strategies with multiple trades in a day. Back tests indicate that the proposed strategies generate excess returns despite the consideration for the transaction costs and risks; the quadruple thrust strategy outperforms the other strategies. However, the experimental results suggest that in the time horizon, the performance in the early years is better than that in the later years, implying that the emerging market becomes efficient as it evolves.

Date of Award	15 Jul 2015
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Leong Chye Andrew LIM (Supervisor)