Does the introduction of stock index futures increase stock market volatility? : evidence from the Chinese stock market crash in 2015

(1)

UNIVERSITY OF AMSTERDAM

AMSTERDAM BUSINESS SCHOOL MSc FINANCE

QUANTITATIVE FINANCE

Does the Introduction of Stock Index Futures Increase Stock

Market Volatility? Evidence from the Chinese Stock Market

Crash in 2015.

Author: Siyun Han

Student Number: 10598006

Thesis Supervisor: Liang Zou

(2)

Statement of Originality

This document is written by Siyun Han, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

The impact of index futures trading on the spot market volatility has been debated throughout the world for many years, but the conclusion is still ambiguous. The main purpose of this thesis is to analyze whether the introduction of the CSI 500 Index and the SSE 50 Index destabilized underlying markets during the 2015 Chinese stock market crash. Additionally, this paper also aims to investigate the regulatory effect on the spot markets volatility. In order to do so, the sample period between April 16, 2013 and September 2, 2017 is applied. Evidence from the GJR-GARCH (1,1) model suggests that neither the index futures effect nor the regulation effect was significant regarding of the CSI 500 Index. However, both the introduction of index futures and the announcement of regulation reduced volatility in the SSE 50 Index during the crash.

(4)

Introduction

On April 16th, 2010, Chinese first index futures, the CSI 300 Index futures was introduced to boost the capital market. As a derivative, the introduction of the CSI 300 Index futures provided investors with more opportunities to hedge the risks and ease fluctuations in the stock market. This development was hailed as a milestone in Chinese financial market, and the introduction of index futures was dominantly viewed as a sufficient tool to stabilize market volatility among investors and regulators at that moment. However, a crash in Chinese stock market led to more than 30% fall in Shanghai stock market in June 2015, five years after the introduction of index futures, and government attributed the crash to index futures’ short-mechanism in the end.

To be specific, two new index futures were just introduced two months ahead of the crash: the CSI 500 Index futures and the SSE 50 Index futures. The CSI 500 Index selects 500 largest remaining A-Share stocks of good liquidity from Shanghai and Shenzhen security markets after excluding both the CSI 300 Index constituents and the largest 300 stocks. The index reflects the price fluctuation and overall performance of small-mid cap A-shares on both Shanghai and Shenzhen security markets. By contrast, the SSE 50 Index only consists of 50 largest A-share stocks of good liquidity in Shanghai security market and reflects the situation of these large and high-quality enterprises. The World Federation of Exchanges ranked China’s two futures markets, the CSI 300 Index and the CSI 500 Index, the most active ones in the world since these two index futures allowed day-trading (Miao et al., 2017). Therefore, regulators started investigating malicious trading in short positions in the index futures market.

Soon on September 2nd_{of that year, the China Securities Regulatory Commission}

(CRSC) announced several restrictions on index futures trading. Curbs included raising margin requirements for non-hedging purposes, imposing higher transaction fees, placing limits on same-day trading, and suspending trading in different company shares that accounted for nearly 40% of market capitalization (Miao et al., 2017). Under such regulation, volume in the CSI 500 Index futures market plunged around 99%, according to Kyoungwha Kim’s report from Bloomberg (2015). Afterward, restrictions on futures

(6)

trading have been eased twice that expanded the maximum limit on daily stock index future trading from 10 lots to 20 lots and reduced margin requirements for non-hedging transactions. However, with the concern of restricted trading on index futures in China, words on the third loosening of futures trading have never stopped since 2017, and this conjecture raises the interest of this thesis.

According to Ross (1976), volatility as a measure of information flow in a derivative instrument has always been an indispensable tool for trading strategies. Therefore, it is essential for investors to have an accurate estimate of volatility. A considerable amount of studies has been focused on how the introduction of index futures influenced the spot market volatility. From a global perspective, however, the main focus of previous studies was on developed markets, and most studies that focused on the Chinese stock market only considered the CSI 300 Index. The conclusion of the index futures effect is mixed. One school, including Bologna and Cavallo (2002), Chang, Cheng, and Pinegar (1999) and Yilgor and Mebounou (2016), maintains the view that trading index futures has a stabilizing effect on the volatility of the spot market. This is due to derivatives’ function of price discovery, which can improve the market depth and informativeness. Moreover, traders can hedge risks through reverse operations and thereby achieve the effect of market stabilization by reducing the volatility of stock prices in the spot market. The opposite argument holds that index futures trading will severely increase the instability of the spot market. This is because the index futures market is more likely to attract uninformed investors and promote speculation activities due to high leverage. In addition, the trading volume in both the index futures market and the underlying market will be exponentially multiplied on the expiration date of the contract. This statement is supported by various results from previous empirical researches, such as Bhamra and Uppal (2009), Antoniou and Holmes (1995), and Gulen and Mayhew (2000).

As both arguments seem to be possible, there is no conclusive result to prove that limiting index futures trading is an effective way to restore the health of the spot market. Therefore, the main purpose of this thesis can be divided into two research questions:

(7)

1). Did the introduction of index futures increase stock market volatility during 2015 stock market crash?

2). How did the volatility in the spot market change after the curbs have been announced?

In order to do so, this paper employs the GJR-GARCH model, an extension model of the Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model, to process daily closing prices from April 16th_{, 2013 to September 2}nd_{, 2017, —— two}

years before the introduction happened and two years after the restriction announced. Empirical results suggest that for the CSI 500 Index, both the introduction of index futures and the announcement of regulation had no significant effect on the spot market volatility. However, when studying the SSE 50 Index, index futures had a positive influence on reducing the spot market volatility during the stock market crash, rather than undermining the stability of the underlying market. Additionally, as expected, the announcement of regulation had an efficient impact on further stabilizing the underlying market.

This thesis contributes to the existing literature in the following two aspects. Firstly, unlike most existed literature which focused on developed markets, this paper will investigate the 2015 Chinese stock market crash by using the CSI 500 Index and the SSE 50 Index as research objects. It is worth paying academic attention to the Chinese market since China is now the world's second-largest economy regarding its GDP and stock market value. Meanwhile, the Chinese stock market has its unique characteristics. Most listed firms in China are state-owned enterprises, and only a small portion of their shares are available for trading, which may increase the probability of speculation. In addition, retail investors who are more likely to be noise traders are still the dominant force that drives stock market movements. Furthermore, futures market allows traders intraday trading while spot market regulators maintain the T + 1 trading mechanism, which also facilitates daily speculative trading (Chen et al., 2013). In this case, the introduction of index futures seems to increase the volatility of the spot market. However, the Chinese stock market has strict government supervision and prohibits

(8)

short selling, which are conducive to stock market stabilization through decreasing volatility. Base on this contradiction, this paper aims to figure out whether index futures is the main cause of the 2015 Chinese market crash.

Secondly, this paper purposes to examine the efficiency of index futures introduction beyond the traditional focus on volatility to investigate whether it is necessary to further relax the curb on index future. As it is widely known, stock returns do not always conform to the normal distribution in the reality, third and fourth moments of returns will be considered in this paper to study the deviation of stock returns and provide a better interpretation of spot market volatility. To my best knowledge, when investigating the impact of index futures effects on the spot market volatility, only a few articles have considered skewness and kurtosis.

The remainder of the paper is divided into five sections. Section II will present a brief review of the previous empirical studies regarding the relationship between index futures and the spot market volatility, the relationship between index futures and the stock crash, and the relationship between index futures and restrictions. Section III will describe the chosen data, and the methodology that will be employed in this paper will be illustrated in Section IV. Analysis of empirical results and further suggestions are provided in Section V. The final section offers conclusions.

II. Literature review

Futures contracts are binding agreements to buy or sell a particular item on a future date for a predetermined price. A “long” position holder expects rising values of the underlying asset and consequently profits from rising futures prices. On the contrary, a “short” position holder expects declining values of the underlying asset and then profits from decreasing prices. Similar as other derivative instruments, futures contracts provide investors with a broader range of investments, more possibilities of the market as well as nonmarket components of risk, and return separation in their portfolios (Figlewski, 1984).

(9)

The first stock index futures contract was the Value Line contract, introduced by the Kansas City Board of Trade on February 24, 1982 (Gulen &Mayhew, 2000). Since the stock index future is a futures contract based on the value of a particular stock market index, the index futures market and the stock market naturally correlate with and affect each other. Therefore, from the emergence of the index futures contract, a considerable number of studies have been done on the interaction and influence between futures contracts and underlying assets. Most studies focused on developed countries, with two main bodies of debates on the index futures effect on the spot market volatility.

One group of researches states that index futures have a positive impact on stabilizing spot market. In theory, compared with spot markets, trading in the futures market has advantages concerning price discovery, lower volatility, higher liquidity and lower costs. Futures are generally regarded as an excellent vehicle for hedging risks and decreasing volatility in stock market. More specifically, Powers (1970) stated that futures trading have a positive impact on the spot market as it improves market depth and informativeness. In consistent with this argument, Bologna and Cavallo (2002) argued that index futures contract has an immediate effect on stabilizing underlying cash market and increasing market efficiency in their research on Italian market. By investigating in Istanbul market, Yilgor and Mebounou (2016) concluded that the introduction of futures contract had caused a significant decrease of the spot price volatility. Chang, Cheng, and Pinegar (1999) found similar evidence in the Nikkei stock market that spot market volatility was decreased when Nikkei futures were introduced on the Osaka Securities Exchange. Other studies, including Drimbetas et al. (2007), Baldauf and Santini (1991) and Schwert (1990), concluded that the introduction of index futures did not significantly change volatility in spot markets.

However, another group argues that index futures have a negative impact on the spot market. Although index futures are closely correlated to the underlying index, they are not identical. Index futures, as a derivative, trade on margins and have a high degree of leverage, often attracts irrational investors and uninformed speculators in reality. Such investors often behave irrationally and create noise in the market, which consequently would result in price fluctuation and underlying market destabilization.

(10)

Bhamra and Uppal (2009) considered a dynamic, general-equilibrium, information-efficient and endowment economy in which only agents differed in risk aversion and showed that introduction of a non-redundant derivative increased the volatility of stock market return. Similarly, Antoniou and Holmes (1995) examined the impact of trading in the FTSE-100 Stock Index Futures on the volatility of the underlying spot market by using the GARCH model. Results indicate that futures trading has led to increased volatility in UK market, but the nature of volatility has not changed post-futures, which implies that the introduction of futures has improved the speed and quality of information flowing to the spot market. After examining the international stock market in 25 countries, Gulen and Mayhew (2000) found the same significant increase in both US and Japan markets while no significant effect in other countries. Hence, the conclusion so far is still controversial for developed markets.

When looking at previous literature that studied on the Chinese stock market, Chen et al. (2012) applied the panel data evaluation approach and found that index futures significantly decreased the volatility of the Chinese stock market. On the contrary, Xie and Huang (2014) employed a set of GARCH models to investigate the impact of index futures trading on the volatility of the CSI 300 Index and concluded that the launch of index futures did not significantly decrease the volatility of the spot market. However, there was a decrease in sensitivity to new information while sensitivity to historical details increases after the introduction of the CSI 300 Index futures.

II.2 Research on index futures and stock market crash

On October 19th_{, 1987, a significant crash happened in US stock market, which led to}

22.6% one-day loss in DJIA Index. The 1987 crash event revealed the weakness of the trading systems themselves and how they could be strained and come close to collapse in extreme conditions (Carlson, 2006). Afterward, a presidential task force named as Brady Commission was set up to figure out the factors contributed to the crash. According to Brady Report, two program trading strategies in the futures market, “portfolio insurance” and “index arbitrage”, were the main problem of the volatility during the crash (Brady Rport, 1988). Ultimately, Brady Report imposed “circuit breaker” device to prevent the spillover effect and enhance liquidity in the equity

(11)

market. However, empirical researches so far have not reached any conclusive evidence on the impact of index futures trading on spot market volatility.

Antoniou and Garrett (1993) used minute by minute values of the FTSE 100 index and minute by minute prices for the index futures contract to study to what extent did index futures contribute to the October stock market crash. They concluded that links between two markets were not provided by arbitrage. The result is consistent in stable time periods. Hence, the blame for the crash did not lie in the future market. Similarly, Antoniou et al. (1998) stated that derivatives trading transmitted news into prices and transferred asymmetries from the spot market to the futures market, rather than resulted in market turbulence. Other researchers, including Karmara, Miller, and Siegel (1992), Becketti and Roberts (1990) and Darrat (2002), found no significant evidence that the introduction of index futures would increase spot market volatility during the crash includes.

The 2015 Chinese stock market crash began with the popping of the stock market bubble in June 2015, which led to a total loss of one-third of A-shares value within one month on the Shanghai Stock Exchange. Even though the Chinese government immediately started to take a series of measures to recover the market, the index encountered another 8.49% drop on August 24th_{, 2015. Afterward, the China Financial}

Futures Exchange (CFFEX) announced two rounds of measures to reduce speculative trading in the index futures market in August and September, separately (Han & Liang, 2017). With regard to the role of index futures during the 2015 Chinese stock market crash, Hou and Nartea (2017) stated that the index futures market played a long-term leading role in sharing information, which then improved stock market efficiency during the crash. They also found that government intervention on futures trading impaired price discovery in the futures market. Similarly, Han and Liang (2017) adopted a difference-in-difference approach using index component stocks as the treatment group and non-component shares as the control group to identify the net effect of index futures trading on spot market efficiency. They claimed that the quality of the spot market has dropped significantly after the index futures trading were abandoned.

(12)

When surveillance is inefficient, the authority may raise position limits and price limits that traders can carry into the final settlement to reduce market volatility (Dutt & Harris, 2004). However, existing literature points out that such limits are not enough to reduce market volatility. Gastineau (1992), Telser (1993), and Grossman (1993) argued that the use of position limits to prevent market manipulation was not a wise choice and did not adequately prevent market manipulation. Gastineau (1992) and Telser (1993) believed that supervision and enforcement were still required to avoid market manipulation. Grossman (1993) pointed out that the implementation of a position restriction on financial futures would move traders to the rest of the market instead of reducing the overall risk.

II.4 Skewness and Kurtosis

It is widely known that stock returns do not always perfectly follow the normal distribution in reality. According to Roon et al. (2012), investors dislike fat tails along with negative skewness which may subject them to significant potential loss. Therefore, higher moments of return are worth studying in addition to the impact on volatility and expected performances. In the normal distribution, the value of skewness should be zero while the value of kurtosis should be three. Analysis and interpretation tend to focus on downside risk. Negative skew means that the left tail is longer than the right tail, which suggests that investors are more likely to have extreme outcomes. Also, excess kurtosis (i.e., kurtosis>3) indicates that increase in volatility is due to extreme values. Therefore, the abnormal values of skewness and kurtosis indicate that risk may be underestimated due to deviations in volatility. Corrado (1997) tested the Black-Scholes option pricing model that assumes stock returns to be log-normally distributed, considered skewness and kurtosis deviation. They observed negative skewness and positive excess kurtosis when calculating volatility of options. Results suggested that additional skewness- and kurtosis-adjustments terms significantly improved accuracy and consistency when pricing deep in-the-money and deep out-of-the-money options.

(13)

As reported in the introduction section, the central purpose of this thesis is to analyze whether the introduction of index futures will increase the volatility in the underlying market in China and whether the volatility will decline after the regulation announced. Since the crash happened in June 2015, just two months after the introduction of the CSI 500 Index Futures and the SSE 50 Index Futures, these two indices will be used to study the volatility behavior of the market. And the sample period to be used is from April 16th_{, 2013, to September 2}nd_{, 2017. A total of 1070 daily data are collected for}

each index.

Moreover, in order to directly compare the index futures effect and regulatory effect, the entire period will be divided into three sub-periods based on the dates of the two major events. These two dates are the date of index futures introduction and the date of limitations announcement. Index futures trading was officially introduced on April 16th_{, 2015, while curbs were announced on September 2}nd_{, 2015. Hence, the first}

period is the period before the introduction of index futures, spanning from April 16th_,

2013, to April 16th_{, 2015. The second period covers the period between the introduction}

of index futures and the announcement of curbs, spanning from April 17th_{, 2015, to}

September 1st_{, 2015. The third period is the period after the publication of regulations,}

spanning from September 2nd_{, 2015, to September 2}nd_{, 2017. According to the Chinese}

government’s actions, it is expected that volatility will increase from the first period to the second period and then decrease in the third period.

Daily data for both the CSI 500 Index and the SSE 50 Index, including the daily trading data and fundamental information (i.e., closing price, trading volume, turnover), are collected from Businessinsider.com. The thesis initially converts the closing price data to daily compounded returns by taking the first log difference to decrease the variation of the time series. That is, the return at time t is given by R =ln (Pt/Pt-1) *100, where Pt is the closing price for day t.

IV. Methodology

The main hypotheses in this thesis are formulated in the following way:

1. H0: Volatility of the underlying markets after the CSI 500 Index and the SSE 50 Index futures introduction did not increase.

(14)

Index futures introduced increased.

2. H0: Volatility of the underlying markets after regulation announced did not decrease. H1: Volatility of the underlying markets after regulation announced decreased. Antoniou and Holmes (1995) claimed that investigation in different time periods may significantly alter the volatility. Therefore, unless information remains constant, volatility must be time-varying, even on a daily basis. According to previous literature (Gulen & Mayhew, 2000; Yilgor & Mebounou, 2016; Bologna & Cavallo, 2002), the most common way to capture the time-varying nature of volatility is to model the Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model.

The GARCH (p, q) model has been developed by Bollerslev (1986), a stable form of Autoregressive Conditional Heteroscedasticity (ARCH (p)) model that was introduced by Engle (1982), which allows a much flexible lag structure. In contrast to the estimation of regression equations by OLS, which requires the error term to be homoscedastic, the GARCH model models the conditional variance of the error term as a linear function of the lagged squared residuals and the lagged residual conditional variance. Another advantage of a GARCH model is that it captures the tendency in financial data for volatility clustering. A model with errors that follows a GARCH (p, q) process is represented as follows:

ht = 𝛼0 +∑$_%&'𝛼i*𝜀t-i2+∑*_%&'βi*ht-i,

where ht is the conditional variance at time t, 𝛼0 is the constant, 𝛼i is the coefficient of

past value of the squared residuals, βi is the coefficient of past period of volatility, and

residual 𝜀t-I follows the normal distribution.

GARCH (1,1) is the simplest but most popular model in empirical studies. In this case, conditional variance depends on the value of past return and variance that only lag one day back. GARCH (1, 1) model:

ht = 𝛼0 + 𝛼1*𝜀t-12+β1*ht-1,

where 𝛼0>0, 𝛼1>=0, β1>=0.

GARCH model assumes that news impact is symmetric, that both positive news and negative news have an equal effect despite the sign. However, it has been improved

(15)

in the stock market that the negative shock may have a bigger impact than the positive shock due to the leverage effect. If so, the GARCH model may be mispriced, and volatility inference from this model will be misleading correspondingly. In order to consider the asymmetric impact, GARCH has been expanded to several forms, such as EGARCH and GJR-GARCH models. In general, the likelihood function for the GJR model is higher than for the EGARCH model (Antonious et al., 1998). So, the GJR-GARCH model found by Glosten, Jagannathan, and Runkle (1989) will be employed in this paper to assess whether there has been an increase in the volatility after the introduction of index futures. The GJR-GARCH model contains a dummy variable I that takes the value of 1 when the shock is negative and 0 otherwise. As the main purpose of this paper is to figure out the impact of the introduction of index futures, this paper will compare volatility before the index futures introduction, after the regulatory announcement and in the intermediate period. Therefore, an additional dummy d is added to present the stage in the sample period. The GJR-GARCH (1,1) model is formed as follows:

(1) Testing index futures effect:

ht = 𝛼0 + 𝛼1*𝜀t-12+β1*ht-1+𝛾It-1𝜀-12+𝛿1*d1

where: It-1=1 if 𝜀t-1<0, It-1=1 if 𝜀t-1>=0;

d1=0 before index futures introduced, d1=1 afterward (2) Testing regulatory effect:

ht = 𝛼0 + 𝛼1*𝜀t-12+β1*ht-1+𝛾It-1𝜀-12+𝛿2*d2

where: It-1=1 if 𝜀t-1<0, It-1=1 if 𝜀t-1>=0;

(16)

V.

Results and Analysis

In this section, empirical results are analyzed. In order to find evidence of how the introduction of index futures may affect the underlying market volatility, Stata program is first employed to provide descriptive statistics of the CSI 500 Index and the SSE 50 Index, such as daily return, trading volume, and turnover. Afterward, Python program is used to establish the GJR-GARCH model to estimate the conditional volatility during the sample period.

V.1 Descriptive Statistics

Daily return, trading volume and turnover are graphed from April 16th_{, 2013, to}

September 2nd_{, 2017, to directly show the changes of fundamental statistics in both}

indices. Figures can be found in Appendix 1. It can be seen that there is no significant fluctuation in the logarithm daily return of the CSI 500 Index and the trend is relatively stable. However, extreme values of daily returns sometimes appear and are concentrated in a specific period. This phenomenon is called volatility clustering, which means that losses tend to be followed by losses while gains tend to be followed by gains. As volatility clustering is a “non-parametric” property of time series variables, the existence of clustering will be discussed along with the GARCH model.

As mentioned earlier, the whole period is divided into three subperiods to test the index effect and the regulatory effect. As can be seen from Figure 1a., the yield rate of spot market return experienced a significant decrease to more than 8 percent after the introduction of the CSI 500 Index futures, and then recovered to a new high of about 5 percent before the regulation was announced. After the announcement of limitation on trading index futures, the rate of return continuously decreased to around zero in the later stage of the sample period. The evidence of loss on returns can also be found from the decline in daily trading volume and daily turnover presented in Figure 1b. and Figure 1c. Although trading volume and turnover had already started to rise before the introduction of index futures, the increase became more significant after the new index futures traded. However, after the crash happened in June, both figures dropped sharply and remained low afterward. Data in Appendix 2 shows the graph of elementary statistics of the SSE 50 Index. A similar trend existed in the SSE 50 Index. However, index futures effect in the SSE 50 Index seems to be smaller than the CSI 500 Index.

(17)

The trading volume and turnover before the index futures introduced were not much different from those after index futures proposed, but both statistics have dropped sharply due to the regulatory curbs. For a more detailed analysis of the impact of the index futures introduction and restrictions, Exhibit 1 and 2 give descriptive statistics of return, trading volume, and turnover of the CSI 500 Index and the SSE 50 Index, respectively. 1070 daily time series observations were collected for each index.

V.1.1 CSI 500 Index

With regard to daily returns, Panel 1 shows that the CSI 500 Index has a mean of around 0.08% with a standard deviation of 1.863% for the whole period. As mentioned earlies, the entire sample period is divided into three parts, pre-futures period (i.e., before April 16th_{, 2015), post-futures but pre-regulation period (i.e., April 16}th_{, 2015-Sep 1}st_{, 2015)}

and post-regulation period (i.e., after Sep 2nd_{, 2015). Figures show that statistics}

changed a lot among three subperiods. The mean of daily return for the CSI 500 Index decreased from 0.19% to -1.7% after the index futures was introduced but recovered to 0.27% with the announcement of regulation. Standard deviation, which measures the market return volatility, increased during the post-futures period from 1.36% to 3.71% and decreased to 1.74% after the restriction. These numbers show that the introduction of index futures reduced daily return and exacerbated spot market volatility. Although the announcement of regulation eased this situation, the mean of daily return was still much lower than pre-futures period while volatility was almost a little bit higher. As can be seen from Panel B and Panel C, daily trading volume and turnover experienced a similar trend as the rate of return. However, it is worth noting that means of daily volume and turnover in the third subperiod was higher than that in the first subperiod with a lower standard deviation, which suggests that index futures brought liquidity to the spot underlying market even under governmental control.

To analyze the validity of volatility, this paper further examines the third and fourth moments of return. The standard value of skewness is equal to zero while standard kurtosis is equal to three in the normal distribution. Results show that the daily return of the CSI 500 Index had a fat tail since skewness in all periods were negative and kurtosis in most periods were significantly higher than three. Negative skew indicates that investors may have more extreme low outcomes. Skewness for daily return was firstly skewed to the left at a magnitude of -0.76 in the pre-futures period, the absolute

(18)

value of skewness decreased after the index futures has been introduced but increased to over one during the post-regulation period. Hence, the introduction of index futures reduced the probability of extreme negative return, while instability became higher after the restriction was announced, which is opposite to what is expected in this paper. However, when observing trading volume turnover, tails are positively skewed, so extreme positive outcomes are more likely to exist instead of a loss.

Distribution with kurtosis greater than three means that instability may be raised by outliers. Evidence suggests that changes in kurtosis for all observed variables almost followed the same trend as skewness. The fat tail initially existed in the pre-futures period and disappeared after the introduction of index futures. But the tail became even fatter after restrictions were announced. Hence, results of skewness and kurtosis show that index futures trading provided benefits for reducing extreme negative returns while the announcement of limitation had an opposite effect. This result is consistent with the conclusion of Bohl et al. (2012).

V.1.2 SSE 50 Index

Exhibit 2 presents the results from the SSE 50 Index. As the SSE 50 Index only consists of 50 largest A-share stocks in Shanghai security market, the trading size tends to be smaller than the CSI 500 Index, which results in a lower value of daily return, trading volume, and turnover. Nevertheless, the trend of index futures effect and regulatory effect in the SSE 50 Index seems to be similar as that in the CSI 500 Index. Index futures introduction caused extreme negative outcome and reduced averaged daily return into higher volatility. Conversely, adoption of index futures expanded spot market trading and thus increased trading volume and turnover. However, after the restrictions were applied, the daily rate of yield was restored, and the transaction size became smaller in the meanwhile. It should be noticed that since both skewness and kurtosis were significantly different from the standardized value after the announcement of supervision, the volatility of outcomes were unreliable and extreme outliers may be attributed to the regulatory effects.

For a more reliable examination, GJR-GARCH model that considers new impact is applied in the following section to estimate daily volatility in the sample period to test the index futures effect and the regulatory effect. Before that, two more tests to evaluate characteristics of time series variable will be first applied.

(19)

EXHIBIT 1

HISTORICAL CSI 500 INDEX RETURN STATISTICS

CSI 500 daily descriptive statistics for entire sample period spanning from April 16, 2013, through September 2, 2017, and three sub-periods. Panel A statistics are based on logarithm daily returns calculated as log(Pt/Pt-1), where Pt denotes an index value at the end of month t. Panel B statistics are based on daily trading volume in a million and Panel C statistics are based on daily turnover in a million. 95% confidence intervals for standard sample skewness and kurtosis coefficients are ± 0.438 and 3 ± 0.877, respectively.

PANEL A: DESCRIPTIVE STATISTICS OF THE CSI 500 INDEX DAILY RETURN

PANEL C: DESCRIPTIVE STATISTICS OF THE CSI 500 INDEX DAILY TURNOVER (IN MILLION)

Obs. Mean Std. Dev. Skewness Kurtosis Entire period 1,070 104212.5 76147.85 1.974409 7.480653 16 April, 2013 -16 April,2015 486 70801.48 52902.87 2.276848 8.499091 17 April, 2015 - 1 Sep, 2015 97 200549 84114.97 0.5681683 2.852324 2 Sep, 2015 - 2 Sep, 2017 488 104319.1 40632.95 1.556045 5.802035 Obs. Mean Std. Dev. Skewness Kurtosis Entire period 1,070 0.0008005 0.0186323 -1.061153 7.125022 16 April, 2013 -16 April,2015 486 .0018644 0.0136349 -0.7575747 4.376113 17 April, 2015 - 1 Sep, 2015 97 -0.017237 0.0371134 -0.5056942 2.393559 2 Sep, 2015 - 2 Sep, 2017 488 0.002697 0.0174085 -1.162217 8.496811

PANEL B: DESCRIPTIVE STATISTICS OF THE CSI 500 INDEX DAILY TRADING VOLUME (IN MILLION)

Obs. Mean Std. Dev. Skewness Kurtosis Entire period 1,070 82.46077 41.7395 1.275717 4.701535 16 April, 2013 -16 April,2015 486 65.12899 35.59081 1.539684 5.146071 17 April, 2015 - 1 Sep, 2015 97 163.0058 39.92356 0.503825 2.506154 2 Sep, 2015 - 2 Sep, 2017 488 83.83609 25.73457 1.090773 4.417279

(20)

EXHIBIT 2

HISTORICAL SSE 50 INDEX RETURNS STATISTICS

SSE 50 daily descriptive statistics for entire sample period spanning from April 16, 2013, through September 2, 2017, and three sub-periods. Panel A statistics are based on logarithm daily returns calculated as log(Pt/Pt-1), where Pt denotes an index value at the end of month t. Panel B statistics are based on daily trading volume in a million and Panel C statistics are based on daily turnover in a million. 95% confidence intervals for normal sample skewness and kurtosis coefficients are ± 0.438 and 3 ± 0.877, respectively.

PANEL D: DESCRIPTIVE STATISTICS OF THE SSE 50 INDEX DAILY RETURN

PANEL E: DESCRIPTIVE STATISTICS OF THE SSE 50 INDEX DAILY TRADING VOLUME (IN MILLION)

Obs Mean Std.Dev Skewness Kurtosis Entire period 1,070 49.04794 50.00417 2.411843 9.828119 16 April, 2013 -16 April,2015 486 45.81705 46.30539 1.969166 6.184041 17 April, 2015 - 1 Sep, 2015 97 149.1489 62.80366 1.308656 5.269366 2 Sep, 2015 - 2 Sep, 2017 488 32.62035 18.16779 2.591372 14.40023

PANEL F: DESCRIPTIVE STATISTICS OF THE SSE 50 INDEX DAILY TURNOVER (IN MILLION)

Obs Mean Std.Dev Skewness Kurtosis Entire period 1,070 56477.45 67738.49 2.365988 8.582118 16 April, 2013 -16 April,2015 486 49644.49 60906.5 2.011246 6.018171 17 April, 2015 - 1 Sep, 2015 97 200549 84114.97 0.5681683 2.852324 2 Sep, 2015 - 2 Sep, 2017 488 35015.54 20189.79 2.548354 14.09174

Obs Mean Std.Dev Skewness Kurtosis Entire period 1,070 0.000540935 0.01652557 -0.4689276 9.045807 16 April, 2013 -16 April,2015 486 .0013449 .0157425 0.0516784 4.557762 17 April, 2015 - 1 Sep, 2015 97 -.0026969 .0327641 -0.2677443 3.623594 2 Sep, 2015 - 2 Sep, 2017 488 0.000470082 0.01191376 -0.7977577 9.317416

(21)

V.2 Covariance Stationary test and autocorrelation test

When the AR model is applied to forecast time series variables (i.e., daily volatility), three assumptions must be satisfied, no autocorrelation, no conditional heteroskedasticity, and covariance stationary. Otherwise, the estimation of future volatility may have no economic meaning. As the GARCH (1,1) model employed, it is assumed that there is a conditional heteroskedasticity. Therefore, only the other two tests must be completed before studying the evidence derived from the GARCH model.

V.2.1 Covariance Stationary test

Theoretically, if there is a random walk, the time series is said to be not covariance stationary. Hence, Dickey-Fuller type unit root strategy is used to examine whether the null hypothesis that the time series has a unit root and is not stationary should be rejected. If the lag coefficient is equal to one, the time series is said to have a unit root. By conducting 1070 daily observation in STATA program, results show that ADF of the CSI 500 Index equals to -22.671 and the ADF of the SSE 50 Index equals to -23.684. Both values are smaller than the critical value of -3.96 at 1% significance level. Therefore, the time series of both indices are covariance stationary to process further empirical tests.

V.2.2 Autocorrelation test

The autocorrelation shows the similarity between the expectation value of residual terms as a function of the time lag between them. Positive autocorrelation means that the positive error in one period will lead to another positive error in the subsequent period. Negative autocorrelation implies that positive error in one period will result in negative error later. Therefore, similar to the random walk, existence of autocorrelation means that OLS estimates and any forecast based on those estimates are inappropriate. To eliminate the autocorrelation and improve the estimation, extra lagged values should be added if the autocorrelation exists. The test is conducted in Python program, and results prove that both indices do not have autocorrelation.

(22)

V.3 GJR-GARCH Analysis

In order to measure the impact of the introducing futures and options, the GJR-GARCH (1,1) model including skewed generalized error distribution is applied to estimate volatility of return. Coefficients indicate how past information is influencing the current volatility of stock returns. This paper examines these parameters for all three sub-periods. Since the news impact is not immediate, the second period and the third period are combined to test against the pre-futures period and post-regulation period. If the value of these coefficients is significantly higher in latter periods than that in the pre-futures period, the introduction of index pre-futures increases volatility; if the value falls from the second period to the third period, regulatory announcements reduce instability. More specifically, a dummy variable presenting the stage in the sample period is added in the model to test the futures and regulatory effects among the subperiods.

V.3.1 CSI 500 Index

Figure 1 shows the conditional volatility estimated by GJR-GARCH (1,1) model during the entire sample period; it is evident that the volatility increased significantly from the beginning of 2015 till September, then dropped sharply. At the beginning of 2016, volatility increased again and remained low since then.

Exhibit 3 gives a detailed value of the coefficients of the constant α0, the lagged return α1, the lagged volatility β1 and the leverage effect 𝛾1. Results show that during the entire sample period, all coefficients are significantly different from zero at the one percent significance level, which implies that today’s conditional variance is highly dependent on past information. Since ARCH coefficient α1 is the "news" component, a significant α1 explains that recent news has a greater impact on price changes, that is, a large shock on day t-1 will lead to a large variance on day t. The GARCH coefficient β1 measures the impact of "old news." A relatively higher value of β1 in this context implies a greater memory of shocks in this model. The sum of the coefficients α1 and β1 tends to be close to 1, indicating a high degree of persistence. The significant negative coefficient of the leverage effect indicates the existence of an asymmetric behavior, that is, negative shocks tend to have a more significant impact than the positive shocks.

(23)

Regarding of the CSI 500 Index, the sum of ARCH and GARCH coefficients is almost equal to 1 for the whole period, and the overall conditional variance is stable. However, when divided into three subperiods, it can be found that the conditional variance in the pre-futures period does not significantly rely on news. However, 𝛼1 is significantly declining while β1 is significantly increasing afterward. Therefore, the conditional variance is more dependent on past news than recent news. Surprisingly, the coefficient of futures effect is not significant at any significance level while the coefficient of regulatory effect significantly reduces volatility by around 6% at 10% significance level. These results mean that neither index futures trading nor supervision has significant impact on the volatility of the spot market.

V.3.2 SSE 50 Index

Figure 2 shows evidence of the conditional variance of the SSE 50 Index. It can be seen that volatility had increased before the index futures introduction and climbed again with the launch of the new financial product. Similar to the CSI 500 Index, the figure experienced a drop after the announcement of regulation and then remained low.

Most of the coefficients in Exhibit 4 indicate an analogous result for the conditional variance shown in Exhibit 3, except for the coefficients of the leverage effects, the index futures effect, and the regulatory effect. The significant positive coefficient γ suggests no asymmetric behavior. Hence, positive and negative shocks have indifferent effects on the conditional variance. Since 𝛿1 and 𝛿2 are significantly negative at the 1%

significant level, the introduction of index futures significantly reduces the volatility of the SSE 50 Index by 1.14, and supervision announcement decreases another 1.27 of the volatility. This result is contrary to that expected earlier in this paper.

In general, findings provide answers for hypotheses, suggesting that the introduction of the index futures did not destroy the stability of two underlying markets, but reduced the volatility instead. And as expected, regulation had a more significant impact on decreasing the volatility of the spot market.

(24)

Figure 1. Conditional volatility in the CSI 500 Index estimated from modeling GJR-GARCH (1,1) model during the whole sample period

EXHIBIT 3

RESULTS ON CONDITIONAL VOLATILITY OF THE CSI 500 INDEX

The table reports the estimated coefficients in the CSI 500 Index from a GJR-GARCH model during the entire sample period and three subperiods. 𝛼1 represents the coefficient of lagged return, β1 represents the coefficient of lagged variance, and 𝛾1 represents the coefficient of leverage lags. 𝛿1 is the coefficient of dummy that equals one after index futures have been introduced. 𝛿2 is the coefficient of dummy that equals one after the regulation has been announced.

t statistics in parentheses *, **, *** denote significance at the 0.10, 0.05, and 0.01 level, respectively.

(1) (1) (2) (3)

Entire Period Pre-futures Post-futures Post-Regulation 𝛼0 1.57e-06*** 8.33e-05* 1.20e-06*** 1.17e-06***

(3.68e-07) (4.33e-05) (2.98e-07) (3.57e-07)

𝛼1 0.0452*** 0.143* 0.0494*** 0.0419*** (0.00759) (0.0864) (0.00912) (0.00786) β1 0.947*** 0.495* 0.953*** 0.959*** (0.00541) (0.259) (0.00584) (0.00728) 𝛾1 -0.00416*** -0.197* -0.0234*** -0.0235*** (0.0132) (0.107) (0.0167) (0.0154) 𝛿1 𝛿2 -.3185919 (.3331478) -.5712934* (.3034078) Observations 1,070 486 584 488

(25)

Figure 2. Conditional volatility in the SSE 50 Index estimated from modeling GJR-GARCH (1,1) model during the whole sample period

EXHIBIT 4

RESULTS ON CONDITIONAL VOLATILITY OF THE SSE 50 INDEX

The table reports the estimated coefficients in the SSE 50 Index from a GJR-GARCH model during the entire sample period and three subperiods. 𝛼1 represents the coefficient of lagged return, β1 represents the coefficient of lagged variance, and 𝛾1 represents the coefficient of leverage lags. 𝛿1 is the coefficient of dummy that equals one after index futures have been introduced. 𝛿2 is the coefficient of dummy that equals one after the regulation has been announced.

t statistics in parentheses *, **, *** denote significance at the 0.10, 0.05, and 0.01 level, respectively.

(1) (2) (3) (4)

Entire Period Pre-futures Post-futures Post-regulation 𝛼0 7.33e-07*** 3.37e-06** 7.33e-07** 5.42e-07**

(2.84e-07) (1.43e-06) (3.18e-07) (2.51e-07)

𝛼1 0.0594*** 0.0559*** 0.0645*** 0.0370*** (0.00800) (0.0118) (0.0126) (0.0105) β1 0.934*** 0.922*** 0.933*** 0.948*** (0.00533) (0.0127) (0.00658) (0.00688) 𝛾1 0.0108*** 0.0195*** -0.00698*** 0.0166*** (0.0106) (0.0151) (0.0192) (0.0172) 𝛿1 𝛿2 -1.14015*** (.3497582) -1.266956*** (.3172527) Observations 1,070 486 584 488

(26)

VI. Conclusion and Discussion

Since the introduction of the first index futures contract, the relationship between index futures and the spot market volatility has been discussed for many years. However, the conclusion of index futures effect is still ambiguous based on previous literature. By studying the 2015 Chinese stock market crash, the main purpose of this thesis is to analyze the changes in the spot market volatility after the introduction of the CSI 500 Index futures and the SSE 50 Index futures. Additionally, the China government announced restrictions on index futures trading to release the instability of the spot market. Hence, this thesis further investigates the regulatory impact on the volatility of the underlying market.

The sample period chosen is from April 16th_{, 2013, to September 2}nd_{, 2017. A total}

of 1070 observations are collected for each index. Stata program is first employed to provide descriptive statistics, including the third and fourth moments of daily return, to examine the validity of the volatility. Results suggest that returns of both indices are negatively skewed, that is, the increase in the volatility may be attributed to extreme losses. When dividing the entire period into three subperiods, results of skewness and kurtosis show that index futures trading provided benefits for reducing extreme negative returns while the announcement of restrictions had an opposite effect. This conclusion is consistent in both indices.

Afterward, the GJR-GARCH (1,1) model is applied to forecast the conditional variance during the sample period. In order to clarify the index futures effect and regulatory effect among the subperiods, an extra dummy presenting the stage of the sample period is added into the model. Empirical evidence suggests that neither the introduction of index futures nor the regulations increased the volatility in the CSI 500 Index. However, these two effects significant stabilize the volatility of the SSE 50 Index during the crash.

Since the empirical results are contrary to the Chinese government’s view, this thesis believes that index futures trading benefits the underlying market through increasing liquidity and reducing volatility. Therefore, it is feasible to further relax

(27)

restrictions on index futures trading. However, since Chinese stock market has its own characteristics, it still requires supervision to ensure the health of stock market trading.

(28)

Bibliography

Antoniou, A., Holmes, P., & Priestley, R., 1998, “The Effects of Stock Index

Volatility: An Analysis of the Asymmetric Response of Volatility to News," The Journal of Futures Markets. Vol. 18(2), pp. 151-166.

Antoniou, A. and Garrett, I., 1993, “To What Extent Did Stock Index Futures Contribute to the October 1987 Stock Market Crash”, The Economic Journal, pp.1444-1461.  

Becketti, S. and Roberts, D.J., 1990, "Will Increased Regulation of Stock Index Futures Reduce Stock Market Volatility," Economic Review, pp. 33-46.   Bhamra, H. S., and Uppal, R., 2009, “The Effect of Introducing a Non-redundant

Derivative on the Volatility of Stock-Market Returns when Agent Differ in Risk Aversion," Review of Financial Studies, Vol. 22(6), pp. 2303-2330.

Bologna, P. and Cavallo, L., 2002, “Does the Introduction of Stock Index Futures Effectively Reduce Stock Market Volatility? Is the ‘Futures Effect’ Immediate? Evidence from the Italian Stock Market Exchange Using GARCH.”, Applied Financial Economics, Vol. 12(3), pp.183-192.

Bollerslev, T. 1986. “Generalized Autoregressive Conditional Heteroscedasticity.”, Journal of Econometrics, Vol. 31, pp. 307-327.

Carlson, M. A., 2006. "A Brief History of the 1987 Stock Market Crash with a Discussion of the Federal Reserve Response.", Finance and Economics Discussion Series 2007-13, Board of Governors of the Federal Reserve System (U.S.).

Chang, E. C., Cheng, J. W., and Pinegar, J. M., 1999, “Does Futures Trading Increase Stock Market Volatility? The Case of the Nikkei Stock Index Futures Markets.”, The Journal of Banking and Finance, Vol. 23, pp.727–753.

(29)

Corrado, C.J., 1997, “Implied Volatility Skews and Stock Index Skewness and Kurtosis Implied By S&P 500 Index Option Prices.”, Journal of Derivatives, pp.8-19.

Darrat, A.F., Rahman, S. and Zhong, M., 2002, “On the Role of Futures Trading in Spot Market Fluctuations: Perpetra- tor of Volatility or Victim of Regret.”, The Journal of Financial Research, No. 3, pp. 431-444.  

Drimbetas, E., Sariannidis, N., and Porgiris, N., 2007, “The Effect of Derivatives Trading on Volatility of the Underlying Asset: Evidence from the Greek Stock Market.”, Applied Financial Economics, Vol. 17, pp.139-148.

Dutt, H. R., and Harris, L. E., 2004, “Position Limits for Cash-settled Derivative Contracts.”, The Journal of Futures Markets, Vol. 25 (10), pp.945-965

Engle, R. F. 1982, “Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation.”, Econometrica, Vol. 50, pp.987-1007.

Figlewski, S., 1984, “Hedging Performance and Basis Risk in Stock Index Futures.”, The Journal of Finance, Vol.39(3), pp.657-669.

Gastineau, G.L., 1992, “Option Position and Exercise Limits: Time for a Radical Change.”, The Journal of Portfolio Management, No. 1, pp. 92-96.

Glosten, L., Jagannathan, R., and Runkle, D. (1993) “On the Relationship between the Expected Value and the Volatility of the Nominal Excess Return on Stocks” The Journal of Finance, Vol. 48(5), pp.1779-1801.

Grossman, S.J.,1993, “The Case for Eliminating Position Limits on Financial Futures.”, The Journal of Financial Engineering, No. 1, pp.39-42.

Gulen, H., and Mayhew, S., 2000, “Stock Index Futures Trading and Volatility in International Equity Markets.”, The Journal of Futures Markets, Vol. 20(7), pp.661-685.

(30)

Han, Q. and Liang, J.F., 2017, “Index Futures Trading Restrictions and Spot Market Quality: Evidence from the Recent Chinese Stock Market Crash.”, The Journal of Futures Markets, Vol. 37(4), pp.411-428.

Hou, Y. and Nartea, G., 2017, “Price Discovery in the Stock Index Futures Market: Evidence from the Chinese Stock Market Crash.”, MPRA Paper, No. 81995. Miao, H., Ramchander, S., Wang, T. and Yang, D., 2017, “Role of Index Futures on

China’s Stock Markets: Evidence from Price Discovery and Volatility Spillover.”, Pacific-Basin Finance Journal, Vol. 44, pp.13-26.

Karmara, A., Miller, T.W.,and Siegel, A.F., 1992, “ The Effect of Futures Trading on Stability of Standard and Poor 500 Returns.”, The Journal of Futures Markets, Vol. 12(6), pp.645-658.

Kim, K., 2015, “China Just Killed the World’s Biggest Stock-Index Futures Market.”, Bloomberg Article.

Powers, M. J., 1970, “Does Futures Trading Reduce Price Fluctuations in the Cash Markets?”, American Economic Review, Vol. 60 (4), pp. 460-464.

Roon, F.D., Eiling, E., Gerard, B. and Hillion, P., 2012, “Currency Risk Hedging: No Free Lunch.”, Journal of International Money and Finance, Vol.31(5), pp.1249-1278.

Ross, S. A., 1976, “The Arbitrage Theory of Capital Asset Pricing”, Journal of Economic Theory, Vol. 13, pp.341-360.

Telser, L.G., 1993, “A Review of the Case for Position Limits on Agricultural Futures.”, The Journal of Financial Engineering, No. 2, pp.33-38.   Xie, S.Q.and Huang, J.Z., 2014, “The impact of index futures on spot market

volatility in China.”, Emerging Markets Finance and Trade, Vol. 50, pp.167-177.

(31)

Yilgor, A. G., and Mebounou, C. L. C., 2016, “The Effect of Futures Contracts on the Stock Market Volatility: An Application on Istanbul Stock Exchange.”, Journal of Business, Economics, and Finance, Vol. 5(3), 307-317.

(32)

APPENDIX A.

Figure 1a. Historical Daily Return of CSI 500 Index.

Figure 1b. Historical Daily Trading Volume of CSI 500 Index.

Figure 1c. Historical Daily Trading Turnover of CSI 500 Index. -10.00% -8.00% -6.00% -4.00% -2.00% 0.00% 2.00% 4.00% 6.00% 8.00% 04/ 16/ 2013 06/ 05/ 2013 07/ 25/ 2013 09/ 10/ 2013 11/ 05/ 2013 12/ 20/ 2013 02/ 13/ 2014 04/ 01/ 2014 05/ 21/ 2014 07/ 08/ 2014 08/ 22/ 2014 10/ 16/ 2014 12/ 02/ 2014 01/ 20/ 2015 03/ 13/ 2015 04/ 30/ 2015 06/ 17/ 2015 08/ 04/ 2015 09/ 22/ 2015 11/ 13/ 2015 12/ 30/ 2015 02/ 23/ 2016 04/ 11/ 2016 05/ 27/ 2016 07/ 15/ 2016 08/ 31/ 2016 10/ 26/ 2016 12/ 12/ 2016 02/ 03/ 2017 03/ 22/ 2017 05/ 11/ 2017 06/ 29/ 2017 08/ 15/ 2017

CSI 500 Index Daily Return

After Regulation Annocement Before Index Futures Introduction 0 50 100 150 200 250 300 04/ 16/ 2013 06/ 04/ 2013 07/ 23/ 2013 09/ 05/ 2013 10/ 30/ 2013 12/ 13/ 2013 01/ 29/ 2014 03/ 21/ 2014 05/ 09/ 2014 06/ 25/ 2014 08/ 08/ 2014 09/ 24/ 2014 11/ 14/ 2014 12/ 30/ 2014 02/ 16/ 2015 04/ 09/ 2015 05/ 26/ 2015 07/ 10/ 2015 08/ 25/ 2015 10/ 19/ 2015 12/ 02/ 2015 01/ 18/ 2016 03/ 09/ 2016 04/ 25/ 2016 06/ 13/ 2016 07/ 27/ 2016 09/ 09/ 2016 11/ 03/ 2016 12/ 19/ 2016 02/ 09/ 2017 03/ 27/ 2017 05/ 15/ 2017 06/ 30/ 2017 08/ 15/ 2017

CSI 500 Index Trading Volume (million)

Before Index Futures Introduction After Regulation Announcement 0 100,000 200,000 300,000 400,000 500,000 600,000 04/ 16/ 2013 06/ 05/ 2013 07/ 25/ 2013 09/ 10/ 2013 11/ 05/ 2013 12/ 20/ 2013 02/ 13/ 2014 04/ 01/ 2014 05/ 21/ 2014 07/ 08/ 2014 08/ 22/ 2014 10/ 16/ 2014 12/ 02/ 2014 01/ 20/ 2015 03/ 13/ 2015 04/ 30/ 2015 06/ 17/ 2015 08/ 04/ 2015 09/ 22/ 2015 11/ 13/ 2015 12/ 30/ 2015 02/ 23/ 2016 04/ 11/ 2016 05/ 27/ 2016 07/ 15/ 2016 08/ 31/ 2016 10/ 26/ 2016 12/ 12/ 2016 02/ 03/ 2017 03/ 22/ 2017 05/ 11/ 2017 06/ 29/ 2017 08/ 15/ 2017

CSI 500 Index Turnover( in million)

Before Index Futures Introduction

After Regulation Announcement

(33)

APPENDIX B.

Figure 1a. Historical Daily Return of SSE 50Index.

Figure 1b. Historical Daily Trading Volume of SSE 50 Index.

Figure 1c. Historical Daily Trading Turnover of SSE 50 Index. -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 16/ 04/ 2013 31/ 05/ 2013 17/ 07/ 2013 28/ 08/ 2013 18/ 10/ 2013 29/ 11/ 2013 13/ 01/ 2014 03/ 03/ 2014 15/ 04/ 2014 29/ 05/ 2014 11/ 07/ 2014 22/ 08/ 2014 13/ 10/ 2014 24/ 11/ 2014 07/ 01/ 2015 25/ 02/ 2015 09/ 04/ 2015 22/ 05/ 2015 06/ 07/ 2015 17/ 08/ 2015 30/ 09/ 2015 18/ 11/ 2015 30/ 12/ 2015 18/ 02/ 2016 31/ 03/ 2016 16/ 05/ 2016 29/ 06/ 2016 10/ 08/ 2016 23/ 09/ 2016 11/ 11/ 2016 23/ 12/ 2016 13/ 02/ 2017 27/ 03/ 2017 11/ 05/ 2017 26/ 06/ 2017 07/ 08/ 2017

SSE 50 Index Daily Return (%)

Before Index Futures After Regulation Announcement 0.00 50.00 100.00 150.00 200.00 250.00 300.00 350.00 400.00 16/ 04/ 2013 05/ 06/ 2013 25/ 07/ 2013 10/ 09/ 2013 05/ 11/ 2013 20/ 12/ 2013 13/ 02/ 2014 01/ 04/ 2014 21/ 05/ 2014 08/ 07/ 2014 22/ 08/ 2014 16/ 10/ 2014 02/ 12/ 2014 20/ 01/ 2015 13/ 03/ 2015 30/ 04/ 2015 17/ 06/ 2015 04/ 08/ 2015 22/ 09/ 2015 13/ 11/ 2015 30/ 12/ 2015 23/ 02/ 2016 11/ 04/ 2016 27/ 05/ 2016 15/ 07/ 2016 31/ 08/ 2016 26/ 10/ 2016 12/ 12/ 2016 03/ 02/ 2017 22/ 03/ 2017 11/ 05/ 2017 29/ 06/ 2017 15/ 08/ 2017

SSE 50 Index Trading Volume(million)

Before Index Futures Introduction After Regulation Announcement 0 50000 100000 150000 200000 250000 300000 350000 400000 450000 2 0 13 -0 4-1 6 2 0 13 -0 6-0 5 2 0 13 -0 7-2 5 2 0 13 -0 9-1 0 2 0 13 -1 1-0 5 2 0 13 -1 2-2 0 2 0 14 -0 2-1 3 2 0 14 -0 4-0 1 2 0 14 -0 5-2 1 2 0 14 -0 7-0 8 2 0 14 -0 8-2 2 2 0 14 -1 0-1 6 2 0 14 -1 2-0 2 2 0 15 -0 1-2 0 2 0 15 -0 3-1 3 2 0 15 -0 4-3 0 2 0 15 -0 6-1 7 2 0 15 -0 8-0 4 2 0 15 -0 9-2 2 2 0 15 -1 1-1 3 2 0 15 -1 2-3 0 2 0 16 -0 2-2 3 2 0 16 -0 4-1 1 2 0 16 -0 5-2 7 2 0 16 -0 7-1 5 2 0 16 -0 8-3 1 2 0 16 -1 0-2 6 2 0 16 -1 2-1 2 2 0 17 -0 2-0 3 2 0 17 -0 3-2 2 2 0 17 -0 5-1 1 2 0 17 -0 6-2 9 2 0 17 -0 8-1 5

SSE 50 Index Turnover (million)