CHAPTER 3 Circuit Breaker and CSI300
3.3 CSI300
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
price triggered the threshold. Fused and continuous means that when the price hits the first threshold, the trade can still continue for a subsequent period of time, but the quote must be limited to the blow threshold and the price volatility is limited.
3.3 CSI300
Stock index futures can not only enrich the types of financial transactions but also improve the utilization rate of funds. In the meantime, it can also avoid the systemic risks of the market and ensure the stable operation of the macroeconomics. Therefore, it has become one of the most important and successful instruments in the financial market. The development of the stock index has deeply promoted the global financial market changes and improved market mechanisms. Along with the continuous development of the Chinese stock market, Shanghai and Shenzhen have formed two separate securities markets. Neither the Shanghai Composite Index nor the Shenzhen Component Index can fully reflect the trend of the entire A-share market. However, investors urgently need a unified index that can truly reflect the changes in the securities market as a direction for investment.
On April 8, 2005, the securities markets in Shanghai and Shenzhen officially released the CSI300 Index to the market, based on the 1000 points on December 31, 2004. As an indicator that can effectively reflect the market state, the CSI300 Index will gradually become the forward-looking indicator for investors. The select criterion was based on large-scale and highly liquid stocks. The average daily transaction of sample stocks was ranked from high to low in the most recent year by excluding the bottom 50% of the stocks, ranking the remaining stocks according to the daily average market value from
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
high to low, and selecting the top 300 stocks as sample stocks. In principle, the index constituents are adjusted semi-annually, generally in early January and early July, and the adjustment will not exceed 10%. Nevertheless, a company making lose money in the last financial report will not be selected to the new sample.
If the sample company withdraws from the stock market, it will be excluded from the index sample and replaced by another stock, which was the highest among the past candidate samples. In addition, the latest rules are adjusted that the newly issued stocks, which meet the sample conditions and whose total market capitalization ranks in the top 10, will be quickly selected to the index and enter the CSI300 after the end of the tenth trading day. Meanwhile, the ranking will exclude the stocks in the original sample list that ranked last in the most recent year.
The next chapter will analyze the statistical data from before and after implementing the circuit breaker mechanism; then, it will explain why we selected the DID model.
The regression analysis is conducted through the DID model in Chapter 5.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
CHAPTER 4 Data and Methodology
Chapter 4 integrates and analyzes the data group. Also, it introduces the details of the DID model architecture and the regression formula. Through the analysis, it provides a better understanding of the regression results in Chapter 5.
4.1 Data
Tables 4.1.1, 4.1.2, and 4.1.3 explain the terms of regression data. This paper examines the key indicators of the circuit breaker mechanism as changes in stock prices and volume. Therefore, the intraday stock price index, the transaction amount, and the trading volume are selected. They were used to measure the effectiveness of the circuit breaker mechanism on the stock market, to encourage a more accurate and comprehensive analysis of the circuit breaker mechanism.
Table 4.1.1 The control variables used in regression
Referring to the empirical method of the past literature, the following control variables were selected: total assets, leverage, profits, and earnings ratio.
Table 4.1.2 The price variables used in regression
Variable Meaning Definition
log_tas Total assets Log(Total assets)
lev Leverage Total debt /Total assets
prof Profit Net profit/Total assets
Gain Earnings ratio Stock price/EPS
Variable Meaning Definition
Clsprc Close price Daily close price
Opnprc Open price Daily open price
Loprc Lowest price Daily lowest price
Hiprc Highest price Daily highest price
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
To determine the mechanism’s effect on stock price, the close price, open price, lowest price, and highest price are selected as the outcome variables. The changes in the long-term analysis and short-long-term analysis are observed before and after implementing the circuit breaker mechanism.
Table 4.1.3 The volume and value variables used in regression
In addition to the outcome variables of price, we also wanted to obtain the difference between the values for the long-term analysis and the short-term analysis of the daily volume and value of stock traded, we also selected daily trading volume and daily trading value as outcome variables.
Because the circuit breaker mechanism is based on the CSI300 Index, the constituent of the CSI300 Index is selected as the treatment group. Due to this treatment group selection, it is necessary to select the data of the approximate treatment group as the control group. We choose the relatively large capital in the Shanghai and Shenzhen stock market to be the control group. The control group will be selected from the top 300 listed companies in the Shanghai and Shenzhen stock market by eliminating the stocks in the treatment group.
Table 4.1.4 The time period of the long-term and short-term groups
Variable Meaning Definition
Dnshrtrd Daily trading volume Dnvaltrd Daily trading value
Long-term data Short-term data
Pre June 2015–December, 2015 November 20, 2015–December 3, 2015 Post January 2016–March, 2016 January 4, 2016–January 7, 2016
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
When analyzing the stock price index in the session, we use long-term analysis and short-term analysis to measure the indicator across the two sets of data. The long-term analysis group selected the time range from June 1, 2015, to March 31, 2016, which is mainly due to the stock market disaster in China's A-share market since June 2015. At that time, the Shanghai Composite Index fell from 5,178 points to 2,850 points in only two months. Subsequently, the Chinese government announced in September 2015 that it intends to follow the foreign -recommended circuit breaker mechanism to stabilize the stock market. In January 2016, the follow-up resolution officially implemented the fuse mechanism. Therefore, the time period in the long-term group will not consider the time before the stock market crash in June 2015, but it will consider the long-term research scope for the three months after the stock market crash and the fuse mechanism.
The short-term data come from the 10 trading days before the China Financial Futures Exchange issued the “Related Regulations on Index Circuit Breaker” because it is speculated that its release psychologically affects the market. Therefore, as a sample for data analysis, the possible bias must be eliminated. The time point before December 4, 2015, was selected as the pre-period of the short-term analysis. Furthermore, the four trading days after implementing the mechanism were selected as the short-term observation group. Because the circuit breaker mechanism was removed four days after it was implemented, the amount of reference data is insufficient. Therefore, for the short-term group, we chose to use the data period that included the week before the mechanism as the control group.
‧
Table 4.1.5 Long-term statistics
Table 4.1.5 illustrates the statistics of outcome variables in the long-term analysis. The
“time-dummy zero” value means the period before implementing the circuit breaker mechanism, while “time-dummy one” means the period after implementing mechanism.
Also, the “treatment zero” indicates the control group, which does not implement the circuit breaker mechanism, and “treatment one” indicates the treatment group, which is implements the mechanism. When the circuit breaker was implemented, the daily trading volume of the components or non-components of CSI300 stocks fell by 50%
and 40%, respectively. The stock prices also generally fell after implementing the circuit breaker mechanism. This showed that both the components and non-components of the CSI300 stocks were affected by the circuit breaker. The standard deviation in the post period is generally larger than in the per period, indicating that the fluctuations between the shares after the policy is relatively large after the policy.
Variable Mean Std. Dev
‧
Table 4.1.6 Short-term statistics
The statistics of the outcome variables in the short-term analysis are listed in table 4.1.6.
Both the statistic in the long-term and short-term analysis achieved similar results. After implementing the circuit breaker mechanism, the mean of the stock prices and trading volume fell, and the mean of the stock price and trading volume in the experimental analysis were also higher than those of the control group. Further regression analysis is included in Chapter5.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
4.2 Methodology
The statistical analysis method will choose to adopt the difference in differences (DID) model. When some interventions such as the implementation of the fuse mechanism cannot study the effectiveness through random assignment, DID is a method of pseudo-experimental design. By using the DID model, we can find a group of people who have not been affected by intervention to use as a control group and consider the group’s trend as a benchmark to see if there is a significant difference in the trend of the intervention group. This benchmark is used to analyze data from two time points. The dynamic perspective can be used to present changes that occur after the implementation of the policy and to reduce or eliminate the impact of unobservable characteristics in the empirical results. As Smith and Todd (2015) explained, “by analyzing the policy effects of the time before and after the implementation, the effects of time-invariant unobservable characteristics are eliminated.”
Most of the other matching methods in the common model are one-to-one matching or one-to-several matching. Moreover, most of the methods are based on the tendency to divide and find one or two control group samples closest to the treatment group samples as paired samples. According to their pairing design, an experimental group sample may find one or two similar control group samples, which are included in the paired sample, but do not control the distance between the treatment group and the control group. Therefore, this pairing is not very efficient in practical applications. The DID model also weights several paired samples within a certain range according to the tendency to divide, for use as a counterfactual sample and uses this process to provide information on the distance between the experimental group and the control group.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The concept of “interaction” looks at whether the trends in the two groups are different.
We can construct the counterfactual trend in outcomes of the treatment group and in the absence of treatment using the trend in outcomes of the control group. Then, we can use the divergent trends in the outcome of treatment group to represent the causal effect of treatment. We are most concerned with whether the coefficients of the interaction term have reached statistically significant differences.
In the DID method, we can use the divergent trend in the outcome of the treatment group to represent the causal effect of treatment. Also, we can construct the counterfactual volatility of the circuit breaker mechanism in the control group and the treatment group.
The following regression equation (1) is constructed by the DID model:
Yit = 𝜇 + 𝛽χi + 𝛾𝐷i + 𝛿𝑃𝑂𝑆𝑇t + 𝛼 (𝐷i · 𝑃𝑂𝑆𝑇t) + 𝜀s t. (1)
Let Yt denote the stock price, average trading volume, and market value after the policy exemption at time t. 𝛽χ is a vector, representing a set of control variables in the model and expressed as 𝛽1*χi1+ 𝛽2*χi2+ 𝛽3*χi3…. . Di is a dummy variable indicating the treatment group (treatment). POSTt is a dummy variable indicating post-treatment period (timedummy). γ captures differences across groups that are constant over time.
δ captures differences over time that are common to all groups. αis the coefficient of interest (the causal effect of treatment). If the analysis model does not control the Di, there will be a large gap between the treatment group and control group.
‧
CHAPTER 5 Regression analysis
The Chapter 5 selects the price, the daily number and value of the stocks traded in as the outcome variables. Through the DID model, we got the regressions in the long-term and short-term analysis.
5.1 Long-term analysis
Tables 5.1.1 and 5.1.2 explain the coefficient of regression in the long-term group. The following tables analyze the price, volume and value of the stocks.
Table 5.1.1 Long-term regression result of stock price
Note. The values in parentheses are t-value.
*** and ** indicate significant level under one percent and five percent, respectively.
Variables Clsprc
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The regression results of the stock price indicators in long-term trading are listed in Table 5.1.1. There are four models in the long-term analysis. In the table, the treatment coefficient in table 5.1.1 can be found to have significant regression results and the coefficients are all positive, meaning that the stock price in the treatment group is higher than the control group after considering the impacts from other variables. Then, notice that the time-dummy coefficient is negative in the long-term group, meaning that the price index in the post-period will be lower than that in the pre-period, including the close price, open price, lowest price, and highest price.
The interaction results of the stock price index in the long-term group all have significant regression results. When the inter value is negative it means that, after the circuit breaker mechanism, the treatment group will decrease in price index compared to the control group. There has been a significant effect on the treatment group. After the circuit breaker mechanism is implemented, while controlling other factors, the policy indeed reduces the day trading stock price of the A-share constituents, including the opening price, the closing price, the highest price, and the low price. In addition, the log (total assets), leverage, profit, and earnings ratio are used as control variables.
In the long-term, a higher total assets and leverage will make the stock price fall; a higher rate of return and earnings ratio will make the stock price rise.
In conclusion, we can clearly find that the circuit breaker mechanism for CSI300 has indeed led to the day trading stock price index decline relative to the other A shares that have not implemented the policy. From the regression results, this mechanism may exacerbate the instability of the stock market. Referring to the research literature of Han Ao Shuang (2018), considering the empirical results that reveal the sell order was withdrawn, the implementation of the circuit breaker mechanism has aggravated the
‧
degree of unbalance of the place orders. Shuang also believes that the policy has increased the instability of the stock market.
Table 5.1.2 Long-term regression result of the daily stock market
Note. The values in parentheses are t- value.
*** and ** indicate significant level under one percent and five percent, respectively.
Table 5.1.2, illustrates the circulation status of the daily stock market, and we have four models in the long-term analysis. We also use the DID statistical regression method to perform the analysis and research. The treatment coefficient in Table 5.1.2 may be positive, which means that the coefficient of the daily stocks traded and their daily trade value in the treatment group is higher than the control group after considering impacts from other variables.
Variables Dnshrtrd
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Moreover, one may notice that the time-dummy coefficient has a negative number in the long-term group. The posting period will decrease the daily number and value of stocks traded from the pre-period.
It is important to note that the different inter-coefficients of the explanatory variables have had different effects. In the long-term group, only the daily number and value of stocks traded have a significant effect when the circuit breaker mechanism is implemented. First, based on the data of model (Ltv1), when the inter-coefficient is negative it means that, after the policy is implemented, the treatment group will decrease the daily number of stocks traded compared to the control group. This significantly affects the treatment group. Second, in the data of model (Ltv2), the inter-coefficient is the same as the negative value. It also means that the treatment group will decrease the daily value of stocks traded after the policy when compared to the control group.
Furthermore, the log (total assets), leverage, profit, and earnings ratio are used as control variables. In the long-term, the higher leverage and earnings ratio will make the number of daily stocks traded and the daily trade value fall; the higher total assets and rate of return will make the number of daily stocks traded and the daily trade value rise.
In the long run, it has indeed reduced the daily number and value of stocks traded in CSI300 constituents by controlling other factors after implementing the policy. This means that the circuit breaker mechanism weakens the coefficient of trading shares and market liquidity during the day. The results of the long-term group showed that the liquidity index has indeed decreased and did not exacerbate the investors’ behavior of disregarding the cost while selling. Subsequent selection will include short-term data
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
groups to further analyze the regression study of the daily circulation status under the circuit breaker mechanism.
The reference literature, Han Ao Shuang (2018), shows that "the fuse mechanism has aggravated the imbalance of the sell order, but at the same time, the rate of withdrawal is also significantly improved. This reflects the possibility that orders remain in market manipulation." Considering the fact that the order remains in control, Shuang continues to analyze the impact of the fuse mechanism on the sell order withdrawal ratio and to find that the circuit breaker mechanism does increase the withdrawal rate.
Then, according to the regression model in (Ltv1) and (Ltv2) in table 5.1.2, the daily number and value of stocks traded all fell after the implementation of the fuse mechanism. Compared to the withdrawal mentioned in the reference, we can reach the same conclusion. As the rate of withdrawal is increased, the number of shares traded in the market will be reduced. Furthermore, Shuang also indicated that the liquidity indicators improved through the regression analysis and believed that the behavior of the investor to disregard the cost was not aggravated.
‧
Tables 5.2.1 and 5.2.2 explain the coefficient of regression in the short-term group. The following tables analyze the price, volume and value of the stocks.
Table 5.2.1 Short-term regression result of stock price
Variables Clsprc
Note. The values in parentheses are t- value.
*** and ** indicate significant level under one percent and five percent, respectively.
Table 5.2.1 shows the regression results of the stock price indicators under the circuit breaker mechanism in short-term trading, and we obtain four models in the short-term analysis. The biggest difference is that the treatment coefficient in model (Stp1), model
Table 5.2.1 shows the regression results of the stock price indicators under the circuit breaker mechanism in short-term trading, and we obtain four models in the short-term analysis. The biggest difference is that the treatment coefficient in model (Stp1), model