THE EFFECT OF RUMOR CLARIFICATION ON CHINESE STOCK MARKETS Jun Wang, School of Economics Information Engineering, Southwestern University of Finance and Economics, Chengdu, P.R. China, [email protected] Yan Chen, School of Economics Information Engineering, Southwestern University of Finance and Economics, Chengdu, P.R. China, [email protected] Yang Tang, School of Economics Information Engineering, Southwestern University of Finance and Economics, Chengdu, P.R. China, [email protected] Qing Li, School of Economics Information Engineering, Southwestern University of Finance and Economics, Chengdu, P.R. China, [email protected]

Abstract Rumor and stock markets are coexisted. Most previous studies have focused on the effect of rumor on stock movements. However, few investigated the reactions of investors when a rumor is verified by authorities. Since Chinese Securities Regulatory Commission requires the listed firms to clarify the relevant rumors in time, it is of great interests to understand the effect of rumor clarification on Chinese stock markets. In this study, we first obtained the information of rumor clarification which is announced by rumor confirmation column in the www.p5w.net, one of the largest rumor clarification websites in China, using text mining techniques, and then study its direct impact on stock movements. Our main findings are that the reactions of investors are strongly related with the types of rumors and the responses of the relevant listed firms. In particular, the positive or negative response to good rumors tends to further raise up or drag down the stock price, while either the positive or negative response to bad rumors is not able to stop the downward trend of stock prices. In addition, the stock market is quite sensitive to the rumors of restructuring, mergers and acquisitions, which leads to the violent fluctuation of stock prices. Keywords: Rumor Clarification, Stock Movements, Stock Market, Stock News



Corresponding Author

1

INTRODUCTION

Wherever are stock markets, there are rumors. Rumor is also referred as hearsay, and believed as a kind of paradoxical organization information which is not confirmed by listed firms and characterized by high distortion, fast transmission, high false reliability, rapid feedback, high selectivity, and strong purpose (Buckner 1965). The rumor in stock markets includes grapevine information, analysts’ subjective recommendations, and even news (Bommel 2003). Similar to tumors in human bodies, rumors have never been cut off or cured in capital markets, which bring massive troubles to investors and market regulators. For example, there is a rumor tweet about wounded Obama on April 23rd in the year 2013. It says that the White House had been witnessed two explosions and the president of America Barack Obama was injured. This short twitter message dragged S&T stock index down sharply. Three minutes after its releasing, the Dow Jones Industrial Average fells down 145 points and the market value of U.S. stocks evaporated $200 billion. Such negative effects of rumors are more critical in immature capital markets like china which lacks of efficient and powerful regulatory rules. With the technological advancement fertilizing vibrant creation, sharing, and collaboration among Web users, the impact of rumors on stock markets has been increasingly prominent. Specifically, traditional news has evolved into various forms of social media including blogs, tweets/micro-blogs, discussion boards, and social news. With such broad communication channels, investors can rapidly reach more information including rumors. Meanwhile, the adaption of user engagement in social media effectively magnifies the information in the news via comments, votes, and so forth. With such rapid information influx, rumor makers can easily distribute their false information. In fact, the number of rumors has been increased as the application of social media in Chinese market. A series of vital rumors has been witnessed in Chinese markets. On April 2015, two rumors, one about the merger of the two largest oil companies and the other about merging all Chinese central enterprises to 40 syndicates, caused Sinopec, China Petroleum and other 16 relevant listed stocks raise up crazily. On May 2015, the merger rumor of FAW Group and Dongfeng Group lead to an abnormal stock increase. Right after that, a rumor about the increase of stamp duty made Shanghai index fell down 4%. Essentially, the herd behavior caused by the self-publication and peer-to-peer spreading greatly enhances the impact of rumor on stock markets. It is of the great necessity to study the effect of rumor in Chinese stock markets, especially with the widespread of social media. Most previous studies have focused on the effect of media, especially rumors on stock movements and showed the its significant impact (Li et al., 2016, Li et al, 2014a, Li et al, 2014b, Clarkson et al., 2006; Spiegel et al., 2010). However, few investigated the reactions of investors when a rumor is verified by authorities. Since Chinese Securities Regulatory Commission requires the listed firms to clarify the relevant rumors in time, it is of great interests to understand the effect of rumor clarification on Chinese stock markets. For instance, will the clarification information of rumors can further affect the stock movements? Whether the positive or negative verifying cause different shocks on stock markets? Whether the types of rumors affect the impact of rumor clarification? The research on these questions will help us to understand the effectiveness of rumor clarification and provide an important guidance on rumor clarification for authorities.

2

RELATED WORK

Previous studies have been studied on the direct impact of rumor releasing on stock markets without considering the consequence of rumor clarification. The earliest study on the effect of rumor on stock markets can be traced back to the work of Rose in 1951 in which she theoretically described the mechanism of rumor and concluded that rumors are able to cause the changes of stock price but it is hard for investors to get excess returns (Rose 1951). Later on, Diefenbach (1972) showed that Wall Street rumors do not have a significant impact on U.S. stock markets, and rumor-based investment

decisions cannot make significant abnormal returns. According to the efficient market hypothesis (EMH) (Fama 1969), this implied that U.S. stock market is a strong efficient market. However, recent studies show that rumor does affect the stock movements. Peter and Michael (1978) found that there is a significant impact of rumors from the column of Heard on the Street (HOTS) in Wall Street news on stock movements. In their study, it showed that positive rumors bring significant positive abnormal daily returns, and negative rumors can make significant negative abnormal daily returns. This finding is also confirmed by the late studies (Syed et al., 1989; Liu et al., 1990; Barber and Loeffler, 1993). In particular, Syed et al. (1989) studied the rumors from HOTS and divided stocks into study group (rumor involved) and control group (rumor uninvolved). It is observed that stocks involved with rumors showed significant abnormal returns right before and on the rumor releasing day. Liu et al. (1990) studied rumor impact on stocks in terms of trading volume and stock price by analyzing two types of rumors, i.e., one is related with a single listed firm and the other is associated with multiple firms. He found that rumors targeted on a single firm have stronger influence on stocks than ones with multiple companies. Barber and Loeffler (1993) studied the stocks recommended by HOTS, and found that these stocks have obvious abnormal returns on the day of rumor release but do not obtain longterm abnormal returns. Pound and Zeckhauser (1990) took a further by mining the impact of rumors on stocks in terms of rumor content, and found the rumors related with mergers and acquisitions cause fluctuations in stock markets though investors cannot obtain excess returns. Instead of studying the rumors from Wall Street News, researchers have been studied a broader rumor channels. Palmon et al. (1994) investigated the rumors from the IWS column of Business Week, and found that the good rumors affect stock trading volume significantly while bad rumors do not have. The average abnormal return rates are significant for both bad and good rumors on the day and the next day of rumor releasing. Sarkar and Jordan (2000) carried out an empirical study on the securities recommended by the five regional Wall Street journal publications, and found that a rumor has a significant impact on the relevant stock on the day of releasing. Besides the mature U.S. Markets, the impact of rumors on stocks has also been observed in the emerging economies. Berument and Kiymaz (2001), Kiymaz (2001) applied the event study method to analyze the rumor effect on the Istanbul Stock Exchange, and found that the rumor-involved firms have significant positive abnormal return four days before the rumor releasing. It demonstrated that stock investors can make profit from rumors such as executing arbitrage before rumor and sell stocks on its releasing. In other words, published rumors are useless to investors. In addition, different types of rumors have distinct impact on stocks. Rumors related with earnings expectations and the foreign acquisition are expected to have strong impact. Similar findings are discovered in Chinese stock markets. Liu and Li (2003) found that rumor-makers can obtain abnormal returns by taking the information disclose 2 or 3 days before the rumor releasing. However, it is hard for such speculators to earn extra returns in a long term. The aforementioned studies focus on the rumors from news or journal columns which are a small portion of rumors circulating on the financial markets. Some researchers have been studied rumor effect from a broader channel. For example, Dan et al. (2009) collected rumors from three major business magazines (Business week, Forbes, Fortune) published between 2000 and 2003, and explore the impact of rumor on stock markets in terms of releasing time, rumor content and writing style. It was also found that the rumors related with management and acquisition have significant impact on stock movements. Zhao et al. (2010) investigated the rumor news in Chinese stock markets, and found most of them are good news which have a strong impact on Chinese stock movements. Most previous studies have focused on the effect of rumor on stock movements. However, few investigated the reactions of investors when a rumor is verified by authorities. Huberman and Regev (2001) found that the price of listed firm involved with bad rumors is hard to recover to represent its real market value even though the rumor is clarified. Marshall et al. (2009) presented that stock price can be rallied within 5 days after clarifying the relevant rumors. Zhao et al. (2010) also found that stock prices are hard to restored to its previous level after the listed firms issued the clarification announcement. However, rumor clarification brings high attentions and trading volume increase.

However, these studies only focuses on the official clarifications on critical rumors which has limited samples. In this study, we utilized the official and unofficial clarifications from www.p5w.net, one of the largest rumor clarification websites in China, to study its impact on stock movements. We also interested in the distinct impact caused by rumor content.

3

RESEARCH HYPOTHESES

In this study, we are of particular interests on the effectiveness of rumor classification in Chinese markets. In other words, we aim to verify whether rumor classification has a significant impact on the fluctuations of stock prices. In particular, we investigate such impact according to rumor content and clarification type. As shown in Figure 1, we have four hypotheses for this study as follows: First of all, intuitively, the positive reaction to good rumors encourages investors and raises up the prices of relevant stocks, while the negative response to good rumors adds downward pressure on stocks. Therefore, our first hypothesis is Hypothesis H1a: Confirmed responses of good rumors further raise up stock prices, while the negative responses make stock prices down. Second, bad rumors depress investors psychologically, and investors’ expectations on future stock prices keep low. Even positive responses to bad rumors are hard to cheer up considerate investors. Our second hypothesis is as Hypothesis H1b: Both positive and negative responses to bad rumors cannot stop prices falling. Furthermore, we investigate the effect of different rumor types. Based on previous studies, regardless of positive or negative clarification, good rumors tend to raise up stock prices, and bad rumors always have downward pressure on stocks. Here, our hypotheses come as Hypothesis H2a: Regardless of positive or negative responses, relevant contents of good rumors tend to bring upward pressure for relevant stocks. Hypothesis H2b: Regardless of positive or negative responses, relevant contents of bad rumors tend to bring downward pressure for relevant stocks.

Figure 1.

Research frame for our study

4

METHOD

4.1

Experimental Data

In this study, we utilized the official and unofficial clarifications from www.p5w.net website, one of the largest rumor clarification websites in China, to study its impact on stock movements. To achieve this goal, we applied a focused crawler to collect rumors and its clarification information from Panorama Network website. We have collected 11485 rumor-clarification pairs released between December 31, 2013 to June 30, 2015. To remove the noise information, we selected rumorclarification pairs according to the follow criteria: 

Delete pairs associated with the stocks with *ST and ST tags, because these special treatment stocks have high probability to be removed from stock markets.



Delete pairs associated with the stocks whose name was changed within our investigating period for consistence.



Delete pairs associate with stocks lack of transaction records within our investigating period.

Finally, we obtained 6459 rumor-clarification pairs involved with1834 listed firms. We obtained the transaction records, such as stock return, volatility, trading volume, and market value, of these listed firms from CSMAR1 Financial Research Database. 4.2

Research Method

The rumors can only affect the stock price for a short period, for investors with long-term goal, it is almost no effect. Therefore, this study applied event study method to investigate the rumor clarification on stock movements for the short period (Brown, 1980). Essentially, an event study is a statistical method to assess the impact of an event on the value of a firm, cycle time of event study method include estimated period and event period (Figure 2).

Figure 2.

cycle time of event study method

Here, the event date is defined as the date for rumor clarification. To estimate the expected rate of stock return (Rit ) , we adopted the classic market adjustment model (Brown and Warner 1980). It is defined as

Rit   i  i Rmt   it ,

(1)

th where, Rit is the revenue of stock i on the t day during estimated period, Rmt 2 is market profitability th on the t day during estimated period,  it is zero mean distribution,  i is a constant term,  i is a regression coefficient, t is the day calculated during estimated period, t3 = -130, ..., -22.

1

website: http://www.gtarsc.com/Home

The abnormal returns ( AR ) of stock i on the t th day (ARit ) during event period is defined as

ARit  Rit  ( i  i Rmt ) ,

(2)

where t is the day measured relative to the event, t 4 = -10, -9, ..., -1, 0, 1, ..., 9, 10(the rumor clarification date is set to be t = 0), Rit is market profitability on the t th day during event period,

( i  i Rmt ) is the prospective earnings on the t th day during event period. Average ARs for trading day within the event window are calculated by

ARt 

1 N  ARit , N j 1

(3)

th where ARt is average abnormal return on t day, N is the number of stocks with ARs during day t,

t  [10,10] .The cumulative average abnormal returns (CART1 ,T2 ) between T1 and T2 , are formed by summing average ARs over time as follows: T2

CART1 ,T2   ARt .

(4)

t T1

In order to test the significance of ARt and CART1 ,T2 , the t-test statistic is applied to examine whether the average abnormal return and the cumulative average abnormal returns are significantly convinced. T (t , t ) and T (T1 , T2 ) 5are t statistic for ARt and CART1 ,T2 and defined as

T (t , t )  SARt  N ,

(5)

T (T1 , T2 )  SCAR  N ,

where, SARt 

N

ARit 1 1 , SCAR   N N i 1 Si

N

T2

 i 1 t T1

ARit Si T2  T1  1

(6)

, S i is the residual standard deviation of

stock i obtained from the regression of the market model.

2

Rate of returns on stocks and market during estimated period consider cash dividends reinvested on daily stocks returns and market returns.

3

Estimated period refers to the event has not yet occurred, it is a relatively normal situation for a period time before or after event period. Estimated period should long enough than event period, Boehmer E. et al. (1991) selected t ＝ - 249 ～ -11; Cowan, A.R. (1992) used t ＝ - 225 ～ - 1 as estimated period. This research select t ＝ 130 ～ - 22 as estimated period, which based on previous reference. 4

Event day is the first time, for market, to achieve some information, it is not the time when the event happens. Event period is a period time before or after event happen, which include event day. This paper refers to previous research, event period selects t  [-10,10] (Binder, J., 1998; Mitchell, M.L. and E. Stafford, 2000; Andrade, G et al., 2001). 5 T (t , t ) is selected period time of average abnormal returns during event period; T (T1 , T2 ) is selected period time of cumulative average abnormal returns during event period.

5

RESULT OF RESEARCH

5.1

Statistics for Rumor-Clarification Pairs

Table 1 shows the descriptive statistics of the sample in our study. There are several interesting findings from these descriptive statistics. In particular, 

Rumors can be divided into good, bad and neutral, good rumors take a great portion up to 88.5% of all the rumors in markets. The distribution of rumors is lopsided. This might because that Chinese stock markets lack of shorting mechanism, which makes malicious whisperers to be difficult to earn profit from bad rumors, and market speculators only use good rumors to raise stock prices and obtain abnormal returns.



87% rumors and clarifications are associated with Shenzhen stock exchange (SZSE) which is characterized with the firms of small and medium size, and the rest is with Shanghai stock exchange (SSE). In other words, rumor speculators favor small capital corporations, this might be that the market funds tend to speculating in the firms of small and medium size, in which whisperers are easier to obtain excess returns. For the market capitalization of large stocks, more money is required for raising up share prices, which is not conducive to the operation of funds and short-term arbitrage.



Rumor clarifications can be categorized into three group, i.e., positive, negative, and no answer. Negative clarification is about 58%, which means most rumors carry false information.



We divide rumors into three categorizations in terms of content, that is, restructuring, mergers and acquisitions (M&A), and production investment (PI). Therefore, we study the clarification impact of different rumor content. Rumor speculators prefer to take the advantage of rumors related with mergers and acquisitions, and this type of rumors is up to 47.8%. Rumors

Stock Exchange

Clarification

Rumor Content

Good

Bad

SZSE

SSE

Positive

5713

698

5618

841

1989

3766

704

231

30.8%

58.3%

10.9%

19.0%

88.5% 10.8% 87.0% 13.0%

Table 1. 5.2

Negative No Answer Restructuring M&A 581

PI 404

47.8 % 33.2%

Total 6459 100.0%

Descriptive statistics for rumor-clarification pairs Impact of Clarification in terms of Good and Bad Rumors

Table 2 shows the average abnormal returns and cumulative average abnormal returns of the 10 trading days before and after the announcement date of information disclosure. It can be observed that positive clarification to good rumors do not shake the prices up on the clarification day, but it brings significant excess returns in the next day ( AR t 1 ) = 0.11% (p