Does Overreaction exist?
An Empirical Research of the Chinese
Stock Market
University of Amsterdam Amsterdam business school MSc Business Economic: Finance track
Master Thesis By JUN LU
Student number: 10897496 Supervisor: Zou Liang
2
Statement of Originality
This document has been written by Jun Lu (student number 10897496), who takes full responsibility for the content 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.
3
Abstract
The paper is aiming to examine the overreaction effect in China’s Stock Market. By taking data from CSMAR database from 1997 to 2014, the study conducted a market model and using the reversal performance as the indicator to test the overreaction hypothesis. By subtracting average cumulative abnormal return of winners portfolio from losers portfolio, the study did not find significant abnormal return. As a result, we concluded that overreaction effect does not exist in China Stock Market.
4
Table of Content
Introduction ... 5
Literature Review ... 8
Data and Methodology ... 15
Empirical Results ... 21
Conclusion ... 27
References ... 29
5
Introduction
Decades ago, the efficient market hypothesis (EMH) was widely accepted by financial economists. It was generally believed that the stock market is extremely efficient in reflecting the information of individual stocks (Fama 1970). When news breaks, the information spreads very fast and is incorporated into stock prices immediately. The EMH is related to an idea, namely random walk. The explanation of random walk is that if information can flow freely in the market, then the change of price must be random and unpredictable (Malkiel 2003). Therefore, EMH holds that information is already reflected in prices and neither technical analysis nor fundamental analysis would enable investors to generate abnormal returns.
However, since the 1980’s, the emergence of contradictory evidence of so-called market anomalies has resulted in critiques of EMH. Plenty of studies on anomalies carried out suggest that the price movement is at least partially predictable in the stock market. Market anomalies demonstrated a price behavior which created an opportunity of abnormal returns. One anomaly that appears to offer this opportunity is called Overreaction Effect. The logic of Overreaction Effect is that investors tend to give excessive weight to new information when conducting their portfolios, which results in an inappropriate effect on price. Consequently, stock prices tend to move drastically following new events and temporarily diverge from their true value. However, the price swing from overreaction does not last long. Once investors
6
realize that the stock is overpriced or underpriced, stock price will eventually move back to its true value. With regards to the price movement, abnormal returns can be obtained by taking a long position of stocks that have witnessed the lowest returns (loser), and meanwhile selling short stocks that have witnessed the highest returns (winner).
As is pointed out by Power and Lonie (1993), Overreaction Effect should be considered as one of the most important anomalies for study. They listed three reasons to support their point of view. First, the abnormal return earned by using contrarian price movement is much more than other anomalies. Second, Overreaction Effect has a more intuitive appeal, since it can be easily explained by the investigating mean-reverting patterns in stock returns. Third, the overreaction hypothesis is supported by cognitive psychology, which has proven that economic agents are likely to exaggerate unanticipated information (Kahneman and Tversky 1977). The overreaction hypothesis has been discussed for decades, mostly by using data from developed countries. As one of the most famous anomalies, studies of Overreaction Effect on emerging markets started in recent years. As is pointed out by Griffin (Griffin et al 2010), abnormal returns earned in emerging countries are significantly larger than developed markets1 because of less efficiency. A different explanation for the greater return in developing markets is proposed by Vadar and Okan (2008). They mentioned that trading barriers and underdeveloped communication technology may also cause stronger reversals.
1
Note that the finding is prior to transaction cost. In fact, the contrarian profit is smaller in emerging countries after considering transaction cost.
7
China’s Stock Market was established in 1990 and is formed by the Shanghai Stock Exchange and Shenzhen Stock Exchange. From the initial 10 firms to currently more than 2500 firms, it has become the biggest equity market in Asia and a member of the top 5 markets worldwide in trading volume. Can overreaction be found in China stock market? Answering the question without any empirical data is difficult. On one hand, we have to acknowledge that the Chinese stock market is still young. Serious of problems such as insider-trading and price manipulating could haunt investors and trigger panic, hence increasing the possibility of overreaction. On the other hand, the Chinese financial market is highly regulated and managerial tools such as price limitation are imposed2. Price limitation requires that stock trading must be halted after price reaches the floor or ceiling. As a result, investors have more chances to review information and be more rational to respond to shocks. In other words, price limitation policy prevents investors from overreaction. Based on the description above, Chinese investor overreacting is confusing, and the question motives me to explore the issue. The purpose of this thesis is therefore: finding the empirical evidence of the overreaction phenomenon in China’s stock market. By using the monthly return data of the CSMAR database, I first attempt build up winner and loser portfolios. Then, I will investigate if loser groups outperform winner groups in different time frames to seek evidence of overreaction.
2
Price limitation indicates that stock trading will be halted after the stock reaches its maximum or minimum. In China, the allowable floating range is 10 percent up or down in one day.
8
The reminder of the paper is organized as follows. The next section will review the relevant literature, which elaborates on the concept, existence and potential disturbing factors of overreaction. Meanwhile, I will point out the research design through comparison of the basis of the review and evaluation of the relevant studies. Section 3 will introduce the data source and data processing. Section 4 will give an empirical analysis of overreaction. In the last section, a brief conclusion will be offered.
Literature Review
Explanations of Overreaction
Barberis (Barberis et al 1998) defines overreaction from a statistical perspective. He assumes that the announcement is made in time Zt and it can be either bad or good, Zt=G (good) or Zt=B (bad). The overreaction appears if the average returns after a series of negative news reports are higher than the average returns after a series of positive news reports. The statistical evidence is described by the formula below (BSV Model):
( | ) J is equal or higher than one. The reason for overreaction is that after a series of positive news reports, investors become over optimistic. They expect that announcements from the company
9
will also be good in the future and therefore overreact. However, the subsequent news of the company may contradict their expectation and results may be low in return.
From a psychological perspective, the motivation of overreaction can be interpreted by overconfidence and attribution bias. First, psychological evidence shows that people are overconfident about their abilities and knowledge (Baruch Fischhoff et al 1977). In general, individuals make decisions based on the most recent and available information but not making use of all useful information. When facing a complex task such as the assessment of stock returns, people appear to instinctively simplify the assessment procedure and ignore the potential risk (Odean, 1988). Therefore, people who actively trade in financial markets are likely to be more overconfident, which causes the stock price to overreact. Moreover, Daniel (Daniel et al 1998) interprets overreaction as not only a result of overconfidence but also a consequence of self-attribution. Self-attribution points out that investors tend to be more confident when their personal information signals are believed to be the same as the public’s information, while confidence does not fall when private information contradicts the public’s information. Briefly, investors appear to applaud themselves for prior success but blamed the failure on back luck. In other words, investors tend to overreact to private information and underreact to public information, which consequently produces short-term momentum and long-term reversal.
10
Fama (1998) criticized the psychological explanation of overreaction. He asserts that events used in psychological study are selective. The psychological theory does not stand up to data, and it can only explain a small fraction of anomalies.
Hong and Stein (1997) conducted a HS model by dividing investors into two groups, namely newswatchers and momentum traders. A newswatcher makes prediction base on signals that they privately observed from markets but do not condition on current or pass price. On the contrary, momentum traders do predict price based on historical price data. The underreaction will present when only newswatcher active in market, because they do not extract information from market. After momentum traders join the market, they arbitrage any underreaction left behind by newswatchers and price will be corrected. It turns out that momentum trader would enhance the market efficiency.
DeBont and Thaler (1985) explain overreaction as a result of belief revision. Specifically, Individuals tend to give extra weight to recent information or undervalue prior data. Consequently, the value they select will violate the Bayer’s rule and therefore causes extreme of prediction. Consequently, overreaction is occurring.
Evidence of Long-Term Overreaction
DeBondt and Thaler (1985) conducted the first systematic study on the issue of overreaction. By taking monthly returns of NYSE stocks from 1926 to 1982, they computed the Cumulative Excess Returns (CAR) for all stocks in 16 non-overlapping formation periods.
11
Portfolios are formed by ranking CAR from low to high. The highest 35 stocks are assigned to the winner portfolio and the lowest 35 stocks are transferred to loser portfolio. The average cumulative abnormal returns of winners and losers in the subsequent 36 months are then tested. They found that, loser portfolios outperform the market by 19.6% on average while winner portfolios earn 5% less than the market. Therefore, a significant abnormal return of 24.6% can be created by short selling the winners and going long of losers. Later, they increased the formation period to 3 and 5 years and repeated the test. The result showed an even stronger reversal performance. There are three main findings concluded: First, prior extreme movements in stock prices will cause price movements in the opposite direction. Second, a greater initial price movement leads to a greater subsequent adjustment. Third, the P/E anomaly can be explained by Overreaction Effect.
In their later study, De Bont and Thaler (1987) reexamined overreaction after controlling for company size, risk and seasonality. They concluded that although losers are usually smaller firms, a company’s size cannot fully explain Overreaction Effect. Further, the reversal of performance cannot be blamed on changes in risk as indicated by CAPM betas, and excess returns in January are negatively correlated to performance in the formation period.
Motived by the evidence that losers are usually small firms, Zarowin (1990) again investigated the issue of size effect in the US market. He conducted two sets of tests to examine how company size influences overreaction. First, he divided each winner and loser into 5 subgroups based on company size and conducted the size-matched Jesen performance
12
test. The result showed that the return divergence disappeared in most of the months during the test period, except in January. Therefore, the return divergence should be attributed to the January effect but not investor overreaction. Moreover, in search of more convincing evidence, Zarowin performed an additional test. The logic of the second test is that if the effect of size is responsible for winner-loser performance, then we should expect losers to outperform winners when losers are small, and winners to outperform losers when winners are small. The test result reveals that reversal performances are a result of size effects, since portfolios with small firms always outperform portfolios with big firms over a 36 month test period.
Contrary to the findings of Zarowin, Chopra (Chopra et al 1992) found evidence consistent with the overreaction hypothesis. After controlling for company size, they presented the significance degree of overreaction around 5% per year. They also confirmed that small firms show a stronger overreaction effect compared to big firms. Ahmad and Hussain (2001) also criticized the notion that reversal performance is a result of size effect. They doubted that the size matched the Jesen performance test adoped by Zarowin (1990), making the firm ranking biased.
The influence of risk is also discussed in various studies, mostly by looking at CAPM-betas. In order to test whether reversal performance disappears after taking market risk into account, De Bondt and Thaler(1987) examined arbitrage portfolios by regressed CAMP-betas respect to the annual return during the testing period. They contend that CAMP-betas cannot explain that the arbitrage profit though the betas in the loser’s portfolio is higher than the
13
winner’s. Chan (1988) computed the CAMP-betas during the formation period and found the loser’s betas increased following a period of abnormal loss, and the winner’s beta decreased after a few years of abnormal earning. The abnormal return is simply the compensation of market risk rather than an evidence of overreaction. Ball and Kothari (1989) found similar results as Chan after allowing betas to change during the testing period.
Evidence of Short-Term Over-Reaction
Unlike most research concentrating on overreaction in long run, Donald and Power (1992) conducted a study of short-team overreaction based on weekly returns in the UK stock market. By using the same model as De Bondt and Thaler, they found evidence of losers outperforming winners over 12 weeks. Wang (2004) tested the short run overreaction effect in both A shares and B shares in the Chinese stock market. He found the reversal performance in A shares are more significant than in B shares. He suggested that the significance level of the test result is highly dependent on which form of analysis is adopted.
Caporale (Caporale et al 2014) confirmed the existence of overreaction in short run first, then examined the abnormal price changes the day after overreaction. They concluded two findings: First, the evidence of counter-movement in short-term can only be found in the US stock market and the oil market. Second, the inertia anomaly is confirmed. Specifically, price tends to moves in the same direction the previous day. They interpret the inertia anomaly as
14
being a result of prices fully incorporating new information or finishing the process of overreaction. Hence, the investment strategy as a chasing-trend would be profitable.
Sabri (Sabri et al 2015) conducted a short-term overreaction test with respect to specific events including terrorist attacks, the formation of governments, tensions in the Middle-East region and the announcement of the privatization of state-owned enterprises (SOEs). By taking data from the EGX market, they confirmed that both terrorist attacks and tensions in the Middle-East region lead to positive and significant abnormal returns within one week following the event window. On the contrary, the formation of new governments and (SOE) does not influence the average abnormal returns post event in EGX.
Besides the above studies, there are plenty of other studies examing the long-term or short-term overreaction phenomenon worldwide [In Spain: Alonso and Rubbio (1990), in France: Dubois and Bacmaan (1998), in Turkey: Bildik and Gulay (2002), in HongKong: Fung (1999), in Japan: Chiao and Hueng (2005)]. To sum up, existing studies concluded three findings. First, by examining the residual returns, the overreaction effect has been confirmed in most stock markets. Three approaches are mostly used in the examination of residual returns: market adjusted excess returns, market model residual and excess returns relative to Sharpe-Lintner CAPM. The observed residuals are not always consistent. The significance of residuals is highly dependent on research approaches and the length of the formation period. Second, there is no consensus on the reason for overreaction. Even though many scholars criticized the methodology adopted by De Bont and Thaler (1995) and assert that observed reversal
15
performance can be driven by various reasons such as risk factors, company size, bid-ask effect and seasonal effects, none of critics have conclude a better methodology to test overreaction. At the meantime, the evidence that they use to attack on reversal performance is not consistent.
Data and Methodology
This paper makes use of monthly price data from China’s stock market, including firms in the Shanghai stock exchange and Shenzhen stock exchange. The research period is from 2007 to 2014. Consistent with other studies, the quantity of stock changes during the research period due to delisting’s and IPO’s. The total number of firms in the beginning of 2007 is 1527 while the number increased to 2564 by the end of 2014, which demonstrated an explosive growth.
An equally weighted index, which is calculated by averaging the return of stocks, is adopted as market returns. The benefit of using an weighted index is that an equal-weighted index is more likely to show a significant abnormal returns when it truly exists, while value-weighted index may cause considerable problems. Specifically, when a methodology such as a control portfolio is used, the null hypothesis can be rejected even when there are no abnormal returns (Brown and Warner, 1980). Moreover, value-weighted index such as SHCOMP gives too much weight to state-owned banks and insurance companies. Therefore, market returns can be easily influenced by financial shocks, further resulting in a biased estimation of abnormal returns.
16
So far, considering that scholars have not yet provided a better methodology of examining overreaction, the thesis will keep using the market model as the indicator to examine the overreaction hypothesis, which is similar to the methodology used by De Bont and Thaler (1985). The detailed statistics are exhibited by the following:
1. Based on the assumption that expected returns of individual stocks are equal to expected returns of the market, the monthly excess return of each stock can be calculated by following the formula:
Uj.t is the excess return of stock j in month t, Rj.t stands for the return of stock j in month t, and Rm.t indicates the market return in month t.
2. According to the monthly excess return, I calculated the cumulative excess returns of each stock in a formation period of 12 months:
∑
The step is repeated 5 times for all non-overlapping 6 month formation periods from January of 1997 up to December of 2011 (January 1997 to December 1997, January 1998 to December 1998, January of 2011 to December 2011). For each formation period, the CUj is ranked from high to low and portfolios are formed. Firms in the top 30 (highest CUj) are distributed to the
17
winner’s portfolio while firms in the lowest 30 (lowest CUj) are distributed to the loser’s portfolio. Therefore, there are 5 winners and 5 losers in total. Following the formation period, the testing period is 36 months (t=-12 to t=0 is formation period, t=1 to t=36 is testing period). There are 5 testing periods in total.
3. Monthly abnormal returns of each portfolio in each month throughout the testing period can be calculated by the formula below:
∑
indicates the specific month in the test period, refers to particular testing period, p denotes either the loser’s portfolio or the winner’s portfolio.
4. Cumulative abnormal returns of all the securities can be computed for portfolios in each 36 month testing period by using the following equation:
∑
5. Based on the CAR for all testing periods, the average cumulative abnormal returns can be computed for each portfolio in each testing month.
18
∑
N represents the total number of testing periods. In this case, N is equal to 5.
6. As the last step, I formed an arbitrage portfolio as the indicator of reversal performance:
Formally, If Overreaction Effect does not exist, then:
If Overreaction Effect exists, then
The null hypothesis above indicates that investors should not earn excess returns by learning from the historical price. If the market is efficient, previous winners and losers cannot generate positive abnormal returns in the subsequent period. The hypotheses H1 ,H2, and H3 state that the past losers should outperform past winners if overreaction exists.
and are examined by a standard t test. Considering the variance of population is unknown, the following statistic is adopted:
√
19
√∑ ( )
To examine the H3, I applied the same statistic as De Bont and Thaler (1985):
√ ⁄
∑ ( ) ∑ ( )
Before implementing the test and analyzing the result, several aspects related to design of this paper are worth to mention. First, the methodology applied in the paper is different from standard event study because there is no specific event appearing (such as earning report). The time frame of the study is composed by formation period and testing period. Considering relative studies have not discussed that in what time frame the reversal performance will occur. To the best of my knowledge, the only suggestion is pointed out by Graham (1959), as the abnormal return should be noticeable after 1.5 to 2.5 years. However, plenty of evidences have showed significant abnormal return can be found in both short run and long run. In the study, the time frame for formation period and testing period is based on learning from other profound overreaction literatures (mainly De Bont and Thaler 1985). The formation period is 12 months and the testing period keeps in 36 months following formation
20
period. Second, during the data processing, numbers of monthly return data of stocks are missing due to the delisting or merger and acquisition over 1997 to 2014. To solve it, the methodology in the paper is that using the stock return in subsequent month replaces the missing data. However, if the return in following month is also missing, I drop the stock out from the sample. The methodology is criticized by scholars as taking out delisting firms from sample may result in a ”Survive bias” because firms who delisted being excluded may potentially categorized as loser and continue to be the worst performer in testing period(Elton et al, 1995). In my defense, taking out of delisting company could cause biased result if data is took from U.S market while it affects little when the study is using data from China’s stock market. Since number of delisting companies in china is rare, which are less than 10 every year i average. While, number of firms being deleted in U.S stock market is about 1000 per year. Third, there is no formal model of overreaction effect. Using abnormal return as the indicator of overreaction effect has been widely accepted by scholars. Although many studies have explained the abnormal return as a result of risk compensation, size effect and many other factors, none of them have provided a well-built mathematic model to support their evidence and there is no consensus yet. Therefore, the study only attempts to examine the existence of overreaction in China’s Stock Market rather than differentiate various psychological and institutional explanations.
21
Empirical Results
Descriptive statistics
The average return and the relative t value of each 5 winners and 5 losers portfolio for 12 month formation period and 36 month testing period from January of 2007 to December of 2014 are displayed in the following table.
TABLE 1
Descriptive statistics of 5 loser portfolios and 5 winner portfolios for 12 months formation period and 36 months
testing period.
Portfolio Formation Period Testing Period Loser Winner Loser Winner
1 Mean t -0.0923 0.0802 0.0014 -0.0359 (-0.12) (0.34) (0.28) (-0.77) 2 Mean t -0.0507 0.0425 -0.0008 -0.0328 (-1.76*) (2.47**) (-0.09) (-0.79) 3 Mean t -0.0348 0.0921 -0.0013 -0.0258 (-0.73) (1.33) (-0.26) (-0.62) 4 Mean t -0.0564 0.0744 -0.0047 -0.0895 (-1.38) (0.95) (-0.85) (-2.01*) 5 Mean t -0.0721 0.0691 -0.0007 -0.1084 (-2.71**) (3.51**) (-0.17) (-2.33**)
Value in brackets () indicates corresponding t-statistics, where
* stands for significance level of 10%; ** stands for significance level of 5%; ***stands for the significance level of 1%
As showed in table 1, the mean return of winners portfolio are all positive while the mean return of corresponding losers portfolio are all negative and statistically significant for four out of ten portfolios in the formation period. Regarding to the testing period, the mean
22
return of all five winner portfolios decreased, while that of losers the mean return raised up. Although scholars such as De Bont and Thaler (1985) pointed out that the reversal performance should be considered as an important indicator of overreaction phenomenon, table 1 only displayed a weak evidence of reversal performance. In portfolio 2 and portfolio 5, return of losers is significantly lower than the return of winners in the formation period, but we can only observe a reversal performance from winners in portfolio 5 in testing period, which shows the mean return drop from 6.91 percent to -10.84 percent with both significance level of 5 percent. Besides portfolio 5, rest of the results in table 1 is insignificant and it is not presenting a reversal performance for losers. Therefore, we cannot conclude that we find the evidence of overreaction effect.
Market Adjusted Abnormal Returns
Table 2 presents the ACARs of all winner portfolios, loser portfolios and arbitrage portfolios as we progressed in the testing period. Figure 1 is demonstrating the movement of all three types of portfolios during 36 months of testing period.
TABLE 2
Average cumulative abnormal return of winner portfolio (ACARW), loser portfolio (ACARL) and arbitrage portfolio along with the relative t value for
month 1 to month 36 in testing period
L 1 0.0338 3.7066** 0.1350 3.5874** -0.1012 -2.6132** 2 -0.0036 -0.1541 -0.0724 -0.7274 0.0688 0.6723 3 0.0056 0.3009 0.0052 0.0592 0.0004 0.0041 4 -0.0085 -0.6481 -0.0233 -0.1982 0.0148 0.1248 5 0.0055 1.0578 0.0321 0.6469 -0.0266 -0.5331
23 6 -0.0053 -0.3464 -0.2384 -1.6765* 0.2331 1.6296* 7 0.0120 0.2728 0.0589 0.4835 -0.0469 -0.3619 8 -0.0137 -0.7578 -0.1693 -1.3927 0.1555 1.2656 9 -0.0020 -0.1451 0.0726 0.6138 -0.0746 -0.6267 10 0.0145 1.1985 -0.0184 -0.1969 0.0330 0.3490 11 -0.0231 -4.2552*** -0.0507 -0.6276 0.0276 0.3412 12 0.0205 1.5851* -0.1294 -0.5990 0.1499 0.6925 13 0.0037 0.1837 0.1631 2.8010** -0.1593 -2.5855** 14 -0.0241 -2.6271** -0.1549 -1.0739 0.1308 0.9049 15 -0.0248 -2.5396** -0.0189 -0.2145 -0.0059 -0.0663 16 0.0121 2.4212** 0.0906 0.7558 -0.0785 -0.6541 17 -0.0004 -0.0321 -0.1015 -1.4317 0.1011 1.4007 18 -0.0200 -1.8476* -0.2996 -2.1164** 0.2796 1.9693** 19 -0.0043 -0.3396 -0.0864 -0.6876 0.0821 0.6499 20 -0.0187 -1.5481* -0.1956 -1.5129* 0.1769 1.3622 21 0.0157 1.5646* -0.0395 -0.7263 0.0552 0.9981 22 0.0275 1.4323 -0.0161 -0.2430 0.0436 0.6335 23 -0.0094 -0.3470 -0.0821 -0.7416 0.0727 0.6377 24 0.0310 3.0981** -0.0561 -0.7147 0.0871 1.1011 25 0.0141 1.2091 0.0969 0.5880 -0.0829 -0.5014 26 -0.0178 -2.3380** -0.2319 -1.3395 0.2142 1.2358 27 -0.0091 -0.6506 0.0394 0.3775 -0.0485 -0.4602 28 0.0026 0.1902 0.1034 0.9292 -0.1008 -0.8991 29 -0.0161 -1.0628 -0.1259 -1.4634 0.1098 1.2573 30 0.0204 1.4158 -0.2415 -5.0001*** 0.2619 5.1962*** 31 -0.0054 -0.3153 -0.0622 -0.4525 0.0568 0.4101 32 -0.0192 -2.0208 -0.2422 -1.5716 0.2229 1.4441* 33 -0.0126 -1.4418 -0.0937 -0.9520 0.0811 0.8209 34 -0.0030 -0.8094 -0.0707 -1.7067* 0.0677 1.6286* 35 -0.0076 -0.4052 -0.1692 -1.6964* 0.1616 1.5919* 36 -0.0148 -1.5748* 0.0870 0.3823 -0.1017 -0.4468 In average -0.0012 -0.0585 0.0573 0.8641
* stands for significance level of 10%; ** stands for significance level of 5%; ***stands for the significance level of 1%
The result showed in Table 2 is inconsistent with the findings of scholars including Debont and Thaler (1985), Alonso and Rubbio (1990) and Dubois and Bacmaan (1998) who found the strong evidence of overreaction effect. As an important evidence of overreaction, they pointed out movements of winners and losers should be divergent over time. Specifically,
24
ACARs of winner would be negative and keep decreasing and ACARs of loser would be positive and keep increasing over the testing period. However, in my study, I did not find the divergent movement. Table 2 shows the ACARs for both winner and loser fluctuates irregularly during the testing period. Regarding to the losers, the ACARs in the first month is about 3.4 percent with significance level of 5 percent. In the subsequent months, ACARs of loser start decrease and drop to -2.31 percent in the month 11, and it finally stops at -1.5 percent with the significance level of 10 percent in the month 36. Concerning to the winners, the ACARs in month 1 is more than 13 percent with significance level of 5 percent. In the following months, ACARs of winner went up to 16.31 percent in month 13, then it starts decreasing and reaches its lowest point of -24.15 percent with significance level of 1 percent in month 30. Finally, it fixed at 8.7% in the last month. As we can see from table 2, neither winners nor losers showed a clear movement of increasing or decreasing. Furthermore, although it is showing that ACARs of losers are positive and ACARS of winners are negative in few months, the result is not stationary and insignificant. Comparing the result to other studies, for example, Vannita and Shalini (2004) who found a significant and consistent upward movement for losers from month 26 to month 36 and a downward movement for winners from month 20 to month 36 in 36 testing months, or Annuar and Taufiq (2009) who found a divergent movement for winners and losers lasted more than 8 months post financial crisis, the result in table 2 is insufficient to support the hypothesis that priori extreme price movement leads to stronger reversal performance in subsequent period (De Bont and Thaler 1985).
25
Moreover, an alternative criterion for overreaction is whether the significant abnormal return can be obtained. As showed in table 2, the abnormal return can only be found in six months including month 6, month 18, month 30, month 32, month 34 and month 35 among 36 testing months. Beside the six months, the arbitrage profit is either significantly negative or positive but insignificant. Moreover, the average arbitrage profit in table 2 is only 5.73 percent and insignificant in the 36 months. The result is different with the findings of Norli and Annuar(2009) who found 3.27 percent short-term abnormal return in first month of testing period and also inconsistent with the findings of De Bont and Thaler (1985) who found a long-term abnormal return of 19.6 percent in 36 testing period. Neither long-long-term nor short-long-term arbitrage profits are found in Chinese stock market.
The ACARs of losers and winners over the testing period are shown in the following graph. As presented by other studies (De Bont and Thaler 1985; Bildik and Gulay 2002; Vannita and Shalini 2004), we would expect a divergent tendency which is an upward slope of the blue line (winner portfolio) and a downward slope of the red line (loser portfolio) in graph if overreaction exists. However, figure 1 shows that the winners and the losers are interweaved but not
divergent. It is clear to see that winners violate more widely while losers is more flat and swing close to 0 over the 36 months. De Bont and Thaler (1985) found the arbitrage profit is
increasing from month 1 to month 36, which is again different as my findings. The ACARs of arbitrage profit fluctuates irregularly in figure 1.
26
Figure 1
ACARs of the losers and winners portfolios of 30 stocks each during three years testing periods
Based on the analysis above, I did find the evidence that losers would outperform winners in testing period. Moreover, the insignificant arbitrage profit tells that contrarian investment strategy does not help in generating abnormal return in Chinese stock market. As a result, the study failed to reject the hypothesis and therefore did not find the evidence of overreaction effect. -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 1 3 5 7 9 11131517192123252729313335 t
ACARS the Losers, Winners and Arbitrage Portfolios during the 5 three-years Testing period
(Between January 1st 2007 to December 31th 2014)
Winner Portfolio Loser Portfolio Arbitrage Portfolio
27
Conclusion
The overreaction effect reported in various countries is not observed in China Stock Market from 2007 to 2014 by using a market model. First, the study did not find the evidence to support the hypothesis that priori extreme movement resulted in price reversal in the subsequent period. On the contrary, the returns of winners exceed the return of losers for few times in testing period. Second, the study only found abnormal returns for six times among the 36 months testing period. On average, the abnormal return is positive but insignificant. Therefore, the study finds neither short-term overreaction nor long-term overreaction.
For future studies, some suggestions are given in follow. First, one limitation of the study is that it did not exam the sensitivity of overreaction to the different length of formation period. The study only uses the formation period of 12 months. De Bont and Thaler (1985) found that the abnormal return obtained in 5 years formation period is significantly more than the abnormal return observed in 2 years formation period. They concluded that overreaction effect is more likely to present along with the increasing of length of formation period. Considering the study did not find the evidence of overreaction effect, I suggest future studies to use different formation periods to test the presence of overreaction. Second, the study did not include the effect of disturbing factors such as risk, size and seasonality. The topic of whether overreaction effect can be explained by disturbing factors is mostly discussed around 90s. In recent years, scholars are more focusing on the examining the overreaction
28
phenomenon is emerging countries and most of them use the market model for the test. However, using of market model will increase the load of work if scholars want to include the effect of disturbing factors. For future studies, adopting alternative mathematical models such Capital Asset Pricing Model or Fama-French 3 factors model may would allow them a simple way to include the effect of disturbing factors when examine the presence of overreaction effect.
29
References
1. Ahmad, Zamri, and Simon Hussain. "KLSE Long Run Overreaction and the Chinese New‐ Year Effect." Journal of business finance & accounting 28.1‐2 (2001): 63-105.
2. Barberis, Nicholas, Andrei Shleifer, and Robert Vishny. "A model of investor sentiment." Journal of financial economics 49.3 (1998): 307-343.
3. Bondt, Werner FM, and Richard Thaler. "Does the stock market overreact?."The Journal of finance 40.3 (1985): 793-805.
4. Boubaker, Sabri, Hisham Farag, and Duc Khuong Nguyen. "Short-term overreaction to specific events: Evidence from an emerging market." Research in International Business and Finance (2014).
5. Caporale, Guglielmo Maria, Luis A. Gil-Alana, and Alex Plastun. "Short-Term Price Overreaction: Identification, Testing, Exploitation." (2014).
6. Chan, K. C. "On the contrarian investment strategy." Journal of Business(1988): 147-163. 7. Chopra, Navin, Josef Lakonishok, and Jay R. Ritter. "Measuring abnormal performance:
do stocks overreact?." Journal of financial Economics 31.2 (1992): 235-268.
8. Bondt, Werner FM, and Richard Thaler. "Does the stock market overreact?." The Journal of finance 40.3 (1985): 793-805.
9. Fama, Eugene F. "Efficient capital markets: A review of theory and empirical work*." The journal of Finance 25.2 (1970): 383-417.
10. Fama, Eugene F. "Market efficiency, long-term returns, and behavioral finance."Journal of financial economics 49.3 (1998): 283-306.
11. Fischhoff, Baruch, Paul Slovic, and Sarah Lichtenstein. "Knowing with certainty: The appropriateness of extreme confidence." Journal of Experimental Psychology: Human perception and performance 3.4 (1977): 552.
30
12. Graham, Benjamin. "intelligent investor." (1959).
13. Griffin, John M., Patrick J. Kelly, and Federico Nardari. "Do market efficiency measures yield correct inferences? A comparison of developed and emerging markets." Review of Financial Studies (2010): hhq044.
14. Hong, Harrison, and Jeremy C. Stein. A unified theory of underreaction, momentum trading and overreaction in asset markets. No. w6324. National Bureau of Economic Research, 1997.
15. J. Wang , B. M. Burton , D. M. Power. "Analysis of the overreaction effect in the Chinese stock market." Applied Economics Letters Vol. 11, Iss. 7, 2004
16. Kahneman, Daniel, and Amos Tversky. Intuitive prediction: Biases and corrective procedures. DECISIONS AND DESIGNS INC MCLEAN VA, 1977.
17. Lehmann, Bruce. "Fads, martingales, and market efficiency." (1988).
18. Lim, Kian-Ping, Muzafar Shah Habibullah, and Melvin J. Hinich. "The Weak-form Efficiency of Chinese Stock Markets Thin Trading, Nonlinearity and Episodic Serial Dependencies." Journal of Emerging Market Finance 8.2 (2009): 133-163.
19. Ma, Shiguang, and Michelle L. Barnes. Are China's stock markets really weak-form efficient?. Centre for International Economic Studies, 2001.
20. Malkiel, Burton G. "The efficient market hypothesis and its critics." Journal of economic perspectives (2003): 59-82.
21. Mobarek, Asma, and Keavin Keasey. "Weak-form market efficiency of an emerging Market: Evidence from Dhaka Stock Market of Bangladesh." ENBS Conference held on Oslo. 2000.
22. Odean, Terrance. "Do investors trade too much?." Available at SSRN 94143(1998).
23. Power, D. M., and A. Alasdair Lonie. "The overreaction effect: anomaly of the 1980s." The British Accounting Review 25.4 (1993): 325-366.
31
24. Schnusenberg, Oliver, and Jeff Madura. "DO US STOCK MARKET INDEXES OVER‐OR UNDERREACT?." Journal of Financial Research 24.2 (2001): 179-204.
25. Tripathi, Vanita, and Shalini Gupta. "Overreaction Effect in Indian Stock Market." Asian Journal of Business and Accounting, Vols (2009): 1-2.
26. Vardar, Gülin, and Berna Okan. "Short Term Overreaction Effect: Evidence on the Turkish Stock Market." EMERGING ECONOMIC ISSUES IN A GLOBALIZING WORLD (2008): 156. 27. Zarowin, Paul. "Short-run market overreaction: Size and seasonality effects." The Journal
of Portfolio Management 15.3 (1989): 26-29.
28. Zarowin, Paul. "Size, seasonality, and stock market overreaction." Journal of Financial and Quantitative analysis 25.01 (1990): 113-125.
32
Appendix
TABLE 1
Descriptive statistics of 5 loser portfolios and 5 winner portfolios for 12 months formation period and 36 months
testing period.
Portfolio Formation Period Testing Period Loser Winner Loser Winner
1 Mean t -0.0923 0.0802 0.0014 -0.0359 (-0.12) (0.34) (0.28) (-0.77) 2 Mean t -0.0507 0.0425 -0.0008 -0.0328 (-1.76*) (2.47**) (-0.09) (-0.79) 3 Mean t -0.0348 0.0921 -0.0013 -0.0258 (-0.73) (1.33) (-0.26) (-0.62) 4 Mean t -0.0564 0.0744 -0.0047 -0.0895 (-1.38) (0.95) (-0.85) (-2.01*) 5 Mean t -0.0721 0.0691 -0.0007 -0.1084 (-2.71**) (3.51**) (-0.17) (-2.33**)
Value in brackets () indicates corresponding t-statistics, where
* stands for significance level of 10%; ** stands for significance level of 5%; ***stands for the significance level of 1%
33
TABLE 2
Average cumulative abnormal return of winner portfolio (ACARW), loser portfolio (ACARL) and arbitrage portfolio along with the relative t value for
month 1 to month 36 in testing period
L 1 0.0338 3.7066** 0.1350 3.5874** -0.1012 -2.6132** 2 -0.0036 -0.1541 -0.0724 -0.7274 0.0688 0.6723 3 0.0056 0.3009 0.0052 0.0592 0.0004 0.0041 4 -0.0085 -0.6481 -0.0233 -0.1982 0.0148 0.1248 5 0.0055 1.0578 0.0321 0.6469 -0.0266 -0.5331 6 -0.0053 -0.3464 -0.2384 -1.6765* 0.2331 1.6296* 7 0.0120 0.2728 0.0589 0.4835 -0.0469 -0.3619 8 -0.0137 -0.7578 -0.1693 -1.3927 0.1555 1.2656 9 -0.0020 -0.1451 0.0726 0.6138 -0.0746 -0.6267 10 0.0145 1.1985 -0.0184 -0.1969 0.0330 0.3490 11 -0.0231 -4.2552*** -0.0507 -0.6276 0.0276 0.3412 12 0.0205 1.5851* -0.1294 -0.5990 0.1499 0.6925 13 0.0037 0.1837 0.1631 2.8010** -0.1593 -2.5855** 14 -0.0241 -2.6271** -0.1549 -1.0739 0.1308 0.9049 15 -0.0248 -2.5396** -0.0189 -0.2145 -0.0059 -0.0663 16 0.0121 2.4212** 0.0906 0.7558 -0.0785 -0.6541 17 -0.0004 -0.0321 -0.1015 -1.4317 0.1011 1.4007 18 -0.0200 -1.8476* -0.2996 -2.1164** 0.2796 1.9693** 19 -0.0043 -0.3396 -0.0864 -0.6876 0.0821 0.6499 20 -0.0187 -1.5481* -0.1956 -1.5129* 0.1769 1.3622 21 0.0157 1.5646* -0.0395 -0.7263 0.0552 0.9981 22 0.0275 1.4323 -0.0161 -0.2430 0.0436 0.6335 23 -0.0094 -0.3470 -0.0821 -0.7416 0.0727 0.6377 24 0.0310 3.0981** -0.0561 -0.7147 0.0871 1.1011 25 0.0141 1.2091 0.0969 0.5880 -0.0829 -0.5014 26 -0.0178 -2.3380** -0.2319 -1.3395 0.2142 1.2358 27 -0.0091 -0.6506 0.0394 0.3775 -0.0485 -0.4602 28 0.0026 0.1902 0.1034 0.9292 -0.1008 -0.8991 29 -0.0161 -1.0628 -0.1259 -1.4634 0.1098 1.2573 30 0.0204 1.4158 -0.2415 -5.0001*** 0.2619 5.1962*** 31 -0.0054 -0.3153 -0.0622 -0.4525 0.0568 0.4101 32 -0.0192 -2.0208 -0.2422 -1.5716 0.2229 1.4441* 33 -0.0126 -1.4418 -0.0937 -0.9520 0.0811 0.8209 34 -0.0030 -0.8094 -0.0707 -1.7067* 0.0677 1.6286* 35 -0.0076 -0.4052 -0.1692 -1.6964* 0.1616 1.5919* 36 -0.0148 -1.5748* 0.0870 0.3823 -0.1017 -0.4468 In average -0.0012 -0.0585 0.0573 0.8641
34
Figure 1
ACARs of the losers and winners portfolios of 30 stocks each during three years testing periods -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 1 3 5 7 9 11131517192123252729313335 t
ACARS the Losers, Winners and Arbitrage Portfolios during the 5 three-years Testing period
(Between January 1st 2007 to December 31th 2014)
Winner Portfolio Loser Portfolio Arbitrage Portfolio
