Do China’s Stock Markets Overreact?

(1)

MSc Program in Business Administration

Do China’s Stock Markets Overreact?

An Inquiry into Winner-Loser Effects on the Shanghai and the

Shenzhen Stock Exchanges

Author: Yimin Qin

Student Number: 1504851

Supervisor: Prof. Dr. Frans M. Tempelaar

Coordinator: Dr. Auke Plantinga

September 22, 2006, Groningen

(2)

Do China’s Stock Markets Overreact?

An Inquiry into Winner-Loser Effects on the Shanghai and the

Shenzhen Stock Exchanges

Author:Yimin Qin

Student Number: 1504851

Supervisor: Prof. Dr. Frans M. Tempelaar

Coordinator: Dr. Auke Plantinga

Program:MSc in Business Administration (Specialization Finance), Faculty of

Economics, University of Groningen

Final Edition: September 22, 2006

Suggested Keywords:China’s stock markets, the winner-loser effects, the

overreaction phenomenon

Correspondence:Faculty of Economics, University of Groningen, P.O. BOX

800, 9700 AV, Groningen, the Netherlands

(3)

ABSTRACT

(4)

ACKNOWLEDGEMENTS

For their help on the final version, I would like to thank my schoolfellows Liu Yao, Nie Zhe and Wang Lei. Liu Yao has provided several suggestions on data resources. Wang Lei and Nie Zhe were particularly helpful on the final draft.

Further, I am very grateful to all the lecturers and staffs of University of Groningen. They offered me an advanced academic education in Finance, and a precious time in the University of Groningen.

My greatest debt of gratitude is to Prof. Dr. Frans M. Tempelaar. Over the past 5 months, his insight into financial markets and thoughtful comments were always inspiring.

Finally, I wish to thank my families in China and my friends in Netherlands for their dedicated and unreserved supports during my study in the Netherlands.

(5)

1. INTRODUCTION

By the end of last century, researchers on the psychology of individual investor decision making have found results deviating from models based on investors’ rationality. In Kahneman and Tversky (1974), they found that when investors make judgments under uncertainty, especially when make subjective assessments of probabilities and predicting values, they would rely on a limited number of heuristic principles, such as representativeness, anchoring and adjustment. People tend to judge the probability of an event by finding a ‘comparable known’ event and assuming that the probabilities will be similar. And, they are inclined to underweight recent information and overweight prior (or base rate) information. When De Bondt and Thaler (1985) examined the market behavior on the New York Stock Exchange, they found evidence for the possible impacts of the above mentioned psychological findings. The so-called winner-loser effects based on the overreaction hypothesis showed that loser portfolio which experienced poor performance over the past three to five years, tends to outperform prior period winner portfolio over the following three to five years. The thoughts behind this finding may be that the market underreacts to recent information on winner’s and loser’s performance; while, continues to overreact to their prior performance. Therefore, the winners and the losers experience a long-term reversal in their abnormal returns as investors adjust for the initial reaction. Later on, other researchers found that not only individual investors exhibited such kind of overreaction behavior, but also security analysts and portfolio managers showed the same bias in decision making. So, in aggregate, this would have influence on market movements.

If such overreaction on market performance is predictable, it would be in contradiction to the weak form efficient market hypothesis. In a weak form efficient market, where security prices fully reflect all past information, no investors can make an excess profit by only using the information based on the past stock prices without bearing any extra risk.

The goal of this study is to examine whether overreaction phenomenon, or so-called winner-loser effects, also appears in China’s two main stock markets: the Shanghai Stock Exchange (SHSE), and the Shenzhen Stock Exchange (SZSE). Using monthly data over the period 1993-2005 of the Shanghai Stock Exchange, and monthly data over the period 1994-2005 of the Shenzhen Stock Exchange, the paper finds that the past loser portfolios significantly outperform the past winner portfolios in the subsequent holding periods. Based on this finding, the paper states that the overreaction phenomenon does exist in China’s stock markets.

(7)

methodology and the data are described; in the fourth section, the results of the primary tests and robustness tests are presented; in the fifth section, the conclusions are discussed; and in the final section, some directions for further studies are suggested.

2. LITERATURE REVIEW

The initial interest of the study is based on one of the most influential findings in De Bondt and Thaler (1985). In their paper, they formed winner and loser portfolios by using monthly returns of stocks listed in the New York Stock Exchange (NYSE) over the period 1926-1982. They first calculated cumulative residual returns (CUis) based on market-adjusted excess

returns, on each stock for each successive non-overlapping three-year period. The stocks were then ranked according to their CUis, and first the top and the bottom 35 stocks were formed

into portfolios of winners and losers. After this, the cumulative average residual returns (CARs) of the two portfolios were calculated over each of the subsequent three-year holding period. The paper found that, on average, the 35 losers outperformed the 35 winners by an average of 24.6%. When the analysis was repeated for a portfolio formation period of five years, the portfolio of losers outperformed the portfolio of winners by an average of 31.9%. Therefore, they suggested in the study that excess returns can be generated from a trading strategy that involves selling short the stocks of the portfolio of winners and holding the stocks of the portfolio of losers, which was also called contrarian investment strategy in other papers.

When researchers studied the post-announcement drift in order to get insight into market reaction to certain events, they also found evidences supporting the overreaction phenomenon. In Howe (1986), he tried to relate the stock’s weekly cumulative average residual (CARs) with the company’s specific favorable and unfavorable events. The results showed that, in the American Stock Exchange (ASE) and the New York Stock Exchange (NYSE), stocks that experienced large positive returns (or had good news) performed poorly in the 50-week period following that event. Besides, this poor performance was not concentrated in the period immediately following the event or announcement, but rather gradually spread out over a period of almost a year. The bulk of the return rebound of bad news stocks occurred after five weeks following the event, and their CARs were significant from week 1 thorough week 20. His findings from another way provided evidences on the existence of short-term overreaction phenomenon.

(8)

from 1926 to 1986, and set the formation period and the holding period both at 5 years. Results turned out to be consistent with a substantial overreaction effect. Using annual return intervals, extreme losers outperformed extreme winners by 6.5% per year. Using monthly return intervals, the spread increased to 9.5% per year. Furthermore, they concluded that the overreaction phenomenon is not just a manifestation of the size effect.

Later, Lakonishok, Shleifer and Vishny (1994) used stock prices from the New York Stock Exchange (NYSE) and the American Stock Exchange (ASE) from the end of April 1963 to the end of April 1990, and tested the contrarian strategy. In the paper, they explained that the contrarian strategy based on past winners’ and past losers’ performance yielded high returns. They argued that this was because the strategy exploited the suboptimal behavior of the typical investors, in which investors’ expectations of future performance appeared to be excessively tied to past performance, but not because this strategy was fundamentally more risky.

Concerning investor’s initial underreaction, Jegadeesh and Titman (1993) evaluated various momentum strategies, a strategy opposite to the contrarian strategy, in the U.S. market. The results showed that such strategies generated significant positive returns over 3- to 12- months holding periods. However, part of the abnormal returns generated by this strategy dissipated in the next two years after portfolio formation. This can be viewed as an indirect evidence of the overreaction phenomenon: short-term trending (momentum) precedes long-term reversal.

After the nineties, also studies have been conducted to inquire the overreaction phenomenon in the stock markets outside of the U.S., Campbell and Limmack (1997) tested for long-term abnormal returns’ reversals of the U.K. stocks which were also classified as 'winners' and 'losers' over the period from January 1979 to December 1990. When the study was extended to cover a five-year holding period following portfolio formation, they found that a reversal in the abnormal returns of the winner and the loser portfolios was experienced over each of years 2-5. In Asia, Chang, McLeavey, and Rhee (1995) stated that they also found a short-term (one year) abnormal returns of the contrarian strategy in the Japanese Tokyo Stock Exchange (TSE). This again can be viewed as an evidence to support the overreaction hypothesis. Recently, Bildik and Gulay (2000) examined both short-term momentum and long-term reversal effects in stock returns on one of the leading emerging markets, the Istanbul Stock Exchange (ISE), between years 1991 and 2000, and the results of long-term contrarian strategy were consistent with the overreaction hypotheses.

(9)

which can be measured by firm size or B/M ratio. They argued that the long-term return anomalies were chance results, by which they meant that the probability of apparent overreaction to information was about as same as underreaction, and that post-event continuation of pre-event abnormal returns was about as frequent as post-event reversal. Furthermore, they stated that such apparent anomalies could be due to imperfect methodology. Most long-term return anomalies tended to disappear with reasonable changes in technique. Zarowin (1990) demonstrated that both the small firm effect and the January effect could explain the abnormal returns reported by De Bondt and Thaler (1985). And Chan (1988) and Ball and Kothari (1989) showed that the estimation of long-term abnormal returns was sensitive to the model and estimation methods used.

With regard to the writer’s opinion, the overreaction phenomenon may exist, since it also has several supporting psychological evidences. Kahneman and Tversky (1974) did many experiments on individual’s judgments under uncertainty. They found that people made estimates by starting from an initial value that is adjusted to yield the final answer. The phenomenon that the different starting points yielded different estimates, which are biased towards the initial values, is called anchoring. In the same paper, they also discussed the representativeness heuristic in individual’s decision making process. Investors who follow this representativeness heuristic may think that they see patterns in truly random sequences, and conclude that the past history is representative of developments in the near future. However, observed consistent pattern of good performance may be nothing more than a random draw for a few lucky firms. As a consequence, investors may over-emphasize the stocks’ past performance and become disappointed in the future when the expectations fail to be realized. In De Bondt and Thaler (1985), this representative heuristic was briefly mentioned as a main psychological finding that generates the overreaction phenomenon. On the other side, Edwards (1986) suggested that human information processing presented conservatism. He stated that individuals were slow to change their beliefs in the face of new evidence. This conservatism was extremely pronounced at underreaction phenomenon. Investors incorporated recent news into stock prices slowly, which caused short-term momentum on winner’s and loser’s stock price development. However, over the long term the past losers would rebound, while the past winners would fail.

(10)

to an individual earnings announcement. In contrast, when the investor believed that earnings followed trends, he/she was subjective to the representativeness heuristic as the investor reacted too much (over-react) to the past performance information.

For more investors’ sentiment, Hirshleifer (2001) stated that over-confident of investors, which was implied by self-deception theory, may caused biases on asset pricing. Besides, investor’s emotion such as fears or other bad moods also affected people’s perceptions of choices with respect to risky future results.

(11)

3. METHODOLOGY AND RESEARCH DESIGN

3.1. Introduction of two stock exchanges

The Shanghai Stock Exchange (SHSE) and the Shenzhen Stock Exchange (SZSE) are two stock exchanges under the direct supervision of the China Securities Regulatory Commission (CSRC). They two together form the tier-markets for securities trading in the Mainland China. The Shanghai Stock Exchange was founded on November, 26, 1990, and in operation on December 19 the same year; while the Shenzhen Stock Exchange was officially set in operation on July 3, 1991. Two stock exchanges have the same variety of functions, such as providing marketplace and facilities for the securities trading; formulating business rules; accepting and arranging listings; organizing and monitoring securities trading; regulating members and listed companies; managing and disseminating market information. A qualified company can choose to be listed on one of the exchanges, or to be listed on both exchanges. Comparing the two exchanges in terms of number of listed companies, number of shares listed, total market value, tradable market value, and stock turnover in value, the Shanghai Stock Exchange is lager than the Shenzhen Stock Exchange (See Table 1). Most national big board companies, key industries in infrastructure and high-tech sectors prefer to list on the Shanghai Stock Exchange. By the end of 2005, the companies listed on SHSE are consisted by 544 industrial sector companies, 132 conglomerates, 83 public utilities, 58 commercial sector companies, and 17 real estates. All these companies can be further divided into 23 categories. Among them, companies in industrial sector and conglomerates cover up to 80% the total stock market value. At the end of 2005, in the Shanghai Stock Exchange, the first 50 listed companies with the largest market capitalization had a combined market capitalization of RMB1, 203.14 billion (Around USD149.07 Billion). These top 50 companies accounted for 52% of the total market capitalization of the Shanghai Stock Exchange.

On the other hand, the Shenzhen Stock Exchange has its own strength and focus, which is called the Small and Medium Enterprises Board (SME Board). The Small and Medium Enterprises Board is designed as an exclusive market segment for small- and mid-caps with pronounced core business, high growth potential or intensive technological contents. Table 2 below provides several important data for the SME Board.

(12)

Table 1: Markets Overview1 (Date: 2005-12-30)

The Shenzhen Stock Exchange The Shanghai Stock Exchange

Listed Companies 544 834

Listed Stocks2 586 878

RMB (¥) USD ($)3 RMB (¥) USD ($)

Stock Issued Capital (Million) 213,364 26,436 502,305 62,237 Stock Tradable Capital (Million)4 93,429 11,576 156,121 19,344 Stock Market Value (Million) 933,414 115,652 2,309,613 286,165 Stock Tradable Market Value

(Million) 387,590 48,023 675,461 83,691

Total Stock Value Traded in 2005

(Million) 1,242,456 153,943 1,924,021 238,390

Notes:

1. Data resource is from the official website of the Shanghai Stock Exchange and the Shenzhen Stock Exchange.

2. The listed stocks are the total of A shares and B shares listed in each stock exchange.

3. According to the historical information of the Bank of China, the currency exchange middle rates for USD and EURO were 1 EURO=9.5624 RMB, and 1 USD=8.0709 RMB on Dec. 30, 2005.

4. The stock tradable capital is estimated by the total capitalization of the tradable shares in the stock exchange, so as the stock tradable market value. The tradable shares here, which are called the free-float shares in other stock markets, are the proportion of shares that are not held by large owners and without sales restrictions. The non-tradable shares, which can be regarded as the restricted shares, are the proportion of shares cannot be traded freely on the stock exchanges. They are mainly state-owned shares and cooperation-owned shares. According to the statistics from the website of CSRC, till the end of February, 2004, around 416.45 billion shares in China’s stock markets were non-tradable shares, almost 64.50% of the total shares issued in the two exchanges.

Table 2: The Shenzhen SME Board Year-end Data (Date: 2005-12-30)1

The Shenzhen SME Board

Listed Companies 50

RMB (¥) USD ($)2

Stock Issued Capital (Million)

5,614 696

Stock Tradable Capital (Million)

2,220 275

Stock Market Value (Million)

48,155 5,966

Stock Tradable Market Value (Million)

18,529 2,296

Total Stock Value Traded in 2005 (Million)

533 66

Notes:

1. Data resource is from the official website of the Shenzhen Stock Exchange.

(13)

3.2. Dataset

The stock shares issued in the two exchanges can be further categorized into A Shares and B Shares. The A Shares are limited to domestic investors only, while B Shares are available to both domestic and foreign investors. By the end of 2005, there were 824 A Shares and 54 B Shares listed on the Shanghai Stock Exchange, and 525 A Shares and 55 B Shares listed on the Shenzhen Stock Exchange. The paper only tests the overreaction phenomena on all A Shares listed before December 31, 2005, since A Shares are shares only tradable by domestic investors, and present more than 90% of the total capitalization of two markets.

The data base consists of monthly returns1 on 824 A Shares in the Shanghai Stock Exchange and 525 A Shares in the Shenzhen Stock Exchange. However, only the stocks which are continuously listing on the exchanges and without any missing return data during the entire formation period are selected for the final tests. The historical data of stocks are downloaded from a Chinese mass portal web-side2.

By the end of 1992, the Shanghai Stock Exchange had only 60 listed stocks, and by the end of 1993, the Shenzhen Stock Exchange had only 27 listed stocks. Therefore, the paper excludes the first year of both stock exchanges. The time interval chosen for the Shanghai Stock Exchange is from January 1993 to December 2005, total 156 months; and for the Shenzhen Stock Exchange is from January 1994 to December 2005, total 144 months. The historical year-end figures for listed securities in the two stock exchanges can be found in Appendix 1 and Appendix 2.

In the previous researches, there are no uniform criteria for the length of the portfolio formation period and the holding period, though the most important papers use 2 years portfolio formation and 3 years portfolio holding period. This may be one of the reasons why the overreaction hypothesis has caused controversy. In this paper, the portfolio formation period for the primary test is: 2 year (n=24, starting in January). Later, the paper conducts robustness tests with the portfolio formation period as: 1.5 year (n=18, starting in January), and 1 year (n=12, starting in January). The holding periods are all set at three years, 36 months after the portfolio formation. In this way, the paper can compare the results of the tests with different portfolio formation periods. Because the stock markets in China do not

1

The monthly returns can be calculated by the month-end closing prices for all A shares listed on the Shanghai and the Shenzhen Stock Exchanges. The prices are adjusted for right issues and stock splits, but not for dividends.

2

(14)

have long histories, the paper applies overlapping periods to all the portfolio formation and holding periods. This enables the research to analyze the returns of the winner and the loser portfolios over 9 series of sample periods for the Shanghai Stock Exchange, and 8 series of sample periods for the Shenzhen Stock Exchange when the portfolio formation period is 2 years and 1.5 years; and over 10 series of sample periods for the Shanghai Stock Exchange, and 9 series of sample periods for the Shenzhen Stock Exchange when the portfolio formation period is 1 year.

3.3. Methodology

As this is the first study of the overreaction phenomenon in China’s stock markets, the paper follows the same standard techniques as used in De Bondt and Thaler (1985). The details are described by following 4-steps procedure used in the primary tests (the portfolio formation period is set at 2 years).

Step 1: Calculating Uit (the market-adjusted excess return) for the sample stocks

In each time series of sample period, for every stock without any missing data (without suspension or de-listing) during the portfolio formation period (month 1 to month 24), starting from month 1, January, the paper estimates next 60 (month 1 to month 60) monthly market-adjusted excess return Uit for the individual sample stocks.

Uit = Rit – Rmt i = 1 ……n, t = 1 . . . 60

Where,

Uit is the market-adjusted excess return on stock i at month t,

Rit is the return on stock i at same month t,

Rmt is an equally-weighted arithmetic average monthly rate of return on all A Shares listed on the

exchange at same month t:

Rmt = ΣRit / n i = 1 ……n, t = 1 . . . 60

The procedure is repeated to 9 series of sample periods for the Shanghai Stock Exchange, starting in January 1993, January 1994 …, up to January 2001, and repeated to 8 series of sample periods for the Shenzhen Stock Exchange, starting in January 1994, January 1995 …, up to January 2001.

Step 2: Calculating CUi (the cumulative excess return) for the sample stocks (see Step 2.1) and portfolio formation (see Step 2.2)

Step 2.1: In each series of sample periods, for every stock, starting in second December (month 24), the cumulative excess return CUi for the prior 24 months of the portfolio

(15)

CUi = ΣUit i = 1 ……n, t = 1 . . . 24

The procedure is repeated to 9 series of sample periods for the Shanghai Stock Exchange, and 8 series of sample periods for the Shenzhen Stock Exchange.

Step 2.2: At the end-month (month 24) of the relevant portfolio formation period, all the CUis

are ranked from high to low. Then, based on the ranking, the extreme CUi’s portfolios, which

are called the winner and the loser portfolios, are formed.

When the research needs to decide how many stocks are assigned to the winner and the loser portfolios, it notices that the authors in previous studies used different criteria. For example, De Bondt and Thaler (1985) first assigned top and bottom 35 stocks to the winner and the loser portfolios; Jegadeesh and Titman (1993) assigned top and bottom deciles; and, Chopra, Lakonishok and Ritter (1992) assigned top and bottom quintiles. This may cause potentially important differences in their findings, particularly with respect to their definitions of the winners and the losers. In the primary tests, the paper assigns the 10% stocks with the highest

CUi (the top decile) to a winner portfolio W, and the 10% stocks with the lowest CUi (the

bottom decile) to a loser portfolio L.

Step 3: Calculating CAR (the cumulative average residual return) for the portfolios In each three-year holding period, for both the winner and the loser portfolios, starting in month 25, month 37, and up to January 2003 (which is month 121 in the Shanghai Stock Exchange, and month 109 in the Shenzhen Stock Exchange), the paper now computes the cumulative average residual return CAR for each portfolio as a whole, which is then denoted as CARW,N,t and CARL,N,t (t =1……36). Here, N is the number of the winner and the loser

portfolios in each stock exchange. Therefore, N = 1……9 in the Shanghai Stock Exchange, and N = 1……8 in the Shenzhen Stock Exchange. If a stock experiences a short-period suspension during the holding period, then the CAR is calculated by the rest of the stocks in the holding period. Furthermore, if a stock experiences de-listing, then it is excluded from calculation forever.

Step 4: Calculating ACARs (the average CARs) for the portfolios over all holding periods Using the CARs for all winner and loser portfolios over the different series of the sample periods, the average CARs (ACARs) over all holding periods are calculated for each month between t=1 and t=36, which are denoted as ACARW, t and ACARL, t, t =1……36.

(16)

sample period for the aggregated market is from 1994 to 2005. In the aggregated market, the numbers of sample stocks are increased, and Rm is an equally-weighted arithmetic average

monthly return rate of all A Shares listed on two exchanges during the same time interval. In the first part’s robustness tests (see section 4.2.), the paper keeps the portfolio formation period at 2-years and the holding period at 3-years, but increases the fraction size of the top and the bottom stocks that assigned to the winner and the loser portfolios to 15%, 20%, 25% and 30%. Then the same test procedure is repeated.

In the second part’s robustness tests (see section 4.3.), the paper conducts the same test procedure after shortening the portfolio formation period to 1.5 years (18 months), and 1 year (12 months). The portfolio formation periods all start in January, and the holding periods are all the same as 3 years (36 months).

3.4. Hypothesis

The overreaction hypothesis built by De Bondt and Thaler (1985) supposes that, over a period of time after the portfolio formation, the past loser portfolios will outperform the past winner portfolios. In other words, the value of (ACARL, t – ACARW, t) over a period of time in the

holding period will be larger than zero. Although the initial researches on the overreaction phenomenon are focus on the U.S. stock markets, new researches are conducted recently on the Asia stock markets and the emerging markets(see Chang, McLeavey and Rhee (1995) and Bildik and Gulay (2000)), and the results supported the overreaction hypothesis. Since China is one of the emerging markets in Asia, it would be expected that the overreaction phenomenon also appear in China’s stock markets.

Therefore, the main hypothesis of the paper is: the past loser portfolios will outperform the past winner portfolios over a period of time after the portfolio formation in China’s stock markets, which include the Shanghai Stock Exchange, the Shenzhen Stock Exchange and the aggregated market.

Bildik and Gulay (2000) found that the profitability of the contrarian strategy in the Istanbul Stock Exchange (ISE) decreased as the portfolio size increased. Based on this argument, in the first part’s robustness tests, this paper inquires whether the long-term abnormal return difference between the winner and the loser portfolios is robust to the changes in the fraction size. The paper supposes that the same result as Bildik and Gulay (2000) will be found in China’s stock markets, which is as the fraction size increases, the long-term abnormal return differences between the winner and loser portfolios decreases.

(17)

(18)

4. RESULTS

The paper collects data information of total 1349 stocks (824 stocks in the Shanghai Stock Exchange, and 525 stocks in the Shenzhen Stock Exchange). The results of the tests are presented in following orders: First (see section 4.1.) are the results of three primary tests. The portfolio formation period is 2 years, the holding period is 3 years, and the top and the bottom decile stocks are assigned to the winner and the loser portfolios; then (see section 4.2.) are the results of the first parts’ robustness tests after changing fraction size to 15%, 20%, 25%, and 30%. At the end (see section 4.3.), the paper presents the results of the second parts’ robustness tests in the Shanghai and the Shenzhen Stock Exchanges after shortening the length of portfolio formation period to 1.5 years (18 months), and 1 year (12 months).

4.1. Results of primary tests

With Rm an equally-weighted arithmetic average monthly rate of return on A shares listed in

the exchange, the portfolio formation period as 2 years, and the top and the bottom decile stocks assigned to the winner and the loser portfolios, the paper gets the first observations in China’s stock markets (see Chart 1, 2 and 3).

From Chart 1 and Chart 2, the paper finds that in the Shanghai and the Shenzhen Stock Exchanges, the loser portfolios have higher ACARs than the winner portfolios immediately after the portfolio formation period. As the holding period extends, the discrepancy of ACARs between the winner and the loser portfolios is enlarging. In the aggregated market (see Chart 3), the return rebound of the loser portfolios is even stronger. During the first 18 months in the holding period, the ACARs of the loser portfolios are increasing sharply. But paper also notices that in the aggregated market, the ACARs of the loser portfolios during the first few months after the portfolio formation are lower than the ACARs of the winner portfolios, which implies a continuance of the momentum effect. All three charts show a clear pattern that over a period of time in the holding period, the loser portfolios have higher ACARs than the winner portfolios, which is supporting the main hypothesis.

(19)

the winner portfolios in Chart 1 and Chart 2 show different performances. The same situation as the Shanghai Stock Exchange is applicable to the aggregated market, since the aggregated market is formed by simply pooling the stocks listed in two exchanges together.

Chart 1: ACARs for the Winner and the Loser Portfolios in the Shanghai Stock Exchange (1993-2005, Portfolio Formation Period: 2 Years)

T h e S h a n g h a i S t o c k E x c h a n g e ( 1 9 9 3 - 2 0 0 5 )

ACARs for the Winner and the Loser Portfolios during 36 Months Test Period Length of the formation period: 2 years

Rm: an equally-weighted market return Portfolio fraction size:10%

-0.10000

0.00000

0.10000

0.20000

0.30000

0.40000

1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

Months after the portfolio formation

ACA

R ACARW

ACARL

Chart 2: ACARs for the Winner and the Loser Portfolios in the Shenzhen Stock Exchange (1994-2005, Portfolio Formation Period: 2 Years)

T h e S h e n z h e n S t o c k E x c h a n g e ( 1 9 9 4 - 2 0 0 5 )

Rm: an equally-weighted market return Portfolio fraction size:10%

0.00000

0.10000

0.20000

0.30000

0.40000

1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

ACAR

(20)

Chart 3: ACARs for the Winner and the Loser Portfolios in the Aggregated Market (1994-2005, Portfolio Formation Period: 2 Years)

T h e A g g r e g a t e d M a r k e t ( 1 9 9 4 - 2 0 0 5 )

Rm: an equally-weighted market return Fraction size:10% -0.10000 0.00000 0.10000 0.20000 0.30000 0.40000 1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

ACAR

ACARW ACARL

To further prove that the long-term abnormal return differences between the winner and the loser portfolios are statistically significant, for all three primary tests, the paper calculates the

T-statistics3 for (ACARL, t – ACARW, t) at each month in the holding periods. And the values of

(ACARL, t – ACARW, t) and T-statistics 1, 12, 24 and 36 months after the portfolio formation

period are presented in Table 3.

3

The formula the paper used is as same as that in De Bondt and Thaler (1985). The pooled estimate of the population variance in CARsare:

(21)

Table 3: (ACAR - ACAR ) and T-statistics 1, 12, 24 and 36 Months after the Portfolio Formation Period (Portfolio Formation Period: 2 Years

L, t W, t

) The Portfolio formation period

2 years starting in January (ACARL, t - ACARW, t ) and T-statistics Months after the portfolio formation period

1 12 24 36

The Shanghai Stock Exchange 0.01920

) (1.38065 0.14860 *** (6.83267) 0.26322 *** (7.23439) 0.33320 *** (7.22698)

The Shenzhen Stock Exchange 0.00097

) ) (0.07096 0.13526 * (2.36324) 0.16024 ** (2.49853) 0.21116 ** (3.26179)

The Aggregated Market -0.01189

(-1.15879 0.23552 *** (4.50558) 0.32037 *** (5.31184) 0.33708 *** (5.68489) Notes:

1. Rm is an equally-weighted market return.

2. Length of the portfolio formation period: 2 years (24 months) starting in January. 3. Length of the portfolio holding period: 3 years (36 months) ending in December.

4. Fraction size of the top and the bottom stocks assigned to the winner and the loser portfolios: 10%. 5. The Shanghai Stock Exchange has 9 paired sample series; the degree of freedom is 8. The Shenzhen

Stock Exchange has 8 paired sample series; the degree of freedom is 7. The aggregated market also has 8 paired sample series; the degree of freedom is 7. Therefore, the crucial T-statistics from

T-distribution with right tail probabilities should be as follows:

DF p=0.05 p=0.025 p=0.005

8 1.859548 2.30600 3.35539

7 1.894579 2.36462 3.49948

6. *** significant at 99% level, ** significant at 95% level, *significant at 90% level.

Table 3 shows, one month after the portfolio formation period, the value of (ACARL, 1– ACARW, 1) in the Shanghai Stock Exchange is 1.92% and in the Shenzhen Stock Exchange is

0.10%, which is positive but not statistically significant. As the same time, the value of (ACARL, 1– ACARW, 1) in the aggregated market is -1.19% and indicates a continuance of

momentum effect. After 12 months in the holding period, the values of (ACARL, t– ACARW, t)

in China’s stock markets are all positive and at least significant at 90% level. In the aggregated market and in the Shanghai Stock Exchange, these values are even significant at 99% level, which provide strong evidence supporting the overreaction hypothesis.

Besides, in the Shanghai Stock Exchange, the increment of (ACARL, t - ACARW, t) from Month

1 to Month 12 is 12.94%, from Month 12 to Month 24 is 11.46%, and from Month 24 to Month 36 is 7%; in the Shenzhen Stock Exchange, the increment of (ACARL, t – ACARW, t)

from Month 1 to Month 12 is 13.43%, from Month 12 to Month 24 is 2.49%, and from Month 24 to Month 36 is 5.10%; and in the aggregated market, the increment of (ACARL, t – ACARW, t)

(22)

also mentioned in De Bondt and Thaler (1985), Campbell and Limmack (1997) and Bildik and Gulay (2000).

Moreover, the values of (ACARL, t– ACARW, t) 12, 24 and 36 months after the portfolio

formation are the highest in the aggregated market and the lowest in the Shenzhen Stock Exchange, which suggests that the winner-loser effects are the most significant in the aggregated market, but the least significant in the Shenzhen Stock Exchange. De Bondt and Thaler (1985) stated that “……stocks that go through more extreme return experiences, the subsequent price reversals will be more pronounced……” Accordingly, this paper provides a new way to generate more extreme observations of the winner-loser effects is to enlarge the sample stocks that enter in the selection. The more the stocks that enter in the selection for the winners and the losers, the more pronounced the return reversal in the subsequent period.

To summarize, the paper states that with 2 years portfolio formation period and decile fraction size, there is overreaction phenomenon in China’s stock markets, which includes the Shanghai Stock Exchange, the Shenzhen Stock Exchange and the aggregated market, and past losers outperform the past winners in the subsequent 36 months in the holding period. Most of the return reversal is happening in the one to three years after the portfolio formation period, however, as the holding period extends, the return reversal shows a downward trend. The winner-loser effects are the most pronounced in the aggregated market, which suggests a new way to generate more extreme observation of the return reversal is to enlarge the sample stocks enter in the selection for the winners and the losers.

4.2. Results of robustness tests after changing the fraction sizes of the top and the

bottom stocks assigned to the winner and the loser portfolios

In three primary tests, the top and the bottom decile stocks are assigned to the winner and the loser portfolios. A question may be raised here: are there too few stocks enter into the observation in the first few series of sample periods, especially in the Shanghai and the Shenzhen Stock Exchanges?

(23)

Moreover, in both two exchanges, the average size of the winner and the loser portfolios over the whole sample periods is 25, which is sufficiently large in the writer’s opinion. In the aggregated market, in spite of the first series of sample period, the portfolio sizes in the following series of sample periods are all above 20. So setting the fraction size as the top and the bottom deciles in three primary tests is proper.

Table 4: Sizes of the Winner and the Loser Portfolios in Each Series of Sample Periods with Different Fraction Sizes (Portfolio Formation Period: 2 Years)

The Shanghai Stock Exchange

Size of Portfolios 10% 15% 20% 25% 30% 1993-1997 3 4 5 7 8 1994-1998 8 12 17 21 25 1995-1999 13 20 26 33 40 1996-2000 16 24 32 40 48 1997-2001 26 39 51 64 77 1998-2002 33 49 65 82 98 1999-2003 38 57 75 94 113 2000-2004 41 62 82 103 123 2001-2005 50 74 99 124 149 Average Size 25 38 50 63 76

The Shenzhen Stock Exchange

Size of Portfolios 10% 15% 20% 25% 30% 1994-1998 5 8 11 13 16 1995-1999 9 14 19 23 28 1996-2000 11 17 22 28 34 1997-2001 20 30 40 51 61 1998-2002 31 46 62 77 92 1999-2003 36 54 72 90 107 2000-2004 41 61 81 102 122 2001-2005 43 65 87 108 130 Average Size 25 37 49 62 74

The Aggregated Market

Size of Portfolios 10% 15% 20% 25% 30% 1994-1998 14 20 27 34 41 1995-1999 23 34 45 56 68 1996-2000 27 41 54 68 81 1997-2001 46 69 92 115 138 1998-2002 63 95 127 159 190 1999-2003 74 110 147 184 221 2000-2004 82 123 164 205 245 2001-2005 93 139 186 232 278 Average Size 53 79 105 131 158

(24)

Stock Exchange (ISE) decreased as the portfolio size increased. To inquire whether the long-term abnormal return difference between the winner and the loser portfolios is robust to the changes in the fraction size, the paper repeats the same test procedure after increasing the fraction size from 10% to 15%, 20%, 25% and 30%. The lengths of the portfolio formation and the holding periods are unchanged as 2 years and 3 years respectively. The trends of (ACARL, t – ACARW, t) with different fraction sizes are presented in the following three charts

(see Chart 4, 5 and 6).

Chart 4: The Trends of (ACAR - ACAR ) with Different Fraction Sizes in the Shanghai Stock Exchange (1993-2005, Portfolio Formation Period: 2 Years)

L, t W, t

T h e S h a n g h a i S t o c k E x c h a n g e

The Trends of (ACARL,t- ACARW,t) with Different Fraction Sizes

Length of the formation period: 2 years starting in January Length of the test period: 3 years ending in December

Rm: an equally-weighted market return

0.00000 0.05000 0.10000 0.15000 0.20000 0.25000 0.30000 0.35000 1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

Months after portfolio formation

ACA RL-ACAR W Fraction size:10% Fraction size:15% Fraction size:20% Fraction size:25% Fraction size:30%

Chart 5: The Trends of (ACAR - ACAR ) with Different Fraction Sizes in the Shenzhen Stock Exchange (1994-2005, Portfolio Formation Period: 2 Years)

L, t W, t

T h e S h e n z h e n S t o c k E x c h a n g e

-0.05000 0.00000 0.05000 0.10000 0.15000 0.20000 0.25000 0.30000 1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

(25)

Chart 6: The Trends of (ACAR - ACAR ) with Different Fraction Sizes in the Aggregated Market (1994-2005, Portfolio Formation Period: 2 Years)

L, t W, t

T h e A g g r e a g a t e M a r k e t

-0.05000 0.00000 0.05000 0.10000 0.15000 0.20000 0.25000 0.30000 0.35000 0.40000 1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

ACA RL-ACA RW Fraction size:10% Fraction size:15% Fraction size:20% Fraction size:25% Fraction size:30%

Chart 5 and Chart 6 show that in the first few months immediately after the portfolio formation, the trend-lines of (ACARL, t - ACARW, t) in the Shenzhen Stock Exchange and the

aggregated market are below zero and both show a short-term continuance of the momentum effect. But this momentum effect disappears 2 months later in the Shenzhen Stock Exchange and 3 months later in the aggregated market. Despite this, one consistent pattern in Chart 4, 5 and 6 is that in the Shanghai Stock Exchange, the Shenzhen Stock Exchange and the aggregated market, the great part of the trend-lines of (ACARL, t - ACARW, t) with different

fraction sizes are stretching above zero in the holding period. This again provides evidence to the winner-loser effects in China’s stock market: the loser portfolios outperform the winner portfolios in the subsequent holding period.

Another clear pattern in Chart 4, 5 and 6 is that the trend-line of (ACARL, t - ACARW, t) declines

as the fraction size increases. In all three charts, over a period of time (around 12 months), the trend-lines of (ACARL, t - ACARW, t) with 10% fraction size are in general the highest; while the

(26)

more stocks with less extreme past return performance are added into the winner and the loser portfolios. As the past performances of the winners and the losers are convergent to the average, the return reversals in the subsequent holding period decrease so as to the winner-loser effects.

In addition, the charts show that with different fraction sizes, the upward trend of (ACARL, t - ACARW, t) in the Shanghai Stock Exchange and the aggregated market last around 18 to 24

months after portfolio formation period; while in the Shenzhen Stock Exchange, the upward trend of (ACARL, t - ACARW, t) lasts only 12 month after portfolio formation period. Later on,

the fluctuant trends appear in all three markets. There is even a downward trend in the aggregated market over 24 months in the holding period. This again affirms that most of the abnormal return reversal in China’s stock markets happens from one to three yeas after the portfolio formation. And as the holding period continues to extend, the abnormal return reversal shows a downward trend.

In the same way, the paper calculates the T-statistics for the (ACARL, t - ACARW, t) at each

month in the holding period. The values of (ACARL, t - ACARW, t) and the T-statistics 1, 12, 24

and 36 months in the holding period reveal the same findings above (see Table 5, 6 and 7). One unexpected finding here is that even after the top and the bottom 30% stocks assigned to the winner and loser portfolios, significant winner-loser effects still exist in the Shanghai Stock Exchange and the aggregated market. With 30% fraction size, over first 12 months in the holding period, the value of (ACARL, 12 - ACARW, 12) is 11.61% in the Shanghai Stock

Exchange and 20.10% in the aggregated market. Another 12 months later, the value of (ACARL, 24 - ACARW, 24) increases to 16.12% in the Shanghai Stock Exchange and 23.88% in

the aggregated market. By the end of 36 months in the holding period, the value of (ACARL, 36

- ACARW, 36) reaches to 17.97% in the Shanghai Stock Exchange and 23.27% in the

aggregated market. The above-mentioned values are all significant at 99% level. These show that the winner-loser effects are quite prevalent in China’s stock markets. These also indicates that when the mass of the Chinese investors process the information, they under-react and over-react on many stocks available in the markets rather than only on a few stocks with the extreme prior performance. However, the origin of such prevalent overreaction behavior in China’s stock markets is unknown yet and needs to be further studied.

(27)

assigned to the portfolios. One surprising finding is that even when the fraction size increases to 30%, the winner-loser effects are still significant in the Shanghai Stock Exchange and the aggregated market. This may suggest that the winner-loser effects are prevalent on many stocks traded in the market.

Table 5: (ACAR - ACAR ) and T-statistics of the Shanghai Stock Exchange 1, 12, 24 and 36 Months in the Holding Period (During 1993 – 2005, Portfolio Formation Period: 2 Years

L, t W, t

)

T

th (A

he fraction size of

e top and the bottom stocks CARL, t - ACARW, t ) and T-statistics

Months after the portfolio formation period

1 12 24 36 1 0.019 (1.38 0.1 (6.8 20 065) 4860 *** 3267) 0.26322 *** (7.23439) 0.33320 *** (7.22698) 0% 1 0.015 (1.28 0.1 (6.5 2 0.018 (1.64 0.1 (6.7 2 0.015 (1.47 0.1 (5.3 3 0.011 (1.14 0.1 (5.1 52 355) 4231 *** 8765) 0.26415 *** (7.61286) 0.26954 *** (6.51493) 5% 31 393) 3939 *** 1930) 0.22133 *** (6.75058) 0.23027 *** (5.65002) 0% 52 028) 1713 *** 1529) 0.18131 *** (5.82851) 0.19693 *** (5.29273) 5% 62 905) 1609 *** 0035) 0.16120 *** (5.14848) 0.17968 *** (4.61158) 0% Notes:

4. The Shanghai Stock Exchange has 9 paired sample series; the degree of freedom is 8. Therefore, the crucial T-statistics from T-distribution with right tail probabilities should be as follows:

DF p=0.05 p=0.025 p=0.005

1.859548 2.30600 3.35539

8

Table 6: (ACAR - ACAR ) and T-statistics of the Shenzhen Stock Exchange 1, 12, 24 and 36 Months in the Holding period (During 1994-2005

L, t W, t

, Portfolio Formation Period: 2 Years)

(ACARL, t - ACARW, t ) and T-statistics

Months after the portfolio formation period The fraction size of

the top and the bottom stocks

1 12 24 36 0.00097 (0.07096) 0.13526 * (2.36324) 0.16024 ** (2.49853) 0.21116 ** (3.26179) 10% -0.00824 (-0.74047) 0.13269 ** (2.44300) 0.15898 ** (2.56657) 0.21302 ** (3.32182) 15% -0.00745 (-0.73681) 0.12932 * (2.17633) 0.14473 * (2.13479) 0.20949 ** (3.15462) 20% -0.00801 (-0.88643) 0.09633 (1.66612) 0.12243 (1.84881) 0.15833 ** (2.43628) 25% -0.00175 (-0.19028) 0.08557 (1.47327) 0.09953 (1.54166) 0.12325 (1.84234) 30% Notes:

4. The Shenzhen Stock Exchange has 8 paired sample series; the degree of freedom is 7. Therefore, the crucial T-statistics from T-distribution with right tail probabilities should be as follows:

p=0.05 p=0.025 p=0.005

DF

1.894579 2.36462 3.49948

7

(28)

Table 7: (ACAR - ACAR ) and T-statistics of the Aggregated Market 1, 12, 24 and 36 Months in the Holding period (During 1994-2005, Portfolio Formation Period: 2 Years

L, t W, t

)

The fraction size of (ACARL, t - ACARW, t ) and T-statistics

1 12 24 36 -0.01189 (-1.15879) 0.23552 *** (4.50558) 0.32037 *** (5.31184) 0.33708 *** (5.68489) 10% -0.00568 (-0.65618) 0.22334 *** (4.29736) 0.29213 *** (4.95219) 0.29572 *** (5.02418) 15% -0.00779 (-0.92408) 0.20850 *** (4.20689) 0.26534 *** (4.62290) 0.27054 *** (4.57455) 20% -0.00949 (-1.13782) 0.19429 *** (4.07799) 0.25317 *** (4.44168) 0.24417 *** (4.06055) 25% -0.00779 (-1.00282) 0.20101 *** (4.10347) 0.23884 *** (4.17975) 0.23269 *** (3.77189) 30% Notes:

4. The aggregate market has 8 paired sample series; the degree of freedom is 7. Therefore, the crucial T-statistics from T-distribution with right tail probabilities should be as follows:

p=0,05 p=0,025 p=0,005

DF

1.894579 2.36462 3.49948

7

4.3. Results of robustness tests after shortening the length of the portfolio

formation period

In order to test if the long-term abnormal return difference between the winner and the loser portfolios is robust to the length of the portfolio formation period in the Shanghai and the Shenzhen Stock Exchanges, the paper now repeats the same test procedure after shortening the portfolio formation lengths to: 1.5 years (18 months, starting in January) and 1 years (12 months starting in January).

Chart 7 and Chart 8 show the trends of (ACARL, t - ACARW, t) with different lengths of the portfolio formation periods in the Shanghai and the Shenzhen Stock Exchanges, and the fraction size is 10%. In two charts, except when the portfolio formation period is 1 year, the trend-lines of (ACAR - ACARL, t W, t) over the first few months present a short-term momentum effect, the great part of the trend-lines of (ACAR - ACAR ) in two exchanges are stretching above zero. This suggests that even after the portfolio formation period is shortened, the winner-loser effects still appear in China’s stock market, and the loser portfolios have better performance than the winner portfolios over a period of time in the holding period.

(29)

Chart 7: The Trends of (ACAR - ACAR ) in the Shanghai Stock Exchange with Different Portfolio Formation Periods (1993-2005, Fraction Size: 10%)

L, t W, t

T h e S h a n g h a i S t o c k E x c h a n g e

The Trends of (ACARL,t - ACARW,t) with Different Portfolio Formation Periods

Length of the test period: 3 years (36 months) Rm: an equally-weighted market return

Fraction size: 10% -0.05000 0.00000 0.05000 0.10000 0.15000 0.20000 0.25000 0.30000 0.35000 0.40000 0.45000 0.50000 0.55000 1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

Months after the portfolio formation period

ACARL-ACARW

Portfolio foramtion period:1 year starting in January

Portfolio formation period:1.5 years starting in January

Portfolio formation period:2 years starting in January

Chart 8: The Trends of (ACAR - ACAR ) in the Shenzhen Stock Exchange with Different Portfolio Formation Periods (1994-2005, Fraction Size: 10%)

L, t W, t

T h e S h e n z h e n S t o c k E x c h a n g e

The Trends of (ACARL,t - ACARW,t) with Different Portfolio Formation Periods

Length of the test period: 3 years (36 months) Rm: an equally-weighted market return

Fraction size: 10% -0.05000 0.00000 0.05000 0.10000 0.15000 0.20000 0.25000 0.30000 0.35000 0.40000 1 4 7 ₁₀ ₁₃ ₁₆ ₁₉ ₂₂ ₂₅ ₂₈ ₃₁ ₃₄

Months after the portfolio formation period

ACARL-ACARW

Portfolio foramtion period:1 year starting in January

Portfolio formation period:1.5 years starting in January

(30)

One consistent pattern in both two charts is that the trend-lines of (ACAR - ACARL, t W, t) with 1.5 years portfolio formation period is always the highest during the entire 36 months holding periods, and the trend-lines of (ACAR - ACARL, t W, t) with 2 years portfolio formation period is always below the trend-lines with 1.5 years portfolio formation period. Besides, especially in the Shenzhen Stock Exchange, the trend-lines of (ACAR - ACAR ) with 1 year portfolio formation period

L, t W, t

are below the trend-lines with 1.5 years portfolio formation period, but across the trend-lines with 2 years portfolio formation period. The above patterns not only appear when the fraction size is 10%, but also appear when the fraction size increases to 15%, 20%, 25%, and 30%. The relevant charts can be found in the Appendix 3 to Appendix 6. Accordingly, the paper states that the long-term abnormal return difference between the winner and the loser portfolios is robust to the changes of the portfolio formation length. However, no consistent relationship can be concluded from these patterns.

De Bondt and Thaler (1985) stated that “…… an easy way to generate more (less) extreme observations on the overreaction hypothesis is to lengthen (shorten) the portfolio formation period……” This statement is not applicable to China’s stock markets. Chart 7 and Chart 8 display that in the Shanghai and the Shenzhen Stock Exchanges, the winner-loser effects become more prominent when the length of the portfolio formation period is shortened from 2 years to 1.5 years. Moreover, when the length of the portfolio formation period is further shortened to 1 year, there is no any sign suggesting that the winner-loser effects become less extreme.

The values of (ACARL, t - ACARW, t) and the T-statistics with different length of the portfolio

formation period 1, 12, 24 and 36 months in the holding period are presented in Table 8 and Table 9 at the end of this sub-section. In Table 8, the portfolio formation period is 1.5 years, which is starting in January and ending in June 18 months later. This means the winner and the loser portfolios are reformed at each July. In the Shanghai Stock Exchange, all the values of (ACARL, t - ACARW, t) over 12 months in the holding period are positive and significant at

99% level. In the Shenzhen Stock Exchange, all the values of (ACARL, t - ACARW, t) are

positive and at least significant at 95% level even only over one month in the holding period. These strongly support the winner-loser effects.

Furthermore, in the Shanghai Stock Exchange, all the values of (ACARL, t - ACARW, t) decrease

from Month 24 to Month 36 except for the 10% fraction size. This again confirms that the long-term abnormal return reversal decreases as the holding period extends. In the Shenzhen Stock Exchange, though the values of (ACARL, t - ACARW, t) increase a little bit from Month 24

(31)

In Table 9, the portfolio formation period is 1 year, which is starting in January and ending in December 12 months later. The results in Table 9 confirm that even when the portfolio formation period is shortened to 1 year, the winner-loser effects still exist in the Shanghai and the Shenzhen Stock Exchanges. In the Shanghai Stock Exchange, the values of (ACARL, t - ACARW, t) with all fraction sizes 12 months after the portfolio formation are positive and at

least significant at 90% level (except the value at Month 12 with 30% fraction size). In the Shenzhen Stock Exchange, all the values of (ACARL, t - ACARW, t) 12 months after the portfolio

formation are positive, and with 10% and 15% fraction sizes, they are at least significant at 90% level.

These results are inconsistent with the findings in De Bondt and Thaler (1985) and Campbell and Limmack (1997), which declared that for a formation period as short as one year, no reversal was observed at all in New York Stock Exchange (NYSE) and the U.K. stock market. On the other hand, such kind of the overreaction phenomenon with short-term (one year) portfolio formation period was also revealed in the research on the emerging market. In Bildik and Gulay (2000), they found the evidence supporting such kind of overreaction phenomenon with short-term portfolio formation period in the Istanbul Stock Exchange (ISE), when they selected stocks based on their returns over the past 12 months. Such inconsistent conclusions may be cause by the different markets the researches focus on. De Bondt and Thaler (1985) and Campbell and Limmack (1997) analyze the overreaction phenomenon on the U.S. and the U.K. stock markets, which are developed market; while this paper and Bildik and Gulay (2000) study on the China and the Istanbul stock markets, which are both emerging markets. On account of that the emerging markets are subjected to high volatility, low liquidity, and small changes in supply and demand can have a dramatic impact on market price, the overreaction phenomenon with short-term portfolio formation period may be a unique characteristic of the emerging markets,

(32)

Table 8: (ACAR - ACAR ) and T-statistics of Two Exchanges 1, 12, 24 and 36 Months in the Holding Period (Portfolio Formation Period: 1.5 Years Starting in January

L, t W, t

) Length of the formation period:

1 5 i i J

(ACARL, t – ACARW, t ) and T-statistics

1 12 24 36 10% 0.03638 * (2.27441) 0.29425 *** (6.20048) 0.42908 *** (7.55802) 0.42909 *** (5.54319) 15% 0.02486 (1.70345) 0.28894 *** (6.61017) 0.41782 *** (7.60915) 0.41574 *** (6.08597) 20% 0.02512 (1.76504) 0.25722 *** (5.95927) 0.38324 *** (6.95885) 0.37043 *** (5.40254) 25% 0.02820 * (1.98247) 0.20283 *** (4.47438) 0.33905 *** (6.20574) 0.29531 *** (4.15547)

The Shanghai Stock Exchange 30% 0.02614 (1.84269) 0.17403 *** (3.71372) 0.28848 *** (5.32671) 0.25651 *** (3.66633) 10% 0.09328 *** (4.96647) 0.22094 *** (4.12168) 0.29584 *** (4.49166) 0.32722 *** (4.55963) 15% 0.07788 *** (4.31547) 0.20905 *** (3.79410) 0.27005 *** (3.60310) 0.28222 *** (3.73088) 20% 0.06713 *** (3.96398) 0.18000 ** (3.25135) 0.22524 ** (3.07911) 0.23667 ** (3.21893) 25% 0.06036 ** (3.43536) 0.14995 ** (2.84070) 0.19420 ** (2.56025) 0.20413 ** (2.71590)

The Shenzhen Stock Exchange 30% 0.05532 ** (3.17270) 0.14794 ** (2.87664) 0.18524 ** (2.41197) 0.18437 ** (2.41897) Notes:

2. Length of the portfolio formation period: 1.5 years (18 months) starting in January. 3. Length of the portfolio holding period: 3 years (36 months) ending in June.

4. The Shanghai Stock Market has 9 paired sample series; the degree of freedom is 8. While, The Shenzhen Stock Market has 8 paired sample series, the degree of freedom is 7. Therefore, the crucial

T-statistics from T-distribution with right tail probabilities should be as follows:

p=0.05 p=0.025 p=0.005 DF 1.859548 2.30600 3.35539 8 1.894579 2.36462 3.49948 7

(33)

Table 9: (ACAR - ACAR ) and T-statistics of Two Exchanges 1, 12, 24 and 36 Months in the Holding Period (Portfolio Formation Period: 1 Year Starting in January

L, t W, t

) Length of the formation period:

1 t ti i J

(ACARL, t - ACARW, t) and T-statistics

1 12 24 36 10% -0.00807 (-0.84015) 0.20796 *** (6.61274) 0.33739 *** (8.90818) 0.37952 *** (7.73647) 15% -0.01170 (-1.18694) 0.13906 *** (5.49800) 0.26813 *** (8.01494) 0.27947 *** (6.52669) 20% -0.01422 (-1.54285) 0.09399 *** (3.58059) 0.20562 *** (6.38046) 0.23122 *** (5.42899) 25% -0.01544 (-1.73472) 0.05005 * (2.04086) 0.16999 *** (5.51050) 0.20493 *** (5.09269)

The Shanghai Stock Exchange 30% -0.01523 (-1.71549) 0.03465 (1.41866) 0.14101 *** (4.41140) 0.17112 *** (4.24366) 10% 0.00375 (0.29389) 0.15386 (2.83420) ** 0.17391 (2.71248) ** 0.24916 (3.56458) *** 15% 0.00993 (0.86702) 0.12866 (2.47338) ** 0.15324 (2.40627) ** 0.21449 (3.20507) ** 20% 0.00104 (0.09781) 0.09519 (1.82910) 0.12153 (1.89170) * 0.15708 (2.44536) ** 25% 0.00112 (0.11501) 0.07887 (1.54186) 0.10483 (1.64697) 0.12590 (1.95766) * The Shenzhen Stock Exchange 30% -0.00017 (-0.01795) 0.05843 (1.16864) 0.08201 (1.28559) 0.09295 (1.44807) Notes:

4. The Shanghai Stock Market has 10 paired sample series; the degree of freedom is 9. While, the Shenzhen Stock Market has 9 paired sample series, the degree of freedom is 8. Therefore, the crucial T-statistics from T-distribution with right tail probabilities should be as follows:

p=0,05 p=0,025 p=0,005 DF 1.833113 2.26216 3.24984 9 1.859548 2.30600 3.35539 8

(34)

5. CONCLUSIONS

Following the methodology used by De Bondt and Thaler (1985), this paper makes an empirical inquiry of the overreaction phenomenon in China’s stock markets. There are four main findings are revealed in this inquiry. First, the overreaction phenomenon does exist in China’s stock markets. Using same portfolio formation method, the paper finds that the overreaction phenomenon is more prominent in the Shanghai Stock Exchange than in the Shenzhen Stock Exchange, and is the most prominent after integrating two stock markets into one aggregated market. This may suggest that a new way to generate extreme observation of the winner-loser effects is to enlarge the sample stocks that enter into selection for winners and losers.

Second, the results of the paper reveal that most of the abnormal return reversal in China’s stock markets is generated from one to three years in the holding period. And, long-term holding period has a higher return reversal than the short-term holding period. But when the holding period continues to extend, a downward trend of long-term abnormal return differences between the winner and the loser portfolios appears, which indicates a weakening of the winner-loser effects.

Third, the paper confirms that the long-term return reversal of the winner-loser effects is robust to the changes in the fraction size. This is consistent with the findings in De Bondt and Thaler (1985) and Bildik and Gulay (2000). With different lengths of the portfolio formation periods (2 years, 1.5 years and 1 year), over 12 months in the holding period, the winner and the loser portfolios formed by the top and the bottom decile stocks generally have the highest return reversal; while the winner and the loser portfolios formed by the top and the bottom 30% stocks generally have the lowest return reversal. One surprising finding is that with 2 years portfolio formation periods, even when the fraction size increases to 30%, the winner-loser effects are still quite significant in the Shanghai Stock Exchange and the aggregated market. This appears again when the portfolio formation period is 1.5 years. The writer supposes that the mass of the Chinese investors exhibits same irrational behavior. They under-react and over-react on many stocks available in the markets rather than only on the stocks with extreme past performances. As a result, the overreaction phenomenon is prevalent in China’s stock markets.

(35)

formation period is further shortened to 1 year, the winner-loser effects are still significant in the Shanghai and the Shenzhen Stock Exchanges. This suggested that in China’s stock market, there is overreaction phenomenon with short-term portfolio formation period.

(36)

6. FURTHER STUDY

The Shanghai and the Shenzhen Stock Exchanges both have very short histories. This fact limits the series of the sample periods that can be created for the observation in this paper. Though it is interesting to make an inquiry on overreaction phenomenon in China’s stock markets at this stage, the reality does impose restriction on the length of portfolio formation period and the holding period in the tests. Besides, most abnormal returns in financial market will disappear when more professionals exploit opportunities on them. So it will be meaningful to conduct continuous inquiries on the overreaction phenomenon in China’s stock markets in the future.

Second, as mentioned in literature review, Zarowin (1990) demonstrated that both the small firm effect and the January effect could explain such long term abnormal return difference between the winner and the loser portfolios. Due to the time constraint, this paper does not make any formal tests on both effects. Following studies on the overreaction phenomenon in China’s stock markets could conduct rigorous tests on both the small firm effect and the January effect.

Do China’s Stock Markets Overreact?

MSc Program in Business Administration

Do China’s Stock Markets Overreact?

An Inquiry into Winner-Loser Effects on the Shanghai and the

Shenzhen Stock Exchanges

Author: Yimin Qin

Student Number: 1504851

Supervisor: Prof. Dr. Frans M. Tempelaar

Coordinator: Dr. Auke Plantinga

September 22, 2006, Groningen

Do China’s Stock Markets Overreact?

An Inquiry into Winner-Loser Effects on the Shanghai and the

Shenzhen Stock Exchanges

Author:Yimin Qin

Student Number: 1504851

Supervisor: Prof. Dr. Frans M. Tempelaar

Coordinator: Dr. Auke Plantinga

Program:MSc in Business Administration (Specialization Finance), Faculty of

Economics, University of Groningen

Final Edition: September 22, 2006

Suggested Keywords:China’s stock markets, the winner-loser effects, the

overreaction phenomenon

Correspondence:Faculty of Economics, University of Groningen, P.O. BOX

800, 9700 AV, Groningen, the Netherlands

ABSTRACT

ACKNOWLEDGEMENTS

TABLE OF CONTENTS

1. INTRODUCTION

2. LITERATURE REVIEW

3. METHODOLOGY AND RESEARCH DESIGN

3.1. Introduction of two stock exchanges

3.2. Dataset

3.3. Methodology

3.4. Hypothesis

4. RESULTS

4.1. Results of primary tests

-0.10000

0.00000

0.10000

0.20000

0.30000

0.40000

0.00000

0.10000

0.20000

0.30000

0.40000

4.2. Results of robustness tests after changing the fraction sizes of the top and the

bottom stocks assigned to the winner and the loser portfolios

4.3. Results of robustness tests after shortening the length of the portfolio

formation period

5. CONCLUSIONS

6. FURTHER STUDY