Student name: Dingyi Chen
Student number: s1741683
Supervisor:
Dr. J. H. Von EijeUniversity of Groningen, Faculty of Economics E-mail address:s1741683@student.rug.nl Date: June 28, 2010
Master Thesis:
Table of content
Abstract ………. 2
Introduction ………...3
Literature Review ……….………..6
Previous Empirical Results……….6
Evidence From the Chinese Stock Market………..7
Underlying Reasons for Holiday Effects………....8
Abstract
This paper studies the impact of long holidays on stock returns at the Shanghai and Shenzhen Stock Exchanges (SSE and SZSE). It is of interest to investigate the Chinese stock markets, since the markets have undergone rapid development in the last decade and have become the forth largest financial market in the world. Besides, I focus on long-holiday effect, since it is unique for China to have long holidays after 2000. In addition, to mitigate the bias from the heteroskedasiticity problem, the GARCH model is employed in this paper. The results indicate that there is a pre-holiday effect in the Chinese stock markets with the stronger effect at the Shanghai stock exchange for the Lunar New Year. The evidence of the post-holiday effect is weak, in particular one corrects for the closing returns during the long holidays and for the high risk after the long holidays.
1. Introduction
Since the seminal work of Fama (1965), the so-called calendar effects have been one of the most extensively investigated subjects in the finance literature. Many studies have provided evidence of calendar anomalies on daily stock returns, including day-of-the-week effects (e.g. Cross, 1973; Keim and Stambaugh, 1984; Board and Sutcliffe, 1988; Chang et al., 1993; Mitta and Jain, 2009), January and other month effects (e.g. Keim, 1983; Jaffe and Westerfield, 1985; Mills and Coutts, 1995; Cheung and Coutts, 1999), and within-month effects (e.g. Ariel, 1987; Ogden, 1990; Kohers and Patel, 1999; Mookerjee and Yu, 1999).
Apart from these fixed calendar anomalies, holiday effects have attracted attention in academia as well (e.g. Pettengill, 1989; Ariel, 1990; Vergin and McGinnis, 1999). These studies indicate that stock returns before and after holidays are abnormal. Pettengill (1989) finds a strong pre-holiday effect in the US stock markets. The pre-holiday return is significantly higher than that of other trading days. This evidence is supported by other studies of Ariel (1990) and Brockman & Michayluk (1998). In addition, certain evidence of a post-holiday effect is discovered. Rogalski (1984) finds that the significant differences in post-holiday returns by weekday and firm size. Pettengill (1989) finds that post-holiday return is high for small-sized firms. On the other hand, Lakonishok & Smidt (1989) provide evidence that the post-holiday return is negative, although it is not statistically significant. The holiday effects are not unique to the U.S. markets. A large amount of international evidence has been found in the last two decades as well (e.g. Cadsby & Ratner 1992; Tonchev and Kim, 2004; Yakob et al., 2005).
This paper aims to explore the holiday effects in the Chinese stock market by investigating both Shanghai and Shenzhen Stock Exchanges. It is of interest to investigate the Chinese stock markets for several reasons. First of all, Chinese stock markets have undergone tremendous development in the past decade and it also has become the fourth largest financial market in the
world1. Secondly, some recent empirical studies show that the calendar effects tend to disappear
(Steeley, 2001; Coutts and Sheikh, 2002). Although the Chinese stock market has enjoyed rapid development, the market is less mature compared with most of the developed and well-structured stock markets like those of the US, UK and Japan. Therefore, this paper may also provide evidence of holiday effects in developing stock markets and in particular a relatively young
1 The Shanghai and Shenzhen Stock Exchanges started operation by issuing eight and five stocks,
market. Thirdly, researchers like Lakonishok and Smidt (1988), Schwert (2002) state that the finding of stock seasonality can be caused by data mining. One type of data mining is that same data set or similar data sets are repeatedly used in different studies. This concern is also one of the motivations to investigate the Chinese stock markets, since little research has been done in showing the holiday effects in the Chinese stock markets. Hence, the analysis on this new data set can add a valuable and independent sample to the body of research. Although previous studies (Mookerjee and Yu, 1999; Gao and Kling, 2005; Cai et al., 2006) provide certain evidence in terms of stock seasonality in the Chinese stock markets (e.g. day-of-the-week effect, monthly effect, turn-of-the-year effect), no extensive study with regard to the holiday effects has been done. Therefore, this study aims to fill in this gap by investigating both Shanghai and Shenzhen Stock Exchanges with a recent data set.
Furthermore, in terms of the choice of holidays, this paper mainly focuses on the long holidays, namely, the Lunar New Year, the Labor Day and the National Day. These three holidays lead to stock market closings for at least seven days2. China is unique in having long-period holidays which are originally meant to stimulate domestic consumption. Therefore, this paper also adds to the literature by investigating long-period holiday effects.
This paper investigates the returns on the Composite Indices at the Shanghai and Shenzhen Stock Exchanges. Furthermore, it is noteworthy that two types of stock returns are used to investigate
holiday effects, namely, open-to-close and close-to-close return.3 Finally, previous studies
indicate that stock seasonality may be attributed to statistical errors like estimated bias due to heteroskedasiticity (e.g. Christie, 1982; Pettengill, 2003). To mitigate this bias, the GARCH model is employed so that errors from heteroskedasticity are controlled for.
The remainder of the paper is structured as follows. In the second part, a comprehensive literature review regarding holiday effects is given. Following that, a brief descriptive of the data used in this paper and the features of the holidays will be presented. The fourth part encompasses the research design and methodology implemented to test the holiday effect. After that, I will report and discuss the main findings of the regression models. The last part contains a summary and the conclusion.
2
Most previous studies analyze holiday effects by investigating the holidays of periods of no more than three days, which are defined as short-holidays in this paper.
2. Literature Review
2.1. Previous Empirical Results
There is a long history for discussing and exploring holiday effects in academia. Previous studies show that the stock returns of the trading days just before and after the holiday are abnormal. This evidence can be traced back to the 1960s for the US stock markets. Merrill (1965) finds abnormal preholiday returns on U.S. stocks by investigating the DJIA (Dow Jones Industrial Average). Similarly, Fosback (1976) reports high preholiday returns in the S&P 500 index. In the late nineties, more evidence was documented for the pre-holiday effects. Ariel (1990) indicates that the mean of pre-holiday returns is statistically higher than that of other trading days return by investigating DJIA from 1963 to 1982. Still focusing on DJIA, Lakonishok and Smidt (1988) add to the literature by extending the data period (from 1897 to 1986) and find consistent results on the pre-holiday effects. Specifically, the pre-holiday rate of return is 23 times larger than the regular daily rate of return, and holidays account for about 50 percent of the price increase in the DJIA. Furthermore, Pettengill (1989) provides strong supporting evidence for the pre-holiday effects in the US markets by controlling for size effects and reports that, for both large and small firms, there are unusually high returns for pre-holiday trading days
The pre-holiday effects are not unique for the US stock market. Cadsby & Ratner (1992) perform a comprehensive study on holiday effects in eight stock markets. They demonstrate that there is a pre-holiday effect in the Canadian, Hongkong and Japanese stock markets. In the other markets (UK, France, Italy, Switzerland and West Germany), the pre-holiday return is higher than that of other trading days, but not statistically significant. Contrary to the findings of Cadsby & Ratner (1992), Kim and Park (1994) provide strong evidence of pre-holiday effects both in the UK and Japanese stock markets. Instead of only focusing on the FT-SE Index, Mills & Coutts (1995)
investigate holiday effects in the UK stock markets by looking into different types of indices4 and
reveal that in finance and consumer groups5, pre-holiday return is higher than other trading days.
Whereas in the industrial and “other” groups indicate that pre-holiday returns are lower than those of other trading days. Besides, evidence of a pre-holiday effect is also shown in stock markets of Hongkong, Australia, Athens and Spain (see more from Yakob et al., 2005; Marrett and Worthington, 2009; Mills et al., 2000; Meneu and Pardo, 2004).
4 Mills & Coutts (1995) investigate various indices including FTSE 100, Mid 250 and 350 indices, and 29
industry baskets.
5
In addition to the evidence of pre-holiday effects found in the developed financial markets, a growing number of researchers have turned their interest to the developing stock markets to see if the same pattern exists in less mature markets. Arumugan (1999) investigates the Indian stock markets by studying the Senative Index during the period 1979 to 1997 and reports a strong pre-holiday effect except for the period 1985 to 1991. A similar result holds for the Thailand stock markets (Holden et al., 2005), where the pre-holiday effects exist with an exception the period 1996 to 1998. Tonchev and Kim (2004) try to extend the evidence to East European stock market, one of the important newly-developing financial centers. However, they find no holiday effect by investigating stock indices from the stock exchanges in the Czech Republic, Slovakia and Slovenia.6
In contrast to the rather consistent empirical evidence of a pre-holiday effect, there is no full consensus on post-holiday effects. Several studies show that there is an abnormal high stock return for the day just after the holiday. Rogalski (1984) finds significant differences in post-holiday returns by investigating DJIA during the period 1974 to 1984. French (1980) studies the S&P 500 for a longer period (1953-1977) and indicates higher than expected returns for post-holiday trading days for every weekday except Tuesday. Pettengill (1989) shows that returns for post-trading days are high only if they occur at the end of the week. Like the pre-holiday effect, a post-holiday effect is not unique for the US stock market. For instance, Yakob et al. (2005) demonstrate significantly positive post-holiday returns for the stock markets in Japan and Singapore. Holden et al. (2005) also report positive post-holiday stock returns for Thailand, except for the period 1996-1998. On the other hand, Lakonishok & Smidt (1989) provide evidence that the post-holiday return is negative, although it is not statistically significant. Other studies, such as Arumugan (1999), Tonchev and Kim (2004) and Marrett and Worthington (2009) either come to inconclusive results or find no post-holiday effect at all. In short, the evidence from the current literature on a post-holiday effect is ambiguous.
2.2. Evidence From The Chinese Stock Markets
Evidence of stock seasonality in the Chinese stock markets can also be found in academic field.
However, due to the comparatively short history of Chinese stock markets7 and due to limited
6 Tonchev and Kim (2004) employ PX-50 and PX-D for Czech stock market; SAX for Slovak stock
market and SBI-20 and SBI-20NT for Slovenian stock market.
7
access to data, the number of studies with regard to stock seasonality is still limited. Mookerjee
and Yu (1999) investigate the day-of-the-week effect, turn-of-the-month effect8, monthly effect9
and the turn-of-the quarter effect10 in both Shanghai and Shenzhen Stock markets during the
initial trading period from 1990 to 1994. They find that the highest daily returns for both exchanges occur on Thursday rather than Friday. They report that daily stock returns for the turn-of-the-month, the monthly pattern and the turn-of-the-quarter are less than returns on other days. Besides, Gao and Kling (2005) study stock seasonality for both Shanghai and Shenzhen Stock Exchanges and reveal that the return on Friday is reported to be significantly positive. Cai
et al. (2006) study the day-of-the-week effect for A shares and B shares at both exchanges and
find that average Monday returns from A share indices are significantly negative during the third and fourth week of a month. They also report that average Tuesday returns on most of the A share and B share indices are negative during the second week of the month. They conclude that even after controlling for spillover from international markets, day-of-the-week effects in the Chinese markets are significant. Finally, in a recent study, Ogunc et al. (2009) finds no significant weekend effect, which contradicts the findings by Cai et al. (2006).
2.3. Theoretical Underlying
Several theories and arguments have been put forward to explain the holiday anomalies. The first explanation for the pre-holiday effect is based on a systematic trading pattern. Keim (1989) suggests that the pre-holiday return may be due to movements from the bid to the ask price. Specifically, if a systematic trading pattern leads to clustering of bid and ask prices for the trading day just preceding holiday closings, then the stock return can be systematically biased. Ariel (1990) points out that pre-holiday effects can be attributed to short-sellers who desire to close short positions but not long positions in advance of holidays or, simply, to some clienteles which preferentially buy (or avoid selling) on pre-holidays. The second reason which may account for a pre-holiday effect is a “closing effect”, which means that the stock return is high at a market closing. This effect is derived form the findings of several studies which show that Friday return is high. Following this logic, since holidays also involve a market closing, it is reasonable to speculate that the pre-holiday effect is due to a “closing effect”. Another explanation of this phenomenon is coming from the viewpoint of behavioral finance, which tries to explain investor
8 As defined by Mookerjee and Yu (1999), turn-of-the-month is taken as the last day of the proceeding
month and the first three days of the current month.
9 Monthly effect is defined as the difference between the stock returns of the first nine days of the month
and the returns of the last nine days of the month according to Mookerjee and Yu (1999).
10
behavior by incorporating a psychological perspective. The core assumption of behavioral finance is that individual investors are bounded rational (contrary to the assumption of “unbounded rational”, according to standard finance theory) and their investment decisions are influenced by cognitive constraints. Moreover, the market stock return is influenced by aggregating the effect of individual investors. This provides the rationale for a high pre-holiday return, since investors are assumed to be consistently more optimistic about the future stock return due to a good mood before the holiday, which in turn leads to the rise of the stock price (Rystorm and Benson; 1989).
There are also several interpretations for post-holiday effects. French (1980) attributes the positive post-holiday return to the time diffusion effect. He acknowledges that post-holiday return involves the return for the non-trading days as well. For instance, if the holiday leads to a two-day holiday closing, post-holiday return actually stands for the return of the three-following holidays after closing. However, Lakonishok and Levi (1982) state that adjusting returns for the time diffusion effect does not eliminate the holiday effect. While the time diffusion effect hypothesis provides a rationale for high post-holiday returns, the theory of behavioral finance provides reasons for negative post-holiday returns. Investor decisions are subject to cognition limitations. Therefore, they tend to hold a less optimistic perception of the stock price due to a “blue” mood after the holiday. This influences their perception of the future movement of stock price in a more pessimistic way. Such reasoning is in line with the “Blue Monday” hypothesis that may explain the weekend effect.
In addition to the reasons cited above, holiday effects could also partly be attributable to other stock anomalies. The first one is the existence of a relationship with other calendar anomalies. Lakonishok and Smidt (1989), Ariel (1990) and Liano et al. (1992) are the first to explain the holiday effect by appealing to other calendar anomalies such as the day-of-the-week effects, the monthly effects11, and the turn-of-the-year effects. Their results indicate that the high returns observed on pre-holidays are not a manifestation of other calendar anomalies. Other studies show that holiday effect is interrelated with size effect. Rogalski (1984) notes that if a holiday falls on a Tuesday, Wednesday, or Thursday, the average two-day calendar return from close to close is greater for small firms than large firms. Similarly, Pettengill (1989) reports that post-holiday returns for large firms are lower on average than other returns; whereas, post-holiday returns for small firms are higher.
11
Lastly, it is worth noting that stock seasonality is likely to be attributable to statistical errors. To be specific, stock seasonality may arise from employing erroneous statistical models (Pettengill, 2003). The simple OLS linear model neglects the nature of most financial time series data that the variance of error terms is not constant (known as heteroskedasticity). For instance, volatility occurs in bursts, which means that large stock return is followed by high volatility, while small volatility usually arises when stock return is relatively low. Chen et al. (2002) find evidence that adjustments for heteroskedasticity reduce the weekday effect. However, Connolly (1989), working with equity securities, and Najand and Yung (1994), working with index futures, find that GARCH analyses fail to reject the hypothesis that average returns are equal across weekdays. For these reasons, evidence of a holiday effect could just be a manifestation of statistical errors and that one should take heteroskedasticity into account.
3. Data Description
In this paper, I investigate holiday effects in the Chinese stock markets during the period January 4, 2000 to December 28, 2007, from both Shanghai and Shenzhen Stock markets. Investigating both stock markets helps to provide a comprehensive picture of the Chinese stock markets. To capture the features of the overall performance of Shanghai and Shenzhen Stock Markets, SSE Composite (Shanghai Stock Exchange Composite) and SZSE Composite (Shenzhen Stock Exchange Composite) are used, which includes all the listed stocks (A stocks and B stocks) at both stock exchanges. The data are retrieved from DATASTREAM.
days (Thursday and Friday). Therefore, people in effect only have three days off for the holiday, but by working beforehand, they are able to enjoy a full week off (Golden Week). The length of the holiday however does not match with the duration of the stock market closing. The stock markets usually close for an even longer period of time for the reason that the stock markets always close on weekends. Therefore, referring to the example above, the stock markets close from the weekend before the “Golden Week” and reopen on Monday just after the “Golden Week”. This means that the stock markets close for 9 days. Starting from 2008, the length of Labor Day has been shortened to one day again. Therefore, only National Day holiday is considered as “Golden Week” after 2008.
Now, I turn to the analysis of the Lunar New Year. The first thing to note is that the occurrence of the Lunar New Year differs every year, since it is defined according to the Lunar calendar, not to the Gregorian calendar. To be specific, the Lunar New Year starts in the beginning of the 12th
lunar month and lasts till the mid of the first lunar month of the next year. Secondly, the length of
stock market closing due to this festival is not fixed. As a matter of fact, unlike Labor Day and National Day, there is no regulation on how long the stock markets should close. Instead, stock exchanges have the authority to decide the exact length of market closing, which leads to the situation that the length of the market closing differs from one year to another. The details of market closing for the Lunar New Year, Labor Day and National Day are shown in table 2 in the
Appendix.
Since this paper focuses on the long-holiday effects, the Lunar New Year, Labor Day and National Day are chosen as the research objectives. Besides, for the reason that the concept of “Golden Week” is introduced in 2000 and also because the length of Labor Day is shortened again after 2008, this paper focuses on the data period 2000 to 2007.
holiday, days after a holiday and all the other days. Pre-holiday return is defined as the daily return occurring on the trading day just before the closing day, post-holiday return is defined as the daily return occurring on the trading day just following a closing day and non-holiday return is daily return for the other trading days. To be specific, the natural log of the relative price is computed for the daily intervals to produce a time series of continuously compounded returns. Open-to-close and close-to-close basis return is defined as follows:
Ropen-to-close= log (popen,t/ pclose, t)*100, where popen,tand pclose, trepresent the opening and closing
price index at t, respectively.
Rclose-to-close = log (pclose,t/ pclose, t-1)*100, where pclose,t and pclose, t-1 represent the closing price
index at t and t-1, respectively.
For the summary of descriptive statistics of the daily returns, I refer to table 3 and table 4 in the
Appendix.
Table 3 reports the comparison of pre-holiday return and non-holiday return for the Shanghai and Shenzhen Stock Exchanges. The table is separated into two panels. Panel A uses close-to-close stock return and Panel B uses open-to-close stock returns. As illustrated in Panel A, the expected return for the trading day just before market closing is surprisingly high for both Shanghai and Shenzhen Stock Exchanges. Table 3 shows that the average pre-holiday return at the Shanghai Stock Exchange is 0.503, which is almost ten times as large as the non-holiday return (0.054). A similar result is found for the Shenzhen stock market. The pre-holiday return is on average 0.536 compared with 0.061 for non-holiday returns. I also separately calculated the return for the day just prior to the Lunar New Year, Labor Day and National Day. The result shows pre-holiday returns for all the three holidays are higher than the non-holiday return. This indicates that the high pre-holiday return is not the result of one particular holiday. What also deserves attention is that the standard deviation for pre-holiday return is lower compared with non-holiday trading days for both markets. This suggests that the high pre-holiday return cannot be contributed to the high risk. Panel B shows that the results for open-to-close stock returns are similar to the close-to-close of Panel A. The Shanghai and Shenzhen stock markets enjoy high pre-holiday returns and this phenomenon is consistent across all the holidays. In short, results in table 3 provide evidence that there is strong pre-holiday effect in both stock markets.
open-to-close return. Panel A shows that an average post-holiday return is higher in both Shanghai and Shenzhen stock markets. Furthermore, this phenomenon is consistent across all the holidays. What is different form the results of table 3 is that standard deviation for the day just after the holiday is higher compared to other trading days. This indicates that high post-holiday returns may result from higher risk. Panel B reports similar results as Panel A. However, here the post-holiday return is not consistently high across the holidays. As shown in the last three rows in Panel B, post-holiday returns for Labor Day and National Day are negative both in Shanghai and Shenzhen stock markets.
4. Methodology
Observations on stock seasonality may be caused by statistical errors due to heteroskedasticity. Christie (1982) presents that the volatility in the rate of returns on equity and the current value of the equity is negatively correlated. Fama (1965) and French and Roll (1986) indicate that variances are higher following holidays (including weekends) than on other days. Ignoring such non-trading day effects would lead to specification bias in the volatility model, and hence bias the estimates. Autoregressive conditional heteroskedasticity (ARCH) models and the GARCH models allow variances to be dependent upon its own lags and these approaches are now widely used in researching effects based on high frequency time series data. Therefore, to control for the bias triggered by the inconstant volatility, this paper employs the GARCH model in testing the holiday effects.
1. Basis Equation
As discussed above, the holiday effect may be interrelated with stock seasonality. Among these calendar anomalies, the day-of-the-week effects are showed to be possibly related to issues to holiday effects. Lakonishok and Smidt (1989) and Ariel (1990) therefore incorporate day-of-the-week effects when testing holiday effects. To ensure that the holiday effect is not a manifestation of day-of-the-week effects, dummies for every weekday are taken as control variables in my test. The day-of-the-week effect is tested by employing a GARCH (1, 1) model. The model is specified as follows:
1 2 3 4 5
...(1)
t t
R
Mon
Tue
Wed
Thu
Fri
2 2 2
0 1 1 1...(2)
t t t
Equation (1) is the parameter estimation equation, where Rt represents the daily Composite
open-to-close returns. Mon, Tue, Wed, Thu and Fri account for the dummies for each weekday. For instance, when it is Monday, the dummy Mon takes the value of 1, and 0 otherwise.
1,2
,
3,
4and
5are coefficients which indicate the expected return for Monday, Tuesday,Wednesday, Thursday and Friday, respectively.
tis the error term.Equation (2) is the conditional variance model, where
t2is known as the conditional variancesince it is dependent on relevant information from the past.
0 is the long-term averagevariance.
1 t21provides information about volatility during the previous trading day.
t21 presents the fitted variance from the model during the previous trading day and it is a summary measure of all of deriving from the estimations before the trading day.2. Holiday effects
The holiday effects are formally tested by employing the model discussed in this subpart. Similar with the model used to test the day-of-the-week effect, a GARCH Model is implemented. Furthermore, in the equations, PRE and POST are added as two dummies to capture the pre- and post-holiday effects. To be precise, PRE takes the value of 1 if the trading day is just one day ahead of the holiday closing; otherwise it takes the value of 0. POST takes the value of 1 if the
trading day is immediately followed by the holiday closing; otherwise it takes the value of 0.
1and
2are the coefficients to be estimated. The dummies for each weekday are taken as controlvariables to control for the day-of-the-week effect. In the conditional variance model, PRE and POST are added since the variance of these two days are expected to be different from the variance of other trading days as noted above in table 3 and table 4. The following model is specified:
1 2 3 4 5 6 7
....(3)
t t
R
PRE
POST
MON
TUE
WED
THU
FRI
2 2 2
0 1 1 1 2 3 ...(4)
t t t PRE POST
5.1. Close-to-close Return
Following the method used by other researchers in investigating the holiday effect, I will first test the holiday effects by employing close-to-close stock return. The result is shown in table 5. Table 5 reports the results for both Shanghai and Shenzhen stock markets. For each market (e.g. Shanghai), the basis equation will be shown first in Column (1). Then the pre- and post-holiday dummies are included in the models and the results are illustrated in Column (2). The same is applied to the Shenzhen stock market. Another point worth mentioning is that the table is vertically separated into two parts. The first part (Parameter Estimation) presents the estimated coefficients of the control variables and independent variables. In the second part (Variance
Estimation), the coefficients of the variance are presented. I will focus my discussion on the first
part.
First, let us take a close look at the empirical results in the Shanghai stock market. As indicated in the Column (1) in table 5, the coefficient for Tuesday is 0.17 and it is statistically significant, which indicates that the index return on Tuesday is significantly positive. Another statistically significant coefficient is for Thursday; however the sign is negative, which implies on average a negative stock return on Thursday. This contradicts with the previous research conducted by Mookerjee and Yu (1999). They find positive and significant high stock returns on Thursday. The magnitudes of coefficients for the other weekdays are not statically significant. Moving to
Column (2), we see that a strong pre-holiday effect is found at the Shanghai Stock Exchange. The
coefficient of PRE is 0.551 and the magnitude is statistically significant at the 1% level. It demonstrates that the pre-holiday return is on average 0.551% higher on the day before a long holiday, compared other trading days. The post-holiday effect is however relatively weak. Although the coefficient is positive, it is insignificant, implying that the average stock return following market close is higher, but not significantly higher than the returns of other trading days.
Now, we turn to the analysis of the Shenzhen stock market. The Shenzhen stock market presents a similar result as the Shanghai stock market except for the post-holiday effect. The results in
Column (3) are in the same direction as Column (1). We also observe statistically significant
post-holiday effect is found in the Shenzhen stock market. The coefficient of the post-holiday is 0.508 and significant, showing that the post-holiday return on average is statistically higher than the returns of other trading days. Lastly, let us take a look at the result of the variance estimation. It is shown that all the coefficients in the variance estimation are statistically significant for both stock markets, which means that the volatility is not constant; instead, it is correlated with previous errors and variance. Besides, the coefficient for the POST is positive and significant, which is in line with the findings by Fama (1965) and French and Roll (1986), who find that the variance of the days after weekend or a holiday is higher than that of other trading days.
5.2. Open-to-close Return
The results shown in table 5 are subject to one problem. By using close-to-close stock returns, the post-holiday returns captures not only the return for the trading day just following the holidays but the return during the holidays as well. As stated by French (1980), stock returns after weekends or holidays might be higher than that of other trading days, since the returns also cover the non-trading period. To avoid this bias and to capture the real trading behavior on the day just following the holiday, I use open-to-close returns to test the holiday effects again, especially to check the post-holiday effects. The empirical results are shown in table 6 in the Appendix.
By examining the evidence in the Shanghai stock market, similar conclusion can be drawn. With regard to the day of the week dummies, Tuesday enjoys a significantly positive return and negative return is found for Thursday. In terms of the holiday effects, the Shanghai stock market witnesses a strong pre-holiday effect and no post-holiday effect is reported. All these findings are in line with the results shown in table 5.
Turning our attention to the Shenzhen stock market, we find a crucial difference on the post-holiday effect. As shown in the Column (4) in table 6, the coefficient of post-holiday is 0.067 and is insignificant, which means no post-holiday effect is found in the Shenzhen stock market by using open-to-close return. This finding demonstrates that the evidence of the strong post-holiday
effect found in table 5 is mainly due to the non-trading period13. Other results found in table 6 are
consistent with those in table 5. The coefficient for pre-holiday is significantly positive, which implies that a strong pre-holiday effect still exits in the Shenzhen stock market and it can be attributed to the trading behaviors of the last trading day before the holiday.
13
5.3. Robustness Test
As shown in table 5 and table 6, a strong pre-holiday effect is found both in the Shanghai and Shenzhen stock markets. It is worth to note that the coefficient for the pre-holiday in the variance equation is significantly negative, which indicates that there is less risk involved in a pre-holiday. This is in line with the results in table 3. The standard finance theory suggests that stock return is influenced by risk, namely, higher risk leads o higher stock return. However, in my precious estimates, I did not control for these effects and it means that it is likely that by controlling for the risks of pre- and post-holiday returns, the estimations of the pre- and post-holiday effects will differ. In fact, if the risk effects are significant, I expect that the pre-holiday effects will become larger than measured in table 6 and the post-holiday effect becomes smaller. In this subpart, I therefore use the GARCH-in-mean model to test the holiday effect again.
GARCH-in-mean allows the stock return to be influenced by its risk by adding the conditional variance of stock return into the conditional mean equation. The model is specified as follows:
2
1 2 3 4 5 6 7 1 ...(5)
t T T
R PRE POST MONTUEWEDTHU FRI
2 2 2
0 1 1 1 2 3 ...(6)
t t t PRE POST
In equation (5),
is the estimated coefficient which represents the influence of the conditionalvariance on the conditional mean. Rt represents the open-to-close stock returns. Other variables
and coefficients remain the same as discussed in the methodology part.
In addition, the positive post-holiday return for the Lunar New Year reported in Panel B of table 4 triggers the additional test in this subpart. To be specific, on average post-holiday return for the Lunar New Year is higher, compared with the post-holiday returns for Labor Day and National Day in both stock markets. Moreover, I also find that high post-holiday return for the Lunar New Year is coming with a high standard deviation. Therefore, it is of interest to employ a GARH-in-mean model to check whether post-holiday return for an individual holiday is statistically different when controlling for the risk. The model is specified as follows:
1 2 3 4 5 6
2
1
...(7)
t
T T
R
PRELNY
POSTLNY
PRELAB
POSTLAB
PRENAT
POSTNAT
MON
TUE WED THU
FRI
2 2 2 0 1 1 1 2 3 4 5 6 7...(8)
t t t
PRELNY
POSTLNY
PRELAB
POSTLAB
PRENAT
POSTNAT
trading day before the Lunar New Year, Labor Day and National Day and 0 otherwise. Correspondingly, POSTLNY, POSTLAB and POSTNAT are dummies which take the value of 1 if it is, respectively, the trading day after the Lunar New Year, Labor Day and National Day and zero otherwise.
is the estimated coefficient which represents the influence of the conditionalvariance on the conditional mean. Rtrepresents the open-to-close stock returns. For the results, I
refer to table 7 in the Appendix.
Table 7 shows that the coefficients for
t21 are positive and statistically significant in both markets, which implies that increased risk, given an increase in the conditional variance, leads to a rise in the mean return. As expected, after applying for the correction for the lower risk rises the coefficient for the pre-holiday effect. For the Shanghai stock market, the coefficient increases from 0.523 in table 6 to 0.563 in table 7 and for Shenzhen from 0.585 to 0.656, respectively. Also in line with the expectation, correction for the higher risk decreases the coefficients for the post-holiday effect. For the Shanghai stock market, the coefficient decreases from -0.126 in table 6 to -0.205 and for Shenzhen from 0.067 to -0.026, respectively. In short, evidence in table 7 supports my expectation that after controlling for the risk, the pre-holiday effect becomes stronger and the post-holiday effect becomes even weaker.Column (2) and Column (4) in table 7 report the individual holiday effect in the Shanghai and
Shenzhen stock market. We find from a comparison of for example Column (1) and Column (2), that all individual pre-holiday effects are positive. However, only the Lunar New Year pre-holiday effect is significant. In addition, as we can see from Column (2), the coefficients of the post-holiday returns for all the three holidays are negative and insignificant. This evidence supports the belief that the positive post-holiday return (1.157) for the Lunar New Year shown in Panel B of table 4 is resulted from the high risk. The return turns negative when risk is controlled for. The results on the Shenzhen stock market shown in Column (4) are similar as the Shanghai stock market, providing evidence of negative but insignificant post-holiday returns for all the three holidays, except for the fact that now the coefficient of the pre-holiday effect of the Lunar New Year is not significant.
Table 5 and table 6 show evidence of day-of-the-week effects in the Chinese stock markets. First of all, both stock markets witness significantly positive stock return on Tuesday. This finding is also consistent for two types of stock return. Secondly, a strong Thursday effect is found. As we have discussed above, the expected stock return on Thursday is negative and statistically significant. It has to be noted that this phenomenon does not disappear even after taking holiday effect into account. This finding is interesting since it contradicts the findings by Mookerjee and Yu (1999) which states that Thursday return is positive. I suggest that the contradictory results may be due to the different data period. Mookerjee and Yu use data from 1990 to 1994, which is the beginning period of the Chinese stock markets. However, this paper implements more recent data covering the period 2000 to 2007. Compared with the current stock markets with larger market capitalization, higher trading turnover more listed stocks and better market regulation, the stock markets at its beginning period are in its infancy with regard to limited listed stocks and relatively poor legislation. Therefore, it is reasonable to draw different conclusion for different periods.
Now, let us turn to the two variables in the model. I find that there is a positive holiday-effect in the Chinese stock markets. This is consistent across both stock markets and also across close-to-close as well as open-to-close return. The results also show that after adjusting for the risk, the pre-holiday effect still exists. The possible interpretation for this effect is the argument based on behavioral finance. These three holidays have a great influence on people’s daily lives in China. The Lunar New Year is a festival as important for Chinese people as Christmas for Western People. As the Lunar New Year is approaching, people completely clean their houses, buy necessities and decorate the door panels with Spring Festival couplets, wishing good luck and a bright future. In all, people are immersed in happiness and festival atmosphere. People also attach great importance to Labor Day and National Day after the introduction of the “Golden Week”. Since people could have the full week off, they often make schedules in advance for their traveling during the holidays or plan to visit their relatives. Furthermore, various promotions in big shopping malls to attract consumers start just before these “Golden Weeks”. As explained above, since people are in a good mood as the holidays are approaching, individual investors are more likely to hold positive perception of stock prices, which lead to a higher and pre-holiday stock return.
both types of stock returns. For the Shenzhen stock exchange, strong post-holiday effect is found by using close-to-close stock return. However, this effect disappears when implementing open-to-close return. One explanation could be provided for this phenomenon. Close-to-close post-holiday return actually captures not only the return on the day just following the holiday, but the return for the non-trading period as well. This could lead to one possibility that a post-holiday effect is just a manifestation of the same stock return during the non-trading period (time diffusion effect put forward by French (1980)). The disappearance of post-holiday effect by using open-to-close stock return validates this conjecture, since open-to-close post-holiday return is a better indicator for the real trading behavior on the post-holiday day. If we, finally, controls for the additional risk after the holiday, the post-holiday return even becomes on average negative, though not statistically significant. This finding implies that the evidence of post-holiday effect in the Shenzhen stock market is weak and lacks robustness.
In addition, evidence of individual holiday effect shown in table 7 shows that after correction for risk. The pre-holiday effects are still positive, but the effects are only significant for the Lunar New Year pre-holiday effect at the Shanghai Stock Exchange. The post-holiday returns for all the three holidays are negative, though not significant. The negative post-holiday return for individual holiday provides further evidence that the post-holiday effect is weak even if we take a close look at each individual holiday.
7. Conclusion
I investigate holiday effects for both the Shanghai and Shenzhen Stock Exchanges during the period January, 2000 to December, 2007. Unlike most of the previous studies which only study close-to-close return, this paper employs two types of stock return, namely, close-to-close and open-to-close return to better capture the feature of long-holiday effects. This paper focuses on long-period holidays, namely the Lunar New Year, Labor Day and Nation Day, which is a unique feature of China and which allows to study long-holiday effects in China. Research on the long-holiday effect is also a supplement to the current literature regarding holiday effect, since the majority of the literature studies short-holiday effect only. Furthermore, the GARCH model is implemented in this paper to control for the statistical errors derived from heteroskedasticity and GARCH-in-mean model is employed to correct for difference in risks pre- and post-holidays.
different types of stock returns. Besides, pre-holiday effects do not disappear after the adjustment for risk. This result is in line with the, which report high pre-holiday returns in most of the developed stock markets (Lakonishok and Smidt, 1989; Ariel, 1990; Brockman and Michayluk, 1998). However, only for the Lunar New Year at the Shanghai Stock Exchange, I find a positive and significant coefficient for the pre-holiday effect. In absence of trading cost, only during this day are suggested to buy at the opening of the stock exchange and sell at the close.
With regard to the post-holiday effect, we find no evidence of post-holiday effect in the Shanghai stock market. For the Shenzhen Stock Exchange, post-holiday return is significantly higher than other trading days when using close-to-close stock return. However, this effect disappears when using open-to-close stock return instead. This can be explained by the fact that close-to-close stock return is not a good manifestation of the real trading behavior on the day just following the holiday, since it includes the return during the non-trading period as well. Therefore, the significantly high coefficient of post-holiday is just a manifestation of high stock return during the holiday. Moreover, the post-holiday effect becomes even negative if we correct for the higher risks in the post-holiday trading day and this evidence is consistent across all the holidays when I test the post-holiday effects for the individual holiday. In brief, evidence of the post-holiday effect is weak and inconsistent and does not provide trading incentives.
References:
Arumugan, S. (1999) Focus on high holiday returns before holidays: New evidence form India,
Journal of Financial Management and Analysis, 12-2, 69-84.
Ariel, R. (1987) A monthly effect in stock returns, Journal of Financial Economics, 18, 161-174. Ariel, R. A. (1990) High stock returns before holidays: existence and evidence on possible causes,
Journal of Finance, 45, 1611-1626.
Board, J. L. G. and Sutcliffe, C. M. S. (1988) The weekend effect on UK stock market returns,
Journal of Business Finance and Accounting, 15, 1611-1626.
Brockman, P. and Michayluk D. (1998) The persistent holiday effect: additional evidence,
Applied Economics Letters, 5, 205-209.
Cadsby, C. B. and Ratner, M. (1992) Turn-of-month and stock returns: Some evidence, Journal of
Banking and Finance, 16, 497-509.
Cai, J., Li, Y. and Qi, Y. (2006) The day-of-the week effect new evidence from the Chinese stock market, The Chinese Economy, 39, 71-88.
Chang, E. C., Pinegar, J. M. and Ravichandran, R. (1993) International evidence on the robustness of the day-of-the-week effect, Journal of Financial and Quantitative Analysis, 28, 497–513.
Chen, C., Lee C. and Wang, A. (2002) A note on stock market seasonality: the impact of stock price volatility on the application of dummy variable regression model, The Quarterly
Review of Economics and Finance, 42, 155-162.
Cheung, K. C., Coutts, J. A. (1999) The January effect and monthly seasonality in the Hang Seng index: 1985–1997, Applied Economics Letters, 6, 121-123.
Christie, A. A. (1982) The stochastic behavior of common stock variances: value, leverage and interest rate effects, Journal of Financial Economics, 10, 407-32.
Connolly, R. (1989) An examination of the robustness of the weekend effect, Journal of
Financial and Quantitative Analysis, 24, 133-170.
Coutts, J. A. and Sheikh, M. A. (2002) The anomalies that aren't there: The weekend, January and 0pre-holiday effects on the all gold index on the Johannesburg Stock Exchange 1987-1997,
Applied Financial Economics, 12, 863-871.
Cross, F. (1973) The behavior of stock prices on Mondays and Fridays, Financial Analysts
Journal, 29, 67-69.
French, K. R. (1980) Stock returns and the weekend effect, Journal of Financial Economics, 8, 55-69.
French, K. R. and Roll, R. (1986) Stock return variances: the arrival of information and the reaction of traders, Journal of Financial Economics, 17, 5-26.
Gao, L. and Kling G. (2005) Calendar effects in Chinese stock market, Analysis of Economics and
Finance, 6, 75-88.
Holden, K., Thompson, J. and Ruangrit, Y. (2005) The Asian crisis and calendar effects on stock returns in Thailand, European Journal of Operational Research, 163, 242-252.
Jaffe, J. and Westerfield, R. (1985) The weekend effect in common stock returns: the I nternational evidence, Journal of Finance, 40, 433-454.
Keim, D. B. (1983) Sized-related anomalies and stock return seasonality, Journal of Financial
Economics, 12, 13-32.
Keim, D. B. (1989) Trading patterns, bid-ask spreads, and estimated security returns: the case of common stocks at calendar turning points, Journal of Financial Economics, 25, 79-97. Keim, D. B. and Stambaugh, R. F. (1984) A further investigation of weekend effect in stock
returns, The Journal of Finance, 34, 819-835.
Kim, C. W. and Park, J. (1994) Holiday effects and stock returns: further evidence, Journal of
Financial and Quantitative Analysis, 29,145-157.
Kohers, T. and Patel, J. B. (1999) A new time-of-the-month anomaly in stock index returns,
Applied Economics Letters, 6, 115-120.
Lakonishok L. and Levi M. (1982) Weekend effects on stock return: a note, Journal of Finance,
40, 347-349.
Lakonishok, J. and Smidt, S. (1989) Are seasonal anomalies real? A ninety-year perspective,
Review of Financial Studies, 1, 403-25,
Liano, K., Marchand, P. H. and Huang, G. C. (1992) The holiday effect in stock returns: Evidence from the OTC market, Review of Financial Economics, 2, 45-54.
Marrett, G. J. and Worthington, A.C. (2009) An empirical note on the holiday effect in the Australian stock market, 1996–2006, Applied Economics Letters, 16, 1769-1772.
Meneu, V. and Pardo, A. (2003) Pre-holiday effect, large trades and small investor behavior,
Journal of Empirical Finance, 11, 231-246
Merrill, A. A. (1965). Behavior of Prices on Wall Street. New York: The Analysis Press.
Mills, T. C. and Coutts, J. A. (1995) Calendar effects in the London stock exchange FT-SE indices,
Mills, T. C., Siriopoulos C., Markellos, R. N. and Harizanis, D. (2000) Seasonality in the Athens stock exchange, Applied Financial Economics, 10, 137-142.
Mittal, S. K. and Jain, S. (2009) Stock market behavior: evidence from Indian market, The Journal of Business Perspective, 13, 19-29.
Mookerjee, R. and Yu, Q. (1995) Capital market reform on the road to a market oriented economy: the case of stock markets in China. Journal of Developing Areas, 30, 23-40.
Mookerjee, R. and Yu, Q. (1999) Seasonality in returns on the Chinese stock markets: the case of Shanghai and Shenzhen, Global Finance Journal, 10, 93-105.
Najand, M. and Yung, K. (1994) Conditional heteroskedasticity and the weekend effect in S&P 500 index futures, Journal of Business Finance and Accounting, 21, 603-612.
Ogunc, A. K., Nippani, S. and Washer, K. M. (2009) Seasonality tests on the Shanghai and Shenzhen stock exchanges: an empirical analysis, Applied Financial Economics, 19, 681-692.
Ogden, J.P. (1990) Turn-of-the-month evaluations of liquid profits and stock returns: a common explanation for the monthly and January effects, Journal of Finance, 55, 1259-1272.
Pettengill, G. N. (1989) holiday closing and security returns, Journal of Financial Research, 12, 57-67.
Pettengill, G. N. (2003) A survey of the Monday effect literature, Quarterly Journal of Business &
Economics, 42, 3-28.
Rogalski, R. J. (1984) New findings regarding day-of-the-week returns over trading and non-trading periods: a note, Journal of Finance, 39, 1603-1614.
Rystorm, S. and Benson, D. (1989) Investor psychology and day-of-the-week effect, Financial
Analysis Journal, 45, 75-78.
Schwert, G. W. (2002) Anomalies and market efficiency, Simon School of Business Working Paper No.FR 02-13. Retrieved on June 15, 2010 from:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=338080
Steeley, J. M. (2001) A note on information seasonality and the disappearance of the weekend effect in the UK stock market, Journal of Banking and Finance, 25, 1941-56.
Tonchev, D. and Kim, T. (2004) Calendar effects in Eastern European financial markets: evidence from the Czech Republic, Slovakia and Slovenia, Applied Financial Economics, 14, 1035-1043.
Vergin, R. C. and McGinnis, J. (1999) Revisiting the holiday effect: is it on holiday?, Applied
Financial Economics, 9, 477-82.
Yakob, N. A., Beal, D. and Delpachitra, S. (2005) Seasonality in the Asia Pacific stock markets.
Appendix
Table 1 Summary of the main studies regarding holiday effects. US Stock Markets (Panel A)
Author Stock Index Data period Pre-holiday effect Post-holiday effect Merrill (1965) Dow Jones
Industrial Average
1897-1965 Positive and significant
-French (1980) S&P 500 Index 1953-1977 - Positive for every
weekday except for Tuesday.
Rogaiski (1984) Dow Jones Industrial Average October 1,1974-April 30,1984 - Inclusive; The post-holiday effect differs across weekdays, size of the firm and types of stock return.14 Lakonishok & Smidt (1989) Dow Jones Industrial Average 1897-1986 Positive and significant
Negative but not significant
Pettengill (1989) S&P 500 Index and CRSP tapes(excluding 500 largest firms at the beginning of the year) July 1962-December 1986 Positive and significant Inconclusive (the result is
dependent on the size of the firm and the date the holiday falls on)
Ariel (1990) Dow Jones Industrial Average 1963-1982 Positive and significant -Brockman & Michayluk (1998)
NYSE and AMEX (1963 – 1993); NASDUQ (1972 -1993).
Positive and significant across all size-based and price-based portfolio
-Other Stock markets (Panel B) Cadsby
& Ratner
(1992)
Toronto Stock Exchange(January 3, 1975 to December 31,1987); Nikkei Index(January 5, 1979 to December 28, 1988); FT 500 Index(August 16, 1983 to June 13, 1988); Hang Seng Index; All Ordinaries Index for Australia; Banca Commerciale Index; Swiss Bank Corporation Industrials Index; Commerz-bank Index; Compagnie des Agents de Change General Index (January 2, 1980 to August 1, 1989)
Positive and significant return for stock markets in Canada, Hongkong and Japan. Positive but not significant effects in UK, France, Italy, Switzerland and West Germany.
-Kim & Park(1994) Nikkei-Dow and U.K. FT Index July 1, 1972-June 30,1987
Positive and significant for Japanese stock markets; positive but not
14
significant for UK stock markets. Mills & Coutts(1995) FTSE 100, Mid 250 and 350 indices, and industry baskets
1986-1992 Positive and significant for all indexes
-Arumugan (1999) Bombay Stock Exchange Senative Index April 4, 1979-March 31,1997
Positive across the whole time period.(positive and significant except for the sub-period 1985-1991) Inclusive(cont rary results for different sub-periods) Mills et al. (2000) Athens Stock Exchange General Index October 1986 to April 1997
Positive and significant
-Meneu and Pardo (2004) Spanish Stock Exchange; individual stocks from Telefo´nica, Banco Bilbao Vizcaya rgentaria (BBVA), Banco Santander Central Hispano (SAN), Repsol YPF and Endesa
January 1990 to December 2000.
Positive and significant across all the samples.
-Tonchev and Kim (2004)
Czech PX-50 and PX-D and the Slovak SAX (From 1 January 1999 to 18 July 2003); Slovenian SBI-20 and SBI-20NT (From on 4 July 2000 to 18 July 2003). No pre-holiday effect is found. No post-holiday effect is found. Yakob et al. (2005) Composite indices for Australia, China, Hong Kong, India, Indonesia, Japan, Malaysia, Singapore, South Korea and Taiwan
From the beginning of January 2000 to the end of March 2005
Positive and significant for stock markets in Hongkong and Australia.
Positive and significant for stock markets in Japan and Singapore. Holden et al. (2005) Thai Stock Market Index January 3,1995 –December 29, 2000
Positive except for the period May 29,
1996–September23, 1998
Positive except for the period May 29, 1996– September23, 1998 Marrett & Worthington (2009) All Ordinaries Index; Small Ordinaries index and ASX/S&P industry indices15 at Australian Securities Exchange. September 9, 1996 – November 10, 2006
Positive and significant return for all ordinaries and small ordinaries indexes. No evidence found for individual industries except for retailing. No post-holiday effect is found. 15
Table 2 Opening and Closing days of the stock markets for the Lunar New Year, Labor Day and National Day
The Lunar New Year Labor Day National Day
Year Pre-holiday Post-holiday Length of the market closed Pre-holiday Post-holiday Length of the market closed Pre-holiday Post-holiday Length of the market closed
2000 Jan.28 Feb.14 13 days Apr.28 May 8 9 days Sept.29 Oct. 9 9 days
2001 Jan.19 Feb.5 16 days Apr. 30 May 8 7 days Sept.28 Oct.8 9 days
2002 Feb.8 Feb.25 17 days Apr. 30 May 8 7 days Sept.27 Oct. 8 10 days
2003 Jan. 29 Feb. 10 11 days Apr. 30 May 12 11 days
Sept.30 Oct. 8 7 days
2004 Jan. 16 Jan.29 12 days Apr. 30 May 10 9 days Sept.30 Oct.8 7 days
2005 Feb.4 Feb.16 11 days Apr. 29 May 9 9 days Sept.30 Oct.10 9 days
2006 Jan.25 Feb.6 11 days Apr. 28 May 8 9 days Sept.29 Oct.9 9 days
2007 Feb.16 Feb.26 9 days Apr. 30 May 8 7 days Sept.28 Oct.8 9 days
Table 3 Descriptive statistics for the day before the long holidays
PRE is taken as the day before all the holidays; PRELNY is taken as the day before the Lunar New Year; PRELAB is taken as the day before the Labor Day and PRENAT is taken as the day before the National Day; Non-holiday represents the return for other trading days. Ropen-to-close= log (popen,t/ pclose, t)*100, where
popen,t and pclose, t represent the opening and closing index price at time t, respectively. Rclose-to-close = log (pclose,t/ pclose, t-1)*100, where pclose,tand pclose, t-1represent the closing index price at t and t-1, respectively.
Panel A: Comparison of close-to-close basis returns of stocks at Shanghai and Shenzhen Stock Exchanges for all trading days before holidays with those for all other trading days for composite index for returns computed between 4 January 2000 and 31 December 2007.
SH-Composite SZ-Composite
Number Mean Std. deviation Mean Std. deviation
Non-holiday 1877 0.054 1.487 0.069 1.567 PRE 24 0.503 1.033 0.584 1.039 PRELNY 8 0.747 0.931 0.843 0.814 PRELAB 8 0.834 0.924 0.428 1.066 PRENAT 8 0.429 1.251 0.480 1.277
Panel B: Comparison of open-to-close basis returns of stocks at Shanghai Stock and Shenzhen Exchanges for all trading days before holidays with those for all other trading days for composite index computed between 4 January 2000 and 31 December 2007.
SH-Composite SZ-Composite
Number Mean Std. deviation Mean Std. deviation
Table 4 Descriptive statistics for the day after the long holidays
POST is taken as the day after all the holidays; POSTLNY is taken as the day after the Lunar New Year; POSTLAB is taken as the day after the Labor Day and POSTNAT is taken as the day after the National Day; Non-holiday represents the return for other trading days. Ropen-to-close= log (popen,t/ pclose, t)*100, where popen,t
and pclose, trepresent the opening and closing index price at time t, respectively. Rclose-to-close = log (pclose,t/
pclose, t-1)*100, where pclose,tand pclose, t-1represent the closing index price at t and t-1, respectively.
Panel A: Comparison of close-to-close basis returns of stocks at Shanghai and Shenzhen Stock Exchanges for all trading days after holidays with those for all other trading days for composite index for returns computed between 4 January 2000 and 28 December 2007.
SH-Composite SZ-Composite
Number Mean Std. deviation Mean Std. deviation
Non-holiday 1877 0.054 1.487 0.069 1.567 POST 24 0.790 2.503 0.791 2.656 POSTLNY 8 1.542 3.148 1.954 3.328 POSTLAB 8 0.326 2.480 0.237 2.583 POSTNAT 8 0.249 1.740 0.180 1.764
Panel B: Comparison of open-to-close basis returns of stocks at Shanghai and Shenzhen Stock Exchanges for all trading days after holidays with those for all other trading days for composite index for returns computed between 4 January 2000 and 28 December 2007.
SH-Composite SZ-Composite
Number Mean Std. deviation Mean Std. deviation
Table 5: Regression on close-to-close Composite index returns at the Shanghai and Shenzhen Stock Exchanges, 2000-2007
Shanghai Shenzhen
Parameter Estimation Column(1) Column(2) Column(3) Column(4)
PRE 0.551* (0.225) 0.564** (0.185) POST 0.349 (0.235) 0.508* (0.237) MON 0.039 (0.054) 0.019 (0.056) 0.012 (0.056) -0.016 (0.058) TUE 0.170* (0.070) 0.164* (0.070) 0.184* (0.073) 0.178* (0.073) WED 0.104 (0.058) 0.101 (0.058) 0.077 (0.062) 0.076 (0.062) THU -0.140* (0.058) -0.133* (0.057) -0.172** (0.059) -0.168** (0.058) FRI 0.001 (0.062) -0.011 (0.063) -0.032 (0.065) -0.039 (0.066) Variance Estimation ARCH (1) 0.108** (0.010) 0.111** (0.011) 0.110** (0.011) 0.114** (0.011) GARCH(1) 0.878** (0.010) 0.867** (0.012) 0.876** (0.010) 0.867** (0.012) PRE -0.637** (0.202) -0.706** (0.170) POST 1.407** (0.320) 1.574** (0.326) Observations 1925 1925 1925 1925 R2 0.0026 0.0063 0.003 0.0067 Notes: 1) **p<0.01, *p<0.05.
2) Column (1) and Column (3) report the results by using the regressions shown as follows:
1 2 3 4 5
t t
R MonTueWedThuFri (Parameter Estimation) 2 2 2
0 1 1 1
t t t
(Variance Estimation); Column (2) and Column (4) report the results by using the regressions shown as follows: Rt 1PRE2POST 3MON4TUE5WED6THU7FRIt(Parameter Estimation), 2 2 2
0 1 1 1 2 3
t t t PRE POST
(Variance Estimation), where Rt represents close-to-close return. Dummy PRE takes the value of 1 if it is the trading day before the holiday and 0 otherwise. Dummy POST takes the value of 1 if it is the trading day after the holiday and zero otherwise. Dummy variables take the value of 1 if the day is, respectively, a Mon, a Tue, a Wed, a Thu and a Fri and 0 otherwise. t21 represents the variable for ARCH (1) and t21 represents the variable for GARH (1)
Table 6 Regression on open-to-close Composite index returns at the Shanghai and Shenzhen Stock exchanges, 2000-2007
Shanghai Shenzhen
Parameter Estimation Column(1) Column(2) Column(3) Column(4)
PRE 0.523** (0.202) 0.585** (0.190) POST -0.126 (0.228) 0.067 (0.223) MON 0.003 (0.055) 0.003 (0.056) 0.012 (0.056) 0.004 (0.058) TUE 0.134* (0.067) 0.133* (0.068) 0.153* (0.072) 0.150* (0.073) WED 0.064 (0.057) 0.064 (0.057) 0.067 (0.060) 0.065 (0.061) THU -0.133* (0.057) -0.129* (0.056) -0.162** (0.059) -0.157** (0.058) FRI 0.016 (0.060) -0.006 (0.061) -0.008 (0.063) -0.027 (0.064) Variance Estimation ARCH (1) 0.085** (0.007) 0.087** (0.010) 0.089** (0.010) 0.090** (0.011) GARCH(1) 0.898** (0.010) 0.892** (0.011) 0.897** (0.011) 0.892** (0.011) PRE -0.470** (0.209) -0.611** (0.206) POST 0.908* (0.283) 1.150** (0.297) Observations 1925 1925 1925 1925 R2 0.001 0.003 0.002 0.003 Notes: 1) **p<0.01, *p<0.05.
2) Column (1) and Column (3) report the results by using the regressions shown as follows:
1 2 3 4 5
t t
R MonTueWedThuFri (Parameter Estimation) 2 2 2
0 1 1 1
t t t
(Variance Estimation); Column (2) and Column (4) report the results by using the regressions shown as follows: Rt 1PRE2POST 3MON4TUE5WED6THU7FRIt (Parameter Estimation), 2 2 2
0 1 1 1 2 3
t t t PRE POST
(Variance Estimation), where Rtrepresents
open-to-close return. Dummy PRE takes the value of 1 if it is the trading day before the holiday and 0 otherwise. Dummy POST takes the value of 1 if it is the trading day after the holiday and zero otherwise. Dummy variables take the value of 1 if the day is, respectively, a Mon, a Tue, a Wed, a Thu and a Fri and 0 otherwise. t21 represents the variable for ARCH (1) and t21 represents the variable for GARH (1)
Table 7 Robustness test on open-to-close return at the Shanghai and Shenzhen Stock Exchanges, 2000-2007
Shanghai Shenzhen
Parameter Estimation
Column(1) Column(2) Column(3) Column(4)
