Seasonal Anomalies on the
Chinese stock market
-- A study of seasonality effects in
Shanghai and Shenzhen stock exchange
Groningen, August 2009
Rijksuniversiteit Groningen
Faculty of Economics and Business
MScBa Finance
Author: Bing Chen
Student Risk & Portfolio Management
Contact info: B.Chen@student.rug.nl
Abstract
Preface
It is my pleasure to thank those who helped me to make this thesis possible. My profound appreciation and sincere thanks go to, Drs. Marc M. Kramer, my supervisor who was always encouraging me and giving useful feedbacks on my thesis from the preliminary to the concluding level. I would also like to make a special reference to my friend Mr. Raven and Kwin who read my thesis and gave me critical reflection on improvement.
Table of contents
PREFACE... 3
1. INTRODUCTION... 5
2. THEORETICAL FRAMEWORK ... 6
2.1EFFICIENT MARKET HYPOTHESIS AND CALENDAR ANOMALIES... 6
2.2THE DAY OF THE WEEK EFFECT... 7
2.3THE MONTH OF THE YEAR EFFECT... 10
2.4THE HOLIDAY EFFECT... 14
2.5RESEARCH QUESTION... 19
3. RESEARCH DESIGN ... 19
3.1METHODOLOGY... 19
The day of the week and the month of the year effect ... 19
The holiday effect ... 20
3.2DATA COLLECTION... 23
4. EMPIRICAL RESULTS ... 27
4.1THE DAY OF THE WEEK EFFECT... 27
4.2THE MONTH OF THE YEAR EFFECT... 33
4.3HOLIDAY EFFECT... 37
5. CONCLUSIONS AND RECOMMENDATIONS... 43
1. Introduction
Stock prices are shown to have various seasonal anomalies in stock markets in many countries. The most documented seasonal effects found in previous studies are the day-of-the-week effect, the January effect and the Pre-holiday effect. Besides, there is also a number of studies that focus on the impact of weather conditions on investors’ behavior. Cao & Wei (2004) identify the relationship between temperature and stock returns. They find a temperature anomaly by examining nine international stock indexes which belong to eight financial markets. The results suggest an overall negative relationship between temperature and stock market returns. Studies about seasonal anomalies in security markets are as important to academic researchers as to market participants. The existence of seasonalities in stock returns contradicts the efficient market hypothesis, indicating that investors can make profits by studying the past pattern of stock returns and applying proper strategies.
These effects are primarily reported in the U.S. stock markets, followed by studies of many major stock markets all over the world. Due to the fact that the Chinese stock market has a relatively short history and is not as mature as western stock markets in its development, very few studies exist on Chinese stock prices. However, along with the development of China, a large emerging market, the Chinese stock market has grown rapidly in the last decade. It is considered as one of the big and influential stock markets in Asia. Enormous capital source is allocated and traded in the market by both domestic investors and international investors. Therefore, to study whether the seasonal anomalies exist in the Chinese stock market is crucial to both academic researchers and investors. This paper focuses on examining three seasonal anomalies in the Chinese stock market, naming the day of the week effect, the month of the year effect and the holiday effect. The main research question is to find out whether some specific seasonal anomalies exist in the Chinese stock market, and if, why?
provides research method and data collection. Data analysis and discussion are presented in Section 4, which is followed by summary and concluding remarks in the final section.
2. Theoretical framework
2.1 Efficient Market Hypothesis and calendar anomalies
The Efficient Market Hypothesis was first developed by Fama (1970). It is stated that the prices of any securities traded in the stock market fully reflect available information. Investors buying securities in an efficient market should expect to obtain an equilibrium rate of return. Under the efficient market hypothesis (EMH), transaction cost is negligible, information is unbiased and it is not possible to predict the future returns by using historical price, which means the prices should be random in the future. Stocks should be traded at fair prices, investors should not speculate by purchasing the underpriced stocks or selling the inflated ones. There are three forms of the efficient market hypothesis: weak-form, semi-strong form and strong-form.
Weak-form EMH states that stock prices reflect all the information from historical prices; excess returns cannot be earned by applying the “pattern” of history prices since there is no “pattern”, thus future price movements are unexpected and random.
Semi-strong form EMH implies that share prices reflect all publicly available information in the market, and the prices adjust to new information rapidly and correctly, thus investors cannot earn excess returns by trading on that information.
However, in the financial markets, a number of ‘seasonal anomalies’, also called ‘calendar effects’, have been noted. The term “seasonality” in the field of finance and investment is defined as the changes in business and economic activities which occur repeatedly and regularly in certain periods. When seasonal effects are applied to stock markets, we mean the stock returns tend to display abnormal returns on specific days, weeks, or months. There are various seasonal effects that have been found in previous studies. The best-known ones are: the day of the week effect, the month of the year effect, the January effect, the holiday effect, and the New Year effect.
A great number of empirical researches have been carried out to examine stock market anomalies and its possible causes. In this section, I will discuss some main theories and findings related to the topic in this field.
2.2 The day of the week effect
The day of the week effect indicates that on particular days of a week, significantly different stock returns exist on stock markets. The most common one is the Monday effect, meaning that the average return on Monday is significantly lower than any other day in the week. French (1980) defines the Monday effect as the returns from the closing of trading on Friday to the closing of trading on Monday. The Fridays normally show the highest return in most of the exchanges in the world. The negative Monday returns and the positive high Friday returns are named as “Monday effect” and “Friday effect” respectively by many researchers, and they are jointly called the ‘weekend effect’. However, Jaffe and Westerfield (1985) find that in Japan the lowest return occurs on Tuesday instead of Monday.
the results demonstrate that the average returns for small firms tend to be higher on Friday, and the smaller the firm, the greater the tendency. At last, by examining actively traded over-the-counter (OTC) stocks, they conclude the day-of-the-week patterns for OTC stocks extremely similar to those for stocks on the Exchanges. In the same year, Rogalski (1984) examines the Dow Jones Industrial Average index every trading day. In order to further understand the Monday effect, both trading and non-trading day returns are examined. He divides the returns from Friday closed to Monday closed into two component parts: one is the Friday closed to Monday open return and the other one is the Monday open to Monday closed return. Rogalski (1984) finds the average negative Monday closed to previous Friday closed index return occurs from Friday’s closed to Monday’s open. Instead of using closed to closed returns, they choose average trading returns. Finally, he concludes that average trading returns are important for understanding day-of-the-week effects and finds evidence to support the hypothesis which was previously rejected.
In another study by Cho et al.(2007), the Monday effect has been confirmed based on the stochastic dominance approach. They observe a large number of major stock indexes on the markets of US, UK and Japan. Modified versions of the tests of Linton et al. (2005) are employed to test the Monday effect. They find that there is a strong Monday effect, which confirms earlier findings of many series over the full sample. In addition, the effect is even stronger in the second half of the month and on days when the previous Friday return was negative, which is opposed to the findings of the other researchers.
is, in principle, forbidden. These funds have to be paid back before weekends, thus can only be used for short-term speculations. A year later, Zhang & Li (2006) investigated in the same stock indices covering a longer period of time, from 1991-2004. They find very interesting results that are different from Gao & Kling’s findings (2005). They detect a positive Friday effect at an early stage by dividing the period from 1993 to 2004 into six periods(every two years are grouped as one period), but it seems to have disappeared since 1997; since then a Tuesday effect started to appear, this observation has never been made in previous studies of the Chinese stock market. Additionally, the studies of the Chinese stock market, Mookerjee & Yu (1999) document the highest daily returns on both exchanges on Thursday rather than Friday.
Gibbons & Hess (1981) suspect that the returns may be different according to the day of the week, especially the influence of weekends on Monday’s returns. They point out that, ‘Since Monday’s return is calculated over three instead of one calendar day, the mean and variance may well be higher on Monday compared with any other daily return (perhaps three times as large).’ while their findings show a contradictive evidence. The possible explanation is the existence of a settlement period. In reality, most transactions occur a few working days after the transaction date, and then the quotations are spot prices but not forward prices. The forward price is equal to the spot price grossed up by the riskless rate of interest for the length of the settlement period. Since settlement days are calculated in terms of working days, any settlement period that is not a multiple of five will introduce a day of week effect. The settlement period before February 10, 1968 was 4 working days, which means Monday’s price should be grossed up by 4 days of interest, whereas Tuesday’s to Friday’s should be grossed up by 6. These asymmetric settlement periods generate the negative return on Monday in that period. However, the negative Monday effects still exist after 1968, and their explanations have failed to describe the data.
of the negative Monday return and highest positive Friday return is that the information released over the weekend tends to be unfavorable. He gives an example of firms who fear ‘panic selling’ when bad news is announced, they will tend to postpone the announcement until the weekend in order to give more time for the information to be digested. Although this behavior is likely to happen, it would not cause systematically negative stock returns in an efficient market. Instead, investors would expect the release of unfavorable information on weekends and they would discount stock prices appropriately throughout the week.
Jaffe & Westerfield (1985) examine daily stock market returns in the US, the UK, Japan and Australia. They find weekend-effect in all four. Moreover, Tuesday has the lowest mean return in Japan and Australia. They suggest the Time-zone theory as an explanation of the negative Tuesday return. The time-zone difference only explains some of the Australian results but does not apply to Japan. Damodaran (1989) reveals that firms are likely to report bad news at the end of a week, and suggests that the postpone announcements of bad news might cause the negative Monday effect. Various explanations are given in the past studies, but the Monday effect is still an anomaly which needs to be extensively explained.
2.3 The Month of the year effect
The month of the year effect indicates significantly different stock returns on particular months of a year, usually higher in January and lower in December. However, in some studies a February effect is found to have significantly positive returns in Asian countries due to cultural influences. The February effect is defined as an abnormally high mean return in the month of February compared to the average returns of the other 11 months.
find the same pattern afterwards. M. Gultekin & B. Gultekin (1983) examine several stock markets in the major developed countries in the world. Excessively large January returns exist in most countries, except for Australia and an April returns in the UK. Chan et al. (1996) find that January and December both display significantly high returns on the Stock Exchange of Singapore (SES) and the Kuala Lumpur Stock Exchange (KLSE) in their study of seasonality on four Asian stock markets.
Wong et. al. (1990) document the January effect in the Malaysian stock market - Six Kuala Lumpur Stock Exchange (KLSE) sectoral indices from 1970 – 1985. They test the seasonality effects of the six sectoral indices under three different calendars: Gregorian calendar, Chinese calendar and Muslim calendar separately. The January effect, the Chinese New Year effect and an Aidilfitri effect1 are found. In the Muslin calendar they find a significant negative return in the 10th month comparing to the remaining 11 months in some industrial sectors2, which indicates that the Aidilfitri effect does exist but not so widespread as an January effect. In the data, the Chinese New Year effect appears to be the large positive average return in the last month of a year and negative return in the first month of the new calendar year. As they explain, ‘the possible reasons of the existence of the January effect in Malaysian stock markets are twofold: first, the influence of stock market movement in foreign markets like the US, where January effect is found to exist. The second reason is liquidity. The year-end bonuses are generally paid out in this period, which may cause the stock market to rise in January. As the Chinese New Year effect and the January effect occur around the same time period in a year, reasons suggested for the January effect probably play a part in the existence of Chinese New Year effect.’
Hamori (2001) investigates the seasonal properties of Japanese stock prices on a monthly basis. He analyzes the problem on two aspects. One is the seasonality pattern of
1
Hari Raya, Aidilfitri (also Hari Raya Puasa, literally "Celebration Day of Fasting") is a significant Islamic cultural event. Wong, et. al.(1990), Aidlfitri is the festival which Muslims celebrate after the end of the fasting month (also called “Ramadhan”, the 9th month of Muslims Lunar Calendar), the celebration starts on the first day of the 10th month and may extend up to a month.
2
Japanese stock prices; the other is the relationship between size effects and seasonality. The sample they used is the large-sized, medium-sized and small-sized stock indices of Tokyo Stock Price Index. When the total sample is employed, the monthly effects in the various stock indexes are confirmed. However, when the sample is separated into different periods, he does not find such effects in the latter half of the period. His test results also reject the null hypothesis that the average return is the same for each month at the 10% significance level in the small-sized stock index, but not in the large- and medium-sized stock indices. He concludes that the seasonality of the Japanese stock prices index is found to be deterministic but not stochastic.
On the other hand, many researchers are convinced that from an economic theory point of view, ‘the stochastic dominance rule is more satisfactory than the traditionally used mean-variance rule since it is defined with reference to a much larger class of utility functions/return distributions’ according to Seyhun (1993), who uses a stochastic dominance approach to test the January effect. He claims that, ‘the SD approach provides a clearer test of the market efficiency hypothesis by taking account of omitted risk factors’. The results suggest that January returns in all portfolios stochastically dominate non-January returns. Moreover, the stochastic dominance results presented in this paper provide strong support for the hypothesis that the high January returns for small firms cannot be attributed to omitted risk factors.
Many studies try to explain why the January effect exists, and one of the most discussed reasons is the tax-loss selling hypothesis (Brown et. al. (1983)). According to this hypothesis, investors will normally sell the losing stocks until the end of the tax year. They do this to make use of the opportunity to write-off capital losses against ordinary income in computing their federal income taxes. In consequence, the declining stocks has to face a downward pressure, but at the beginning of the next year the downward pressure will disappear due to the absence of selling pressure, therefore the stock prices can gain their real market value. This phenomenon can generate big stock abnormal returns at the turn of each tax year. In the study of Brounen & Ben-Hamo (2007), they find evidence that high January returns are caused by tax-loss selling in small and young firms in the US. However, the tax-rolling hypothesis is not applicable to countries where no capital gains tax exists. Thus, some researchers suggest that abnormal returns in January are caused by the firms who release new information at the end of the year. The financial earning announcement is generally made in January, which is a very influential factor to boost the stock returns.
Another explanation is the Portfolio Rebalancing hypothesis, which is developed by Haugen & Lakonishok3 (1988). This hypothesis asserts that high returns on risky stocks in January are the results of investors switch from holding high risk stocks to low risk stocks. Portfolio Managers actively engage in this so called “window dressing” activity at the turn-of-the-year. In order to rebalance their portfolios and construct a well-presentable balance sheet, they sell all underperformed-securities and invest in high return and risky securities, including high-risk small firms. This is also one reason that small firms outperform larger firms on the January effect in some stock markets.
We have discussed some causal reasons of the January effect, whereas the January effect is not fully explained by any one of them. For that reason more studies are expected to be carried out to find a better explanation to the January effect.
2.4 The Holiday effect
One of the best-known calendar anomalies is the holiday effect, also called the pre-holiday effect. The pre-holiday effect has been found in a number of studies that report abnormally high returns on the trading day preceding a holiday.
In a study by Lakonishok & Smidt (1988), daily returns on the Dow Jones Industrial Average over the period of 1897-1986 are examined. They find that the return of pre-holiday is 23 times higher than the ordinary daily rate of return, and the average rate of return after holidays is negative for the whole period. Ariel (1990) also documents a pattern of daily stock index returns surrounding holidays: the trading days prior to holidays on average perform high positive returns. In the investigated period of 20 years, more than one third of the total return was grossed on the eight trading days which fall before holidays every year.
Kim & Park (1994) report significant high returns on the pre-holiday trading days in three major U.S. exchange markets. The pre-holiday return is 9.0 times higher than the mean of the non-pre-holiday returns for the NYSE, 27.0 times higher for the AMEX and 10.9 times higher for the NASDAQ. In some recent studies, the pre-holiday effect is also found in the international markets besides the U.S. market (Meneu & Pardo, 2004; Chong et. al., 2005; Picou, 2006), for instance the UK and Hong Kong markets.
finds pre-holiday effect on organized stock markets and asserts, ‘pre-holiday strength can be attributed to short-sellers who desire to close short positions but not long positions in advance of holidays or, simply, to some clientele which preferentially buy (or avoid selling) on pre-holidays’.
Chan et al. (1996) find strong Chinese New Year effects on the Stock Exchange of Singapore (SES) and the Kuala Lumpur Stock Exchange (KLSE). In addition, they find Islamic New Year and Vesak effects4, but no Aidilfitri effect on the KLSE. All three holidays display abnormally high returns on the stock market. These results suggest that cultural holidays support a stronger effect than state holidays. The study also confirmed that seasonality and cultural influences are both important to the investing public.
As mentioned above, we can have some understanding about the stock market seasonal anomalies phenomenon. Some conflicting findings are likely to be the consequence of numerous occurring events, new regulations, new tax laws and cultural influences etc. Nevertheless, when reviewing the findings in most studies, seasonality effects exist in various forms. In addition, in most of the previous studies, the day of the week effect, the month of the year effect and holiday effect are the best known and the most occurring in stock markets all over the world. However, no existing studies examine the possible effects of four official public holidays in the Chinese stock market. Moreover, the findings of existing studies about seasonality effects in the Chinese market are inconsistent. Therefore, the purpose of this paper is to examine the day-of-the-week effect and month-of-the-year effect in the Shanghai and Shenzhen stock indices with a newer set of data, as well as to find out the possible effects and influences of the four official public holidays in these two markets.
Some selected studies are summarized in the following table. Including the data, samples period and methods they use and their key findings:
4
Table 1: Summary of selected studies
Author Researched seasonality effects
Sample Method Findings
The Day of The Week Effect
French (1980) Journal of Financial Economics, The weekend effect
S&P composite portfolio from 1953-1977. Use Daily stock returns.
Dummy regression; F-test; Bayesian analysis.
Negative Monday returns and the highest return in a week is on Friday, also called weekend effect. Gibbons
&Hess (1981)
Journal of Business
Day of the week effects
S&P 500 and the value- and equal-weighted portfolios from Jul. 2, 1962 – Dec. 28, 1978. Use Daily stock returns.
Dummy regression
Negative Monday returns and the highest return in a week is on Friday, also called weekend effect. Keim &
Stambaugh (1984) The
Journal of Finance,
The day of the week effect/ the weekend effect.
S&P Composite Index from 3 Jan. 1928 – Dec. 1982.
Use Daily stock returns.
Dummy regression
They find significantly negative Monday returns, which is so called “the weekend effect”. Chan, Khanthavit & Thomas (1996) Asia Pacific Journal of Management
The day of the week effects, the month of the year effects, Chinese New Year effect, Islamic New Year effect, Aidilfitri effect and Vesak effects.
Four Asian stock markets: KLSE (Jan. 1974 – Dec. 1992), SEB (Apr. 1979 – Dec. 1992), SES (Jan. 1969 – Dec. 1992) and SET (May 1975 – Dec. 1991). Daily returns were calculated as the natural logarithm of the price relative to consecutive closing price index values.
OLS dummy regression analysis
Strong day-of-the-week effects exist in KLSE and SES; strong Chinese New Year effects exist in SES and KLSE; on KLSE, the Islamic New Year and Vesak effects are also evident, but no Aidilfitri effect.
Cho, Linton & Whang (2007)
Journal of Empirical Finance
Monday effect. DJIA and S&P 500 cover the period 1 Jan. 1970 – 31 Dec. 2004; sample of NASDAQ, CRSP indexes and Russell 2000 is 1 Jan. 1988 – 31 Dec. 2004; and they also examine the Nikkei 225 and the FTSE 100 during the period 1 Jan. 1990 – 31 Dec. 2004. Daily return is calculated as the natural logarithm of the price relative to the closing price of the day prior.
Stochastic dominance approach.
They find strong evidence of a Monday effect in many cases under the stronger criterion which they based on – the stochastic dominance. Mookerjee & Yu (1999) Global Finance Journal
The day of the week effects, the turn of the month and the month of the year effect, the
Shanghai securities index (19 Dec. 1990 – 11 Apr. 1994) and Shenzhen securities index (3 Apr. 1991- 11 Apr. 1994). Use Daily
Dummy regression
turn of the quarter effect.
stock returns. turn of the quarter are less than returns on other days. Zhang & Li (2006) Journal of Chinese Economic and Business studies
Day of the week effect, small firm effect, turn of the month effect.
Shanghai Composite Index from 19 Dec. 1991 – 30 Apr. 2004; Shenzhen Composite Index from 3 Apr. 1991 – 30 Apr. 2004; Zhongxin Small firm Index from 4 May 1994 – 30 Apr. 2004. Use Daily closing price data, the returns are taking as natural logarithm. Dummy variables, Garch (1, 1)-GED. GED is short for Generalized Error Distribution.
Friday effect existed at early stage, but it seems to have disappeared since 1997; Tuesday effect started to appear since then. Small firm January effect was found with high volatility; they also found that the turn-of-the- month effect has also disappeared in the Chinese stock market since 1997.
The Month of The Year Effect
Seyhun (1993)
Journal of Financial and Quantitative Analysis
January effect. Data are combined from NYSE-AMEX files and NASDAQ files from 1926–1997. Both weekly and monthly stock returns are used.
Stochastic dominance approach.
Large January returns can be attributed to omitted risk factors. M. Gultekin & B. Gultekin (1983) Journal of Financial Economics The month of the year effect
Stock indices in major industrialized countries, for instance the US, the UK, Norway, Austria and France, etc, reported in CIP, the series returns cover the period from Jan. 1959 to Dec. 1979. Monthly stock market returns are computed as percentage changes in the monthly price indices without dividend yields
Non-parametric test.
Disproportionately large January returns are found in most countries and April returns in the UK, except for Australia. Khalid & Moosa (2005) Applied Financial Economics The month of the year effect
Global Market Index of the Kuwait Stock Exchange over the period of Jan. 1984- Dec. 2002. Monthly stock returns.
Stochastic dummies.
A July effect is found in stead of the better-recognized January effect, also called “summer holiday effect”. Gao & Kling
(2005)
Journals of Economics and Finance
The month of the year effect and the day of the week effects, year-end effect.
Shanghai and Shenzhen stock index from 1990 – 2002. Index data and individual stock returns.
Dummy regression
Fridays show significantly positive average returns; the Chinese year-end effect is found to be in February. Hamori (2001)
Japan and The world
Economy
The month of the year effect.
Tokyo Stock Price Index (TOSPI); the large-sized, medium-sized and small-sized stock indices. All indices are Value-weighted and represent end-of-the-month value. Modified ANOVA test- Levene test (Levene (1960)) and Brown-Forsythe test
Forsythe (1993)) Wong, Neoh, et. al. (1990) Asia Pacific Journal of Management January effect, Chinese New Year effect and Aidilfitri effect.
Six Kuala Lumpur Stock Exchange (KLSE) sectoral indices from 1970 - 1985; Neoh’s Market Return is used as a comparison which dividends are taken into account. Monthly natural logarithm of the returns is measured according to the Gregorian, Chinese and Muslim calendars.
Linear model. Different types of
seasonality are found when using different types of calendar. January effect, Chinese New Year effect and Aidilfitri effect are found.
The Holiday Effect
Lakonishok and Smidt (1988) Review of Financial Studies Pre-holiday effect
Dow Jones Industrial Average over the period of 1897-1986. Daily stock returns
Dummy regression
The pre-holiday rate of return is 23 times larger than the regular daily rate of return. Ariel (1990) The Journal of Finance Pre-holiday effect CRSP indices during 1963 – 1982. Use daily stock returns Dummy regression
The trading days prior to holidays on average perform high positive returns.
Kim and Park (1994) Journal of Financial Economics Pre-holiday effect
NYSE, AMEX and NASDAQ. Use daily stock returns
Dummy regression
High pre-holiday returns.
Meneu and Pardo (2004) Journal of Empirical Finance Pre-holiday effect
Spanish stock exchange from Jan. 1990 –Dec. 2000. Use daily stock returns
Dummy regression
High pre-holiday returns.
Chong et. al. (2005) Journal of International Money and Finance Pre-holiday effect
The UK and Hongkong markets during the period 1991-2003. Use daily stock returns
Dummy regression
High pre-holiday returns.
Picou (2006) Managerial Finance Pre-holiday effect Australia’s all ordinances (AUS), Japan’s Nikkei 225 (NI), Hong Kong’s Hang Seng index (HSI), the UK’s financial times stock exchange (FT), Canada’s Toronto stock exchange (TSE), and the US’s S&P 500 (S&P). Use daily stock returns
Dummy regression
2.5 Research question
Proceeding on my discussion thus far, to find out whether seasonal anomalies exist and in which forms they do exist in the Chinese stock market, the core research question in this investigation will be whether some specific seasonal anomalies exist in the Chinese stock market and why. I focus on three well-documented seasonalities for hypothesis testing:
1. The day of the week effect 2. The month of the year effect 3. The holiday effect
Further more, I will examine whether these effects change over time by splitting sample period into four sub-periods and examine size effect by taking size of firms into consideration.
3. Research design
3.1 Methodology
The day of the week and the month of the year effect
& Ben-Yamo (2007), the analysis is conducted using dummy variables. Thus, the following regression is used to estimate possible day of the week effect:
=
t
r α1Mont +α2Tuet+α3Wedt+α4Thut +α5Frit +εt (1)
Where rt is the daily return in period t for the stock indexes, Mont through Frit are
dummy variables from Monday through Friday, α is the coefficient of dummy
variables.εtis a disturbance or error term, which is assumed to be independently and
identically distributed as normal distribution. All estimates are made using OLS, applying White’s (1980) heteroskedasticity consistent standard errors. Resulting test statistics are asymptotically appropriate, whether or not the excess returns have a constant variance or are normally distributed.5
I examined possible month of the year effect by using a similar method, and the following regression is applied:
=
t
r α1Jant +α2Febt +α3Mart +α4Aprt +α5May5+α6Junt +α7Jult+α8Augt +α9Septt
+α10Octt +α11Novt +α12Dect +εt (2)
Where rtis the monthly return in period t for the stock indexes, Jant through Dect are
dummy variables from January through December, and εtis a random error term. Again,
all estimates are made using OLS, applying White’s (1980) heteroskedasticity consistent standard errors.
The holiday effect
A holiday is defined as one or some days-off on which the stock market is closed. I have taken into account the following four Chinese official holidays on which the stock markets are closed.
New Year's Day (January 1)
There is not much celebration in China on January 1st as it is in other parts of the world, since the upcoming Chinese New Year about a month away overshadows it. However, employees will enjoy 1 paid day-off. Stock markets close one day on this specific holiday.
Lunar New Year (the 1st day of the 1st month according to lunar calendar, the date in solar calendar differs every year)
It is also called The Chinese Spring festival, and is the greatest and the most celebrated festival in China and in part of East and Southeast Asia. Employees enjoy 3-7 paid days-off. New Year Eve is a time when all members in a family should get together to have supper and celebrate the beginning of the New Year. Stock markets close for 7-9 days which varies every year on this specific holiday period.
International Labor Day (May 1st – 3rd)
No less celebrated than the New Year's Day. Employees will enjoy three paid days-off. Celebration parties in parks take the place of this day’s parades. While these days-off
have been extended to 7 days since 2000. Stock markets close 3-7 days on this specific holiday period.
National Day (October 1st – 3rd)
It is the anniversary of the founding of the People's Republic of China in 1949 in the wake of routing the Nationalists who have since taken refuge in Taiwan. Employees enjoy 3 paid days-off. It is also a good occasion for many people to take a short excursion to enjoy the beauty of the golden fall. Stock markets close 3-7 days on this specific holiday period6.
I focus on examining the above mentioned four holidays and group the holiday dummy variables as discussed below. The trading days in the sample period are separated in three subgroups: one day before a holiday, one day after a holiday and all the other regular days. Pre-holiday returns are defined as daily close-to-close returns occurring on the day prior to a holiday; post-holiday returns are defined as daily close-to-close returns
occurring on the day following to a holiday; and the non-holiday returns are all the remaining close-to-close returns excluding pre-holidays’ and post-holidays’.7 The dummy variable was set to 1 for the day prior to a holiday (pre-holiday return) and the day following a holiday (post-holiday return), otherwise 0. The days-off of every holiday differ every year, and the government sometimes compensates more than two working days for getting a 9-days’ holiday (including two weekends), so called “a golden holiday week”. In the year from 1992-1996, there was only one day off on the Labour Day; in 1997-1999, there were three days off; from 2000 – 2008, the days off on Labour’s Day have been extended to seven days. On the National Day, people enjoyed three days-off from 1992-1998, and the seven days’ holiday starts from 1999 – 2007 in the data. To estimate whether there is a holiday effect, the following regression is used:
t
r= αi,preDpre+αi,postDpost +αi,otherDother+εit (3)
Where ri is the return in period t for the stock indexes;αiis the coefficient of the dummy
variables; Dpre is a dummy variable taking the value 1 on the day before a holiday and 0
otherwise; and Dpost is a dummy variable taking the value 1 on the day following a
holiday and 0 otherwise; Dother is a dummy variable taking the value 1 of ordinary days
other than pre-holidays and post holidays, and 0 otherwise. εit is a disturbance or error
term, which is assumed to be independently and identically distributed as normal distribution.
I use daily closing prices for testing three different effects. The daily rate of return is calculated as the natural logarithm of the closing price index divided by the closing price of the preceding day. I use dividend-adjusted prices to calculate the returns:
100 ) ( − 1 × = − t t t LnP LnP r (4)
Where rt is the return in period t; Pt is the closing price in period t; Pt-1 is the preceding
closing price in period t-1; Ln is the natural logarithm. All returns are in percentage.
All estimates are carried out using OLS, applying White’s (1980) heteroskedasticity consistent standard errors. Resulting tests are asymptotically appropriate, whether or not the excess returns have a constant variance or are normally distributed. I first give an overview of the characteristics of the sample, these results are presented in data collection. In the next section, I use equation 1 to test whether the mean of the tested day is different from all the other days in a week, use equation 2 to test whether the mean of the tested month is different from all the other months in a year, and use equation 3 for testing the holiday returns. Following Jaffe & Westerfield and French, I construct an F-test for equality of mean return across days of the week, months of the year, and holidays by computing equation 1-3. The hypothesis is α 1 = α 2 = ··· =αi for
respective equations. In addition, stocks are categorized in three sizes, taking firm size as a control variable while examining the three seasonal anomalies. Further, a difference of the means statistical t-test is performed by comparing average return of every single day with the average return of the remaining days in a week. The same manner is applied to compare the average return of every month with the average of the remaining months of a year.
3.2 Data collection
There are two stock exchanges in the Chinese stock market – the Shanghai stock exchange (SHSE) and the Shenzhen stock exchange (SZSE). SHSE started trading on 19th December 1990. After several years' operation, the SSE has become the most preeminent stock market in Mainland China. December 2007 ended with over 71.30 million investors and 860 listed companies. The total market capitalization of SSE hit RMB 26.98 trillion (US$ 3.95 trillion). A large number of companies from key industries, infrastructure and high-tech sectors have not only raised capital, but also improved their operation mechanism through listing on the Shanghai stock market.8 Shenzhen SE started trading on 3rd July 1991 officially. Since its creation in 1990, Shenzhen SE has
blossomed into a market of great competitive edges in the country, with a market capitalization around RMB 1 trillion (US$ 146.5 billion). A broad spectrum of market participants, including 740 listed companies, 35 million registered investors and 177 exchange members, create the market.9 Shanghai SE is considered to be the main indicator of the stock market and widely adopted by both domestic and foreign investors in evaluating the performance of the Chinese stock market. However, in order to examine seasonality effects on the Chinese stock markets, it is more complete to include both of these two stock exchange markets rather than to include only Shanghai stock exchange. Figure 1 shows the overall performance of Shanghai and Shenzhen stock exchange from 1992-2008. As we can see, the movements of both line charts are quite similar. The overall performance for both exchanges increased dramatically and reached the peak in the year 2007.
Stock Exchanges Overall Performance 1992-2008
0 20000 40000 60000 80000 100000 120000 1 9 9 2 1 9 9 3 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 9 9 2 0 0 0 2 0 0 1 2 0 0 2 2 0 0 3 2 0 0 4 2 0 0 5 2 0 0 6 2 0 0 7 2 0 0 8 Year V a lu e i n M il li o n s o f R M B Shenzhen Shanghai
Figure 1: overall performance of Shanghai and Shenzhen Stock Exchange from 1992- 2008
I use daily stock price indexes from 1992-2008 for both Shanghai and Shenzhen SE collected from DataStream for all regressions. This paper focuses on A shares, since A share is traded in RMB for Chinese citizen only, which therefore better reflects the behavior of Chinese investors. Stock returns are calculated using equation 3. There is some evidence that returns before and after holidays are relatively higher than returns on ordinary days.10 To avoid confounding results when observing the day of the week effect, equation (1) is tested twice by including returns of one day prior to and after a holiday and excluding returns of those days. Since the mean returns do not show much difference in two results, I report the result including those pre- and post- holiday returns. In both stock exchanges, the listed companies are categorized by market capitalization (number of listed shares multiply by share price). The large-sized stocks have market capitalization over 2 billion US dollars. The medium-sized stocks have market capitalization from 300 million - 2 billion US dollars. The small-sized stocks have market capitalization less than 300 million US dollars. Among all listed companies in Shanghai and Shenzhen SE, some firms do not share information in DataStream. Moreover, firms shift in sizes due to expansion. Thus, stocks are categorized on yearly basis. In Shanghai SE, large-sized stocks are dominated by 581 companies, followed by 406 medium-sized stocks and only 84 small-sized stocks. In Shenzhen SE, 432 medium-sized stocks are dominant, there are 372 large-sized stocks and 72 small-sized stocks.
Table 2 gives an overview of the characteristics of the sample for days, months and pre-holiday, post-holiday returns for the Shanghai and Shenzhen stock exchanges. Friday’s return is higher than Monday’s on both stock exchanges. Mean returns of total pre-holiday days are much higher than mean returns of total ordinary days. Summaries statistics of stocks in large size, medium size and small size in two stock exchanges for all frequencies are presented in the appendix I - VI.
Table 2 Summary statistics of the Shanghai and Shenzhen Stock Exchanges (1992-2008)
Table shows the summary statistics of percentage rates of return by day of the week, month of the year and holidays from 1992-2008 in two stock exchanges. All abbreviations used in the table are explained as follow: Pre-NY = pre-New Year days, Post-NY = Post- New Year days, Pre-LNY = Pre Lunar New Year days, Post-LNY = Post - Lunar New Year days, Pre–Lab = pre- Labor’s days, Post-Lab = Post- Labor’s days, Pre-Na. = pre-National days, Post – Na. = Post-National days, All Pre-days = the days before holidays in total, All Post-days = days after holidays in total, Ordinary days represent all the days other than holidays, Std. Dev. = standard deviation, Prob.= probability, Obs.= observations.
4. Empirical results
This chapter describes the empirical work and the results of this study about the seasonal anomalies in the Chinese stock market: the day of the week effect, the month of the year effect, the holiday effect. Taking size as control variable, I also present the results of large-, medium- and small-sized stocks.
4.1 The day of the week effect
The OLS results using equation (1) for the day of the week effect are reported in table 3. The findings are similar to several academic researchers’, but some new effect is also discovered in the Chinese stock exchanges. In the whole sample, mean returns on Fridays are the highest in Shanghai SE, but Wednesdays display the highest return in Shenzhen SE instead of Fridays. Mondays show negative returns in both stock exchanges, but the results are not significantly different from zero. Returns show similar patterns in all three sizes portfolios as the returns under the whole sample size portfolio. Monday, Tuesday and Thursday display negative returns, Wednesday and Friday show high positive returns.
Academic researchers have assumed that Monday’s return is the highest in a week, since Monday’s return is calculated over three instead of one calendar day, the mean and variance should display higher value compared to any other day in a week. However, Gibbons and Hess (1981) discover that mean returns on Monday is unusually low or even negative. French (1980), Keim & Stambaugh (1984) and Lakonsnishok & Smidt (1988) also document the same evidence to support low Monday’s return. I find a negative Monday return in the whole sample and in three different portfolios in both Shanghai and Shenzhen SE-s, but none of the result is statistically significant.
both the Japanese and Australian indices. The negative Tuesday return is the lowest in Shanghai SE and is significant at a level of 10% in the whole sample. This tendency is clearer in the large-sized stocks and medium-sized stocks but not found in the small-sized stocks. The negative Tuesday return in Shenzhen SE is not found to be significant in any sample size. Wednesday returns are significantly higher than other days in the week at 10% level in Shanghai SE and 5% significance level in Shenzhen SE, respectively.
Table 3 The day of the week effect for daily price returns in Shanghai and Shenzhen SE (1992-2008)
This table presents the results of the regression: =
t
r α1Mont+α2Tuet+α3Wedt+α4Thut+α5Frit+εt, where rt
is the daily return in period t for the stock indexes, Mont through Frit are dummy variables from Monday through
Friday, α is the coefficient of dummy variables,εtis a disturbance or error term, which is assumed to be independently and identically distributed as normal distribution. All estimates are made using OLS, applying White’s (1980) heteroskedasticity consistent standard errors. Rates of return are in percentage.
Shanghai Monday Tuesday Wednesday Thursday Friday
Whole sample (full size and period)
-0.114 (-1.139) -0.144* (-1.750) 0.147* (1.775) -0.139 (-1.111) 0.204** (2.501) Large size -0.103 (-9.222) -0.236** (-2.462) 0.141** (2.027) -0.159 (-1.197) 0.108* (1.719) Medium size -0.138 (-1.285) -0.172* (-1.898) 0.123 (1.289) -0.172 (-1.355) 0.130* (1.800) Small size -0.080 (-0.723) -0.051 (-0.573) 0.104 (1.212) -0.150 (-1.642) 0.068 (0.798) Shenzhen Whole sample (full size and period)
-0.075 (-0.771) -0.078 (-0.994) 0.168** (2.149) -0.126 (-1.563) 0.064 (0.845) Large size -0.143 (1.307) -0.136 (-1.589) 0.138* (1.655) -0.169 (-1.519) 0.051 (0.571) Medium size -0.113 (-1.026) -0.071 (-0.840) 0.144* (1.669) -0.208 (-1.386) 0.059 (0.705) Small size -0.119 (-1.091) -0.149 (-1.563) 0.208** (2.354) -0.153 (-1.632) 0.028 (0.285)
Note: *indicates significance at 10% level, **indicates significance at 5% level and ***indicates significance at a 1% level. T-statistics are reported in parentheses.
effects of stocks. They further propose settlement procedure as an explanation of the difference in returns of a week.
Table 4 t- Test statisticb for average return of five week-days in Shanghai and Shenzhen SE
This table presents the significance test results of the day of the week effect by comparing mean return of one day between the average return of the other four days, test statistics are t-value.
Shanghai Mondaya Tuesday Wednesday Thursday Friday
Whole sample -1,259 -1,891 2,121 -1,158 2,785 Large size -0.511 -2.158 2.438 -1.201 1.954 Medium size -0.935 -1.514 2.066 -1.379 2.083 Small size -0.241 -0.317 1.745 -1.481 1.201 Shenzhen Whole sample -0,717 -0,941 2,642 -1,497 1,105 Large size -0.909 -1.054 2.593 -1.394 1.310 Medium size -0.741 -0.418 2.449 -1.039 1.302 Small size -0.500 -1.143 3.265 -1.171 0.283 a t(.10, ∞) > 1.64; t(.05, ∞) > 1.96; t(.01, ∞) > 2.33.
Following an OLS regression test, a difference of means statistical test is performed by comparing the average return of one day with the average return of the remaining days of a week, the t-statistics are reported in table 4. Several days have statistical significance at a reasonable confidence level on both stock exchanges. Wednesday is significantly different from other days in a week under all portfolios. Besides a significant Wednesday effect, no other difference is found in Shenzhen SE. In Shanghai SE, Friday shows significant results in three portfolios but not in the small-sized stocks. Tuesday is significant in the whole sample and in the large-small-sized stocks. As Jaffe & Westerfield and French did, I further constructed an F-test for equality of mean return between the days of the week by computing equation 1. An F-statistics is computed for each regression and reported in appendix VII. The t-test and F-test results both indicate that inequality between means does exist in the coefficient estimates.
delivery. Thus, the observed quotations are then forward prices but not spot prices. These forward prices equal the spot prices grossed up by the riskless rate of interest for the length of the settlement period. If the settlement procedure is five business days, Monday’s price should be grossed up by 4 days of interests, whereas Tuesday’s through Friday’s prices should be grossed up by 6. As they documented, this asymmetry in settlement periods creates a low and, perhaps, negative return on Monday.
Settlement period differs in countries. It normally takes 5 business days in Canada and the United States; 3 business days in Japan; 1 to 10 business days in Australia. In Shanghai and Shenzhen SE, A shares are traded and settled in RMB. All A shares and exchange-traded instruments are settled on trade date (T) with cash settlement on T+1.11 In other words, a settlement period in Chinese stock market is 1 business day. Delivery of securities and receive of cash payment on Shanghai and Shenzhen SE occur on the second business day after the trading day. According to this transaction period, Monday’s through Thursday’s prices should be grossed up by 1 day of interests, whereas Friday’s price should be grossed up by 3 days of interests, including two days of the weekend. Thus, the settlement procedure suggests that Chinese common stocks should have high expected return on Friday. Friday shows the highest return in Shanghai SE in the whole sample, so the settlement procedure is a supportive explanation in this case. However, the highest return occurs on Wednesday in Shenzhen SE, and Wednesday’s return in Shanghai SE is statistically high as well. Since the settlement procedure does not affect the expected return on other weekdays, the high return on Wednesday in both stock exchanges is still puzzling. Possible explanations for the presence of Wednesday effect are left for future research.
Some finance literatures documented that the Monday effect has been reversed (from negative to positive) or disappeared over time.12 Brusa et. al (2000) find a relationship between the ‘reverse’ weekend effect and firm size. Their results reveal a ‘traditional’ weekend effect for small firms, and a ‘reverse’ weekend effect for large
11
http://gmi.rbcdexia-is.com
firms. Mehdian & Perry (2001) report a large reversal of the Monday effect in large-cap stocks over time in the USA, however, the results are statistically insignificant. In order to find out whether the day of the week effect is consistent over time, I divided the whole sample period (1992-2008) into four sub-periods: [A] 1992 -1995, [B] 1996 -1999, [C] 2000 -2003, [D] 2004 -2008 and examined equation (1) in all sizes. The results are in table 5.
Table 5 The day of the week effect for daily price returns in Shanghai and Shenzhen SE in 4 sample periods and different sizes
This table presents the results of the regression: =
t
r α1Mont+α2Tuet+α3Wedt+α4Thut+α5Frit+εt, where rt
is the daily return in period t for the stock indexes, Mont through Frit are dummy variables from Monday through
Friday. α is the coefficient of dummy variables,εtis a disturbance or error term, which is assumed to be independently and identically distributed as normal distribution. All estimates are made using OLS, applying White’s (1980) heteroskedasticity consistent standard errors. The whole sample period (1992-2008) is divided into four sub-periods: [A] 1992 -1995, [B] 1996 -1999, [C] 2000 -2003, [D] 2004 -2008. The whole sample size Stocks, Large-sized stocks, Medium-sized stocks and Small-sized stocks are all tested in four sub-periods. Rates of return are in percentage.
Shanghai Monday Tuesday Wednesday Thursday Friday Sub-period [A]:1992-1995
Whole sample size -0.631** (-2.067) -0.449* (-1.741) -0.017 (-0.068) 0.007 (0.029) 0.635*** (2.464) Large size -0.729** (-2.084) -0.885*** (-2.863) 0.046 (0.165) 0.041 (0.147) 0.489* (1.713) Medium size -0.811*** (-2.411) -0.529* (-1.851) 0.034 (0.109) -0.208 (-0.714) 0.369 (1.264) Small size -0.330 (-0.940) -0.224 (-0.796) -0.059 (-0.225) -0.182 (-0.661) 0.368 (1.417) Sub-period [B]:1996-1999
Whole sample size 0.112 (0.659) -0.247* (-1.769) 0.403*** (2.823) -0.130 (-0.866) 0.338** (2.365) Large size 0.154 (0.819) -0.264* (-1.777) 0.428*** (2.806) -0.187 (1.181) 0.253 (1.645) Medium size 0.169 (0.879) -0.257 (-1.608) 0.457*** (2.796) -0.207 (-1.148) 0.483** (2.312) Small size 0.086 (0.528) -0.117 (-0.847) 0.346*** (2.956) -0.081 (-0.568) 0.257* (1.903) Sub-period [C]: 2000-2003
Whole sample size -0.065 (0.544) 0.168* (1.762) -0.046 (-0.515) -0.036 (-0.363) -0.037 (-0.464) Large size -0.084 (-0.672) 0.162 (1.615) -0.116 (-1.215) -0.111 (-0.971) -0.085 (-0.954) Medium size -0.162 (-1.175) 0.087 (0.780) -0.173 (-1.520) -0.049 (-0.364) -0.060 (-0.653) Small size -0.071 (-0.637) 0.176** (1.971) -0.022 (-0.259) -0.028 (-0.262) -0.073 (-0.924) Sub-period [D]: 2004-2008
Table 5 (continued)
Shenzhen
Sub-period [A]: 1992-1995
Whole sample size -0.771*** (-2.535) -0.336 (-1.413) 0.013 (0.055) -0.054 (-0.229) 0.178 (0.779) Large size -0.973*** (-2.718) -0.321 (1.296) -0.073 (-0.270) -0.124 (-0.488) 0.035 (0.126) Medium size -0.978*** (-2.747) -0.233 (-0.948) -0.102 (-0.348) -0.225 (-0.886) 0.159 (0.642) Small size -0.691** (-2.098) -0.254 (-1.024) 0.175 (0.699) -0.124 (-0.488) 0.161 (0.619) Sub-period [B]: 1996-1999 Whole sample size 0.245
(1.284) -0.174 (-1.046) 0.429*** (2.865) -0.157 (-0.913) 0.298* (1.768) Large size 0.212 (0.967) -0.260 (-1.379) 0.369** (2.359) -0.299 (-1.583) 0.314 (1.482) Medium size 0.265 (1.224) -0.211 (-1.148) 0.426*** (2.888) -0.207 (-1.123) 0.227 (1.195) Small size 0.137 (0.732) -0.286 (-1.620) 0.383** (2.326) -0.163 (-0.930) 0.349* (1.933) Sub-period [C]: 2000-2003 Whole sample size -0.086
(-0.698) 0.160 (1.634) -0.063 (-0.666) -0.068 (-0.624) -0.051 (-0.606) Large size -0.104 (-0.930) 0.044 (0.406) -0.090 (-1.041) -0.045 (-0.419) -0.039 (-0.465) Medium size -0.052 (-0.393) 0.210 (1.496) -0.046 (-0.517) -0.105 (-0.902) -0.045 (-0.509) Small size -0.049 (-3.63) 0.161 (1.372) -0.030 (-0.283) -0.015 (-0.127) -0.055 (-0.553) Sub-period [D]: 2004-2008 Whole sample size 0.147
(0.939) -0.011 (-0.088) 0.248* (1.847) -0.198 (1.489) -0.120 (1.018) Large size 0.103 (0.621) -0.048 (-0.347) 0.284** (1.975) -0.196 (-1.369) -0.086 (-0.665) Medium size 0.117 (0.692) -0.070 (-0.496) 0.238* (1.684) -0.283 (-1.519) -0.069 (-0.521) Small size 0.005 (0.023) -0.215 (-1.049) 0.280 (1.601) -0.279 (-1.455) -0.461 (-1.214)
Note: *indicates significance at 10% level, **indicates significance at 5% level and ***indicates significance at a 1% level. T-statistics are reported in parentheses.
A similar pattern applies to the Friday effect; evidence of significant high returns only appear in sub-period A and B, this effect disappears in sub-period B and it is weakened when it reappears in sub-period D in Shanghai SE. There is little evidence for small-firm effect: Tuesday effect in sub-period C in Shanghai SE and Friday effect in sub-period B in Shenzhen SE.
To sum up, the day of the week effect changes over time in the Chinese stock market, there is reversal Monday return in two sub-periods but not statistically significant, small-firm effect has little evidence. Most of these anomalies disappear or attenuate as time passes. One explanation is the economic development of the Chinese stock market and the improvement of regulations in two stock exchanges. Information system has become more transparent and investors can obtain public information more smoothly. The Chinese stock market has become more efficient compared to 17 years ago. Thus, less price irregularities appear on the Chinese stock market.
4.2 The month of the year effect
Table 6 The month of the year effect for monthly price pattern in Shanghai and Shenzhen SE (1992-2008)
This table presents the results of the regression: =
t
r α1Jant+α2Febt+α3Mart+α4Aprt+α5May5+α6Junt+α7Jult+α8Augt+α9Septt+α10Octt+α11Novt+α12Dect+εt where rtis the
monthly return in period t for the stock indexes, Jant through Dect are dummy variables from January through December. αis the coefficient of dummy variables and εtis a random
error term. Again, all estimates are made using OLS, applying White’s (1980) heteroskedasticity consistent standard errors. Rates of return are in percentage.
Shanghai Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Whole sample (full size and period)
0.112 (0.937) 0.225 (1.419) 0.005 (0.045) 0.205 (1.339) -0.049 (-0.321) -0.021 (-0.174) -0.168 (-1.488) -0.073 (-0.479) -0.090 (-0.710) -0.209 (-1.272) 0.069 (0.594) -0.070 (-0.537) Large size 0.252 (1.449) 0.216 (1.228) -0.099 (-0.741) 0.235 (1.258) 0.033 (0.189) -0.152 (-1.045) -0.246 (-1.013) -0.064 (-0.380) -0.115 (-0.878) -0.228 (-1.298) 0.015 (0.122) -0.207 (-1.545) Medium size 0.039 (0.257) 0.269* (1.955) -0.043 (-0.952) 0.214 (1.078) -0.001 (-0.008) -0.154 (-1.157) -0.169 (-1.289) -0.055 (-0.309) -0.114 (-0.862) -0.322 (-1.619) 0.053 (0.449) -0.218* (-1.655) Small size 0.080 (0.701) 0.213* (1.881) 0.088 (.0664) 0.132 (1.034) 0.039 (0.225) -0.057 (-0.434) -0.161 (-1.361) -0.131 (-0.837) -0.071 (-0.255) -0.342 (-1.608) 0.117 (1.017) -0.136 (-1.000) Shenzhen Whole sample (full size and period)
Table 7 presents t-test statistics of the month of the year return by comparing mean return of one month between the average return of the other 11 months. The results are consistent with the results of the regression. The February returns show significant difference in three out of four portfolios in Shenzhen SE, but only in the medium-sized and small-sized stocks in Shanghai SE. The t-statistics support the results in Shanghai SE at a 1% and 5% significance level, respectively. In Shenzhen SE, t-statistics are significant at 5% level for all three portfolios. The December returns in the medium-sized stocks show significant results in two stock exchanges at a 10% level.
Table 7 t- Test statistic for average return of all months for Shanghai and Shenzhen SE
This table presents the significance test results of the month of the year effect by comparing mean return of one month between the average return of the other 11 months, test statistics are t-values.
Shanghai Jana Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Whole sample
(full size and period) 0.991 1.508 0.386 1.335 -0.399 -0.259 -1.262 -0.448 -0.815 -1.392 0.533 -0.635
Large size 1.094 1.497 -0.459 1.137 0.438 -0.807 -1.378 -0.152 -0.589 -1.117 0.454 -1.291
Medium size 0.558 2.251 -0.014 1.208 -0.374 -1.032 0.918 -0.107 0.473 -1.430 0.680 -1.695
Small size 1.291 1.993 0.823 1.205 0.333 -0.324 -1.264 -0.779 -0.438 -1.618 1.181 -0.919
Shenzhen
Whole sample
(full size and period) 0.479 1.967 0.263 1.217 -0.347 -0.872 -0.521 0.321 -0.393 -0.706 -0.093 -1.092
Large size 1.461 2.057 1.054 0.504 -0.370 -1.333 -0.193 0.463 -0.009 -1.117 -0.373 -1.263
Medium size 0.067 2.282 0.932 0.940 -1.477 -0.089 -0.929 -0.776 0.109 -1.236 0.603 -1.798
Small size -0.113 1.588 0.399 0.742 0.449 -0.065 0.420 0.019 0.262 -1.051 -0.338 0.083
a t
(.10, ∞) > 1.64; t(.05, ∞) > 1.96; t(.01, ∞) > 2.33.
D. In Shenzhen SE, I find significant results of high February returns in the medium-sized and the small-medium-sized stocks in sub-period D. These results reveal that the February effect has started to appear on the Chinese stock market only in recent years. More surprisingly, there is a March effect in some portfolios in sub-period B on both stock exchanges. Starting from 1992, the Chinese People’s Political Consultative Conference is held for two weeks in March every year, for making important national level political decisions.13 The decisions that are made have great influence on the economic development in China. Thus, it is possible that investors are influenced by new legislations. However, the March effect appears only in the sub-period B on both stock exchanges. Moreover, March’s return is insignificant in the data with full sample size and complete sample period, the t-test statistics in table 7 do not have evidence that March’s return is statistically significant than other months either. Thus, the Chinese People’s Political Consultative Conference cannot fully explain the existence of a March effect in the period 1996-1999. Reasons of this specific effect are remain ambiguous in this period and leave a great room to future research.
Many previous studies about the month of the year effect of the western stock markets have confirmed the January effect. Gultekin & Gultekin (1983) find that the disproportionately large January return exists in stock markets over 17 major industrial countries. Brounen & Ben-Hamo (2007) also find January return is superior to most other months, but their results lack statistical significance. Considering Christmas in December is the end-of-the-year-holiday in most of the western countries, some literatures14 suggest that the high return in February in the Chinese stock market is equivalent to the January effect in other stock markets in the world. Hamori (2001) provides several reasons for the January effect. One reason is an increase in January cash flows due to holiday bonuses and pensions; which is applicable to the case in China. In tradition, the most celebrated holiday in China is the Lunar New Year instead of Christmas. Chinese Lunar New Year falls most of the time at the end of January or at the first half of February. In other words,
13
http://www.npc.gov.cn/
14
Gao & Kling (2005), Wong, et. al (1990) propose that ‘The highest returns can be archived after the
February is the end of an old year and the beginning month of the New Year in most cases of the lunar calendar. Thus, February in the lunar calendar is, to a certain extend, equivalent to January in the solar calendar. Second reason is the sale of non-profitable stocks for tax reasons at the end of the year and subsequent re-investment in January. However, as mentioned above, tax-loss selling is not applicable in China, due to the fact that there are no taxes on capital gains. A third reason is financial manager’s attempts to show better end-of-year portfolio structures, as mentioned above the ‘window dressing’ activities. The last reason is the fact that January is the month in which governments, firms and individual investors form budgets for their future investments. The third and the fourth reasons are also possible explanations for the presence of Chinese year-end effect in February.
Keim (1983) shows that the January effect is mainly a small-firm effect. However, this is not proved by the results of the whole sample portfolio in Shenzhen SE under table 6, because the February effect does exist in the small-sized stocks. Although the result of large-sized stocks in Shanghai SE is not significant, I cannot conclude that the small firm effects pronouncedly in Shanghai SE, because the medium-sized stock displays significant high return on February as well. Assuming that the high February return is closely related to the Chinese Lunar New Year, I examined the holiday effect in the Chinese stock markets and see whether the results are supportive in the following part. Moreover, I test the assumption by adding February dummy into the regression of the Lunar New Year holiday returns.
