The holiday effect in Chinese stock markets – evidence from Shanghai and Shenzhen
The holiday effect is one form of the calendar effect, which means the abnormal return or abnormal volatility of the stock market before and after the holiday. The holiday effect can be explained by the behavioral finance perspective. Investors could adopt an active investment strategy to take advantage of the holiday effect and achieve excess returns. This paper examines the holiday effect on the Chinese stock markets of Shanghai and Shenzhen between January 2008 and August 2022. Standard OLS, GARCH-M, and EGARCH-M models are used to observe the existence of holiday effects in the mainland Chinese market before and after the Chinese New Year and the National Day. The results suggest that a significantly positive Chinese New Year effect exists for 2 days before, 3 days before, and 2 days after the Chinese New Year in the Shanghai market. There is no or very few holiday effect in the Shenzhen market. The positive National Day holiday effect exists 1 day before, 1 day after, and 2 days after the holiday. Besides, the results show that 1 day after the holiday is more volatile than other days around the holiday in Shanghai and Shenzhen stock markets. In addition, the study found evidence of the mainland China market reacting asymmetrically to positive and negative news.
Keywords: Holiday effect, Chinese holiday, Chinese markets, GARCH-M, EGARCH-M.
The holiday effect is defined as the tendency for asset returns in stock markets to display systematic modeling before and after a particular holiday. According to Fama (1970)’s efficient market hypothesis (EMH), it is impossible to attain consistent abnormal profits since share prices reflect all information. Nonetheless, the stock market cannot be entirely explained by EMH (Lakonishok and Smidt, 1988; Ritter, 2003). They argued that investors could achieve abnormal returns through active market timing strategies if the seasonal pattern exists, including timing the January effect, day-of-the-week effect, and other calendar effects (Mitchell & Ong, 2006). Nonetheless, less emphasis has been placed on the holiday effect. Stock prices before a holiday tend to be higher compared to other trading days. Gao and Lin (2011) agree that holiday effects exist globally. In as much as its earlier theory was based on western countries; Asians are more susceptible to the holiday effects (Yates et al., 1997), a behavior that has been evident in the Chinese financial markets (Brown and Mitchell, 2008). Chen and Chien (2011) and Huang et al. (2021) demonstrate the holiday effect during the Chinese New Year in the Taiwanese market. Notably, more companies have entered into and invested in the Chinese financial market; nonetheless, fewer studies have been done on this phenomenon. As a result, this empirical study will investigate the impact of the holiday effect on China’s stock market.
Since the 1980s, empirical studies on the holiday effect have been published in the economic and financial sectors and have attracted the attention of investors. McGuinness and Harris (2011) found significant evidence for the holiday effect. They found that the pre-holiday average returns were around 23 times higher than the returns on other trading days. Pantzalis and Ucar (2014) used S&P500 between 1946 and 2000 to find similar results. The opposite view provide by Hudson et al. (2002), that suggested that pre-holiday effects are actually negative. In addition, Mitchell & Ong (2006) found there were no significant pre-holiday effects in the stock market in Spain. Eidinejad and Dahlem (2021) used the Swedish stock market to show there is no evidence that the pre-holiday market exists.
Compared to Western holidays, only a few studies have examined the holiday effect of Chinese New Year on stock returns. With more and more countries that have large Chinese populations in the Asia Pacific region starting to celebrate the Chinese New Year, the Chinese New Year holiday effect is being increasingly appreciated by investors (Chia et al., 2015). Bergsma and Jiang (2016) point out that the Chinese New Year holiday effect is significantly observed in the Chinese and Hong Kong markets. This is inconsistent with Yuan and Gupta’s (2014) suggestion that the Chinese New Year holiday effect only exists in the Hong Kong market.
McGuinness and Harris (2011) state that the Chinese New Year holidays effects have a more significant effect than other holidays in China. Cao et al. (2007) confirmed this view using OLS regressions to test the Chinese New Year, Labour Day, National Day, and New Year’s Day. The results demonstrate that the impact of Chinese New Year is statistically and economically significant compared to other holidays where stock returns show minimal seasonal behavior. The recent confirmation comes from Casalin (2018), showing that the holiday effects are significantly positive, time-varying, and have no downward trend over time. Keef and Roush (2005) acknowledge that the importance of the Chinese New Year holiday effect is demonstrated by the fact that it does not diminish in the same way as the holiday effect in the United States. The above studies have focused on the pre-holiday effects of the Chinese New Year. Chai et al. (2015) used the Hang Sheng index to show the existence of a post-Chinese New Year effect, and suggest that the inconsistency of the previous post-Chinese New Year effect literature may be due to differences in the definition of the Chinese New Year. The impact of the Chinese New Year holiday effect can be determined by analyzing stock returns for each trading day that straddles the Chinese New Year holiday.
The OLS method was generally used to find out the holiday effect. However, the residual term in the OLS model does not have a normal distribution and has a fat tail in the real finance market, which may result in unreliable outcomes (Yuan and Gupta, 2014; Chiu, 2020). Engle (1993) proposed the autoregressive conditional heteroscedastic (ARCH) model to address the problem of the assumptions on the distribution of residual terms in the OLS model and allow for more robust explanations. Bollerslev (1986) extended Engle (1993)’s ARCH model and proposed the Generalized Autoregressive Conditional Heteroscedastic (GARCH) model which allows fewer parameters in the model. Yuan and Gupta (2014) argued that a GARCH model would be more suitable for testing seasonalities than an OLS model, and it is better for addressing autocorrelation and time-varying variance in the data. Engle (1993) developed a GARCH-M model based on the standard GARCH model, which to capture the time variation in the variance of the error term. Subsequently, The EGARCH and TGARCH models respectively to examine the possible asymmetric behavior and leverage effects of the stock market. The GARCH model is used to explain the time-varying volatility of the stock return to examine the pre and post-Chinese New Year holiday effect in the stock market by using GARCH-M, TGARCH-M, and EGARCH-M. GARCH models have been commonly used in recent studies of holiday effects (Sifeng, 2008; Tangjitprom, 2010, Yuan and Gupta, 2014, Chai et al., 2015). Hence, it is reasonable to believe that using several GARCH models would provide more accurate results.
This study examined the impact of the holiday effect in China by using the daily Shanghai Composite Index (SSEC) and Shenzhen Composite index (SZI) which would be representative of the Shanghai Stock Market and the Shenzhen Stock Market for a total of 3548 trading days from January 2008 to August 2022. This allows the market to have 15 Chinese New Year and 14 National Day holidays to be examined. On 14 December 2007, the Chinese State Council amended the ‘holiday arrangements for holidays and commemorative days in China’, adjusting the vacation time for the New Year, Labour Day, and other holidays in China. It means that starting from 2008, National Day on 1 October become the only longest holiday in China (7 days) apart from the Chinese New Year. Therefore, it is reasonable to believe that data from January 2008 to August 2022 (a total of 3548 trading days) would provide a more reliable result.
To avoid the effect of extreme values, daily stock market returns adopted Eidinejad and Dahlem (2021) ’s method. The daily returns rate is as：
Rt = 100 Pricet- Pricet-1 Pricet-1(1)
The Rt is the return on time t, and pricet and pricet-1 are the stock composite index in time t and t-1 respectively. There are non-trading days in the stock market and the index for the day before the non-trading day is used to calculate the return for the day after the non-trading day.
Holidays are defined in this study as public holidays involving trading breaks. In order to extend McGuinness and Harris (2011)’s study, and more deeply observe the impact of the holiday effect in China. The study investigated the three days before the market closure and two days after the market closure. In more detail, this study defined Pre1-CNY as one day before, Pre2-CNY as two days before, Pre3-CNY as three days before to examine the pre-holiday effect for the Chinese New year, and Post1-CNY as one day after, Post2-CNY as two days after to examine the post-holiday effect for the Chinese New year. Similarity, using Pre1-NH, Pre2-NH, Pre3-NH, Post1-NH, and Post2-NH to examine the National Day effect. The rest of the trading day is treated as other trading days.
Table SEQ Table * ARABIC 1. Statistics for the Shanghai and Shenzhen Composite Index.
Table 1 shows the descriptive statistics for the daily returns of the Shanghai and Shenzhen Composite index for the pre and post-holiday effect of the Chinese New Year and National Day, including the mean, maximum, minimum, and standard deviation, Jarque-Bera normality test, and Augmented Dickey-Fuller (ADF) test. The returns data of the composite indexes of the Shanghai and Shenzhen as a whole showed a p-value less than the 1% level of statistical significance in the ADF test. This indicates the smoothness of the data and does not require an additional transformation of data. The Jarque-Bera normality test rejects the null hypothesis of normal distribution at the 1% significance level, showing the daily stock returns are not normally distributed. The assumption of normality required by classical linear regression models was violated. It is consistent with the results of most of the previous studies.
standard ordinary least square
Most of the previous literature typically used the ordinary least squared (OLS) regression to examine whether stock returns are different on trading days before and after holidays than on other trading days. The simply OLS regression equation with dummy holiday variables can be stated below:
Rt = ɑ1 + β1DChinese New Year holiday effect + β2DNational Day holiday effect + ηt + εt(2)
Where Rt is the daily market return (Shanghai and Shenzhen) at time t. DChinese New Year holiday effect and DNational Day holiday effect represent the dummy variables for the Chinese New Year and National Day. DChinese New Year holiday effect included Pre1-CNY, Pre2-CNY, Pre3-CNY, Post1-CNY, and Post2-CNY. DNational Day holiday effect included Pre1-NH, Pre2-NH, Pre3-NH, Post1-NH, and Post2-NH. The dummy variables take the value of 1 if the daily return is on a corresponding day, and take the value of 0 on other trading days. ɑ1 is the intercept term that shows the percentage average return for other trading days. β1 and β2 are the estimated coefficient, ηt is the monthly fixed effect and εt is the error term. If the estimated coefficient of the dummy variables is significant, which shows the holiday effect in mainland china exists. The daily market returns before and after the Chinese New Year and National Day are significantly different from other trading days.
The main problem with using the OLS model is that the OLS model requires that the error terms on the stock returns data are homogeneous, normally distributed and continuously uncorrelated. It is difficult to achieve in reality and may result in unreliable conclusions (Chien et al.,2002, Yuan and Gupta, 2014, Connolly, 1989, Chiu,2020). Moreover, the OLS model cannot capture the time-varying volatility in the stock market (Chai et al., 2015). Hence, using the OLS model may not be suitable for examining seasonality in stock markets.
To overcome the drawbacks of the OLS model, the study used GARCH-M model. Compared to the standard GARCH model, the GARCH-M model associates stock returns with volatility and could capture the time variation in error term variance. Moreover, according to Bollerslev et al. (1992), the GARCH model is adequate for most financial series when the highest order of p and q is equal to 1. Therefore, the GARCH-M (1,1) model is adopted in this study and the equation writes as:
Rt = ɑ1 + β1DChinese New Year holiday effect + β2DNational Day holiday effect + β3σ2t + εt(3)
σ2t = γ1 + β1*DChinese New Year holiday effect + β2*DNational Day holiday effect + γ2σ2t-1 + γ3 ε2t-1(4)
Compared to Eq.(2), the return rate (Rt) in Eq.(3) depends on the conditional variance and error term εt with a mean of zero. β3 values the ratio of return to risk. In addition, Eq.(4) shows the conditional variance in a linear regression term. γ1 is the interception，γ2 and γ3 are the coefficients to capture the heteroskedasticity present in the daily stock return.
Engle and Ng (1993) indicated that markets tend to react asymmetrically to positive and negative shock, where the market tends to have higher volatility to negative news than to positive news. Hence, the study used the EGARCH model which proposed by Nelson (1991) and to obseve the asymmetrically relationship between returns and volatility. The EGARCH-M model is considered to provide more accurate results. The equation of the EGARCH-M model is specified as follows:
Rt = ɑ1 + β1DChinese New Year holiday effect + β2DNational Day holiday effect + logβ3σ2t + εt(5)
The conditional variance in the EGARCH-M model is nature logarithm, and express as:
logσ2t= γ1 + β1*DChinese New Year holiday effect + β2*DNational Day holiday effect + γ2σ2t-1 + ( γ3|εt-1σt-1 – 2/π | + φ1 εt-1σt-1)(6)
Where γ2 explains the persistence, γ3 explains the volatility clustering, and φ1 explains the leverage effects. φ1 equal to 0 if the impact of external shocks on the stock market is symmetric, and not equal to 0 if the impact is symmetrical. In addition, the negative φ implies that financial markets are more affected by negative news than positive news and that leverage effects exist.
Results and Discussion
Table 2 shows the OLS results for the Chinese New Year and National Day effect in mainland China. For the Chinese New year holiday effect, the estimated coefficients of the dummy variables are significantly positive on 2 days before and after the holiday in both Shanghai and Shenzhen stock markets, and the results can pass the 10% significant level. Moreover, the Pre-Chinese New Year holiday effect is more significant than the post-Chinese New Year, with higher coefficients. For the Chinese National Day, the 1 day before the holiday and 2 days after the National Day have shown significantly positive coefficients. The coefficient of the positive significant pre-holiday effect is higher than the post-holiday in the Shanghai market, whereas the Shenzhen shows opposite results. The results from the ARCH-LM test reject the null hypothesis that there is no ARCH effect in the model, indicating the untreated return volatility in the OLS model has ARCH effects. Hence, it is necessary to adopt the GARCH model to consider such volatility in order to receive a clearer understanding of the holiday effect on the Chinese stock market.
Table SEQ Table * ARABIC 2 The OLS results
From Table 3, the β5 is not significant at a 10% significant level in the Shanghai and Shenzhen market in both GARCH-M and EGARCH-M models. This presents the mean return may not increase with the volatility increase. For the Chinese New Year, the EGARCH-M model shows that 2 days before, 3 days before, and 2 days after the holiday provided significantly positive results in the Shanghai market. The results can pass the 5% significant level. The return on those days may be higher than on other trading days. Compare with the GARCH-M, only 2 days after the Chinese New year had a significant holiday effect. However, in the Shenzhen market, there are no significant results in the EGARCH-M model, and only post2-CNY had positive significant results with a 10% significant level in GARCH-M. In addition, the post-holiday effect is stronger than the pre-holiday effect in both markets, with higher coefficient figures. This is different from the OLS model, indicating that the pre-holiday effect is diminished after considering the ARCH effect in the model.
For the Chinese National Day, the estimated coefficients in the EGARCH-M model can pass the 1% significant level on 1 day before, 1 day after, and 2 days after in the Shanghai market, and can pass the 5% on 1 day before and after the National Day in the Shenzhen market, indicating the National Day effects are statistically significant in the Shanghai and Shenzhen stock markets. Similar to the Chinese New Year, the post-holiday effect is more significant and stronger than the pre-holiday effect. Moreover, the Shenzhen stock market provides a more significant National Day effect than the Shanghai stock market on 1 day before and after the holiday. The positive estimated coefficients are higher than the Shanghai. Further, the National Day holiday effect only shows the post-holiday effect in the GARCH-M model, without any significant results in pre-holiday effects.
Table 3 The GARCH-M and EGARCH-M model results (return equation)
Table 4 presents the conditional variance equation. In EGARCH-M, the result showed that Post1-NH have the highest volatility in Shanghai and Shenzhen market, and the lowest volatility in the two markets are associated with Post2-CNY and Post2-NH respective. In addition, the day after the holiday (Chinese New Year and National Day) had the highest volatility than other days around the holiday, and was more volatile than the pre-holiday effect. Besides, the Shenzhen market only showed statistically significant in post-Chinese New Year and National Day (Post1-CNY, Post2-CNY, Post1-NH, and Post2-NH). It demonstrated significant volatility on post-holiday.
Table 4 The GARCH-M and EGARCH-M model results (variance equation)
In the variance equation, the negative ARCH (γ2) parameter is allowed because the EGARCH model uses the logarithm sigma-square to confirm that the sigma-square is positive. Furthermore, the significantly positive φ1 shows the impact of the positive shocks and negative shocks on the stock market is asymmetrical, and the positive shock has a greater impact on volatility rather than the negative shocks of the same magnitude in both the Shanghai and Shenzhen stock market. Positive persistence (γ2) also indicates that investors are more likely to be susceptible to positive news, compared to negative news. This means that the volatility premium is asymmetric in the mechanism.
Table 5 The Ljung-Box and ARCH LM Tests
The diagnostic test is shown in Table 6, both GARCH-M and EGARCH-M do not have the remaining ARCH effect. All p-value is not significant at the 10% level, indicating the GARCH-M and EGARCH-M models are better than the OLS model to test the holiday effect.
Compared to the previous studies, the estimated coefficients of the 2 days before, 3 days before, and 2 days after the Chinese New Year holiday effects are statistically significant in the Shanghai stock market, which provides more evidence to show the Chinese New Year holiday effects exist pre and post-holiday. The findings of the pre-Chinese New Year holiday effect (Pre2-CNY and Pre3-CNY) are consistence with other studies such as Yuan and Gupta (2014). However, there is no positive significant holiday effect has been found on 1 day after the holiday This result is inconsistent with Chai et al. (2015)’ finding that 1 day after the Chinese New Year shows a significantly higher return than on other trading days. Nonetheless, their results are based on the Hong Kong stock market, not mainland China. Many other literatures also show the Chinese New Year holiday effect exists on 1 day before and 1 day after the holiday such as McGuinness and Harris (2011), but this study is not shown a significant positive holiday effect on 1 day before the holiday either. One possible explanation for the difference is that the holiday effect in the mainland Chinese market changed in recent years. Moreover, the result shows the post-holiday effect is more significant than the pre-holiday effect on the Chinese New Year and National Day in Shanghai and Shenzhen markets. Furthermore, the Chinese New Year holiday effect in the Shanghai market is stronger than in the Shenzhen market. There is no significant holiday effect in the Shenzhen market in the EGARCH-M model and a less significant holiday effect in the GARCH-M model. This shows that investors in mainland China do not have the same investment behavior. Indicating that studying the different stock markets in China separately is necessary for the investor’s investment strategies.
There is relatively less literature on the holiday effects of National Day. Cao et al. (2007) suggest that the holiday effects of Chinese New Year are more significant than those of other Chinese holidays (Labour Day, National Day, and New Year’s Day). Similar findings were obtained in this study. The Chinese New Year holiday effect on the Shanghai stock market is more significant than the National Day. The estimated coefficient 2 days after the Chinese New Year is stronger than other significantly positive coefficients in the Shanghai market. Nonetheless, the Shenzhen market shows an opposite phenomenon, in which the National Day is more significant than the Chinese New Year.
Most all variables in the variance equation in the EGARCH-M model show a significant result in the Shanghai stock market. Indicating the stock return had significant volatility during the straddling Chinese New Year and National Day period. The 2 days after the holiday are normally associated with the lowest volatility, and 1 day after the holiday is associated with the highest volatility. Moreover, the volatility on the first day after the holiday stock market is more volatile than the per-stock market. It is further confirmed by Chai et al. (2015)’ finding. They argue that as a result of Chinese superstition and the house money effect, individual investors preferred higher-risk stocks after the Chinese New Year. The results of volatility can be used to explain the attitude of investors to risk.
From a behavioral finance perspective, holiday stock returns contradict the EMH theory. Holiday-related anomalies may benefit investors. China’s vacation impact on stock returns is important for theory and practice. This helps comprehend how Chinese stock market volatility influences investor decisions. Using OLS, GARCH-M, and EGARCH-M, this research examined Shanghai and Shenzhen stock returns from 2008 through 2022. The main findings are that Shanghai had greater atypical returns before and after holidays, with more positive consequences. Second, only the Shanghai stock market shows a positive holiday impact for Chinese New Year, consistent with past results. More than National Day, Chinese New Year influences Shanghai’s market. After Chinese New Year and National Day, volatility increases. Positive shocks impact China’s volatility more than negative shocks, according to EGARCH-M. Investors must be wary with high-risk investments.
Bergsma, K. & Jiang, D. (2016) Cultural New Year Holidays and Stock Returns around the World: Cultural New Year Holidays and Stock Returns. Financial management, 45 (1), 3–35.
Bollerslev, T. (1986) Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307–327.
Brown, P. & Mitchell, J. (2008) Culture and stock price clustering: Evidence from The Peoples’ Republic of China. Pacific-Basin finance journal, 16 (1), 95–120.
Cao, Z. et al. (2007) Seasonality in the returns, volatility and turnover of the Chinese stock markets. Finance Letters, 5, pp. 1-11
Casalin, F. (2018) Determinants of holiday effects in mainland Chinese and Hong-Kong markets. China economic review, 4945–67.
Chia, R. C. J. et al. (2015) PRE AND POST CHINESE NEW YEAR HOLIDAY EFFECTS: EVIDENCE FROM HONG KONG STOCK MARKET. Singapore economic review, 60 (4), 1–1.
Chien, C.-C. et al. (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 (1), 155–162.
Chong, R. et al. (2005) Pre-holiday effects: International evidence on the decline and reversal of a stock market anomaly. Journal of international money and finance, 24 (8), 1226–1236.
Chiu, C.-N. (2020) Holiday effects on stock prices of the restaurant industry. Current issues in tourism, 23 (9), 1109–1121.
Engle, R. F. (1993) Statistical Models for Financial Volatility. Financial analysts journal, 49 (1), 72–78.
Fama, E. F. (1970) Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of finance (New York), 25 (2), 383–.
Gao, X. & Lin, T.-C. (2015) Do Individual Investors Treat Trading as a Fun and Exciting Gambling Activity? Evidence from Repeated Natural Experiments. The Review of financial studies, 28 (7), 2128–2166.
Huang, Y.-S. et al. (2022) The effect of Chinese lunar calendar on individual investors’ trading. Pacific-Basin finance journal, 71101694–.
Keef, S. P. & Roush, M. L. (2005) Day-of-the-week effects in the pre-holiday returns of the Standard & Poor’s 500 stock index. Applied financial economics, 15 (2), 107–119.
McGuinness, P. B. & Harris, R. D. F. (2011) Comparison of the ‘turn-of-the-month’ and lunar new year return effects in three Chinese markets: Hong Kong, Shanghai and Shenzhen. Applied financial economics, 21 (13), 917–929.
Mitchell, J. & Ong, L. (2006) Seasonalities in China‘s stock markets: Cultural or structural? International monetary fund working paper.
Pantzalis, C. & Ucar, E. (2014) Religious holidays, investor distraction, and earnings announcement effects. Journal of banking & finance, 47102–117.
Ritter, J. R. (2003). Behavioral finance. Pacific-Basin Finance Journal, 11(4), 429–437.
Sifeng, LLL. (2008). Are there holiday effect in China’s stock market? Journal of Financial Research, 2008-2.
Tangjitprom, N. (2010). Pre-holiday returns and volatility in Thai stock market. Asian Journal of Finance & Accounting, 2(2), 41–54.
Yuan, T. & Gupta, R. (2014) Chinese Lunar New Year effect in Asian stock markets, 1999–2012. The Quarterly review of economics and finance, 54 (4), 529–537.