The Chinese New Year Effect on Stock Markets
Bachelor Thesis
June 2015
BSc Economics and Business
Name:
Y. Vlutters
Student Number: 10465146
Supervisor:
J.J.G. Lemmen
ii
Preface and Acknowledgements
This is my Bachelor Thesis, which covers the subject of the Chinese New Year and its effects on stock markets in the world, for the BSc Economics and Business, specialization Finance, at the University of Amsterdam.
I would like to thank Dr. J.J.G. Lemmen for his continued guidance and supervision of this thesis.
iii
Abstract
This paper concerns the effect of the Chinese New Year on several countries’ large cap stock markets, focusing primarily on the U.S. and China itself, by means of an event study that looks at abnormal returns. Different event windows are employed and tests on 2015 only are conducted as well as tests over the last five years. The output shows that the Chinese New Year does have a considerable impact on China’s stock market itself and that other countries are not significantly influenced by this holiday.
Keywords: Chinese New Year, anomalies, stock market effect, China, event
study
iv
Table of Contents
Preface and Acknowledgements ... ii
Abstract ... iii
Table of Contents ... iv
1. Introduction ... 1
2. Literature Review ... 3
3. Methodology & Data ... 7
4. Results ... 11 5. Discussion ... 20 6. Conclusion ... 24 References ... 25 Appendix A ... 26 Appendix B ... 29 Appendix C ... 31
1
1. Introduction
Over the past few decades there have been many empirical studies that have displayed the existence of stock market anomalies, such as the Size effect, Value effect, Momentum effect, January effect, Turn-of-the-Month effect and many more that distort the efficient market hypothesis, which states that stock returns should behave randomly. Fama and French (1992), for instance, showed that small and value companies on average outperform large and growth companies respectively and Haug and Hirschey (2006) display the consistent reoccurrence of the so-called January effect, which are important discoveries for investors who wish to estimate future security prices. If new stock market anomalies are discovered, investors can use these to their advantage by adopting investment strategies that speculate on these events to generate profits, hence anomalies are interesting for both theoretical and practical uses. A conspicuous observation, however, that one could make is that there is relatively little research on cultural anomalies and their effects on international stock markets.
One interesting culture to observe is the Chinese, because China is population-wise the biggest country in the world and their presence in the world is becoming stronger by the day; there are Chinese restaurants galore, Chinatowns are situated in many big cities and big shopping malls are starting to provide guidance in Mandarin. Furthermore, China is growing wealthier rapidly and is currently the second largest economy in the world1, which means that Chinese individuals have more capital to invest with and are more influential in the world economy. As a result, the impact of Chinese investments could currently be quite prominent in non-Chinese stock markets. Consequently, this paper aims to examine the existence of the Chinese New Year effect in the world and the countries reviewed will be China itself, the Netherlands and the top three countries with the most Chinese immigrants as of 2013 not counting Hong Kong and Singapore, which are the U.S., Japan and South Korea2. Firstly, the effect will be tested in the year 2015 and subsequently the effect will be tested over a period of five years. The results of this study could be used by investors to predict stock market prices more correctly during future Chinese New Years.
The simple intuition behind this anomaly is that people feel more optimistic around this festive holiday, gifts and money are flowing around and thus more people decide to invest more in stocks; in short there is a positive holiday sentiment. Additionally, some Asian
1 Source: http://money.cnn.com/news/economy/world_economies_gdp/
2 Source:
2 companies adhere to the Chinese calendar as their fiscal year, as opposed to the conventional January 1 - December 31 fiscal year, resulting in earnings announcements and bonus payouts around the Chinese New Year, which could potentially incentivize investors to buy additional stocks. When we plot the graphs of the stock markets of the countries that will be researched, an upward trend starting approximately ten days before the Chinese New Year of 2015 is noticeable, except in South Korea, which starts its upward trend at around five days before the holiday, as shown in the figure below.
Figure 1. Market index levels for the observed countries.
The y-axis shows the value of the market index and the x-axis the discrete number of days prior to the Chinese New Year of 2015.
Whether or not this effect is actually consistent and caused by the Chinese New Year will be explored in the remainder of this paper, which is structured as follows. First, previously published literature concerning stock market anomalies, particularly holiday-related and cultural stock market anomalies, will be reviewed. Second, the methodology used for this research is elaborately explained as well as information on the source of the data. Third, the results of each individual observed country will be displayed and discussed and lastly a conclusion of the results will be presented.
17000 17200 17400 17600 17800 18000 18200 18400 -10 -8 -6 -4 -2 U.S. Japan 2950 3000 3050 3100 3150 3200 3250 3300 -10 -8 -6 -4 -2 China 440 445 450 455 460 465 470 -10 -8 -6 -4 -2 Netherlands 244 246 248 250 252 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 South Korea
3
2. Literature Review
Fields (1934) noted the existence of holiday effects early on and stated that “it will be seen
that the preholiday index relative to the mean of the two adjacent days was in fact more frequently higher and less often lower”. He then goes on to explain that during the period of
June 17 1901 until July 8 1932 he examined eighty holidays and that 65 per cent of the means of these holiday indices were significantly higher than the mean of the two adjacent days and just 18.8 per cent were significantly lower. Furthermore he shows that since 1916 until 1932 in fact 71.1 per cent of the holiday indices outperformed their adjacent counterparts while only 15.9 per cent underperformed. Additionally, Fields describes that during the Great Depression of 1929-1932 “the preholiday index displayed remarkable strength”. The
preholiday index was significantly higher than the mean of the two adjacent days in 85.7 per cent of the cases. However, although the evidence seems to be favorable for holiday
anomalies, Fields implores that practitioners be careful with the interpretation of the results.
Half a century later, Ariel (1990) conducted a similar experiment over the period 1963 until 1982 to test the existence of a holiday anomaly. He also concluded that the mean of pre-holiday returns significantly outperformed the average non-pre-holiday returns. Moreover, he discovered that the pre-holiday returns’ risk (i.e. variance) is not greater than the variance of the returns of the other days. In fact, not only does the pre-holiday variance not increase proportionately with the return, it actually exhibited a decrease, implying an improved risk-return tradeoff. Ariel explains that the decreased variance is an indicator of the fact that higher pre-holiday returns are not rewards for bearing extra risk. Additionally, Ariel found that the holiday effect predominates only on the single trading day prior to the certain holiday. He subsequently strived to dive into even more detail by looking at hourly stock prices on the day preceding the holiday. He discovered that stocks open significantly higher on these days than the day earlier with an average overnight return of 0.09%, which has a t-value of 3.16 and is therefore statistically significant. In particular Ariel found that last hour returns are high, are more often than not accountable for more than one fourth of the total return prior to the holiday and “display a disproportionate frequency of positive returns”.
Keim and Stambaugh (1984) try to find an underlying reason for this extreme last-hour stock price increase. They suspect that it might have to do with brokers and investors buying relatively much at the ask price close to market closure which could stimulate stock price increases. Consequently, the extreme last-hour pre-holiday return could be a result of a disproportionate frequency of late stock purchases. However, after conducting a research,
4 they reject this hypothesis and conclude that the so-called bid-ask effect is not an important driving force behind the holiday effect. At the same time, the fact that pre-holiday stock prices already start moving positively overnight also indicates that last minute transactions do not contribute much to this effect.
Ariel (1990) then goes on to discuss the segregation of holiday anomalies and other calendar anomalies. He starts off by invalidating those who claim that pre-holiday returns are
significant only because the high return of the January effect contributes to the pre-New Year return, which is one of the included holidays in Ariel’s sample. He applied a regression using a pre-holiday dummy and afterwards adding an additional pre-New Year dummy variable and finds highly significant test statistics that disprove the relation between the January effect and the strength of stock returns prior to New Year. Similarly, he also statistically proves that the holiday effect is not driven by the weekend effect and that there is not enough evidence to infer that returns accrue only to small firms as a result of the holiday effect.
McGuinness (2006) tested the existence of the Chinese New Year effect on the Hang Seng Index in Hong Kong and he found clear evidence of a reoccurring pre-Chinese New Year effect that was reportedly more stable than other anomalies such as the day-of-the-week and turn-of-the-month effect. Additionally, Cadsby and Ratner (1992) tested local holiday effects in ten different countries. Their results showed that the holiday anomaly is real around the world except for Europe, at least, none of their European sample observations showed significant holiday returns. However, they do note that only local holidays were taken into account.
In contrast to Ariel’s claim that pre-holiday returns are not higher in risk, Yuan and Gupta (2014) find that high pre-Chinese New Year returns in China are in fact rewards for high risk, whereas other markets, in particular Hong Kong and Singapore, exhibit pre-holiday abnormal returns that are solely caused by their holiday dummy regression variable. This suggests that the high holiday returns, whether positive or negative, are not anomalies, but rather rewards for running additional conditional risk. Furthermore, Yuan and Gupta disentangle the myth that the Chinese New Year effect is influenced greatly by the turn-of-the-month effect. They explain that this holiday generally begins at different dates every year, usually somewhere between the last week of January and the first week of February. As a result, as described by Yuan and Gupta, a pre-Chinese New Year effect is inconsistent with turn-of-the-month effects, although they do acknowledge that some portion of the turn-of-the-month effect could be
5 transferred to this holiday effect. Moreover, Yuan and Gupta argue that a potential answer for the reason that the Chinese New Year only impacts China may be found in behavioral finance. They claim that the Chinese stock market is saturated with mostly irrational retail investors and traders, who tend to be more concerned with short-term gains rather than long-term gains and often ignore long-term investment objectives and opportunities. Adding to this effect is the fact that China has long public holidays due to the Chinese New Year, which can also be seen from this study’s event window for China which commences on February 17 but is followed up on February 25. Yuan and Gupta illustrate that given this relatively long trading break, retail investors are apt to trade heavily in anticipation of this break. Furthermore, Yuan and Gupta clarify that prior to the Chinese New Year there are often political reforms, or at least political activity, and restructuring of Chinese companies which increases the volatility of the Chinese stock market. This argument could be one of the explaining factors of the high significance of pre-holiday returns.
Another contradiction in previous literature is that between Keim and Stambaugh (1984), who claim that the bid-ask effect does not have an important impact on holiday effects, and Bhana (1994) who tested the holiday effect on the Johannesburg Stock Exchange. Bhana argues that abnormal trading activity by brokers and investors cannot be dismissed as a reason for highly abnormal holiday returns, because “there is a fairly large bid/ask component to the
pre-holiday return”. He then explains other factors that could possibly explain the return strength
of pre-holiday prices, such as covering by short sellers who wish to close risky short positions just before the inception of a holiday and go back to this strategy after the holiday period. Additionally, Bhana says that changes in stock prices may be induced by brokers who urge their clients to buy, and to avoid selling, before a holiday. A possible explanation for this lies in the instinctive human nature. Bhana illustrates this by mentioning experimental market games that were performed by behavioral psychologists who demonstrated that human beings are behaviorally inclined to bid up stock prices just before relatively long-lasting market closings such as weekends and holidays. This result also helps to understand the day-of-the-week effect as well as the holiday effect.
Ahmed and Atiq (2014) tested the cultural holiday Ramadan on the stock market in Pakistan using monthly data over a period of three years and found that the Ramadan has no effect on stock markets. However, this study is different in many regards compared to other studies, because their time period is short and monthly instead of daily data is used. Furthermore, the Ramadan’s duration is one month whereas other holidays usually only last one or a couple of
6 days. Chougule and Khamborkar (2014) also tested a local cultural holiday, Diwali, on the Indian stock market and also found that it had no significant effect on India’s stock market. In summary, the holiday anomaly seems to be evident worldwide, however, holidays are often taken as a whole entity, for instance in the U.S. the average return of all days surrounding New Year, Christmas, Labor Day, Fourth of July (Independence Day) and so on are taken into consideration, whereas a single holiday has not been extensively tested. Moreover, there seems to be little to no research on the effect of foreign cultures’ holidays on other countries’ stock markets such as the effect of the Fourth of July on Brazilian or Chinese stock markets and little research is done on post-holiday effects. An overview of the reviewed articles is given below. Author (Date) Country/ region Time Period
Holiday Method Estimation Window Event Window Fields (1934) U.S. 1901-1932
American Constant mean return
- (-1,1)
Ariel (1990) U.S. 1963-1982
American Constant mean return - (-1,0) & (-1,1) McGuinness (2006) Hong Kong & U.S. 1990-2005 Chinese New Year Constant mean return - (-1,0) Cadsby & Ratner (1992) 10 countries Different per country Local holidays Constant mean return - (-1,0) Yuan & Gupta (2014) Asia 1999-2012 Chinese New Year GARCH(1,1)- model - (-1,0) Bhana (1994) Russia 1975-1990
Russian Constant mean return - (-1,0) Ahmed & Atiq (2014) Pakistan 2010-2012
Ramadan Dummy var. regression
- 1 month
Chougule & Khamborkar (2014)
India 2013 Diwali Autocorrelation & Ljung-Box test
(-30,30) Event itself
7
3. Methodology & Data
The basis of the methodology employed in this paper is based on A. Craig MacKinlay’s market model from his article Event Studies in Economics and Finance (1997). The event studied is the Chinese New Year and the event window will be set from one day prior to this event until one day after the event (an event window of (-1,1)), so a period of three days. Later on a (-5,5) event window will also be considered. MacKinlay (1997) depicts it as shown in the figure below.
He defines τ = 0 as the date of the event, τ = 𝑡1+ 1 to τ = 𝑡2 the event window, τ = 𝑡0+ 1
to τ = 𝑡1 the estimation window, 𝐿1 = 𝑡1− 𝑡0 the length of the estimation window and 𝐿2 = 𝑡2− 𝑡1 the length of the event window. MacKinlay (1997) explains that more often than not the event window is defined to be longer than the specific event itself as to allow research on periods surrounding the particular event. Subsequently we need an estimation window, which is used to estimate the stocks’ expected normal returns. MacKinlay (1997) argues that when using daily data an estimation window of 120 days prior to the event is a good basis for the calculation of the normal returns. The event and estimation window do not overlap, because the potential abnormal returns in the event window could then possibly affect the estimation window variables, which would distort our findings. Therefore the estimation window commences in fact not 120 days before the actual event, but 120 days before the inception of the event window. Our event, the Chinese New Year in 2015, took place on Thursday
February 19 2015, thus the (-1,1) event window will consist of the returns from February 18 until February 20 and the estimation window reaches from October 21 2014, 120 days prior to February 18 2015, until February 17 2015. The event window differs slightly for China and Korea since the Chinese New Year is accompanied by a national holiday in those two
countries and hence a closure of the stock markets. However, this should not form a problem as we can simply take the returns of the last day before and first day after the holiday to replicate an identical event window as previously discussed.
In order to investigate the existence of the Chinese New Year effect in multiple stock markets, daily stock prices of twenty randomly picked companies that are listed on the most prominent country-specific index as of May 2015 are gathered from Datastream. Since a three day event window is utilized and twenty firms per country are examined, each country will have sixty
8 different daily observations. The companies are randomly picked by means of a random number generator and the indices utilized for this experiment will be the SSE Composite Index, AEX, Dow Jones Industrial Average, Nikkei 225 and the KOSPI 200 for China, the Netherlands, U.S., Japan and South Korea respectively. For this test it is assumed that the stock returns are jointly multivariate normal, independent and identically distributed, which is claimed by MacKinlay (1997) to be a sufficient assumption for the market model to be
correctly specified and empirically reasonable. A student’s t-test could therefore be applied to test the significance of the abnormal returns and we will employ a 5% significance level throughout the entire study.
Subsequently, a test on the event’s impact has to be conducted by testing the significance of the abnormal returns, which is defined as 𝐴𝑅𝑖𝑡 = 𝑅𝑖𝑡− 𝐸(𝑅𝑖𝑡|𝑋𝑡), where 𝐴𝑅𝑖𝑡 is firm i’s
abnormal return for date t, 𝑅𝑖𝑡 firm i’s actual observed return at time t and 𝐸(𝑅𝑖𝑡|𝑋𝑡) the
normal return for firm i at time t, where 𝑋𝑡 is the information conditioned to find the normal
returns. MacKinlay’s market model, which stipulates that there is a linear relation between the market’s and stock’s returns, will be used to acquire the normal returns. The general market model for a stock is defined as 𝑅𝑖𝑡 = α𝑖+ β𝑖𝑅𝑚𝑡+ ε𝑖𝑡, where 𝑅𝑖𝑡 is the proxy for 𝐸(𝑅𝑖𝑡|𝑋𝑡), α𝑖 the constant, β𝑖 the stock’s sensitivity to market return fluctuations, 𝑅𝑚𝑡 the actual market return at time t and ε𝑖𝑡 firm i’s error term or disturbance with a mean of zero and variance of
σ𝑒2𝑖. The above-mentioned stock indices will be employed as the market proxies.
Ordinary least squares (OLS) is applied to estimate the parameters of the market model, which should be an efficient estimator given our normality assumption. Subsequently, firstly, an OLS regression of a stock’s returns on the market’s returns during the estimation window gives us an estimate of that firm’s α and β. Secondly, using these parameters we can calculate our predicted normal returns over the event window conditioned on the market return. Lastly, we obtain the daily abnormal returns by subtracting the daily predicted normal returns from the daily actual returns.
According to MacKinlay (1997) the abnormal returns can be seen as deviations from the market model that are computed outside of the estimation window sample. Under the null hypothesis in this test the abnormal returns are normally distributed with a mean of zero and a variance of σ2(𝐴𝑅̂ 𝑖𝑡) = σ𝑒2𝑖+ 1 𝐿1[1 + (𝑅𝑚𝑡−µ ̂𝑚)2 σ̂𝑚2 ] where σ𝑒𝑖 2 = 1 𝐿1−2∑(𝑅𝑖𝑡− α𝑖 − β𝑖𝑅𝑚𝑡) 2, which is equal
9 to 𝐿𝑆𝑆𝑅
1−2, and 𝐿1 is the amount of observed days in the estimation window. It can be perceived
that the abnormal returns’ conditional variance is made up of two parts; the simple variance of the errors of the returns of a specific firm on the market returns and additional variance
caused by sampling error in αi and βi, which is a regular occurrence for all event window
observations (MacKinlay, 1997). However, as MacKinlay explains, as the size of the estimation window increases, the second component of σ2(𝐴𝑅̂
𝑖𝑡) gradually decreases to zero
due to the disappearance of the sampling error of the parameters, such that the variance of the abnormal returns will become just σ𝑒2𝑖. In this case an estimation window size of 120 days is
large enough to reasonably assume that the additional variance from the abnormal returns is zero.
At this point we deviate from MacKinlay’s method, which finds the cumulative abnormal returns (CAR) of all the observed stocks over the entire event window and tests whether or not this is significantly different from zero. Resultantly, researchers and readers will only know whether or not the event had a significant impact over an entire event window that could potentially be as long as months, whereas this paper will test the individual daily abnormal returns of each individual company to spot more precisely when exactly and how severely the Chinese New Year impacts certain stock markets and allows for distinction between pre- and post-holiday abnormal returns as well as positive and negative returns between companies. In order to test each daily abnormal return of each firm individually over the event window we perform a t-test, which is possible because the aforementioned
assumptions for a t-test are satisfied and the variance is known (simply each firm’s error variance with the market). Although this is not the approach that MacKinlay administers in his paper, he does acknowledge this method and comments on when it can be effective: “One
can consider testing the null hypothesis of the event having no impact using unaggregated security by security data. This approach is applied most commonly when there is total clustering, that is, there is an event on the same date for a number of firms” (MacKinlay,
1997). For the reason that our specific event is a worldwide event that happens simultaneously for all companies, as opposed to company earnings announcement for example, it seems that this method is the appropriate tool to test our null hypothesis with.
However, after applying this method, MacKinlay’s CAR method will also be utilized to find out whether or not the overall event has an impact on stock markets. The accumulation of the returns will be done in two directions; through time and across stocks. The time-accumulated
10 CAR’s will firstly be considered, which we define as 𝐶𝐴𝑅𝑖(𝑡1,𝑡2), which is the sum of the
accumulated abnormal returns for firm i from the beginning of the event window (𝑡1) until
the end of the event window (𝑡2), so 𝐶𝐴𝑅𝑖(𝑡1,𝑡2) = ∑ 𝐴𝑅̂𝑖𝑡
𝑡2
𝑖=𝑡1 . Still assuming a large
estimation window, the variance of the CAR is defined as σ𝑖2(𝑡
1, 𝑡2) = (𝑡2− 𝑡1+ 1)σε𝑖
2.
Under the null hypothesis the 𝐶𝐴𝑅 is 𝐶𝐴𝑅𝑖(𝑡1, 𝑡2) ~ 𝑁 (0, σ𝑖2(𝑡1, 𝑡2)) distributed.
It is however often of more empirical interest to find whether the average return across all the observed firms is statistically significant as opposed to a single firm over the entire event window. In order to test this we define 𝐴𝑅̅̅̅̅𝑡 =𝑁1∑𝑁 𝐴𝑅̂𝑖𝑡
𝑖=1 as the average abnormal return of
all the observed firms at time t. The variance of the average abnormal return across firms will then be defined as σ̂2(𝐴𝑅̅̅̅̅
𝑡) =𝑁12∑𝑁𝑖=1σ̂2(𝐴𝑅𝑖𝑡). The 𝐴𝑅̅̅̅̅ is 𝐴𝑅̅̅̅̅𝑡 ~ 𝑁((0, σ̂2(𝐴𝑅̅̅̅̅𝑡))
distributed under the null hypothesis. Afterwards, the abnormal returns can be aggregated in both directions simultaneously and the average time- as well as stock-cumulated abnormal return is subsequently defined as 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) = ∑𝑡𝑡=𝑡2 1𝐴𝑅̅̅̅̅𝑡 with σ
2(𝐶𝐴𝑅̅̅̅̅̅̅(𝑡
1, 𝑡2)) =
∑ σ2(𝐴𝑅̅̅̅̅ 𝑡) 𝑡2
𝑡=𝑡1 . Equivalently, depending on in which direction the abnormal returns are
aggregated first, 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) =𝑁1∑𝑖=1𝑁 𝐶𝐴𝑅̂𝑖(𝑡1, 𝑡2) and σ2(𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2)) =𝑁12∑𝑁𝑖=1σ𝑖2(𝑡1, 𝑡2).
Under the null hypothesis the 𝐶𝐴𝑅̅̅̅̅̅̅ is 𝐶𝐴𝑅̅̅̅̅̅̅(𝑡1, 𝑡2) ~ 𝑁 (0, σ2(𝐶𝐴𝑅̅̅̅̅̅̅(𝑡
11
4. Results
The (-1,1) event window results analysis commences by looking at the individual stock return behavior in the Chinese market. The abnormal daily returns of twenty random companies were firstly tested on the first available day prior to the Chinese New Year (from now on referred to as Day -1), in the case of China this corresponds to February 17. Then the first available day after the New Year (Day 0) was tested, February 25 for China, and lastly another consecutive day (Day 1), February 26, is included in the sample. We apply similar definitions for these days in the rest of this paper for the other countries, however, the dates may not correspond, which is caused by different stock market schedules (e.g. China’s stock market is closed during this holiday whereas the Dutch market is open) and the fact that the Chinese New Year takes place on a different date every year. The obtained statistics are extremely notable, as depicted in Table 1 on the next page, as almost all returns on Day -1 are statistically insignificant. Specifically, 18 out of the 20 companies, or 90 per cent of the observed companies, exhibited abnormal returns that were statistically insignificant from zero on Day -1. Additionally, the two post-holiday observed days had no observations that were significant at all, which implies reasonable justification to infer that returns around the Chinese New Year in China were not affected by this holiday in 2015. All tests and samples will deliver at least some observations that are statistically significant and at a 95%
confidence level we would expect to find 5 per cent of the observations to be significant. This test only produced 3.33% statistically significant observations, so resultantly, there is no reason to believe that stock returns were extremely abnormal around the holiday.
One striking observation that might surprise some investors is that even though these returns are insignificant, they do not tend to be positively biased as one could expect. In fact, only just more than half of the observations (55%) had positive abnormal returns, even though insignificant. It therefore seems to be the case that the Chinese New Year does not necessarily induce significant abnormal returns nor positive returns in general, at least in China. This result is in accordance with the findings of Yuan and Gupta (2014) who found that Chinese excessive pre-holiday returns are accompanied by additional risk, which is exemplified in our observations by having an approximately equal distribution between positive and negative returns. Consequently, if the anomaly were to be real, speculators would have a difficult time to use this anomaly to their advantage since they do not know which companies will perform well that day and which do not, unless a pattern is discovered between firms in a certain sector or industry and whether or not they end up in the green or in the red after the
pre-12 holiday trading day.
Table 1. The t-values for the Chinese companies’ abnormal returns observed over the event window (-1;1) in 2015.
CN:HPI CN:AIR CN:ITL CN:SDW CN:ARI CN:ZHD CN:TTP
17-02 -0.0247 1.8047 0.4945 0.1524 -0.3863 0.5228 -1.2940 25-02 0.9350 -0.2871 0.5158 -0.2253 0.0218 -0.1017 0.8466 26-02 -0.0070 -0.8851 -0.1454 -0.4015 -0.7604 0.0978 -0.5193 CN:PNM CN:YMR CN:ZCM CN:LMM CN:SVD CN:JHM 17-02 -1.1744 -0.0977 0.8788 0.0972 0.1199 0.1408 25-02 -0.9835 0.6452 0.7512 -0.0560 -1.1718 -1.0120 26-02 0.3851 0.1215 -0.2034 0.0046 0.3229 0.0124
Given our sample size and significance level of 5%, a tvalue greater than approximately 1.99 or smaller than -1.99 is statistically significant. Green shaded t-values represent positive significant returns and red t-values negative significant returns. The top rows display the companies’ code name on Datastream. All values are rounded to four decimals.
Secondly, we look at South Korea, which is the only other country besides China that also has a public holiday during the Chinese New Year, so one would potentially expect similar results as for China. The results are displayed in Table A1 in the Appendix. As depicted in the table, the returns on Day -1 do not indicate that returns were fiercely different from zero at all; only 1 out of 20 observations was statistically significant. The same goes for the two days after the holiday, where each day also only has one significant observation. As a result, the Chinese New Year does not seem to have an impact on the South Korean stock market in 2015 even though a substantial amount of Chinese immigrants, both in absolute and relative terms, reside there and the indigenous population also celebrates this holiday. Still, practitioners should be
CN:DER CN:JGA CN:CFP CN:SON CN:AHH CN:DNG DN:NAM
17-02 0.2398 -0.2111 2.0198 0.1112 0.4806 -0.3270 3.0645
25-02 0.0445 -0.6422 -0.0578 0.5269 0.0007 -1.1098 -0.3670
13 careful with this result. The average 𝑅2, the proportion of the total variation in the dependent
variable (the returns of the companies) explained by the independent variable (the returns of the relevant market) in the regressions of the South Korean firms on the KOSPI 200 was merely 0.1040. This low value could have resulted in inaccurate predictions of the normal returns and consequently abnormal returns that do not appear to be statistically significant.
Next in line is Japan, whose results are shown in Table A2 in the Appendix. Also for this country it seems not to be the case that the Chinese New Year has a significant impact on its stock market in 2015. Only one out of the twenty observations on Day -1 significantly
outperformed and similar disappointing outcomes are present for the other two days. Likewise, the U.S. and Netherlands, Table A3 and A4 respectively in the Appendix, also do not present statistics in favor of the Chinese New Year anomaly in 2015. The Dow Jones merely brings forward one significant return on Day -1 and the Netherlands just two. Naturally, the two post-Chinese New Year trading days do not display alarming numbers. In summary, it does not seem to be the case that the Chinese New Year has a severe impact on stock markets when looking at individual companies’ daily returns over a (-1,1) event window.
Now we turn to the time-accumulated CAR to test whether companies are affected over the entire event window, still (-1,1), as opposed to on a daily basis. The results are depicted in Table 2 on the next page. As can be observed, there is again not much evidence in favor of the Chinese New Year anomaly around the world in 2015 when looking at companies over the entire event window. China, South Korea and Japan all just bring forward one company (5%) that outperformed its expectations and the U.S. and the Netherlands two (10%). These small significant amount of observations point toward a coincidence instead of abnormal returns due to the holiday.
Subsequently we test the significance of the average abnormal returns across stocks. The results are displayed in Table 3. Once again there is not much evidence to prove the existence of the Chinese New Year anomaly in 2015 when looking at firm-average returns. There was just one statistically significant average return which is most likely a coincidence and not a result of the holiday.
14
Table 2. The t-values for all companies’ time-accumulated abnormal returns of all observed countries over the entire event window in 2015.
1 2 3 4 5 6 7 China 0.5215 0.3652 0.4993 -0.2739 -0.6495 0.2996 -0.5581 South Korea -0.8590 -0.3760 -0.1065 -0.3223 0.0616 0.8331 0.3592 Japan -0.0123 -0.4797 0.7405 0.3342 2.0520 -0.3856 0.0690 U.S. -0.1123 0.1089 2.2351 -0.0196 0.0935 -0.2853 0.2811 Netherlands 0.6147 0.9818 1.1771 -0.4924 -0.4392 0.0845 -1.1564 8 9 10 11 12 13 14 China 0.0410 -0.7539 0.4947 -0.0645 1.1709 0.0774 2.1093 South Korea -0.7745 0.1852 0.3451 0.2615 0.6835 0.4307 0.7019 Japan -0.3553 0.6241 0.4577 0.7023 0.0097 1.9909 0.9335 U.S. -0.4216 -0.2807 -0.6669 -0.3470 0.1819 2.1754 -0.6622 Netherlands -0.6420 0.1752 0.1380 -1.7168 0.4966 0.8013 3.9101 15 16 17 18 19 20 China -1.0235 0.3862 0.8237 0.0264 -0.4209 -0.4959 South Korea 0.2588 1.0998 0.2714 0.6833 2.5288 0.2435 Japan 0.4114 -0.1741 1.0890 -1.5181 0.7412 0.4448 U.S. -0.8213 1.0278 -0.3579 -0.1356 -1.4567 -0.2103 Netherlands 0.5712 -0.5584 0.8082 -1.1852 -0.3193 2.8086
All companies are ranked in the same order as in Table 1 and Table A1 until A4, which is captured by the numbers 1-20 on top.
Table 3. The t-values of the firm-average abnormal returns on each day of the event window for all observed countries.
China South Korea Japan U.S. Netherlands
Day -1 1.6209 0.5179 0.6850 0.5943 1.0710
Day 0 -0.1674 0.9593 2.5483 0.0761 1.0217
15 Lastly, the average returns of all companies will be aggregated over the entire event window and tested to ultimately test the impact of the Chinese New Year on stock markets around the world. The results are shown in Table 4. Not to any surprise when taking the previous three tests into consideration, the results show that the Chinese New Year did not have any significant impact on any of the observed countries when looking at the entire accumulated abnormal returns.
Table 4. The t-values of time- and firm-accumulated abnormal returns for all observed countries in 2015.
China South Korea Japan U.S. Netherlands
t-value 0.6103 1.4339 1.4639 0.2465 0.8278
Now it is time to dig deeper into this anomaly by looking at the impact of the Chinese New Year over the last five years in China and the U.S., where the same companies are observed. The first test done previously was on the individual companies’ daily returns, however we will skip this test for the upcoming part and only go back to it if the CAR turns out to be significant to pinpoint on which days exactly the effect was felt the most. The dates for the Chinese New Years were different for each year, but the underlying research method remained unchanged; the event window was set from one day prior to the holiday until one day after and the estimation window captured the 120 days before the event window. The Chinese time-accumulated abnormal returns are tested firstly, then the firm-average abnormal returns and afterwards the time- and firm-accumulated returns. The results are depicted in Table 5, 6 and 7 respectively. Two observations are missing because that specific company was not listed on the stock market during 2011 and 2012.
Table 5. The t-values for all the observed Chinese companies’ time-accumulated abnormal returns over a period of five years employing a (-1,1) event window.
CN:HPI CN:AIR CN:ITL CN:SDW CN:ARI CN:ZHD CN:TTP
2011 -0.2539 -0.3547 -0.0814 -0.4696 0.3574 0.2049 0.2049
2012 0.4866 -1.1927 -1.4625 2.5008 -0.4416 -1.1125 -1.2304
2013 -0.7449 -0.4851 -0.5646 0.8010 1.5966 1.0259 0.5884
2014 -1.1236 0.0200 -0.7993 0.6170 -0.0050 -0.5091 -0.3671
16
Table 5 continued.
CN:DER CN:JGA CN:CFP CN:SON CN:AHH CN:DNG CN:NAM
2011 -0.4342 0.6444 0.1454 0.8968 -0.3160 1.3086 0.3784 2012 -0.2988 2.2708 1.6050 3.9423 0.3160 0.7509 0.2720 2013 -1.2792 -1.0404 -0.9134 0.2767 -0.8340 -0.7361 1.7814 2014 0.3543 0.8526 0.4195 1.1802 0.4885 -0.5215 0.9925 2015 0.0410 -0.7539 0.4947 -0.0645 1.1709 0.0774 2.1093 CN:PNM CN:YMR CN:ZCM CN:SNP CN:SVD CN:JHM 2011 0.1383 - -0.3800 -0.0018 0.5673 -0.0001 2012 1.4640 - 0.0623 1.0809 -0.1139 1.5434 2013 0.2317 -0.3780 -0.6461 -0.0529 -0.0383 0.7460 2014 2.0032 -0.1979 0.7788 0.9722 1.8198 0.2367 2015 -1.0235 0.3862 0.8237 0.0264 -0.4209 -0.4959
Table 6. The t-values of the Chinese firm-average abnormal returns on each day of the event window over five years employing a (-1,1) event window.
2011 2012 2013 2014 2015
Day -1 -1.6494 1.7454 -0.2679 1.9385 1.6209
Day 0 1.3906 3.1404 -0.7173 1.0790 -0.1674
Day 1 1.8475 0.7873 0.7196 0.1955 -0.3963
Table 7. The t-values of Chinese time- and firm-accumulated abnormal returns over the last five years employing a (-1,1) event window.
2011 2012 2013 2014 2015
17 When giving the results a quick overview, the results are not striking. Only 5 per cent of the time-accumulated observations and 6.67 per cent of the firm-average abnormal returns are statistically significant, however, we see that the Chinese New Year as a whole has a
significant impact on the Chinese stock market in 20 per cent of the observations, but this is not a reliable estimate due to the small sample size.
We subsequently employ the same tests on the U.S. stock market over the last five years and its outcomes can be observed in Tables B1, B2 and B3 in the Appendix. The
time-accumulated and firm-average abnormal returns show similar results as for China. 8 per cent of the time-accumulated observations and 6.67 per cent of the firm-average abnormal returns are statistically significant. However, the total effect of the Chinese New Year does not seem to exist in the U.S., whereas in China 20 per cent of the sample showed an impact as a result of the holiday.
To dig even deeper into the Chinese New Year anomaly, a test on the stock markets of China and the U.S. over the last five years will once more be tested, but this time with a (-5,5) event window. The results of the Chinese time-accumulated abnormal returns, the firm-average abnormal returns and the time- and firm-accumulated returns are depicted in Tables 8, 9 and 10 respectively.
Table 8. The t-values for all the observed Chinese companies’ time-accumulated abnormal returns over a period of five years employing a (-5,5) event window.
CN:HPI CN:AIR CN:ITL CN:SDW CN:ARI CN:ZHD CN:TTP
2011 0.0241 -0.5730 -0.2706 0.7120 -0.2427 -0.8599 0.9724
2012 -0.3547 -0.2944 -0.5823 0.5522 -0.2258 -0.4812 -0.4179
2013 -0.3257 -0.7049 -2.1609 0.7839 2.1608 0.8399 0.6696
2014 -0.6743 -1.0055 0.8552 0.3853 -0.3429 -0.0683 2.5826
2015 -0.2059 -0.1524 -0.3849 0.5383 0.5744 0.1634 2.2818
CN:DER CN:JGA CN:CFP CN:SON CN:AHH CN:DNG CN:NAM
2011 -0.1321 0.1249 1.1244 0.4485 2.3290 1.2527 0.1874
2012 0.0537 1.9613 0.7246 2.9898 -0.0493 0.2032 -0.3551
2013 1.2201 -0.5611 0.7898 -0.0588 1.0804 -1.3592 1.6965
2014 0.1912 -0.6407 -0.0031 0.4645 0.1212 0.2287 0.2668
18 Table 8 continued. CN:PNM CN:YMR CN:ZCM CN:SNP CN:SVD CN:JHM 2011 0.4133 - -0.2244 0.2304 0.3261 1.0501 2012 -0.2965 - 1.5968 -0.5584 0.1419 -0.9726 2013 4.4731 1.3010 1.0827 4.1295 1.3876 1.5007 2014 1.6342 0.0027 -0.8093 0.2382 0.7566 0.1397 2015 0.8299 1.2424 0.2102 0.3630 0.8971 -0.5090
Table 9. The t-values of the Chinese firm-average abnormal returns on each day of the event window over five years employing a (-5,5) event window.
2011 2012 2013 2014 2015 Day -5 1.2472 -1.1307 -1.3973 1.2565 0.5747 Day -4 1.2905 0.4557 2.5008 1.4315 0.5276 Day -3 0.6120 -1.0889 1.0222 -1.9071 2.1691 Day -2 0.1490 -0.2903 2.0526 -0.2410 1.7638 Day -1 -1.5950 1.7472 -0.2276 1.9445 1.6385 Day 0 1.4045 3.0506 -0.6567 1.0665 -0.0060 Day 1 1.9067 0.7718 0.8098 0.1632 -0.3678 Day 2 1.1685 0.5731 2.4610 -0.8245 0.6561 Day 3 -0.6492 -1.2379 2.3257 1.2690 0.6112 Day 4 -0.1455 1.0165 2.1140 -1.0811 1.1399 Day 5 2.0314 0.7845 -0.6691 2.4158 -0.0599
Table 10. The t-values of Chinese time- and firm-accumulated abnormal returns over the last five years employing a (-5,5) event window.
2011 2012 2013 2014 2015
t-value 1.7040 1.2229 3.4799 0.9732 2.7535
19 To visualize the effect of the Chinese New Year, the Chinese and American firm-average abnormal returns for the years 2011-2015 over a (-5,5) event window are plotted in Figure 2 below.
Figure 2. Chinese and American abnormal returns for the years 2011-2015 in a (-5,5) event window.
The left graph depicts the Chinese abnormal returns and the right graph the American. The y-axis shows the abnormal returns and the x-axis the day in the event window.
As can be seen from Figure 2, China shows a lot more activity around the Chinese New Year and mostly has a positive character (71 per cent of the graphs are situated above 0), whereas the U.S. shows little extraordinary signs and is mostly centered around the mean return of 0. According to this analysis, the results do seem to point more towards the existence of the Chinese New Year effect in China. Still only 9.2 per cent of the time-accumulated abnormal returns and 16.4 per cent of the firm-average abnormal returns are significant, however, the Chinese New Year as a whole seems to impact China in 40 per cent of our observations and it would even be 60 per cent if a 10% significance level was set. The identical test observations for the U.S. stock market are displayed in Table C1, C2 and C3 in the Appendix. 12 per cent of the time-accumulated abnormal returns and 10.9 per cent of the firm-average abnormal returns exhibit significant behavior, but none of the observed effects of the Chinese New Year as a whole show signs of significance. Additionally, almost all of the significant Chinese observations are positive, whereas the U.S. significant observations are often negative. Even though no clear-cut conclusion can be made, it does seem that the Chinese New Year has a positive impact on the Chinese stock market, although its power can be discussed, while it has no significant impact on the U.S. stock market.
-0,01 -0,005 0 0,005 0,01 0,015 -5 -4 -3 -2 -1 0 1 2 3 4 5 2015 2014 2013 2012 2011 -0,01 -0,005 0 0,005 0,01 0,015 -5 -4 -3 -2 -1 0 1 2 3 4 5 2015 2014 2013 2012 2011
20
5. Discussion
It appears to be that the Chinese New Year does not have such an influence to severely impact the entire world’s large cap stock market as of 2015 or over the last five years (keep in mind that the observed results of this study do not necessarily coincide with the results of a similar test on mid- or small cap stocks). This conclusion seems paradoxical with respect to the articles discussed in the literature review where there was overwhelming evidence in favor of holiday anomalies, especially of pre-holiday anomalies. However, one must take into account that those experiments were conducted solely with holidays that are native to those specific countries. This means that the majority of the population is anticipating those events with all the accompanying effects on the population’s sentiment. On the other hand, in the test of this paper a non-native holiday was transferred to multiple countries and tested negatively. Although when looking at it in this way the results might not be surprising, China’s strong presence in and grip on the world could just as well have shown in the form of stock market fluctuations around the Chinese New Year and it would not come as a surprise if the abnormal returns of other countries will become statistically significant around this holiday in the future if China keeps up its rapid economic growth and development. Another factor that could explain the difference in conclusions is that none of the reviewed articles employed MacKinlay’s market model to test the Chinese New Year, but mostly simply looked at the averages of non-holiday and holiday returns or used advanced methods such as the
GARCH(1,1)-model.
Additionally, the tests in previous literatures were mostly done on longer testing periods than a year. When we incorporated multiple years in our study we, however, still did not find extraordinary evidence in favor of the Chinese New Year, but it can be noted that the average t-values for the Chinese observations are substantially greater than the average t-values for the American observations for all directions of the cumulative abnormal returns. This
eye-catching fact shows that the Chinese stock market is much more affected by the Chinese New Year than the Dow Jones, although its power is ambiguous. There could be many reasons for this difference in conclusion between this test and previous literatures’ tests that may
enlighten the biasedness of these results, beginning with the fact that throughout the entire research we utilized a significance level of 5 per cent. If this would have been 10 per cent then 3 out of the 5 (60% of the sample) observed years in China with a (-5,5) event window would have shown statistically (positive) significant abnormal returns. If this was the case it would
21 not have been so straightforward to conclude that the Chinese New Year anomaly does not exist in China.
Furthermore, the observed period for the estimation window of 120 days is not extremely large, especially when considering that not all of these days, but only around 80 to 90, are trading days where the stock market is actually open. In practice, small sample sizes can result in samples that are not representative of the entire population, yield a low statistical power, could distort the normality assumption and in event studies could result in inaccurate approximations of the market model’s normal returns, but a too high sample may result in inaccurate estimates of current parameters. In the case of the Chinese companies there was on average also a relatively low 𝑅2; 0.24665 compared to 0.4339, 0.4334 and 0.4385 for the U.S.,
Netherlands and Japan respectively. Consequently, one could argue that the predicted returns and therefore also the abnormal returns are estimated with a very wide margin of error, which could be seen as a huge factor in explaining the statistically insignificant holiday returns. Perhaps a larger sample size and a proxy that better tracks the returns of each company would have resulted in more reliable results. MacKinlay (1987) also mentions two disadvantages of testing each daily return of each individual company separately. He says that more often than not the test statistics will have poor finite sample properties and that the test often has little power against economically reasonable alternatives. There is therefore a tradeoff in having large or small sample sizes. Ang and Kristensen (2012) find in their data set that an optimal sample size lies between 1.5 and 8.5 years of monthly data. However, since the Chinese New Year is a yearly occurrence, these returns would be internally correlated and also distorted if we use such a long estimation window, for the same reason as why we make sure the
estimation window and event window do not overlap.
The normality of daily stock returns should also be called into question. Brown and Warner (1985) explain that daily stock returns are often substantially non-normal and have a relatively fat tail, which is not the case when using monthly data. The problem, however, with using monthly data is that there are only twelve observations per year in that case. If we use more, as explained in the previous section, event windows overlap and will affect each other, resulting in biased estimates. Normal distributions have a skewness of approximately 0 and a kurtosis of approximately 3. The skewness of the sample in China, U.S., Japan, South Korea and the Netherlands is respectively 0.2149, -0.5730, 0.6599, 0.0509, 0.0177 and the kurtosis 2.5553, 11.7473, 3.1312, 4.4582, 3.0916, respectively. Based on this the normality
22 explanation for the biased and different results. A possibility to work around this issue could be the implementation of Newey-West’s heteroscedasticity and autocorrelation consistent standard errors. Resultantly, these Newey-West standard errors produce lower standard errors and thus higher t-values, which leads to more statistical significance and possibly a
conclusion stating that the Chinese New Year does have a serious impact on stock markets.
Moreover, even though it was assumed that the event windows for all countries were identical, they are in fact not. In terminological terms all countries had event windows ranging from one day prior to the event until one day after the event, however, in actual calendar dates the event windows for South Korea and China were different than the ones for the Netherlands, U.S. and Japan. Day -1 for South Korea and China took place on February 17 whereas the other countries experienced the similar day on February 18, but this discrepancy should not severely influence our conclusion. When thinking about it and looking at previous literature on this matter, a specific holiday should impact the stock market the most just before the holiday starts. Given that our results for the Netherlands, U.S. and Japan and South Korea do not seem to be statistically significant one day pre-holiday, it is reasonable to assume that the returns two days prior to the holiday are also statistically insignificant. On the other hand, however, the discrepancies of Day 0 and 1 for all the countries could be of importance, especially for China. China’s Day 0 takes place six days after the Chinese New Year and shows statistically insignificant result, but it is uncertain how returns would have behaved one, two or three days after the Chinese New Year. Since China always has public holidays after its New Year it is difficult to estimate what actually would have happened to the stock market if it were open and this likely has much to do with behavioral finance (how long does the Chinese New Year’s positive sentiment persist?).
Additionally, practitioners also have to be wary of the nonsynchronous trading effect, as illustrated by MacKinlay (1987). This effect arises when prices are assumed to be taken at the same time, ignoring the different event windows for now, but are actually not because of time differences in the world. The closing stock prices are consequently taken at different times, which could cause a bias in the event study. MacKinlay explains that another way this effect can arise is irregularities in time intervals. In most event studies, this one included, stock prices are taken at closing time each day, but MacKinlay argues that these closing prices usually do not occur at exactly the same time every day, which causes one to incorrectly
23 assume that the prices are observed at 24-hour intervals. Brown and Warner (1985) argue that nonsynchronous trading biases the results of OLS market model estimates.
24
6. Conclusion
It seems that the positive holiday sentiment around the Chinese New Year is currently not yet powerful enough to impact the entire world’s stock market. Even in China the abnormal returns are not extremely striking in that there are still many abnormal returns that are
statistically insignificant. However, when this holiday was tested using a (-5,5) event window over five years, the results showed that, under a 10% significance level, 60 per cent of the observed years were positively significantly impacted by the Chinese New Year as a whole. This implies that the holiday does have an effect in China, although its impact is not
consistent.
Future research is needed in order to deeply understand the mechanics behind holiday effects in general. For example, future research can be done on the relation between companies that had negative or positive significant abnormal returns and the Chinese New Year. Perhaps investors are induced to invest in certain companies in a specific industry during specific holidays. More future research is needed in general with regard to behavioral finance and holiday effects. This could then shed light on more reasons for the existing holiday effects; is it just a result of a positive holiday sentiment or do other factors contribute as well?
Additionally, one could conduct a similar research as this paper’s, but with a larger estimation window than that of MacKinlay’s and over a testing period of more years. This would allow for better tracking of stock returns and better normal and abnormal return estimates.
Furthermore, future research can be done on the Chinese New Year effect in non-Chinese countries such as the U.S. or the Netherlands over the last fifty years. The results of such an experiment could inform us of the development of the effect of the Chinese New Year in the world. For instance, 50 years ago the pre-Chinese New Year returns in the U.S. might not have been statistically significant at all, while there is a trend of those returns moving toward statistical significance, caused by China’s growing importance and wealth.
Possibly the effects of China’s holiday on the world will become more pronounced over the next years due to China’s explosively growing economy, but as of currently China’s presence is not great enough to influence the entire world’s stock market during Chinese New Year.
25
Reference List
Ahmed, S. N., & Atiq Ur Rehman Shah, S.M. (2014). The Ramadan Effect on Stock Market.
European Academic Research, 1(11), 4712-4720.
Ang, A., & Kristensen, D. (2012). Testing conditional factor models. Journal of Financial
Economics, 106(1), 132-156.
Ariel, R. A. (1990). High Stock Returns before Holidays: Existence and Evidence on Possible Causes. The Journal of Finance, 45(5), 1611-1626.
Bhana, N. (1994). Public holiday share price behavior on the Johannesburg Stock Exchange.
Investment Analysts Journal, 39, 45-49.
Brown, S. J., & Warner, J. B. (1985). Using Daily Stock Returns: The Case of Event Studies.
Journal of Financial Economics, 14(1), 3-31.
Cadsby, C. B., & Ratner, M. (1992). Turn-of-month and pre-holiday effects on stock returns: Some international evidence. Journal of Banking and Finance, 16, 497-509.
Chougule, A., & Khamborkar, A. (2014). A Study of Seasonality in Stock Market: With Special Reference to Diwali Effect. Journal or Business Management & Social
Sciences Research, 3(6), 26-29.
Fama, E. F., & French, K. R. (1992). The Cross-Section of Expected Stock Returns. The
Journal of Finance, 47(2), 427-465.
Fields, M. J. (1934). Security Prices and Stock Exchange Holidays in Relation to Short Selling. The Journal of Business of the University of Chicago, 7(4), 328-338.
Haug, M., & Hirschey, M. (2006). The January Effect. Financial Analysts Journal, 62(5),
78-88.
Keim, D. B., & Stambaugh, R. F. (1984). A Further Investigation of the Weekend Effect in Stock Returns. The Journal of Finance, 39(3), 819-835.
MacKinlay, A. C. (1997). Event Studies in Economics and Finance. Journal of Economic
Literature, 35(1), 13-39.
McGuinness, P. B. (2005). A re-examination of the holiday effect in stock returns: the case of Hong Kong. Applied Financial Economics, 15(16), 1107-1123.
http://dx/doi.org/10.1080/09603100500359575
Yuan, T., & Gupta, R. (2014). Chinese Lunar New Year effect in Asian stock markets, 1999-2012. The Quarterly Review of Economics and Finance.
26
Appendix A
Table A1. The t-values for the South Korean companies observed over the event window (-1;1) in 2015.
KO:SGL KO:SFC KO:SBG KO:DGY KO:IBK KO:GSG KO:HMA
17-02 -0.2454 -0.3075 0.4683 0.3539 0.5554 -0.2370 0.4744
23-02 -0.8733 -0.3320 -0.2232 0.1847 -0.4183 1.4550 -0.6531
24-02 -0.3691 -0.0118 -0.4296 -1.0968 -0.0305 0.2251 0.8009
KO:LCB KO:DSN KO:OTF KO:KFZ KO:HME KO:DDE KO:ABL
17-02 -0.4007 -0.9522 0.1427 -0.8076 -0.5488 -0.1068 1.8528
23-02 -0.5156 1.1411 0.6014 1.1786 -0.4892 0.6315 0.1168
24-02 -0.4252 0.1318 -0.1464 0.0819 2.2220 0.2213 -0.7539
KO:L&L KO:KCZ KO:IJM KO:IUH KO:KPY KO:CCM
17-02 -0.4768 0.3789 -0.8822 -0.0452 2.5986 -0.2154
23-02 -0.9067 1.2350 -0.6043 0.9901 2.0177 0.4103
24-02 1.8318 0.2909 1.9566 0.2386 -0.2364 0.2268
Given our sample size and significance level of 5%, a tvalue greater than approximately 1.99 or smaller than -1.99 is statistically significant. Once again green shaded values represent positive significant returns and red t-values negative significant returns. The top rows display the companies’ code name on Datastream. All t-values are rounded to four decimals.
Table A2. The t-values for the Japanese companies observed over the event window (-1;1) in 2015.
J:AJ@N J:CN@N J:SECR J:DU@N J:FUKU J:HO@N J:KE@N
18-02 0.8857 -1.0675 -0.6121 0.0920 2.1181 -0.4682 -0.5852
19-02 0.4533 0.1163 1.0525 -0.1383 1.0784 -0.0681 0.8040
27
Table A2 continued.
J:BS@N J:KU@N J:MIFU J:OU@N J:NH@N J:OB@N J:MI@N
18-02 1.0237 -1.0679 -0.0202 0.8185 0.4571 0.4529 1.9821
19-02 -0.3564 -0.4732 1.0713 -0.2069 -0.1215 2.4024 0.3856
20-02 -1.2828 2.6222 -0.2584 0.6048 -0.3189 0.5930 -0.7508
J:SHBA J:SO@N J:SMTH J:GC@N J:TOME J:WJR
18-02 -0.1583 -0.0217 0.3968 -0.2576 -0.2691 0.0595
19-02 1.2504 0.3258 0.9431 -0.4905 2.3864 1.1510
20-02 -0.3795 -0.6056 0.5480 -1.8813 -0.8336 -0.4400
Given our sample size and significance level of 5%, a tvalue greater than approximately 1.99 or smaller than -1.99 is statistically significant. Once again green shaded values represent positive significant returns and red t-values negative significant returns. The top rows display the companies’ code name on Datastream. All t-values are rounded to four decimals.
Table A3. The t-values for the American companies observed over the event window (-1;1) in 2015.
U:MMM @AAPL U:BA @CSCO U:KO U:DD U:GS
18-02 0.4356 0.4656 0.6421 0.4393 -0.1077 -0.0660 -0.7924
19-02 -0.0605 -0.1080 1.4573 -0.4030 1.5326 -0.0163 0.9655
20-02 -0.5696 -0.1690 1.7719 -0.0701 -1.2629 -0.4118 0.3137
U:HD U:JNJ U:JPM U:MCD @MSFT U:NKE U:PFE
18-02 -0.0076 -0.4130 -1.2464 0.4254 0.1535 2.611 -0.4681
19-02 -0.8882 1.3725 0.2546 -0.1957 0.3629 0.4666 -0.2275
28
Table A3 continued.
U:PG U:UNH U:VZ U:V U:WMT U:DIS
18-02 1.5155 0.0153 -0.4023 -0.5622 0.3676 -0.2971
19-02 -1.5009 0.4468 0.1347 0.0717 -3.1495 0.0838
20-02 -1.4372 1.3182 -0.3524 0.2556 0.2587 -0.1510
Given our sample size and significance level of 5%, a tvalue greater than approximately 1.99 or smaller than -1.99 is statistically significant. Once again green shaded values represent positive significant returns and red t-values negative significant returns. The top rows display the companies’ code name on Datastream. All t-values are rounded to four decimals.
Table A4. The t-values for the Dutch companies observed over the event window (-1;1) in 2015.
H:ALBI H:AGN H:AH H:AKZO H:MT H:ASML H:BOSK
18-02 0.4650 1.0506 0.3592 -0.5803 0.5794 0.4105 0.4435
19-02 -0.6145 -0.7749 1.0694 -0.6088 -1.2502 0.4771 -0.2419
20-02 1.2142 1.4247 0.6103 0.3363 -0.0899 -0.7413 -2.2046
H:GTO H:HB H:ING H:KPN H:OCIO H:PHIL H:RAND
18-02 0.5963 -1.5771 0.7899 -1.9891 -0.1086 1.1977 -0.2201
19-02 0.4953 1.8215 -0.7581 -0.0319 1.2053 0.1119 5.7338
20-02 -2.2036 0.0590 0.2072 -0.9526 -0.2365 0.0784 1.2588
H:ELS H:RDSA H:TNTE H:UBL H:UNIL H:WSG
18-02 0.6007 0.1740 1.1125 -2.0732 -1.5513 5.4491
19-02 0.3263 -1.6402 -0.3766 1.3883 0.9712 -0.7789
20-02 0.0623 0.4990 0.6640 -1.3680 0.0270 0.1945
Given our sample size and significance level of 5%, a tvalue greater than approximately 1.99 or smaller than -1.99 is statistically significant. Once again green shaded values represent positive significant returns and red t-values negative significant returns. The top rows display the companies’ code name on Datastream. All t-values are rounded to four decimals
29
Appendix B
Table B1. The t-values for all the observed American companies’ time-accumulated abnormal returns over a period of five years employing a (-1,1) event window.
U:MMM @AAPL U:BA @CSCO U:KO U:DD U:GS
2011 0.3379 -0.1835 0.7148 0.7912 -0.9475 0.8877 -0.352
2012 -0.1347 -0.7697 -0.5946 -0.2963 0.6909 -0.3142 0.2560
2013 0.5174 0.1009 -1.3549 -0.8054 -3.4064 0.6216 0.5897
2014 -1.9563 0.0857 -1.3763 0.6879 0.1066 0.7905 0.3633
2015 -0.1123 0.1089 2.2351 -0.0196 0.0935 -0.2853 0.2811
U:HD U:JNJ U:JPM U:MCD @MSFT U:NKE U:PFE
2011 -0.5875 0.5059 -1.5417 0.6600 -0.8715 1.2963 -0.0342
2012 -1.0267 -0.4080 0.4395 -2.4128 2.6854 0.1478 -1.0916
2013 0.5391 0.5040 0.6345 0.1098 1.1823 -0.0493 -0.3966
2014 -0.1412 -0.0685 0.5091 1.6008 0.5412 1.2776 3.3081
2015 -0.4216 -0.2807 -0.6669 -0.3470 0.1819 2.1754 -0.6622
U:PG U:UNH U:VZ U:V U:WMT U:DIS
2011 0.8298 0.0562 -0.1687 0.9380 -0.6174 0.7607
2012 -2.8266 -0.8993 -2.79535 -0.8237 0.5873 -0.7417
2013 -0.8307 -0.2673 -0.2736 -0.6858 0.1472 0.2738
2014 -0.7465 1.4681 -0.5667 0.5463 -0.4723 0.7449
30
Table B2. The t-values of the American firm-average abnormal returns on each day of the event window over five years employing a (-1,1) event window.
2011 2012 2013 2014 2015
Day -1 0.0400 -2.0568 -0.0057 0.5474 0.5943
Day 0 0.3324 -0.4131 1.0232 1.8972 0.0761
Day 1 0.7396 -0.3915 -1.7123 -0.2103 -0.2435
Table B3. The t-values of American time- and firm-accumulated abnormal returns over the last five years employing a (-1,1) event window.
2011 2012 2013 2014 2015
31
Appendix C
Table C1. The t-values for all the observed American companies’ time-accumulated abnormal returns over a period of five years employing a (-5,5) event window.
U:MMM @AAPL U:BA @CSCO U:KO U:DD U:GS
2011 0.1890 0.0694 0.3339 -1.9892 0.1073 2.4257 -0.2987
2012 0.9799 1.6708 -1.2070 -0.0101 0.3591 0.9268 1.6822
2013 0.8480 0.8835 0.1315 -0.0546 0.3424 -0.3177 0.5035
2014 0.1738 -1.5454 -2.2045 0.7455 -0.9347 2.8128 -1.2359
2015 0.0260 0.1971 -0.0383 2.9818 -0.4642 0.5548 0.2322
U:HD U:JNJ U:JPM U:MCD @MSFT U:NKE U:PFE
2011 -1.3271 0.2226 -0.4312 0.6955 -2.7347 1.1037 0.6540
2012 -0.0996 -0.0486 0.0394 -1.5148 1.1698 1.1662 -1.6172
2013 -0.1912 1.2748 0.5856 -0.9724 0.5628 0.3261 -0.7238
2014 -1.3788 0.0060 1.2823 1.3665 -0.1528 1.3269 2.0854
2015 0.5194 0.6572 1.5375 1.5403 0.7792 2.0065 -0.8407
U:PG U:UNH U:VZ U:V U:WMT U:DIS
2011 -0.3392 -0.7232 -0.6401 2.6004 -2.2559 2.0601
2012 -2.9485 -1.0633 -2.0137 -1.0178 0.4060 -0.5593
2013 0.0530 1.2570 0.1586 -0.4574 -0.1030 0.4983
2014 -0.8996 0.4368 -0.3140 -0.1105 0.1977 1.6360
32
Table C2. The t-values of the American firm-average abnormal returns on each day of the event window over five years employing a (-5,5) event window.
2011 2012 2013 2014 2015 Day -5 -1.8017 -0.3404 0.0908 -1.5006 2.5933 Day -4 -0.7821 2.1931 2.3676 -2.4919 -0.7732 Day -3 0.8858 -0.4774 0.9770 0.1568 -0.7112 Day -2 0.0099 -2.0555 0.6071 0.5040 0.6145 Day -1 0.3169 -0.4025 0.0507 1.9485 0.0885 Day 0 0.7312 -0.3818 1.0604 -0.2044 -0.2132 Day 1 -0.3948 0.5230 -1.6480 1.4349 1.6330 Day 2 0.6696 -0.5328 -0.1220 0.6568 -0.3555 Day 3 0.8072 1.5675 0.0508 0.5878 -0.3617 Day 4 -2.7054 0.3105 0.1991 -0.2740 1.3862 Day 5 0.7834 1.1212 -0.2498 -0.3777 1.0012
Table C3. The t-values of American time- and firm-accumulated abnormal returns over the last five years employing a (-1,1) event window.
2011 2012 2013 2014 2015