Seasonality Effect on the Vietnamese Stock Exchange

The Vietnamese Stock Market is a remarkable emerging market, including the two stock markets Ha Noi and Ho Chi Minh Stock Exchange and they have been playing a very important role inVietnamese economy. More and more attention is focused on the emerging Vietnamese market, and investors have been trying to find the opportunity to achieve abnormal returns through the Vietnamese Stock Market. We name this phenomenon market efficiency a nomaly, one pattern of which is seasonality effect. In this study, the topic about the seasonality effect is chosen. We try to test the seasonality in Vietnamese Stock Market by day of the week effect, January effect and t u r n o f t h e month effect. Deductive approach and quantitative research method are used in this thesis. To analyze seasonality effect, the data has been collected from Ho Chi Minh Stock Exchange Co mp o s i t e Index – VN Index and has been tested from 2006 to 2014. Hypothesis and T-test with α=0.05 isused to test the seasonality effect. The results show that seasonal anomalies exist. The above indicates that the Vietnamese Stock Market is not fully efficient yet. Investors may have opportunities to make use of the seasonal anomalies to earn abnormal return. However, the study is based on the historical data, but the future stock price is affected by lots of factors; and like in other invested stock markets, as soon as the seasonal anomalies is certified by the public, the opportunity of making excessive return by profitable trading strategies will disappear at once.


Introduction
Vietnam is a developing country in Asia. Vietnam's GDP is rising dramatically, its foreign direct investment is rapidly increasing, and its stock market is soaring. On November 7, 2006, Vietnam was officially recognized by the international community as the 150th WTO member. Vietnam is Asia'ssecond fastest growing economy after China, and it has been forecasted that Vietnam will become the world's 17th largest economy by 2025. Although the country has been a one party communist state since 1976, Vietnam has eased restrictions on private enterprisesand has been selling off many state owned enterprises (SOEs) to the public. These offerings are not referred to as "privatization" for ideological reasons, but rather are called"equitization". In 1986, Vietnam launched the Renovation intended to transform the country from a centralized economy to a socialist-oriented market economy. These reforms have earned fruits. In 2006, Vietnam's stock market index surged 145% and in the first three months of 2007, the index rose another 60%. It should be understood that the Vietnamese stock market started from a very low base of 22 listed companies at the start with a market capitalization of $144 million.
Vietnam is a remarkable emerging market, while theVietnamese Stock Market is emerging as well. There are two stock exchanges in Vietnam, one in Hanoi and one in Ho Chi MinhCity (formerly Saigon and still called Saigon by many who live there). The Ho Chi Minh Stock Exchange was inaugurated in July 2000 with trading in two equity issues. Today, about 110 companies are listed on Vietnam's two exchanges. With the development of capitalization in Vietnam, more private investors engage in the stock market and the stock market has been playing a more and more important role in the Vietnamese economy.

Turn of the Month Effect.
Turn-of-the-month (TOM) effect is defined as the tendency of stocks yielding sudden change in the period between the end of this month and early next month. Turn of the Month effect is a sign of inefficiency market when the returns gained in this period is affected by the returns of other days of the month. In the literature, the last trading day of the previous month and the first four trading days of the current month (-1 to +4) are usually considered as the turn of the month (TOM. In the recent decades, TOM is an anomaly mentioned in regular and seasonal reports in order to help investors to remove it from the market to avoid risk and profit from exploiting anomalies. TOM therefore seems to be a global effect, rather than the result of sampling errors or data snooping. The definition of the turn-of-the-month effect varies across the studies. The majority of studies either employ the definition of Turn-of-the-month proposed by Ariel (1987), for example Ogden (1990), Gerlach (2007 and Floros (2008), or the definition put forward by Lakonishok and Smidt (1988) like Kunken, Compton and Beyer (2003), McConnell and Wei (2008) and McGuinness and Harris (2011). If TOM effect exists, it will complement the study of the seasonal anomaly when it raises questions about the efficiency of markets and the rationale of investors. A surge profitability and overall market lasted statistically significant when switching months is inexplicable within the framework of discursive are many motivations for conducting behavioral finance research. So far, there has been relatively little research has implications for the presence (or not present) of the TOM effect in both developed and emerging markets.

Research Method
To construct the hypothesis, three ways can be chosen:

Form 1
Form 2 Form 3 H0: µ ≥µ0 H0: µ ≤µ0 H0: µ =µ0 HA: µ <µ0 HA: µ > µ0 HA: µ ≠µ0 As to our study, for example, in order to test January effect, the mean return of January should be compared with the mean return of all the other months. Therefore, the two-sided hypothesis will be suitable as an instrument for our case, which can be described by the following: µ1= the average daily log return of the turn of the month in percentage µ2= the average daily log return of the not turn of the month in percentage.

Model for Day of the Week Effect
This research uses data collected from Ho Chi Minh Stock exchange (HOSE), which is mainly the market index series every day (VN-index). The VN-index is a blend value combined from value of all stocks' price presented on the HOSE. Actually, the index value is market weighted-average index of capitalization values. Data used in the study is gathered over the period of time from 3rd January 2006 until 31st December 2014, which is down loaded from Website of the HOSE (www.hsx.vn) via website: www.cophieu68.vn. Particularly, the material is observed over the 9-year period is the returns of each day within the weeks by using the closed value of VN-index's share price of every trading day over the mentioned period. Afterwards, the changing of the value of the index from the previous day, presented in percentage, would be calculated based on the collected raw data (2243 observations).
With the calculated data, regression models would be established to examine the impact of each day of the week on the stock returns in Vietnamese stock exchange. According to Lewis-Beck (1994) measuring performance should include empirical and theoretical concerning, so the main role of measuring activities is briefed in the following formula: ISSN 1923-4023 E-ISSN 1923 Where: -X: observed score The following formula is used to calculate daily returns R = * 100% Or R = ln * 100% Where: -R t : the return over the period t.
-P t : the daily closed share price index of day t.
-P t-1 : the daily closed share price index of day t-1 Regarding to the formula, Fama (1980) has announced an original equation = ln + / ln = + ; the P t is the closed price of current session; D t : is the dividend of the period; P t-1 is the closed price of the previous day. D t value is on ex-dividend day, so the referring price decrease by the amount of the dividend, which has no relation with the performance of the paying securities.
To test the theory, t-test method is applied in the thesis for independent samples. The dummy variables 1 and 0; the variable will be 1 = average return of the concerning day, and 0 = average return on other days of the week. The value 1 will be applied from dummy variable Monday to Friday. Then the quantitative pattern is presented. The impact of the specific day of the week on the returns of VN-index could be assessed by applying the regression model, Standard Ordinary Least Square (OLS), with dummy variables showing each day of the week. In other words, the regression model is applied to assess the theory that the returns would vary depending on different day of the week. Many empirical researches are utilized when establishing the following OLS formula: R = α + β * MON + β * TUE + β * THU + β * FRI + ε To prevent of the collinearity's trap, the constant of regression formula is minimized. The given formula is utilized to experiment the relative return within a particular day of the week (equal to zero or not) and to assess the variation of daily returns on different day of the week.

Model for the January Effect
The regression model is used to examine the relationship among the variables. The focus of the analysis is on the link between a dependent variable and independent variables. To be clear, the regression analysis helps determine how the value of a dependent variable changes when one of the independent variable is adjusted while the others are static. A variable that is dependent depends on other independent variables which are used to forecast the expected value of the dependent variable. While the independent variables are non-random, the dependent variable is random. Snee (1977) points out that regression analysis is a technique used for the process of forecasting, controlling and learning a mechanism from the data collected.
The monthly return is calculated using the following formula: Rt: the monthly return at the period of t Pt: VN-Index at the period of t The reasons why logarithm returns are chosen are seen to be suitable both theoretically and empirically. Under theoretical perspective, logarithmic returns are easily controlled and they can connect sub-period returns to form returns over long period. Under empirical perspective, there is strong possibility that logarithmic returns are normally distributed, which is a prerequisite for standard techniques in statistics.
To test the presence of monthly effect, in this case, January effect, the following model should be used: Whereas R t is the monthly return. D i is the dummy variable that receives the value of 1 in the month and zero otherwise. For example, D Jan = 1 if the return is in January and 0 otherwise, D Feb = 1 if the return is in February and 0 otherwise, D Mar = 1 if the return is in March and 0 otherwise, D Apr = 1 if the return is in April and 0 otherwise, D May = 1 if the return is in May and 0 otherwise, D Jun = 1 if the return is in June and 0 otherwise, D Dec = 1 if the return is in December and 0 otherwise. The coefficients β 2 to β 12 illustrate the difference between the month of January and the i th month with i runs from 2 to 12.

Model for the Turn of the Month Effect
As a specific definition on TOM interval has been lacking in the literature, the method of defining TOM interval by Xu and McConnell (2008) has been considered for this thesis. Xu and McConnell (2008) defines the interval [-10,+10], but we take into account of the average index returns by day of the month for the last five days of the previous month and the first five days of the next month [-5,+5] because our sample includes some months of 16 trading days only. Hence, the last trading day of the month is day -1, the first trading day of the month is +1, the second day trading day of the month is day +2, etc… Same method is applied to calculate the volumes of buy/sell/gross by day of the month but in the present study, the author decides to adopt the daily volume over each interval [-5,+5] and then divide each observation of the daily volume by this average figure and average across all days of the [-5,+5] interval for the entire period of 2006 to 2013. By this way, the average across all days marked -5, the average across all days marked -4 and so on is found. Following the methods of Booth, Kallunki and Martikainen (2001), the author uses standardized returns and standardized volumes to check the first part of the previous analysis.
We use OLS method (Ordinary least square) to test for a TOM effect returns over the chosen interval. Following (Kunkel, Compton and Beyer, 2003), the following specification is used: Where: If it is sign non-TOM positive va days are l non-TOM

Researc
Results of T

Correlation Matrix
To establish the correlation matrix, we should base on the basic hypothesis that one variable would relate to other variable(s), which might be negative correlation or positive correlation. The Pearson correlation is one the applicable methods to estimate relationship between two variables. There are score and interval or ratio levels. In case of the researches, the presume is that Return should belong to particular day of the week (Mon-Monday, Tue-Tuesday, Wed-Wednesday, Thu-Thursday, or Fri-Friday).
As shown in the correlation matrix, the correlation level between Return variable (-0.057) and Tue variable is lowest compared with other correlations of the variables, and the value is also negative. The value of correlation matches with the discovery found above in the statistic description. Furthermore, correlation between Return and Fri variable (+0.05) is positive and highest one over the concerning group. Besides, there are two negative correlations between Return and Mon and Tue, three positive correlations between Return and Wed, Thu, and Fri. However, the correlation value between Return and Thu is only 0.004. It is quite interesting that the correlation level between pairs of days of the week is nearly stable (around -0.245 to -0.252) except for correlation between Fri and Mon (0.05). In summary, the returns of the VN-Index portfolio seem to decrease at the beginning of the week and reach the bottom on Tuesday then it increase and achieve the peak of the week on Friday. *. Correlation is significant at the 0.05 level (2-tailed). As shown in the ANOVA table above the factor F is ratio between the variance between the dependent variable with independent variables and the variance within the variable Return. In the studying case the F ratio is 2.704. Furthermore total squares value is 6294.323. Degree of freedom -df of regression formula is 4 therefore the mean squares is equal sum of square divided by degree of freedom (7.573).

Results and Interpretations of Multiple OLS Regression Models
The most information and result the analysis process are presented in the following Table. Furthermore, as shown in the table the given model, the variables named TUE and FRI are more significant than the rest. With regards to Tuesday, the most statistically significant variable amongst explanatory ones, we find a negative ratio between Tuesday and the return. The p-value indicates a statistically significant relationship between Tuesday and the return. Furthermore, the coefficient value of Tuesday is around -0.232 that means the relative returns on Tuesday 0.232 less than returns on other days of the week relatively. The result is quite similar to output of researches performed by Truong Dong Loc (2012) in Ho Chi Minh stock market, Vietnam; Jaffe and Westerfield (1985); Dubois and Louvet (1996); Balaban at al. (2001).
Regarding to Friday, the result of the studying shown that Fridays have insignificant impact on the return yields. In particular, Fridays have positive relationship with returns. As can be seen from the table returns on FRI is 0.126, which is relatively higher than return on other days of the week. However, the p-value of Friday is 0.03, which is quite large compared with other days and represents an insignificant relationship. But, the positive result is consistent with precious researches not only in developed markets but in emerging markets, such as Truong Dong Loc (2012) in Ho Chi Minh market, Vietnam; Kiymaz and Berument (2003) in US and Canada; Wong and Yuanto (1999) in Jakata.

Results of Testing the January Effect
After running the regression model, the following result has been produced: Valid N (listwise) 8 The Table 5 provides information about the minimum, maximum and average figures of the months during the period from 2006 to 2014. As can be seen from the descriptive statistics, the mean return of January was 7.14%, which was the highest average monthly return that was recorded for this period of time. Meanwhile, for some other months such as May, June, July, October and November in the year, Vietnamese stock market witnessed negative returns with October suffering from the lowest average return. Therefore, it can be determined that January effect existed in the securities market of Vietnam over the period of 8 years from 2006 to 2014.
The SPSS software uses Durbin-Watson to test the conformity of the factors which affect the monthly returns. The results are shown as below: The Durbin-Watson coefficient aims to investigate the incident of autocorrelation in the model. The condition is that if the coefficient is within the range of from 1 to 3, it means there is no autocorrelation phenomenon; otherwise, the phenomenon exists. According to the Table 6, the value of Durbin-Watson is 1.276, in the range of 1-3; consequently, the model has no autocorrelation phenomenon.

Conclusion
As stated above about the efficiency of the market, and according to the principle of the "random walk" in the efficient market efficiency, the future of securities is unpredictable. Actually, the future price fluctuates in no patterns and which are independent, or they move in random walks (Brealy and Myers, 1996). However, the results of this study are similar to the previous ones conducted in the past. Particularly, according the collected data from 2006 until the end of 2014 from Vietnam stock exchange market, the study reveals that the movement of securities prices in Vietnam exchange market is abnormal or anomalous. That means there is still a chance for arbitragers to create profits based on forecasting in Vietnam stock exchange market. The day of the week effect in Vietnam market demonstrates that the future securities prices are still able to forecast in some extent.
According to the results of studies, there are chances for investors making profit in Vietnamese market by purchasing stocks on Tuesdays and selling them on the afternoon of Fridays. An individual investor would postpone planned purchasing stocks on Thursdays and Fridays until Tuesdays when the price is lowest of the week. On the other hand, the sellers would wait until Fridays to conduct the transactions when the prices get the highest point. From the results of the regression model that has been run in the precious charter, it can be concluded that the January effect had been present when the monthly returns of VN-Index during the period of 2006-2014 were taken into account. This finding about Vietnamese stock market is similar to that of other researches that have been undertaken about other developed and emerging securities markets all over the world. Based on the results, it is undoubted that the efficiency of Vietnamese stock market is weak.