The Weekend Effect Investigation: Evidence from Indonesia Capital Market

In the weak-form market efficiency, investors cannot obtain profits by trading stocks through a positive price difference. Unfortunately, this perspective is doubtful because of the weekend effect theory. The highest and lowest return exists on Friday and Monday, respectively, based on this theory. For that reason, this study is presented to prove the weekend effect by statistically checking the return based on the trading days with the variance analysis model and the Tukey Honestly Significant Difference Testing. Unlike the other research, this study utilizes the shares' return to prove this effect in the bullish and bearish markets. Data are collected by archival method from the Yahoo Finance website based on the names of shares chosen as the LQ45 members. The workdays and holidays are obtained based on the information from the Indonesian stock exchange. This finding shows that the lowermost return exists on Tuesday for two markets. Separately, the reverse Monday and the middle-week effects exist in the bearish and the bullish. The return change in the bullish period is more significant than that in the bearish. Therefore, this study is helpful for public investors to transact their shares to gain a positive return by utilizing these daily patterns in these two markets.


Introduction
The stock price becomes the attention for the investors to analyze because of a capital gain [1]. In this context, investors can do it, either fundamentally or technically. They can recognize the fair price to buy and sell stocks by fundamental analysis. Utilizing technical analysis, they can purchase and sell the stocks based on some market indicators on the chart [2]. These analyses can be applied if the market is inefficient [3]: All price changes are predictable [1]. Many scholars attempt to investigate this market efficiency by serial market return testing, and their results are still various, classified by two groups. The first is the supporting one covering the research proving the random movement, as Capobianco et al. [4], Al-Ahmad [5], Kushwah et al. [6], Camba Jr. and Camba [7], and Sahoo et al. [8] execute. The second is opposing: the studies verifying the forecasted pattern (see Lo and MacKinlay [9], Onour [10], Al-Jafari and Altaee [11], Haque, Liu, and Nisa [12], Mudassar, Ali, Nawaz, and Shah [13], Ananzeh [14], Ananzeh [15], Rehman, Kashif, Chhapra, and Rehan [16], Kelikume, Olaniyi, and Iyohab [17]).

1.
French [29] The United States By employing the S&P 500 market index between 1953 and 1977, this study shows Wednesday provides the highest return, but Monday gives the lowest negative.

Pettway and Tapley [33] Japan
In their study, they use three market indexes, i.e., the Nikkei Dow Jones (NDJ), Tokyo Stock Exchange index based on simple average (TSE/SA), and Tokyo Stock Exchange Index based on weighted average (TSE/WA), from January 4th, 1979, to April 3rd, 1982. After examining the pattern, this study reveals that the lowest return happens on Tuesday, but the highest return occurs on Wednesday.

Tandelilin and Algifari [34] Indonesia
After examining the most 40 dynamic stocks based on trading frequency from January until December 1996 in Indonesia, this study finds the most profound negative return occurs on Monday. Conversely, the most significant positive return is on Wednesday.

Mills et al. [38] Greece
Once investigating the daily market index of the Athens stock exchange between October 1986 and April 1997, this study locates the lowest negative return is on Wednesday. However, the highest positive return is on Friday.
Unlike the previous studies mentioned above, this study wants to prove the weekend based on two market situations: bearish and bullish. According to Jones [2], the bearish is a diminishing trend in the capital market: the market price index cannot exceed its previous uppermost limit or the decline in the market index that passes its lowermost limit. However, the bullish is increasing drift in the capital market, where the market price index can surpass the preceding upper limit. Also, if the market index goes down, it will not outdo its lowest frontier. Furthermore, the shares of the LQ45 index in the Indonesian stock exchange become the focus of this research. This situation exists because the stocks in this index are actively traded by the investors and meet the efficient market-related testing necessity [1].
Besides, the motivation to take this issue is due to the limited literature reviews in Indonesia (see Asnawi et al. [47]) and the non-Indonesia context (see Blose and Gondhalekar [48], Obadele and Muzindutsi [49]). Only Obadele and Muzindutsi [49] prove the weekend effect in the bearish market when verifying the market index in Nigeria and Casablanca in Africa between 1998 and 2017. Unfortunately, the rest cannot prove this effect. Employing the daily indexes in the Indonesian capital market from 2018-2019, Asnawi et al. [47] find no difference in Monday and Friday returns in bullish and bearish markets. Utilizing the international price of the gold market from 1975 to 2011, Blose and Gondhalekar [48] find the return among the days in a week is the same during a bull market. However, in a bullish market, the return on Friday and Monday is lower than that of the weekdays.

Literature Review and Hypothesis Development
Cross [19] investigated the S&P 500 index returns from 1953 to 1970 to prove the weekend effect. His finding demonstrated that the mean return on Fridays exceeded that on Mondays. In his research, French [29] divided the S&P index return into five periods: 1953-1957, 1958-1962, 1963-1967, 1968-1972, 1973-1977. He found that the uppermost average return on Friday and the bottommost average return on Monday only happen from 1958 to 1962 and 1963 through 1967. In their study, Lakonishok and Levi [20] learned the market trading pattern between July 1962 and December 1979 by employing the regression model. They found that the market gave the most resounding negative return on Monday but the most profound positive return on Friday. Moreover, Lakonishok and Maberly [21] explained the reason for the negative Monday return. They argued that the institutional investors transact less, and the individuals aggressively sold their stocks on Monday.
Another effort to validate the weekend effect came from Wong, Hui, and Chan [50] by utilizing the data from the capital market in Singapore, Malaysia, Hong Kong, Taiwan between January 1975 and May 1988, and Thailand between May 1975 and May 1988. After testing the data, they concluded that the weekend effect existed for Singapore, Malaysia, and Hong Kong, reflected by the minimum negative return on Monday and the maximum positive return on Friday. The lowermost negative return was on Tuesday for Thailand, and the uppermost positive return was on Friday. No significant return based on days was available in Taiwan. Abraham and Ikenberry [22] investigated the same trading array using the regression model. Using the data from 1963 to 1973, they confirmed that the bottom negative market return occurred on Monday; however, the peak positive return was on Friday.
Utilizing the data between 1987 and 1994 from the Bombay stock exchange, Poshakwale [23] rejected the market efficiency theory in a weak form by proving that the market index return is not random. Instead, this study confirmed the weekend effect: the highest positive return exists on Friday; unfortunately, the lowest negative return occurs on Monday. Correspondingly, by operating the daily market return in Finland from 1989 to 1990, Martikainen [57] attempted to prove the weekend effect on three indexes of the Shariah-compliant products, i.e., the Kuala Lumpur Shariah Index (KLSI), FBM Emas Shariah Index, and FBM Hijrah Shariah Index. They found that the weekend effect is only available on KLSI: the lowest negative return exists on Monday, and the highest positive return happens on Friday. Similarly, after studying the second-period data from January 2005 to December 2014 in the Karachi stock market, Raza et al. [28] effectively verified the weekend effect.
Mongrut and Delfino [58] stabbed to validate the weekend effect in the capital market in the Latin American countries: Peru, Argentina, Brazil, Colombia, Chile, and Mexico by utilizing the daily market return between January 3rd, 2005, and December 31st, 2014. After examining the data, they found that the Monday and Friday effects only occur in Peru, Colombia, Brazil, Chile, and Mexico, and for the rest, this effect does not exist.
In their research applying the African capital markets, Obalade and Muzindutsi [49] wanted to prove the weekend effect by employing five indexes representing the states: Nigerian stock exchange all-share index (NGSEINDX), JSE Africa All-share index (JALSH), the Stock Exchange of Mauritius all -Share index (SEMDEX), Casablanca stock exchange all-share index (MONENEW), Tunisia Stock Exchange All -Share Index (TUSISE). Once checking the data, they only proved that the weekend effect happens in a bearish market in NGSEINDX and MONENEW.
Assuming the return pattern in the weekend effect happens in the bullish and bearish markets, we propose the first and second hypotheses in this way.
H 1 : If the weekend effect happens in a bearish market, the average Monday return is negative, but the Friday return is positive.
H 2 : If the weekend effect happens in a bearish market, the average Monday return is negative, but the Friday return is positive.

Research Method
This study is designed quantitatively because of the investigated hypothesis [59]. As a result, the research object and the number have to be known as the population [60]. We use the firms periodically to become the LQ45 index constituent from February 1st, 2015, until January 31st, 2019. Based on this situation, 30 firms are available, with their names in Table 2. By denoting Suliyanto [60], we utilize the Slovin formula to get the total representing sample (TRS). Furthermore, this formula includes the total population (TP) and the fault margin (FM), as seen in the first equation. (1) According to the fault margin of 5%, the TRS is = 27.907 ≈ 28 firms. Additionally, This study uses an analysis of variance (ANOVA) model to investigate the weekend effect. According to Ghozali [61], this analysis tests differences in single variables based on certain levels (D_LEVELS). In this context, the single variable is the stock returns of firms becoming the LQ45 Index constituent. The factor intended is five trading days: Monday, Tuesday, Wednesday, Thursday, and Friday. Moreover, the ANOVA model must achieve the assumption of normality and variance homogeneity. The Kolmogorov-Smirnov and Levene tests are to examine the first and second assumptions.

Results
By following the number of stock trading days set on the Indonesia capital market adjusted to this study analysis period, it gets obtained that 968 days exist from February 1st, 2015, to January 31st, 2019, consisting of 192, 195, 198, 193, and 130 days from Monday until Friday. The information covering the average, standard error, standard deviation, and coefficient of variation of relative return based on stock trading days is available in Table 3.
According to the normality test result, the asymptotic significance (2-tailed) on the Z-statistic of the Kolmogorov-Smirnov exhibits 0.000 for bullish and bearish (see Table 4). This condition indicates the null hypothesis rejection because this value is lower than a 5% significance level. Thus, relative returns are not normally distributed. Fortunately, according to Ghozali (2016), this condition can still be tolerated because the ANOVA is robust even though the utilized data do not follow normality distribution. Asymptotic significance (2-tailed) 0.000 0.000 Sig. (F-statistic) 0.000 0.000 Note: If the relative return is above one, the total return is positive, and vice versa. Table 5 exhibits the result of the error variance homogeneity, with the F-statistical significance value of 0.0000 for bullish and bearish markets. This situation indicates that the null hypothesis is disallowed because this value is below α of 5%. Thus, the day group-based residual variance of the relative return is heterogeneous. Although this assumption is violated, the ANOVA model is still acceptable as long as the number of samples is the same for each group [61]. In this research context, the number of pieces is 28 firms. Therefore, the ANOVA model is still relevant to be estimated.
Furthermore, the homogeneous subset is employed to know the same trading day-based relative returns for the bullish and bearish markets based on the Tukey honestly significant difference (HSD) testing. In subset one for a bullish market, one trading day, i.e., Tuesday, exists. In subset two, two trading days are available: Monday and Friday, which means the relative return between these days is the same. In subset three, three trading days are available: Friday, Thursday, and Wednesday. It means the relative return on these days is alike (see Panel A in Table 6).
Moreover, Tuesday and Thursday are present in the first subset for the bearish market: the relative return between these days is similar. Meanwhile, Thursday and Wednesday are available in subset two: there is no difference in the relative return. Furthermore, Wednesday, Friday, and Monday exist in the third subset: the relative returns among these days are the same (see Panel B in Table 6).
According to Rystom and Benson [18], this effect happens when the Monday return is negative and lowermost, and however, the Friday return is positive and uppermost. By standing for this explanation, this study does not support this effect but demonstrates another pattern for a bullish and bearish market.
The highest negative return on Tuesday in the bullish market means the Monday effect transmission from the US capital market to the Asian stock exchange markets, as Aggarwal and Rivoli [63] emphasize. Furthermore, Aggarwal and Rivoli [63] explain that this situation is due to the time difference. Besides, the Indonesian capital market index is dominantly affected by the American market index, such as Dow Jones Industrial Average [64]. On Wednesday, the highest positive return occurs. According to Tandelilin and Algifari [34], this condition exists because numerous investors take profits to realize their investment strategy based on the information obtained in advance. This situation implies that investors can use the return difference to get the positive price difference by purchasing the stocks on Tuesday and selling them on Wednesday.
The highest return on Monday in the bearish market indicates the contrary of this day's effect. Therefore, this study confirms the study of Mehdian and Perry (2001), Raj and Kumari [36], Chia and Liew [37], and Nipani and Greenhut [32] demonstrate without market division. In their study, Mehdian and Perry [31] find a positive effect in the stock index with a large market capitalization.
One of the studies having a similar approach to this research belongs to Asnawi et al. [47]. However, they prove the market is efficient in the weak form in Indonesia; therefore, the weekend effect does not exist: Monday and Friday have similar returns. Although this research cannot prove the weekend effect, it successfully another return pattern based on bullish and bearish markets. The different results perhaps come from the number of time observations and the unit analysis used. Unlike Asnawi et al. (2021), using two years as the period and the indexes as the unit analysis, this research utilizes six years and the shares forming the LQ45 index.

Conclusion
This study discards the efficient capital market theory in the weak shape and the weekend effect by utilizing the LQ45 members with the trading pattern from February 1st, 2015, until January 31st, 2019. Instead, this study finds that Tuesday gives the negative and minimum return in the bullish and bearish market. Meanwhile, the maximum return in the bullish and bearish market exists on Wednesday and Monday, respectively. Additionally, the change in return in bullish is still more significant than that of a bearish market. By this evidence, investors can utilize the return difference to take advantage of trading in two ways. Firstly, they can purchase the stocks on Tuesday. Then, they can sell stocks on Wednesday in bullish and execute a similar thing the following Monday in bearish.
However, despite finding the other configuration, this study is still inadequate in some of the following aspects, leading to some academic implications. The first aspect is short-period utilization. Additionally, to follow up on this condition, the subsequent investigators can lengthen the period to six years or more to capture the weekend effect in the Coronavirus disease 19 pandemic duration. The second aspect is the LQ45 index utilization, where the firm's shares exist. Besides, this research suggests that the following academics use another index with more firms as its constituents, such as the Kompas 100 index, or combine the shares with the quick-moving prices forming the relevant index in the Southeast Asian countries to handle this limitation. The third aspect is that the statistical test is only for the difference in return and does not involve the days-based trading volume. Hence, the other scholars are expected to execute them. The fourth aspect is the ANOVA model utilization, only presenting the return difference test result based on each trading day without the volatility. Hence, the subsequent scholars can utilize the GARCH or ARCH model to examine both of them.