“Semi-monthly effect in stock returns: new evidence from Bombay Stock Exchange”

Semi-monthly effect is a kind of calendar anomalies which is less explored in the fi- nancial literature. The main objective of this paper to investigate the presence of semi-monthly effect in selected sectoral indices of Bombay Stock Exchange (BSE). The study uses the daily stock returns of five sectoral indices viz S&P BSE Auto Index, S&P BSE Bankex, S&P BSE Consumer Durables Index, S&P BSE FMCG Index and S&P BSE Health Care Index for the period of 10 years starting from 1st April 2007 to 31st March 2017. The data were analyzed using two approaches namely calendar days approach and trading days approach. To test the equality of mean returns for the two halves of the month, Mann-Whitney U test is used. The empirical results of the study did not provide any evidence for the presence of semi-monthly effect in the selected sectoral indices. Nevertheless, BSE Auto Index showed significant difference in the mean re- turns of first half and second half of trading month during the study period. of semi-monthly effect of the holiday The present study focuses on the semi-monthly effect in Indian Stock Market which is relatively less explored than other types of calendar anomalies in the literature. Semi-monthly effect refers to the stock returns for the first half of the month is significantly greater than second half of the month and vice versa. This study was carried out to detect the presence of semi-monthly effect in the select sectoral indices viz S&P BSE Auto Index, S&P BSE Bankex, S&P BSE Consumer Durables Index, S&P BSE FMCG Index and S&P BSE Health Care Index of Bombay Stock Exchange for the period between 2007 and 2017. The analysis was done using both calendar days approach and trading days approach. The results of the study showed none of the selected sectoral indices of BSE exhibited significant difference in the mean returns for the first and half of both calendar month and trading month. However, BSE Auto Index showed significant difference in the mean returns of first half and second half of trading month during the study period. The findings of the study indicate anomalies do not exist currently in Indian Stock Market and it is a sign of market efficiency as far as Bombay Stock Exchange is concerned. This study provides a scope for the researchers to explore other kinds of calendar anomalies in the sectoral indices of BSE.


1.
In financial literature we do not find many studies on semi-monthly effect. However, the significant work on semi-monthly effect in stock returns abroad and in India is quoted in this section. Ariel (1987) pioneered the literatures on semi-monthly effect. He studied US equity market and reported that stock returns earned positive average returns around the beginning and during the first half of the calendar month and zero average during the second part. The study conducted by Penman (1987) revealed that the reason for semi-monthly effect may be the firm's announcement of good news in the first half of the month and the bad news in the second half. In a study conducted by Jaffe and Westfield (1989), intra-month effects were found in Australian market, but not for Japanese, Canadian and British markets. Balaban and Bulu (1996) did not find any evidence of semi-monthly effects in an emerging Turkish Stock Market for the study period between 1988 and 1995. However, when individual years are examined separately, the study reported significant monthly effect in 1994. Arsad and Coutts (1997) employed a large sample of daily returns from the Financial Times Industrial Ordinary Shares Index and documented the existence of semi-monthly effect. Karmakar and Chakraborthy (2000) found that mean returns in the first half of the month was significantly greater than that of second half of the month in Indian Stock Market. Mills et al. (2000) documented significantly higher average return during the first fortnight of the month for the ASE General Index for the period 1986 to 1999. Bahadur and Joshi (2005) did not find strong evidence for semi-monthly effect in the Nepalese Stock Market during the study period. Eleftherios Giovanis (2009) examined different types of calendar anomalies in 55 stock exchanges across the globe and found the presence of semi-monthly effect in Indian and Canadian Stock Exchange. Agathee (2009) studied official Mauritian Stock Market and reported the presence of significant higher stock returns for the first half of the calendar month as compared to the second half for the whole sample period. A study conducted by Mangala and Sharma (2007) revealed significantly high mean daily returns for the first half of the trading month. The study used daily closing prices of S&P CNX Nifty for a period between January 1994 through April 2005. Garg, Bodla, and Chhabra (2010) made an attempt to examine whether calendar anomalies still existed in developed and developing markets. They studied calendar effects such as turn of the month effect, semimonthly effect, monthly effect, Monday effect and Friday effect in the Indian and US markets for the period between January 1998 and December 2007. The analysis of the study confirmed the presence of the semi-monthly and turn of the month effect in both the markets. Swami (2011) examined different types of calendar anomalies in South Asian Markets and found semi-monthly effect only in Indian Stock Market during the study period. Nageswari, Selvam, and Gayathri (2011) examined the presence of semi-monthly effect in Indian Stock Market and concluded that the said anomaly was not present during the study period. Using daily returns of S&P CNX FMCG Index, Shakila, Pinto, and Rohit (2015) tested the presence of semimonthly effect in Indian Stock Market for a period from 2007 to 2013 and the findings of the study did not provide any evidence for the said anomaly.
Abraham (2016) who analyzed Singapore Stock Market from 1995 to 2015 revealed that significant semi-monthly anomaly was not present in the market, even though the mean percentage returns during the first and second half show high relative difference.
Shakila, Pinto, and Rohit (2015) examined semimonthly effect in the CNX Pharma Index of NSE, India, for a period between 2001 and 2013. The results of the study confirmed the presence of semimonthly effect under two approaches viz. calendar day approach and trading day approach.

2.
The present study intends to examine the semimonthly effect in the selected sectoral indices of Bombay Stock Exchange (BSE) covering a period of 10 years from 1 st April 2007 to 31 st March 2017.

Hypothesis of the study
The following hypotheses are tested in this study: H0: There is no significant difference between the mean returns of the first half and second half of the month for the selected sectoral indices of BSE.
H1: There is a significant difference between the mean returns of the first half and second half of the month for the selected sectoral indices of BSE.
where U -Mann-Whitney -test; U 1 n -sample size one; The present study analyzes the semi-monthly in a more recent context. To examine semi-monthly effect and turn of the month effect, the present study uses two approaches viz. calendar day approach and trading day approach.

Calendar day approach
The calendar days for the study period have been identified on the basis of working days of the BSE i.e., from Monday to Friday totaling 2.479 calendar days.
Under calendar day approach, first half of the month includes last two calendar days of the previous month they are thirtieth (30 th ) and the thirty first (31 st ) and then the first (1 st ) to thirteenth (13 th ) calendar days of the following month are considered in total fifteen calendar days. The second half of the month takes into consideration fourteenth (14 th ) to the twenty-ninth (29 th ) calendar days of the month in total, sixteen calendar days.

Trading day approach
The trading days for the study period have been identified on the basis of minimum number of trading days available in a month. The study covers a time period of 120 months. The least number of trading days available in a month during the period of study is 16. Therefore, the total number of trading days identified is 1.920.
Eight trading days before the start of each month (-8 to -1) and eight trading days (+1 to +8) after the commencement of month are considered. The mean returns for 16 trading days are calculated.
Under trading approach, the first half of the trading month includes last trading day of the previous month and first seven days of the following month, i.e. (-1 to 7). The second half begins from the eighth day to the second last trading day of the month, i.e. (8 to -2).

3.
3.1. Analysis of daily returns semimonthly wise (calendar day approach) for selected sectoral indices of BSE

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE Bankex
As shown in However, the results of Mann-Whitney test (P = 0.516 > 0.05) confirm that there is no statistically significant difference between mean returns of the first half of calendar month and the second half. Thus, the null hypothesis is accepted as the mean returns for two halves of the calendar month for BSE Bankex do not exhibit any significant difference.

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE Consumer Durables Index
As depicted in The return distribution is negatively skewed for both the periods. The kurtosis measure for return distribution is Leptokurtic for both the periods during the study period.
However, the results of Mann-Whitney test (P = 0.670 > 0.05) confirm that there is no statistically significant difference between mean returns of the first half of calendar month and the second half. Hence, the null hypothesis that there is no significant difference in the mean returns of first half and second half of calendar month in BSE Consumer Durables is accepted.

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE FMCG Index
As depicted in the second half. Therefore, the null hypothesis cannot be rejected as there is no major variation between mean returns for the first half and second half of the calendar month in BSE FMCG Index.

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE Health Care Index
As depicted in

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE Bankex
As illustrated in Table 7, the first half of the trading month for the Bank Index exhibits mean returns of 0.1059 (median = 0.085, minimum = -8.521 and maximum = 11.60), standard deviation 1.965.
Percentile analysis signifies 25% of the days out of 960 days have returns below -0.847. 75 th percentile implies 25% of the days have returns above 1.094. The return distribution is positively skewed for the first half and negatively skewed for the second half of the trading month. The kurtosis measure for return distribution was Leptokurtic for both the periods in the Auto sector during the study period.
The results of Mann-Whitney test (P = 0.185 > 0.05) confirm that the mean returns for the first half of trading month is not statistically significant compared to the second half. Hence, the null hypothesis that there is no significant difference in the mean returns of first half and second half of trading month in BSE Bankex is accepted.

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE Consumer Durables Index
As shown in The return distribution is negatively skewed for both the period of trading month. The kurtosis measure for return distribution was Leptokurtic for both the periods in the Auto sector during the study period.
The results of Mann-Whitney test (P = 0.755 > 0.05) confirm that the mean returns for the first half of trading month is not statistically significant compared to the second half. Hence, the null hypothesis that there is no significant difference in the mean returns of first half and second half of trading month in BSE Consumer Durables Index is accepted.

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE FMCG Index
As described in

Analysis of descriptive statistics and Mann-Whitney U-test results for S&P BSE Health Care Index
As described in The return distribution is negatively skewed for both the periods of trading month. The kurtosis measure for return distribution was Platykurtic for the first half of the trading month and Leptokurtic for the second half in the Health Caresector during the study period.
The results of Mann-Whitney test (P = 0.227 > 0.05) confirm that the mean returns for the first half of trading month is not statistically significant compared to the second half. Hence, the null hypothesis that there is no significant difference in the mean returns of first half and second half of trading month in BSE Health Care Index is accepted. This study was carried out to detect the presence of semi-monthly effect in the select sectoral indices viz S&P BSE Auto Index, S&P BSE Bankex, S&P BSE Consumer Durables Index, S&P BSE FMCG Index and S&P BSE Health Care Index of Bombay Stock Exchange for the period between 2007 and 2017. The analysis was done using both calendar days approach and trading days approach. The results of the study showed none of the selected sectoral indices of BSE exhibited significant difference in the mean returns for the first and half of both calendar month and trading month. However, BSE Auto Index showed significant difference in the mean returns of first half and second half of trading month during the study period. The findings of the study indicate anomalies do not exist currently in Indian Stock Market and it is a sign of market efficiency as far as Bombay Stock Exchange is concerned. This study provides a scope for the researchers to explore other kinds of calendar anomalies in the sectoral indices of BSE.