Bachelor Thesis
The influence of payroll disbursement date on the ‘turn-of-the-month effect’:
a comparison between India and The Netherlands
Name: Ruben van den Eshof Student number: 10252584
Specialization: Economics and Finance Field: Finance
Supervisor: Rob Sperna Weiland
Research question:
Does the date on which salaries are paid in a country determine the time span
that the turn-of-the-month effect occurs?
Contents
1. Introduction ……… 3
2.1 Calendar effect……… 4
2.2 The turn-of-the-month effect……… 5
2.3 Explaining the turn-of-the-month effect……… 6
3. Methodology……… 8
4. Data……… 12
5. Results………. 15
6. Conclusion………. 21
Statement of Originality
This document is written by student Ruben van den Eshof (10252584) who declares to take full responsibility for the contents of this document. I declare that the text and work presented in this document is original and that no sources other than mentrioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsibly solely for the supervision of completion of the work, not for the contents.
1. Introduction
One of the anomalies in stock market returns is the so called ‘turn-of-the-month-effect’. This anomaly refers to the finding that, during the turn-of-the-month (the last day of the current month, and the first three days of the new month) the average return on stock is higher than during the rest of the month. General theory on efficient markets do not support the existence of such anomalies, and there is no widespread consensus about the cause of this effect.
Since its discovery by Ariel (1987), the turn-of-the-month effect has been proven to exist in the market indices of countries in several parts of the world. A popular explanation for this effect during the end of the month is the excess cash balances held by banks and pension funds, caused by remuneration payments during the end of the month by most firms (Ogden 1990). This suggests that the occurrence of the turn-of-the-month effect is correlated with the date of payroll disbursements. Not every modern producing economy has, however, the same general payroll date scheduled. Although most Western European countries have a synchronized time span for these transactions, India's legislation surrounding the payment of employees determines that salaries should be paid after the 10th of every month. These different payment dates enable us to make a statistical comparison between the two countries, and have a closer look on the specific impact of the payment of salaries on the turn-of-the-month effect. We hypothesize that the turn-of-the-month effect will occur during an earlier stage in the month in India than in The
Netherlands. Whereas in The Netherlands it will take place during the end and beginning of the new month, in India it will happen half way during the month.
In order to investigate this matter we will take a look at historical closing prices of the largest market index in India, The Bombay Stock Exchange (BSE). For the sake of comparison, we will do the same for the index of The Netherlands: The
Amsterdam Exchange Index (AEX). In section 2, we will construct a literature framework on past research and potential explanations of the turn-of-the-month
effect. In section 3, we will compose a methodology to statistically investigate the significance of a potential turn-of-the-month effect. After that, we describe the data used in this research and disentangle its relevant details. In the result section,
statistics will show the existence of a turn-of-the-month effect in both countries, and a regression will test its significance. Finally we will conclude our statistical research and debate on our findings.
2.1 Calendar effect
The turn-of-the-month effect is part of a larger group of patterns in stock returns that seem inconsistent with Efficient Market Theory. These patterns are called ‘calendar-effects’. This group of calendar anomalies also includes for example the January-effect, which refers to the tendency of stock prizes of small cap companies being higher than the rest of the index during the month January. The major
explanation for this is the characteristic of investors to sell part of their shares in December, in order to be able to report 'capital losses', and create a tax shield for their investment earnings. In the beginning of January they reinvest this money, and the higher demand for stock creates higher prices (Jaffe & Westerfield, 1985). Another calendar anomaly is the holiday-effect. This can be seen during the last day before a hiatus in trading. During this last day the stock prices will close with higher returns than on normal trading days. These returns are substantially higher, and are seen as the largest calendar-effect that exists. Clarifications for this
phenomenon are mostly sought in the behavioral economics. Good mood of investors due to the upcoming holidays would increase optimism in the market place. (Lakonishok & Smidt. 1988). Closely related to both of these effects, there is the turn-of-the-year-effect. During the first 10 trading days of the new year, average return is 3 times higher than standard daily return. (Lakonishok & Smidt, 1984).
2.2 The turn-of-the-month effect
The turn-of-the-month-effect was recorded for the first time by Ariel (1987). During his research he found a significant positive effect using 19 years of stock data. Although he concludes that the turn-of-the-month-effect is by some fraction correlated with other calendar effects (Ariel, 1987), no complete explanation is presented in his paper that explains the statistical anomaly. One might expect that after the discovery of such a calendar effect, investors start to exploit this to such a degree that this effect ceases to exist. As Fama (1970) describes in his paper, even in its weakest form the Efficient Market Hypothesis dictates that all past prices are absorbed in the prices of today. This would make an mispricing like the turn-of-the-month-effect impossible to exist in the long run.
Some argue that the Efficient Market Hypothesis still holds, and that calendar effect are the result of statistical errors and data snooping. White (2000) describes this practice as: ‘’An extensive specification search that observes a pattern with no forecasting causality, but is instead just luck’’ (White, 2000). More specifically he states that economics, and in particular finance is a domain where data reuse has become a standard practice (White, 2000).
In order to correct for such possible snooping behavior Agrawal and Tandon (1994) conducted a similar empirical study over 18 countries. This research seems free of data reuse, since variables and patterns that are apparent in multiple countries with different market conditions are more likely to be true (Bowers & Dimson 1988). They find strong evidence that in 14 of the 18 countries a significant turn-of-the-month effect is present. Close to 70% of the average monthly returns was accrued during the turn-of-the-month period. They do find, however, that the effect is more apparent in the 1970s and is fading during the end of the 1980s. This decreasing effect could suggest that the Efficient Market Hypothesis holds and the turn-of-the-month effect was exploited by investors after the first papers on the subject were published. Contradictory, Lakonishok and Smidt (1988) found that data
from a 90-year sample of the Dow Jones Industrial Average showed persistent evidence for ‘the- holiday-effect’, ‘the-weekend-effect’, ‘the turn-of-the-month effect’ and the ‘end-of-December-effect’. In addition Jaffe and Westerfield (1989) find (weaker) evidence that some turn-of-the-month effect is present in Australia, the United Kingdom and Canada.
Aside from the data snooping concern, critics suggest that significant calendar-effect results can be caused by erroneous ways of testing. During
examination of the robustness of the weekend-effect, Connolly (1989) states that his specification tests show great differences with results obtained with OLS methods. Also, the strength of the weekend-effect depends on the statistical method of testing (Connolly, 1989). In addition, Sullivan et al. (1998) show that if you adjust p-values for events such as data snooping, no significant calendar effects are existent.
With these thoughts in mind, Kunkel et al. (2003) conducted a non-parametric empirical research for 19 countries (8 European, 6 Far East and 2 Latin American, 2 North American and South Africa) with 12 years of stock returns, correcting for potential OLS mistakes or the use of erroneous statistical approaches. They found that in 15 of the 19 countries a turn-of-the-month effect was still present, and that this effect is good for 87% of the average monthly returns. Around the same period, Booth et al. (2001) proved that the turn-of-the-month effect is still present in
Finland. Instead of returns, they used trading volume and number of trades. Using this approach, they were able to show that the turn-of-the-month effect is caused by huge orders of stock placed during the end of the month.
2.3 Explaining the turn-of-the-month effect
One possible explanation of the turn-of-the-month effect is that the effect is related to the practice that firms often report positive earnings-related news at the
beginning of the month, and negative news about earnings during the second half of the month. The explanation for this pattern is simply that firms show delaying
behavior whenever they have news that might devastate the stock price (Penman S., 1987). This could explain why stock prices have a higher average return during the turn-of-the-month.
The results of Booth (2001) in Finland, however, imply that institutional investors are the force behind the turn-of-the-month effect. These are the only players who can consistently accumulate great balances of excess cash at the end of the month. This is the time salaries are paid in the developed Western world. The salaries will be wired to the banks, leaving them with much more liquid assets than necessary for their liquidity requirement during this time. Also, pension funds receive their monthly contributions and will reinvest these as soon as possible to ensure the highest return. All in all, demand for stock is excessively higher during this period. In the concluding remarks of Lankonishok and Smidt (1988) they assemble the turn-of-the-month to be caused by patterns in cash flows of institutions during the end of the month as well.
The turn-of-the-month effect has been shown to be apparent in several countries in Europe, North and South America and Asia in research done by (Cadsby & Ratner, 1992) and (Kunkel et al, 2003). This suggests that the effect is not a result of data snooping for just one specific country and is not caused by data mining. On the other hand, the absence of the turn-of-the-month effect in for example Japan, Italy and France implies that the effect is not simply caused by spillovers between countries. It underlines that local institutions and their practices in terms of transaction dates are the main cause. This suggests a high correlation between payout dates of salary and the presence of the turn-of-the-month effect (Cadsby & Ratner, 1992).
Therefore, the comparison of the turn-of-the-month effect in India and The Netherlands could result in concluding remarks about its cause. The differences in pay out dates of salary and the resulting effects on stock returns could be supporting evidence that the effect is indeed caused by excess cash balances of institutions. We hypothesize that the turn-of-the-month effect will take place on other days of the
month in India than in the Netherlands. To be precise, the Dutch turn-of-the-month effect will be during the actual turn-of-the-month, and the Indian turn-of-the-month effect will be the period after the 10th of the month (this is the period that salaries are paid in India, and therefore institutional investors have excess cash balances). From this point, we will refer to this period in India after the 10th of the month, as the ‘half-of-the-month’.
3. Methodology
The countries studied will be The Netherlands and India. The choice for these 2 countries has several reasons. First, even though past research has shown that the turn-of-the-month effect is apparent in all parts of the world, their differences in geography, culture, politics and economic system make it interesting if this
phenomenon applies in these two polarized circumstances. The difference in politics and economy are reflected in how the countries are each intertwined in the world economy. Whereas the Dutch economy has been one of the worlds’ leading economies in terms of innovation for decades, India is an upcoming market. It has experienced fast growth after the abolishment of strict economic ruling and
liberalization in the beginning of the nineties. Between 1990 and 2005, the country has even experienced economic growth of 6% annually (Kohli, 2006). Most laws, however, are not modernized in comparison with leading economies, and democracy is not at its peak (Kohli, 2006). This causes limited integration between India and other large economic power in terms of legislation.
According to the Indian Payment of Wages Act 1936, which extends to the whole of India, the payment date of salaries is determined by company size. When a firm has over a thousand workers, the payroll is disbursed on the 10th of the month. If a firm has less than a thousand workers, the payment is done at the 7th of every month. In addition, many companies pay their employees in either cash or by cheque. In The Netherlands salary payments are paid the 24th of every month by
government instances (Wet op Loonbelasting 1964). In addition, this date has been synchronized by almost every employer in the country (which is consistent with most of the Western world). After processing the deposits and retrieving the pension contribution, institutional investors will be left with excess cash balances at the end of every month due to incoming transactions (Booth, 2001).
A study conducted by the CITI group approximated that only 18% of the market capitalization of the Bombay Stock Exchange 200 largest firms is foreign investment. In addition, there is ceiling to foreign institutional investments. Only 24% of all share capital of BSE holdings can be held by foreign institutional investors. This implies that the major part of shares in Indian companies are held by domestic institutions and investors. This is in line with several studies concerning the home bias of investors – the tendency to invest a suboptimal part of their portfolio in domestic equity. The concept of home bias was documented initially by (French & Poterba, 1991). They mostly lay the focus of its cause on investor behavior. Investors tend to rate the risk of foreign investment (incorrectly) higher than domestic
investments, which leads to underdiversified portfolios. Noteworthy is that Indian households tend to invest very little of their income in stock. Only 1.2 billion dollar is invested by all Indian households in the BSE Sensex in 2011, which less than 1% of the market capitalization (Bose & Suchismita, 2012).
In order to emphasize the different time periods in which the presumed ‘turn-of-the-month effect’ will take place in India and the Netherlands, we give the
phenomenon a different name for each country. First, we define this name for India. Since the salaries are paid on the 10th of the month, we expect the effect to take place between the 11th day of the month and the 14th day of the month. We call these days +11, +12, +13 and +14. Because these days are not during the actual turn-of-the-month, we call the potential effect the ‘half-of-the-month effect’ from this point for India.
For the analysis of a possible ‘turn-of-the-month effect’ for The Netherlands and a possible ‘half-of-the-month’ effect for India, we analyze the closing prices per
day of the biggest market indices in The Netherlands and India. First, we will take the simple mean return for the turn-of-the-month (TOM days) in The Netherlands and the ‘half-of-the-month’ days in India (HOM days), and compare them to the rest-of-the-month days (ROM days) in each respective country. Examining the differences in mean return we can already conjecture whether a TOM/HOM effect is there. After this we perform a regression. We use a methodology similar to the one used by Lakonishok & Smidt (1988) and Kunkel et al. (2003). First we want to see if a
significant positive return is actually present on these days. We use the following the 16 trading days around the ‘half-of-the-month’ in the following regression by OLS, where +11, +12, +13 and +14 are the ‘half-of-the-month’ days, and rest are ‘rest-of-the-month’ days.
𝑹𝒕 = (𝛽+5∗ 𝐷+5) + (𝛽+6∗ 𝐷+6) + (𝛽+7∗ 𝐷+7) … + (𝛽+18∗ 𝐷+18) + (𝛽+19∗ 𝐷+19) + (𝛽+20∗ 𝐷+20)
In this equation ‘𝑹𝒕’ is the expected return on day ‘t’. In this equation, ‘t’ can take the value +5, +6, +7 … +18, +19, +20. These values relate to the 5th, 6th, 7th … 18th, 19th and 20th day of the month. In addition, ‘D’ is a binary dummy variable that either takes the value 0 or 1, depended on the day of the month. For example, on the 5th day of the month (day +5), 𝑫+𝟓will take value 1. In every instance, only one of these
dummy variables will take the value of 1, and the others will take the value of 0. The beta’s ′𝜷+𝟓′ … ′𝜷+𝟐𝟎′ are the coefficients of the daily effects. We defined that during day +11, +12, +13 and +14 the ‘half-of-the-month effect’ will take place.
Therefore, if the coefficients of 𝜷+𝟏𝟏, 𝜷+𝟏𝟐, 𝜷+𝟏𝟑 and 𝜷+𝟏𝟒 are significantly positively
different from zero, we have established a ‘half-of-the month effect’ on these days. Results are shown in table 5.
Assisted with the results of this regression, we can now test the ‘half-of-the-month’ days directly against the rest-of-the-month days. We use OLS regression:
Here, ‘𝑹𝒕’ again is the expected return on day ‘t’ and, ‘𝛼’ is the intercept. When ‘t’
corresponds to a ‘half-of-the-month’ day, ‘𝑫𝑯𝑶𝑴’ will take the value 1. In case of a rest-of-the-month day, ‘𝑫𝑯𝑶𝑴 ′ will take the value 0. If the p-value of ‘𝜷’ shows results that are positively significantly different from 0, we have established a definite ‘half-of-the-month effect’ in India. Results are shown in table 6.
For The Netherlands, we specify the turn-of-the-month period in another way. We use 16 trading days around the turn-of-the-month. We call these days -7, -6, -5 …, +8, +9, +10. We define the turn-of-the-month as the one day before the new month, and the first 3 days of the new month. We call these days -1, +1, +2 and +3. These days are also used by Lakonishok & Smidt (1988). The other days are used as ‘rest-of-the-month’ (Day -7, -6, -5, -4, -3, -2, +4, +5, +6, +7, +8, +9, +10). In order to validly test the statistical significance of the average return on the trading days, we estimate the following regression by OLS:
𝑹𝒕 = (𝛽−7∗ 𝐷−7) + (𝛽−6∗ 𝐷−6) + (𝛽−5∗ 𝐷−5) … + (𝛽+8∗ 𝐷+8) + (𝛽+9∗ 𝐷+9) + (𝛽+10∗ 𝐷+10)
In this equation, ‘𝑹𝒕’ is the return. ‘t’ equals a number that indicates the position of the day in the sequence before or after the turn-of-the-month. For example, +1 is the first day of the new month. ‘𝜷’ is the coefficient of the specific day, and ‘D’ is a
binary dummy variable that takes either 0 or 1, dependent on the day of the month. When it is day +9, ‘𝑫+𝟗’ will take the value of 1, and the other D’s will take the value of zero. The beta’s are the expected effect of the specific day of the month. When the beta of a turn-of-the-month day (So 𝜷−𝟏, 𝜷+𝟏, 𝜷+𝟐 and 𝜷+𝟑 ) is significantly positive, we determine that the turn-of-the-month effect is present. Results can be found in table 7.
Once we obtain the results of this regression, we use a second regression to assist us in identifying a significant turn-of-the-month effect. We use a method that was also applied by Cadsby and Ratner (1992). We estimate the following regression equation:
𝑹𝒕 = 𝛼 + 𝛽𝐷𝑇𝑂𝑀+ 𝜀𝑡
Again, ‘𝑹𝒕’ is the return for a specific day. ‘α’ is the average return during a ‘rest-of-the-month’ day. The ‘𝜷’ is the coefficient of the turn-of-the-month effect, and ‘𝑫𝑻𝑶𝑴’ is a qualitative variable that is equal to 1 during a ‘turn-of-the-month’ day, and 0 during a ‘rest-of-the-month-day’. We test if the coefficient of the beta is significantly different from zero. Once we find evidence that this is true, we have proven the presence of the turn-of-the-month effect in The Netherlands. Results on this test can be found in table 8.
4. Data
As used in similar empirical experiments conducted by Cadsby & Ratner (1992) and Ziemba (1991), we use the largest national market index as source of data. We obtain the daily end-of-day index prices of active trading days from
finance.yahoo.com. The indices we use are the Amsterdam Exchange index (AEX) for The Netherlands, and the Bombay Stock Exchange Sensex (BSESN) for India. We prefer the BSESN over the National Stock Exchange India (NSE) because the BSE is one of the the most advanced stock exchanges in Asia with the highest trade volume (leading to larger potential TOM effects), and it is the biggest exchange in India in terms of market capitalization. The BSESN, also called the BSE 30 are the 30 largest and highest trade volume shares of the BSE. This makes this index better to compare with the AEX in terms of size, and a better indicator of the turn-of-the-month effect because of its activity. We will use data from the past 20 years. The turn-of-the-month from December to January will be excluded from the data, since this might bias the results due to the ‘End-of-the-year’ effect (Cadsby & Ratner 1992) and (Agrawal & Tandon 1994). This gives us 220 turn-of-the-months per country.
Table 1. – General information on national stock indices.
In table 1 we can see the comparative statistics on the data used for this research. The BSESN and AEX have similar market capitalization and a fairly similar amount of stock. Both indices were launched during the same time period. Although the initial cornerstones for the Bombay Stock Exchange were already laid in 1855, since 1957 it is strictly regulated by the Indian government (BSEIndia). The BSE Sensex was
launched in 1986 and contains the 30 largest stock in India. It is seen as the most important index and market indicator in the country (BSEIndia). The Amsterdam Exchange index is the leading index in The Netherlands, and has world famous companies listed in its index (Heineken, Shell). The BSE Sensex and the AEX have similar market capitalization and fairly similar amount of listed stock.
Table 2. – Summary statistics on time period 25 May 1995 – 25 May 2015
* Data from time period 25 May 1995 – 25 May 2015 ** Data from time period 1 July 1997 – 25 May 2015
For the AEX we use data from the past 20 years, and for the BSESN for the past 18 years (due to the public availability of daily closing prices). A close look reveals that the BSESN has been outperforming the AEX for the last 20 years. As seen in table 2, its average daily return is about 3.5 times higher than the AEX return. This is partly due to the fact that the AEX index was one of the worst performing indices in the world during the time period 1998-2008 (AmsterdamTrader), whereas India has been
Country Index Start date Nr. of stock Market capitalization
India BSESN 01/01/1986 30 € 421 billion
The Netherlands AEX 03/01/1983 25 € 491 billion
Country Observations Mean (%) StDev (%) Kurtosis Skewness
The Netherlands (AEX)* 5099 0.0083 1.4495 5.9168 -0.3072
one of the largest growing economies for the past 30 years. From 1998 till 2008, the AEX index increased by 19.4%, which is corresponds to only 1.78% per year. In the same time period, the BSE Sensex netted 9.2% per year on average. In addition, The Netherlands was greatly influenced by the credit crunch and the sovereign debt crisis from late 2007 till 2013. This has caused stock prices to be lower than average.
Whereas the AEX averages 6.5% growth per year since 1988, between 2007 and 2013 the AEX decreased 20.46% in total. On the contrary, between the end of 2007 and 2013 the BSE Sensex gained 24.37%. In figure 1 we can see the outperformance of the AEX index by the BSE Sensex. We depart from 1996 where the base of the index is 100, and see its development over the years from that point up until now.
Figure 1. – Cumulative yearly return from 1997 till 2015 for the AEX and BSE Sensex
For our research, we use Stata and Excel for minor adjustments in the data. We obtain the mean returns for turn-of-the-month days (TOM) for The Netherlands, the half-of-the-month days (HOM) for India and compare them to both countries
respective rest-of-the-month days (ROM). For this table, we define
turn-of-the-month days as day -1, +1, +2 and +3 for The Netherlands (corresponding to last day of the current month, and the first 3 days of the new month). For India we define half-of-the-month days as day +11, +12, +13 and +14. We do this to get a more detailed
0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 900.00 1996 2001 2006 2011
BSE Sensex (India) AEX (Netherlands)
picture about the differences. In addition, this works as a preparation for the regressions still ahead.
Table 3. – General statistics on TOM and ROM for The Netherlands.
Country TOM obs. Mean (%) StDev (%) ROM obs. Mean (%) StDev (%)
The Netherlands (AEX) 594* 0.1134 1.4956 4447** -0.0050 1.4871
* TOM = Turn-of-the-month days, these include days -1, +1, +2 and +3 for The Netherlands (Turn-of-the-year is excluded). ** ROM = Rest-of-the-month days, these days include all trading days except TOM days, and turn-of-the-year days.
Table 4. – General statistics on HOM and ROM for India
Country HOM obs. Mean (%) StDev (%) ROM obs. Mean (%) StDev (%)
India (BSESN) 586* -0.0968 1.5860 3770** 0.0396 1.5736
* HOM = Half-of-the-month days, these include day +11, +12, +13 and +14 for India (Turn-of-the-year is excluded). ** ROM = Rest-of-the-month days, these days include all trading days except HOM days, and turn-of-the-year days.
As shown in table 3, the first results indicate that the turn-of-the-month effect is apparent in The Netherlands. The mean return for turn-of-the-month days is much higher than the return for rest-of-the-month days. In fact, during the TOM days the average return is positive whereas during the ROM days the average return is
negative. In table 4 we can see the mean returns for India. So far we cannot observe a positive half-of-the-month effect. The mean return for half-of-the-month days is negative. In the result section we will dissect the individual days for both The Netherlands and India, and run regressions on the data to reach final conclusions.
5. Results
In this section, we will disclose all the results of the analysis of our data. We conclude whether there is significant presence of a turn-of-the-month effect in The
Netherlands and a half-of-the-month effect in India. First, we look whether the effect will indeed take place at another time frame in India. After that we will examine if the turn-of-the-month effect takes place at the classical time period in The
Table 5. – Results on days around the half-of-the-month for India. India SD (%) 𝜷+𝟓 -0,0831768 (1,5298255) 𝜷+𝟔 0,1123084 (1,4976646) 𝜷+𝟕 0,0922049 (1,5256725) 𝜷+𝟖 -0,0436814 (1,4926456) 𝜷+𝟗 0,1710637 (1,4759489) 𝜷+𝟏𝟎 -0,000051 (1,6201001) 𝜷+𝟏𝟏 -0.2532625** (1.5146047) 𝜷+𝟏𝟐 -0.2671553** (1.5369564) 𝜷+𝟏𝟑 -0.0188355 (1.6902105) 𝜷+𝟏𝟒 0.1753546 (1.5739204) 𝜷+𝟏𝟖 -0.1681011 (1.8746342) 𝜷+𝟏𝟔 0.0651208 (1.439194) 𝜷+𝟏𝟕 -0.2097278 (2.0279441) 𝜷+𝟏𝟖 0.033048 (2.0500629) 𝜷+𝟏𝟗 -0.0653337 (1.5819311) 𝜷+𝟐𝟎 -0.1152577 (1.3044765) * = Statistically significant at α=0.1 ** = Statistically significant at α=0.05 *** = Statistically significant at α=0.01
Table 5 contains the results of the first regression on stock data from the BSE Sensex. We hypothesized a significant positive return on day +11, +12, +13 and +14 due to payroll disbursement dates between on the 7th and the 10th of every month in India. This is what we defined as the ‘half-of-the-month effect’. However, opposing our hypothesis, the significant results are negative instead of positive. We test the half-of-the-month days (HOM days) against the rest-half-of-the-month days (ROM days) to confirm these results.
Table 6. – Results for India of directly testing HOM days against ROM days Country α 𝜷 ∗ 𝑫𝑯𝑶𝑴 India 0.03958 -0.13635 ** (1.5859) * = Statistically significant at α=0.1 ** = Statistically significant at α=0.05 *** = Statistically significant at α=0.01
The regression results shown in table 6 reflect the results of testing half-of-the-month days against rest-of-the-half-of-the-month days. The beta in this regression is the
coefficient of the return of half-of-the-month days. Since the value of this coefficient is significantly negatively different from zero at a level of α = 0.05, there is enough evidence to assume a negative effect on returns during the half-of-the-month. A possible explanation for this negative effect might be the fact that salary payment in cash is still a very regular practice in India. This could suggest that instead of excess cash balances after salary payments, banks are actually left with less cash after the payroll disbursements. Note that ‘𝑫𝑯𝑶𝑴’ takes value 1 for days +11, +12, +13 and +14,
and 0 for the other days. Against our expectations, we therefore suspect that the
payment date of salaries have less influence on turn-of-the-month effect than theory prescribes. In order to confirm this, we perform a regression on stock data for The Netherlands, to see whether a turn-of-the-month happens after the payment of salaries at day -1, +1, +2 and +3. After discovering that the half-of-the-month effect does not exist for India after payment of the salaries, we will also perform this
regression on day -1, +1, +2 and +3 for India. If significant positive returns exist during these days in India, we have shown that payment date of salaries is not the driver behind the turn-of-the-month effect.
Table 7. – Results for days around the turn-of-the-month for The Netherlands. Netherlands SD (%) 𝜷−𝟕 0,1222661 (1,5785685) 𝜷−𝟔 0,143789 (1,1705908) 𝜷−𝟓 0,0137423 (1,2796191) 𝜷−𝟒 0,0006588 (1,2891008) 𝜷−𝟑 0,2538723 * (1,5949034) 𝜷−𝟐 -0,092253 (1,483256) 𝜷−𝟏 0,0367435 (1,3516721) 𝜷+𝟏 0,2184688 (1,7440306) 𝜷+𝟐 0,2146764 * (1,5173827) 𝜷+𝟑 -0,000894 (1,3524655) 𝜷+𝟒 0,0927974 (1,5524353) 𝜷+𝟓 -0,078219 (1,448404) 𝜷+𝟔 0,0666013 (1,5233774) 𝜷+𝟕 -0,021317 (1,3141657) 𝜷+𝟖 -0,079733 (1,4424349) 𝜷+𝟗 0,0102459 (1,1055632) 𝜷+𝟏𝟎 -0,179585 (1,6296157) * = Statistically significant at α=0.1 ** = Statistically significant at α=0.05 *** = Statistically significant at α=0.01
As shown in table 7, we find a significant positive return on day +2, which is the second day of the month. This corresponds to the hypothesis of the presence of a turn-of-the-month effect during these days. However, the results are not to the extent we expected. We directly test the turn-of-the-month days against the rest-of-the-month days to see if the effect holds for the entire time period of day -1 till day +3.
Table 8. – Results of direct test on ROM versus TOM days (-1, +1, +2, +3) for The Netherlands. Country α 𝜷 ∗ 𝑫𝑻𝑶𝑴 The Netherlands -0.00501 0.11841 * (1.4956) * = Statistically significant at α=0.1 ** = Statistically significant at α=0.05 *** = Statistically significant at α=0.01
Table 8 shows the results of testing TOM against ROM days. We find that there is a significant positive return during the of-the-month at α=0.10. During the turn-of-the-month days, ‘𝐷𝑇𝑂𝑀’ will take the value 1, and during the rest-of-the-month days it will take the value 0. Curiously, we conclude that for all the days of the month the mean return is negative, except for the turn-of-the-month days. This implies that on average, all the monthly profits on the AEX are accumulated during the turn-of-the-month. We will now test whether the same is true for India. We test if India also exerts a positive turn-of-the-month effect during days -1, +1, +2 and +3.
Table 9. – Results for the days around the turn-of-the-month for India
India SD (%) 𝜷−𝟕 0,1773413 (1,4408282) 𝜷−𝟔 0,0736844 (1,4114025) 𝜷−𝟓 -0,0104 (1,5621633) 𝜷−𝟒 0,0438775 (1,6208864) 𝜷−𝟑 0,1607499 (1,3949821) 𝜷−𝟐 0,119762 (1,4870145) 𝜷−𝟏 0,318153 ** (1,4990512) 𝜷+𝟏 0,2981487 * (1,6349367) 𝜷+𝟐 0,3515491 ** (1,7043881) 𝜷+𝟑 0,0134137 (1,5421151) 𝜷+𝟒 0,3243001 ** (1,5560595) 𝜷+𝟓 -0,0831768 (1,5298255)
𝜷+𝟔 0,1123084 (1,4976646) 𝜷+𝟕 0,0922049 (1,5256725) 𝜷+𝟖 -0,0436814 (1,4926456) 𝜷+𝟗 0,1710637 (1,4759489) 𝜷+𝟏𝟎 -0,000051 (1,6201001) * = Statistically significant at α=0.1 ** = Statistically significant at α=0.05 *** = Statistically significant at α=0.01
Opposing our expectations, we discover a significant turn-of-the-month effect in India. The results can be found in table 9. The effect is even much more apparent than in The Netherlands. During 3 of the 4 turn-of-the-month days there is a significant positive result in India, where 2 are not even rejected at α = 0.05. In the Netherlands there is only one turn-of-the-month day significant, which is day +2. Remarkably, in both countries we find that day +3 does not give a significant positive result (the results are barely positive at all).
Finally, we test the turn-of-the-month days directly against rest of the month days. Whereas in our last test we compared all turn-of-the-month days individually, this way we can test the effect of the entire period against regular days. This
increases sample size and gives are more detailed picture.
Table 10. – Results of direct test on ROM versus TOM days for India
Country α 𝜷 ∗ 𝑫𝑻𝑶𝑴 India -0.00809 0.24859 *** (1.6479) * = Statistically significant at α=0.1 ** = Statistically significant at α=0.05 *** = Statistically significant at α=0.01
The results reported in table 10 oppose our hypothesis. There is a significant turn-of-the-month effect during days -1, +1, +2 and +3 in India and The Netherlands, but in
India it is most significant (α=0.05). We have discovered that even though the disbursement of salaries is between the 7th and the 10th of the month, the turn-of-the-month effect appears in India during the actual turn of the month. Since most investments in the BSE Sensex in India are from domestic investors, the chance that this effect is a result from a spillover from international investors is slim. Although most economic theory relates the turn-of-the-month effect to the payment date of salaries in a country, the results from this research strongly imply that its influence on the effect is small. It is more likely that the true driver behind this effect is the
tendency of positive macro-economic announcements to cluster during the
beginning of the month. These macro-economic changes are of a more global nature, which would explain the international similarities in the time frame in which the turn-of-the-month effect appears.
6. Conclusion
In this paper we investigated whether the turn-of-the-month effect could be caused by the payment date of salaries, like suggested in previous papers on the subject. In order to test this, we found two countries where salaries are paid on other dates. Whereas in The Netherlands the payroll disbursements occur at the 24th of every month, in India this happens at either the 7th or the 10th. We, therefore, hypothesized that if payment date is indeed an important factor for the turn-of-the-month effect, the phenomenon would occur at an earlier stage in the month in India (day +11 till +14, defining it as the ‘half-of-the-month effect’) than in The Netherlands (day -1, +1, +2 and +3).
We obtained daily closing prices of the national stock indices, and
distinguished ‘turn-of-the-month days’ for The Netherlands, ‘half-of-the-month days’ for India, and compared them with ‘rest-of-the-month days’ with assistance of
several regressions. Opposed to our hypothesis, we found no evidence that a ‘half-of-the-month effect’ occurs in India. On the contrary, we actually found evidence that
the there is a negative ‘half-of-the-month effect’ during these days. We did, however, find significant evidence for a turn-of-the-month effect that occurs during the last and first three days of the month in India. For the Netherlands, we found prove for a turn-of-the-month effect at the same period as well, although it was statistically less significant.
Given the fact that the major part of investments in the BSE are domestic, we think it is unlikely that the turn-of-the-month effect in India is simply the result of an international spillover. Therefore, we conclude that the payment date of salaries is not an important influence on the turn-of-the-month effect. The results obtained in this research, give more support to the theory that the effect is caused by the clustering of positive macro-economic announcements at the beginning of the month. This would leads to higher stock prices in this period. In addition, this very same theory also suggests that negative macro-economic announcements are made half way during the month. This would, in turn, be a clarification for the negative returns we found in India during the half-of-the-month. Any future research should focus on elaborating and testing this theory and its influence on the turn-of-the-month effect, in order to potentially solve one of the seasonal anomalies of stock returns.
Literature list
AEX-index profile - Euronext
Agrawal, A., Tandon, K. (1994), Anomalies or illusions? Evidence from stock markets in eighteen countries. Journal of International Money and Finance. Vol. 13, pp. 83-106.
AmsterdamTrader (2009), – In the long run, we're all dead.
Ariel, A. (1987), A monthly effect in stock returns. Journal of Financial Economics. Vol. 18, pp. 161-174.
Booth, G., Kallunki J., Martikainen, T. (2001), Liquidity and turn-of-the-month-effect: evidence from Finland. Journal of Institutional Financial Markets, Institutions
Bowers J., Dimson, E. (1988), Stock Market Anomalies. Cambridge University Press. p13.
Bose, Suchismita (2012). Mutual fund investment, FII investment and stock market returns in India. Money and Finance. pp. 89-110.
BSEIndia – BSEInda.com
Cadsby, C., Ratner, M. (1992), Turn-of-the-month and pre-holiday effects on stock returns: Some international evidence. Journal of Banking and Finance. Vol. 16, pp. 497- 509.
Connolly, R. (1989), An examination of the robustness of the weekend effect. Journal
of Financial and Quantitative Analysis. Vol. 24 (2), pp. 133-169.
Dodge, Y. (2003), The Oxford Dictionary of Statistical Terms.
Fama, F. (1970), Efficient capital markets: a review of theory and empirical work.
Journal of Finance. Vol. 25 (2), pp. 383-417.
French, R., Poterba J. (1991), Investor diversification and international equity markets. National Bureau of Economic Research. Working paper 3609. Jaffe, J. & Westerfield, R. (1985), The weekend effect in common stock returns: the
international evidence. Journal of Finance. Vol. 40, pp. 433-454.
Jaffe, J. & Westerfield, R. (1989), Is there a monthly effect in stock market returns?
Journal of Banking and Finance. Vol. 13, pp 237-244.
Jarque, Carlos, M., Bera, Anil, K. (1981), Efficient tests for normality homoscedasticity and serial independence of regression residuals: Monte Carlo evidence.
Economics Letters. Vol. 4, pp. 313–318.
Kohli, A. (2006), Politics of Economic growth in India 1980-2005. Economic and
Political Weekly. pp. 1251-1370.
Kunkel, R., Compton, W., Beyer, S. (2003), The turn-of-the-month effect still lives: the international evidence. International Review of Financial Analysis. Vol. 12 (2), pp. 207-221.
Indian Payment of Wages Act 1936.
Lakonishok, J. & Smidt, S. (1984), Volume and turn-of-the-year behavior. Journal of
Financial Economics. Vol. 13 (3), pp. 435-455.
Lakonishok, J. & Smidt, S. (1988), Are seasonal anomalies real? A ninety-year perspective. The Review of Financial Studies. Vol. 1 (4), pp. 403-425.
Lowry, R. (2010), Concepts and applications of inferential statistics. Chapter 12: t-test for correlated samples.
Nikkinen, J., Sahlstrom, P., Aijo, J. (2006), TOM and intramonth effects: Explanation from the important macroeconomic news announcements. Journal of Futures
Markets. Vol. 27 (2), pp, 105-126.
Ogden, J., (1990), TOM evaluations of liquid profits and stock returns: A common explanation for the monthly and January effects. Journal of Finance. Vol. XLV (4).
Penman, S. (1987), The distribution of earnings news over time and seasonalities in aggregate stock returns. Journal of Financial Economics. Vol. 18, pp. 199-228. Sullivan, R., Timmermann, A., & White, H. (1998), Dangers of data-driven inference: a
case of calendar effects in stock returns. San Diego: Department of Economics, University of California. Discussion Paper 98-16.
Wet op de Loonbelasting 1964.
White, H. (2000), A reality check for data snooping. Journal of the Econometric
Society. Vol. 68 (5), pp. 1097-1126.
Ziemba, T. (1991), Japanese security market regularities: Monthly, turn-of-the-month and year, holiday and golden week effects. Japan and the World Economy. Vol. 3, pp. 119-146.
