Applied Financial Economics, 2003, 13, 553–563

Stability of the day of the week effect in

return and in volatility at the Indian capital
market: a GARCH approach with proper
mean specification
K A U S H I K B H A T T A C H A R Y A * , NI T Y A N A N D A S A R K A R { and
DE B A B R A T A MU KH O P A D HY A Y{
Monetary Policy Department, Reserve Bank of India, Shahid Bhagat Singh Road, Fort
Mumbai – 400 001, India and {Economic Research Unit, Indian Statistical Institute,
203 B.T. Road, Calcutta- 70035, India

This paper examines the stability of the day of the week effect in returns and vola-
tility at the Indian capital market, covering the period January 1991–September
2000. The paper specifies a generalized autoregressive conditional heteroscedasticity
(GARCH) model on returns and introduces separate dummies for days in alternate
weeks in the specification of both the mean and the conditional variance to examine
the robustness of the day of the week effect in return and in volatility within a
fortnight. Results are compared to those based on ordinary least squares (OLS)
procedure to examine how erroneous the inference on day-level seasonality could
be when the aspect of volatility is ignored. The paper finds evidence in favour of
significant positive returns on non-reporting Thursday and Friday, in sharp contrast
to the finding of significant positive returns only on non-reporting Monday by OLS
procedure. Separate subperiod analyses reveal that there have been changes in daily
seasonality in both returns and volatility since the mid-1990s at the Indian capital
market, manifested in the opposite signs and changes in the level of significance of
some similar coefficients across periods. These findings on the day of the week effects
along with its variation within a fortnight suggest that stock exchange regulations
and the nature of interaction between the banking sector with the capital market
could possibly throw valuable insights on the origin of the day of the week/fortnight
effect in returns, while interexchange arbitrage opportunities due to differences in
settlement period could lead to a seasonality in volatility.

I. INTRODUCTION consistently experienced significant negative and positive

returns on Mondays and Fridays respectively. The obser-
Seasonalities in security market returns have been exten- vation once again started receiving increasing attention
sively documented. Among the different seasonal effects during the 1980s (French, 1980; Gibbons and Hess, 1981;
observed in stock markets, an interesting one is the season- Lakonishok and Levi, 1982), especially when it was discov-
ality across the days of a week. Its discovery goes back to ered that capital markets of many other countries also
Fields (1931). Fields observed that the US stock market experience similar seasonality (Jaffe and WesterField,

* Corresponding author. E-mail: [email protected]

Applied Financial Economics ISSN 0960–3107 print/ISSN 1466–4305 online # 2003 Taylor & Francis Ltd 553
http://www.tandf.co.uk/journals
DOI: 10.1080/0960310021000020924
554 K. Bhattacharya et al.
1985; Peiro, 1994; Agarwal and Tandon, 1994).1 This ‘day ables or due to structural changes and parameter instabil-
of the week effect’, in sharp contrast to the theories of ity. In fact, recent studies reveal that returns from almost
efficient market, was considered a puzzle and despite dif- all capital markets are weakly related to their past values
ferent theories,2 so far the puzzle has not been satisfactorily (Campbell et al., 1997). It is somewhat surprising that the
resolved yet. impact of autocorrelation on the estimated coefficients per-
As more and more empirical evidences are obtained from taining to the day of the week effect has not been examined
different stock markets all over the world, the puzzle – far systematically in the literature. It is well-known that in the
from being solved – seems to have increased. Using a long presence of significant autocorrelation, ordinary least
time series from 1962–1993, Wang et al. (1997) have shown squares (OLS) method would produce biased and ineffi-
that for the US market, the well-known Monday effect cient estimates of the coefficients, which in turn might dis-
occurs primarily in the last two weeks of a month. For tort the ‘true’ day of the week effect. Moreover, Woolridge
the US market, Peiró (1994) has, in fact, obtained a posi- (1991) and Lumsdaine and Ng (1999) have shown that in
tive Monday returns. Of late, studies have incorporated the framework of ARCH/GARCH model, failure to model
volatility of returns in the framework of analysis (Ho and the conditional mean correctly can lead to erroneous infer-
Cheung, 1994; Choudhry, 2000). However, no satisfactory ence. In fact, Woolridge has considered a more general case
explanations have been found for the observed day of the where possible misspecification of conditional variance has
week effect in volatility either. also been allowed for.
Following Wang et al. (1997), this paper examines the Based on daily data on stock returns of BSE 100 index
stability of the day of the week effect at the Indian capital from January 1991 to September 2000, in this paper the
market. India becomes an interesting case study because stability of the day of the week effect in returns and in
since the early 1990s, some very fundamental changes volatility – if any – across weeks and across different time
have taken place at the Indian capital market. These periods is examined. Appropriate GARCH specifications
include the birth of the Securities and Exchange Board of are considered by incorporating lagged returns as explana-
India (SEBI) as a regulator of the Indian capital market, tory variables in the conditional mean, so that there are no
the birth of the National Stock Exchange (NSE) as a com- specificational inadequacies either in the conditional mean
petitor of the Bombay Stock Exchange (BSE), introduc- or in the conditional variance of returns. Based on the
tions of computerized screen based trading at both the findings it is asserted that the day of the week effect
exchanges and dematerialization of shares.3 These changes could be different for the first and the second week in a
have led to substantial improvement in market capitaliza- fortnight. The analysis also suggests that stock exchange
tion, liquidity and efficiency of the Indian capital market, regulations, and the linkage between the banking sector
especially during the second half of the 1990s. The pattern and the capital market could play important roles in gen-
of equity returns on different days in a week at the Indian erating seasonality of this kind. It may be relevant to state
capital market has been examined by many researchers here that in India, the cash reserve ratio (CRR) is main-
(Chaudhury, 1991; Poshakwale, 1996; Arumugam, 1997; tained by the banking sector on average basis for a fort-
Goswami and Angshuman, 2000; Choudhry, 2000). All night and reported to the Reserve Bank of India (RBI)
these studies, except Choudhry (2000), relate to the period every alternate Fridays called Reporting Fridays. A strong
varying between mid-1980s and early-1990s, and use con- difference is observed in the Friday effects between report-
ventional techniques, e.g., first-order autocorrelation test, ing and non-reporting weeks at the Indian capital market.
linear regression with dummy variables representing sea- Further, the effect appears to have changed significantly
sonalities, etc. It is only Choudhry (2000) who has examined during the 1990s. It is difficult to identify the precise reason
seasonality in returns and volatility prevailing at the Indian behind the change in the day of the week effect in returns.
capital market under a unified framework. The main short- However, the observations could potentially narrow the
coming of his work relates to possible misspecification of search for a plausible explanation. So far as seasonality
the conditional mean (of the underlying model) either due in volatility is concerned, based on the empirical evidence,
to omission of lagged values of returns as explanatory vari- the paper argues that its change during the 1990s could be

1
For example, the lowest and negative return falls on Tuesday in Japan and Australia and a significant positive return falls on Tuesday in
Luxembourg. Moreover, Agarwal and Tandon (1994) find that the day of the week effect exists in 16 out of 18 countries with either
Monday or Tuesday being the day of lowest returns.
2
Some of the theories on the emergence of the ‘day of the week effect’ are different trading patterns of individual and institutional
investors (Lakonishok and Maberly, 1990), the daily seasonality in the release of new information (Penman, 1987), concentration of
certain investment decisions (Miller, 1987), country-specific settlement procedures (Solnik, 1990), risk-return tradeoff (Hoa and Cheung,
1994) and a spill-over effect from the US or other large markets (Jaffe and Westerfield, 1985; Agarwal and Tandon, 1994).
3
Misra (1997) provides a brief survey of the historical transition of the Indian capital market from 1947, the year of India’s indepen-
dence, to the first half of the 1990s.
Stability of the day of the week effect 555
explained by interexchange arbitrages at the Indian capital of the account period settlement cycles on returns and
market. volatility.
The plan of the paper is as follows: Sections II and III The settlement process at the Indian capital market is
present the analytical framework and the empirical results further complicated because of badla. Badla is a carry-for-
respectively. Finally, Section IV summarizes the main find- ward system that exists at the Indian capital market for
ings with some concluding comments. specific individual stocks. By this mechanism, a buyer or
a seller can carry forward the transaction for a limited
period through a financier. In the past badla trading at
the Indian capital market was almost always singled out
II. THE ANALYTICAL FRAMEWORK for criticism for ‘excessive’ speculation.
On 12 December 1993, SEBI banned badla trading on
In this paper, the day of the week effect in returns is studied the BSE and at the Calcutta, Delhi and Ahmedabad stock
using a GARCH framework. Also included are proper exchanges. In October 1995, SEBI introduced a modified
lagged values of returns as explanatory variables so that badla system which began on the BSE in January, 1996 and
the conditional mean part of the model is appropriately subsequently revised it further in October 1997. Taking
specified. The GARCH model by Bollerslev (1986), gener- advantage of this ban, some researches were conducted to
alized from a seminal paper of Engle (1982) that introduced assess the impact of badla trading on returns, volatility and
the ARCH model, has been found to be a model that could market efficiency, revealing that badla trading had no
incorporate the underlying volatility of financial variables impact on stock price volatility and that it was, in fact,
adequately (Bera and Higgins, 1995). slightly beneficial for short-horizon market efficiency.4
The model used in this paper is specified as follows: The study of Goswami and Angshuman (2000) also
revealed that badla trading had no impact on the day-of
X
m X
5 X
5
rt ¼
k rtk þ 1j D1jt þ 2j D2jt þ "t the-week pattern of returns.
k¼1 j¼1 j¼1 The model specified in Equations 1–3 is estimated by the
method of maximum likelihood using TSP software. It is
t ¼ 1; 2; . . . ; T ð1Þ well-known that the standard asymptotic results would
hold and usual asymptotic tests may be used to carry out
"t j t1  Nð0; ht Þ ð2Þ relevant hypotheses of interest. Insofar as the determina-
tion of the appropriate value for the number of lags m in
X
p X
5 X
5
the conditional mean is concerned, we use the usual infor-
ht ¼  0 þ i hti þ 1l D1l t þ 2l D2lt mation-based criteria like AIC and BIC and/or Hall’s
i¼1 l¼1 l¼1
(1994) procedure. With regard to the choice of p and q,
X
q the maximum value of each of p and q is fixed at 3 and 2
þ i "2ti ð3Þ respectively and the six possible combinations are consid-
i¼1 ered. In both these cases, appropriate diagnostic tests are
where rt ¼ ln pt  ln pt1 represents the continuously com- finally used to check the appropriateness of the specifica-
pounded rate of return for holding the (aggregate) securi- tion of both the conditional mean and variance.
ties for one day, pt is the log of stock price index at time t,
1j and 2j ’s [ j ¼ 1 (Monday), 2 (Tuesday), 3 (Wednesday),
4 (Thursday), 5 (Friday)] respectively denote the day of the I I I . E M P I R IC A L R E S U L T S
week effect in a reporting and non-reporting week, D1jt and
D2jt ’s are corresponding values of the dummy variables In an earlier study on day of the week effect in returns at
taking values 0 and 1, t1 is the information set at time the Indian capital market, Chaudhury (1991) examined the
t  1 and ht ’s are conditional variances specified as a behaviour of the Bombay Stock Exchange (BSE) Sensitive
GARCH(p; q) process. ht is also assumed to be affected Index (SENSEX) between June 1988 and January 1990. He
by day of the week effects – both reporting and non- found that average return pertaining to Monday was sig-
reporting – represented by 1l ’s and 2l ’s ðl ¼ 1; 2; . . . ; 5Þ. nificantly negative and highest returns were usually on
It may be noted at this stage that the specification in Fridays. Poshakwale (1996) studied the BSE National
Equation 1 entails whether the so-called ‘day of the week Index between January 1987 and October 1994. He found
effect’ should actually be called ‘day of the fortnight effect’. that mean returns except for Monday and Wednesday were
In the Indian context such an approach could be more positive and that weekend effects on returns support the
meaningful because of the possibility of an indirect impact presence of first order autocorrelation. While Chaudhury’s

4
Endo (1998) provides a discussion on the issue.
556 K. Bhattacharya et al.
0.2000
results were based on the Kruskal–Wallis test,
Poshakwale’s findings were based on various autocorrela- 0.1500

tion tests. Arumugam (1997), using regression analysis,

found that Friday returns at the Indian stock market 0.1000

were significantly positive, with no significant returns on

Mondays except in bear phase. Recently, Goswami and 0.0500

Return
Angshuman (2000), using disaggregate data on 70 individ- 0.0000

ual stocks from 1991 to 1996, obtained a positive return at

the BSE on Friday. Their study revealed that the excess -0.0500

returns at the Indian capital market were related to firm

size. For post-1994 data, Goswami and Angshuman (2000) -0.1000

also obtained a significantly negative Tuesday returns. In

-0.1500

another recent study, Choudhry (2000) examined the week-

1041

1106

1171

1236

1301

1366

1431

1496

1561

1626

1691

1756

1821

1886

1951

2016

2081

2146

2211
131

196

261

326

391

456

521

586

651

716

781

846

911

976
66
1
day patterns in return and volatility of seven emerging Time (Day 1: January 3, 1991)

markets including India. Choudhry specified a GARCH Fig. 1. Daily return in BSE100
model of returns for all countries. For the Indian market,
using the daily data of returns from January 1990 to June Table 1 indicates substantial changes in the distribu-
1995, he obtained a positive Friday effect in returns and a tional structure of the returns between Period 1 and
positive Thursday effect in volatility. Period 2. Both Period 1 and Period 2 reveal the presence
In this paper, the daily data on the BSE 100 index are of zero mean returns with near identical variance.
used from January 1991 to September 2000, comprising However, during Period 1, the returns used to be positively
2222 observations. The BSE 100 Index, formerly known skewed, indicative of a less developed market moving
as the BSE National Index is a market value weighted asymmetrically to ‘news’. The distribution was also highly
index of 100 stocks. The BSE started publishing the leptokurtic, i.e., the return series used to have a thicker tail
index from 3 January 1989. The index was initially and a higher peak than a normal distribution, indicating
intended to reflect stock price movements on a national the presence of very large movements of share prices on
scale and for that purpose the 100 stocks were selected either side. In the Indian context, these movements were
from the five major stock exchanges (viz., Mumbai, typical products of ‘euphoria to despondency cycles’
Calcutta, Delhi, Ahmedabad and Madras) at that time. (Gupta, 1997, p. 3). However, for Period 2, the return series
However, fast advancement in communication as well as appears to be symmetric and its kurtosis, while still higher
trading technologies quickly diminished the price differ- than that pertaining to a normal distribution, has dropped
ences among the Indian stock exchanges. Consequently, down considerably.
the BSE redesigned the index based only on prices quoted
on the BSE and renamed it as the BSE 100 Index, effective
from 14 October 1996. OLS-based results
Figure 1 shows the plot of the return data based on BSE-
100 index covering the aforesaid period. It is clear from this Table 2 presents the mean and the standard deviations of
plot that the data exhibit strong volatility. In Table 1 some returns for each working days of the week. As in Table 1,
of the basic features of the returns series are presented. To the results are presented separately for Period 1 and Period
examine the implications of some of the phenomenal 2. Table 2 reveals some interesting features. First, it reveals
changes which took place at the Indian capital market dur- that the significant positive mean return on Fridays at the
ing the early 1990s, the features for the periods of 1991– Indian market obtained by many researchers primarily
1995 and after 1995 are separately examined. These two occurred on non-reporting Fridays. While for the Entire
subperiods are respectively called Period 1 and Period 2. Period, Table 2 reveals the existence of significant positive

Table 1. Basic statistics for returns

Period No. Mean Variance Skewness Kurtosis

Period 1 1065 0.0010 0.00034 0.8093 # 13.8836 #

(January 1991–December 1995)
Period 2 1156 0.0003 0.00035 0.1140 2.9973 #
(January 1996–September 2000)
Entire period 2221 0.0006 0.00035 0.4308 # 7.9150 #

Note: # implies significance at 1% level.

Stability of the day of the week effect 557
Table 2. Mean and standard deviation of returns based on days of a week/fortnight

Mean return Standard deviation of returns

Period 1 Period 2 Period 1 Period 2

(Jan 1991 to (Jan 1996 to (Jan 1991 to (Jan 1996 to
Day of the week Dec 1995) Sep 2000) Entire period Dec 1995) Sep 2000) Entire period

Reporting Monday 0.0019 0.0020 0.0019 0.0255 0.0242 0.0248

Non-reporting Monday 0.0011 0.0058# 0.0037# 0.0190 0.0251 0.0225
All Monday 0.0016 0.0039# 0.0028# 0.0226 0.0246 0.0237
Reporting Tuesday 70.0005 70.0013 70.0009 0.0196 0.0180 0.0187
Non-reporting Tuesday 0.0004 70.0003 0.0000 0.0199 0.0132 0.0167
All Tuesday 70.0001 70.0008 70.0004 0.0197 0.0158 0.0177
Reporting Wednesday 70.0016 0.0030{ 0.0007 0.0155 0.0171 0.0164
Non-reporting Wednesday 0.0009 0.0014 0.0012 0.0168 0.0186 0.0177
All Wednesday 70.0004 0.0022{ 0.0009 0.0162 0.0178 0.0171
Reporting Thursday 70.0021 70.0008 70.0014 0.0156 0.0171 0.0164
Non-reporting Thursday 0.0027 70.0014 0.0005 0.0139 0.0174 0.0160
All Thursday 0.0002 70.0011 70.0005 0.0150 0.0172 0.0162
Reporting Friday 70.0001 70.0016 70.0009 0.0193 0.0175 0.0184
Non-reporting Friday 0.0071# 70.0037* 0.0017 0.0136 0.0158 0.0156
All Friday 0.0035# 70.0026* 0.0004 0.0170 0.0167 0.0171
Reporting week 70.0005 0.0003 70.0001 0.0194 0.0190 0.0192
Non-reporting week 0.0025 0.0004 0.0014 0.0169 0.0187 0.0179
All days of all week 0.0010 0.0003 0.0006 0.0183 0.0188 0.0186

Note: #, * and { imply significance at 1%, 5% and 10% levels respectively.

returns only on Mondays (especially non-reporting set of regressions were run with an additional explanatory
Mondays), subperiod-wise break-ups present a different variable, namely a dummy named BADLA which took the
story. During 1991–1995, the returns were significantly value unity during the badla period and zero elsewhere. It
positive for non-reporting Fridays. However, post-1995 may be noted that as badla was reintroduced during
returns pertaining to non-reporting Fridays were signifi- January 1996, the variable BADLA took the value unity
cantly negative. Also returns corresponding to Mondays, for almost all the observations in Period 2. For Period 2,
especially non-reporting Mondays, were significantly posi- therefore, the variable BADLA was not considered.
tive in Period 2. When the regressions were actually done, the presence of
Table 2 also reveals the changing pattern of volatility significant autocorrelations in the residuals was noted.
across the days of the week at the Indian market. Autocorrelations were high at lag one for the Entire
Volatility, measured in terms of standard deviation of Period and both the subperiods while they were moderate
returns, appears to be stable in the aggregate across and significant – either for the Entire Period, Period 1 or
Period 1 and Period 2. However, after 1995 the volatilities Period 2 – at lags 3, 7, 8, 9, 10, 12, 15 and 19. The auto-
corresponding to Monday, Wednesday and Thursday have correlation structures were fairly similar in all the equa-
increased and that for Tuesdays has reduced considerably, tions. In all the regressions, the estimated coefficient of
the direction of change being generally similar for report- BADLA turned out to be insignificant. The specifications
ing and non-reporting weeks. were, therefore, changed to include these lags of return in
It may be noted that as D11 , D12 ; . . . ; D15 , D21 ; . . . ; D25 the regression equations for the Entire Period as well as for
defined in Section II are orthogonal, an OLS regression of both Period 1 and Period 2 and drop BADLA from all the
returns involving only these ten dummies as independent equations.
variables (and obviously without consideration to con- Table 3 presents the results of these regressions. In the
ditional heteroscedasticity) would yield identical estimates estimated regression equations, most of the lags chosen
corresponding to day of the week/fortnight effect of returns turn out to be significant for the Entire Period as well as
in Table 2. From Table 2, therefore, inferences could be for both Period 1 and Period 2. The first lag of return, in
drawn on results from simple dummy variable regressions, fact, turns out to be highly significant in all the regressions.
a tool frequently employed by previous researchers to esti- Estimated values of Qð40Þ were 51.02, 24.10 and 28.97 for
mate the nature and the extent of the day of the week effect Period 1, Period 2 and the Entire Period respectively. Thus,
in returns (Solnik, 1990; Peiró, 1994; Arumugam, 1997; while the process of inclusion of lags led to improved
Wong et al., 1999). To analyse the effect of badla, another Ljung–Box Q statistic for all periods, the estimated re-
558 K. Bhattacharya et al.
siduals for Period 1 still showed evidences of autocorrela- term appear as highly unstable (cf. Table 3). Given the
tion. However, significantly, it was found that the esti- highly volatile and unpredictable nature of returns, such
mated day of the week/fortnight effects in returns in a result is not surprising. Individually, most of the other
Table 3 are similar to those in Table 2. estimated coefficients, however, are found to be stable.
So far as the stabilities of the estimated equations are Instability is more prominent in the equation for the
concerned, Hansen’s tests (Hansen, 1992) confirm the pres- Entire Period, where coefficients pertaining to non-
ence of instability in all of them. Hansen’s joint statistics reporting Thursday, non-reporting Friday and the seventh
for the estimated equations corresponding to Period 1, and eighth lags are found to be unstable. Among the esti-
Period 2 and the Entire Period are 5.73, 6.62 and 8.53 mated coefficients for Period 1 and Period 2, however, only
respectively, all of them being significant at the 1% level. the coefficient of non-reporting Monday in Period 1
In all the equations, the estimated variances of the error appears as significantly unstable.

Table 3. OLS regression results with lagged returns as explanatory variables

Coefficients Period 1 Period 2 Entire Period

11 0.000 087 909 0.002 053 126 0.001 314 256
(0.05) (1.20) (1.08)
21 0.001 830 825 0.005 652 397 0.003 740 247
(1.00)} (3.26)# (2.96)#
12 70.001 606 508 70.001 334 133 70.001 378 323
(70.94) (70.79) (71.14)
22 0.000 571 861 70.000 698 549 70.000 257 421
(0.32) (70.40) (70.21)
13 70.002 235 668 0.002 938 875 0.000 613 461
(71.29) (1.72){ (0.50)
23 0.000 965 633 0.001 817 136 0.001 005 163
(0.55) (1.05) (0.81)
14 70.001 530 459 70.001 328 062 70.001 245 198
(70.89) (70.78) (71.02)
24 0.002 400 949 70.001 364 826 0.000 272 632
(1.34) (70.79) (0.22) }
15 70.000 157 713 70.001 482 032 70.000 765 034
(70.09) (70.87) (70.62)
25 0.006 193 142 70.003 856 865 0.001 215 452
(3.56)# (72.19)* (0.98) {

t1 0.152 341 502 0.088 685 026 0.120 401 726
(4.97)# (3.02)# (5.70)#

t3 0.089 316 783 0.040 055 582 0.059 003 041
(2.91)# (1.37) (2.80)#

t7 70.039 043 992 0.060 311 833 0.014 579 367
(71.27) (2.04)* (0.68) {

t8 70.063 059 940 0.054 291 692 70.004 358 117
(72.03)* (1.83){ (70.20) {

t9 0.100 801 869 0.065 757 400 0.089 439 412
(3.23)# (2.21)* (4.16)#

t10 0.044 199 788 0.052 587 665 0.053 254 677
(1.43) (1.77){ (2.48)*

t12 0.043 179 999 0.005 935 379 0.019 112 645
(1.42) (0.20) (0.90)

t15 0.051 546 086 70.016 362 352 0.024 649 818
(1.70){ (70.55) (1.17)

t19 70.029 027 741 70.111 524 263 70.070 243 037
(70.95) (73.79)# (73.31)#
Observations 1065 1156 2221
R2 0.074 2 0.057 2 0.042 4
R-BAR2 0.058 0 0.042 3 0.034 5

Notes: (1) The bracketed figures are values of t-statistics. (2) #, * and { represent sig-
nificance at 1%, 5% and 10% levels respectively. (3) { and } indicate that Hansen’s
stability test statistic for the coefficient is significant at 1% and 5% levels respectively.
Stability of the day of the week effect 559
GARCH-based results reporting Wednesday (negative), non-reporting Thursday
(positive) and non-reporting Friday (positive). The day-
The empirical findings are now discussed of the complete
level seasonality corresponding to Period 2 has been
model incorporating GARCH process for the conditional
variance, as specified in Equations 1–3. The estimates of found to be of the similar type namely, significant non-
the parameters along with their t-statistic values for the reporting Monday (positive), reporting Wednesday (posi-
Entire Period as well as for Period 1 and Period 2 are tive) and non-reporting Friday (negative). While in Period
presented in Table 4. Table 4 reveals that returns based 1, non-reporting Thursday has also been found to be sig-
on the entire sample have significant positive effect on nificant, what is very contrasting to note is that the signifi-
both non-reporting Thursday and non-reporting Friday, cant day effects are of opposite signs across Period 1 and
the t-statistic values being 2.244 and 2.305, respectively. Period 2. Possible explanations are provided for such find-
The findings for Period 1 and Period 2, however, reveal a ings later. What is, however, relevant to note at this stage
different story. In Period 1 four significant coefficients have are (i) the importance of ‘reporting’ and ‘non-reporting’
been observed, namely non-reporting Monday (negative), aspects of market microstructure of Indian stock market

Table 4. The estimated GARCH model Table 4. Continued

Coefficients Period 1 Period 2 Entire period Coefficients Period 1 Period 2 Entire period

1 0.3080 0.1415 0.1810 1 0.193 0.209 0.187

(8.537)# (4.276)# (7.748)# (8.218)# (5.571)# (10.136)#

2 70.0850 – – 2 7 0.133 –
(72.341){ (3.901)#

3 0.0780 – 0.0450 3 7 0.0842 –
(2.171){ (1.932){ (3.197)#

6 – 70.0830 70.0610 1 0.084 – 0.098
(73.58)# (72.94)# (2.3)# (3.281)#

8 – 0.0515 – 2 0.716 – 0.665
(1.836){ (16.959)# (18.006)#

10 – 0.0511 0.0500 11 0.0000234 0.000295 0.0000656
(2.123){ ( 2.461)# (1.552) (5.95)# (6.164)#

11 – 70.0470 70.0410 21 * 0.000436 0.0000806
(72.046){ (72.027){ (8.650)# (6.014)#

13 – – 70.0390 12 * 0.000050 *
(71.98){ (1.685){

18 – 70.0439 – 22 * 0.000010 *
(71.831){ (0.382)

19 – 70.0860 – 13 * 0.000063 *
(73.41)# (1.827){
11 70.001150 0.002980 0.001000 23 * 0.000150 *
(70.937) (1.662) (1.138) (3.369)#
21 70.003380 0.006260 0.001000 14 0.0000958 0.000098 0.0000246
(72.906)# (2.925)# (1.295) (0.462) (3.043)# (2.007)#
12 70.000097 70.000925 70.000045 24 0.00001174 0.000097 *
(70.065) (70.698) (70.419) (0.788) (2.71)#
22 70.000478 70.001610 70.001100 15 0.00000143 0.000061 *
(70.404) (71.22) (71.052) (0.133) (1.984){
13 70.002680 0.003280 70.000325 25 0.00000484 0.000024 *
(71.96){ (2.377){ (70.305) (0.512) (0.000)
23 0.001110 0.001770 0.001590 L 3025.74 3086.85 6016.89
(0.865) (1.125) (1.643)
14 70.001790 70.000740 70.000960 Notes:
(71.724) (70.521) (70.995) 1. Bracketed figures are values of t-statistics.
24 0.003170 0.0005170 0.002080 2. – indicates that the corresponding parameter is not relevant for
(2.79)# (0.372) (2.244){ the series.
15 0.000375 70.000881 70.000610 3 * shows that the computed value of the coefficient is so small
(0.395) (70.683) (70.734) that its value and the corresponding value of the t-statistic are
25 0.005640 70.002970 0.002080 treated as zero.
(5.767)# (72.287){ (2.305){ 4. #, { and { imply significance at 1%, 5% and 10% levels respec-
0 * 0.000065 * tively.
(3.008)# 5. L represents values of the log-likelihood function.

Continued
560 K. Bhattacharya et al.
on the day of the week effect on returns, and (ii) the difference in day of the week effect in volatility in Period 2
observed differences in day-level seasonality between as compared to Period 1.
Period 1 and Period 2 leading thereby to the conclusion Finally, the values of the Q-test of the estimated models
that there has indeed been a change effected in mid-1990s. are reported to find that the residuals of these models are
The results in Table 4 vis-à-vis Table 3 reveals how the day significant in the three models based on entire sample and
of the week effect could be found to be erroneous if due those for Period 1 and Period 2. In each case 40 lags have
consideration to volatility is not taken care of in the model. been considered and only the Q-test statistic values with the
Insofar as the day of the week effect on the volatility i.e., lags 10, 20, 30 and 40 have been reported. From the Q-test
conditional variance is concerned, it is observed that in the statistic values in Table 5 it is noted that for standardized
full-sample period, both reporting and non-reporting residuals, no significant autocorrelation was observed in
Monday and only reporting Thursday are found to have Period 1, Period 2 and Entire Period. Standardized-
significant effect. The t-statistic values are 6.164, 6.014 and squared residuals, however, displayed evidences of auto-
2.007, respectively. The results concerning Period 1 and correlation only in Period 2.
Period 2 are, as in the case of conditional mean, quite Also reported in Table 5 are the values of skewness and
varying. While no significant day of the week effect has kurtosis coefficients of the above three residuals, and the
been observed in Period 1, significant positive effect has LM test statistic value for testing for ARCH/GARCH. It
been found to be present on both reporting and non- has been observed, as reported in Table 4 that
reporting Monday, reporting Tuesday (at the 5% level GARCH(1,2) has been found to be the most adequate
only), both reporting (at the 5% level only) and non- GARCH specification for the data corresponding to the
reporting Wednesday, both reporting and non-reporting Entire Period and to Period 1; ARCH(3) has been found
Thursday and reporting Friday (at the 5% level only) in to be the best specification for Period 2. The LM test sta-
Period 2. It is thus observed that day of the week effect is tistic values based on the standardized residuals of the
an important factor in explaining observed volatility at the three models are 0.059, 0.412 and 0.079 respectively.
Indian capital market and that there has been a significant Thus it is clear that the hypothesis of no ARCH/

Table 5. Properties of the residuals from the estimated GARCH models

Residuals Period 1 Period 2 Entire period

1. Non-standardized ð"t Þ
Skewness 1.424 0.028 0.574
Kurtosis 18.339 2.918 8.626
Qð10Þ 65.028# 20.155 38.774#
Qð20Þ 77.555# 30.935 56.778#
Qð30Þ 93.238# 35.265 67.876#
Qð40Þ 120.076# 40.994 80.694#
ARCH 66.630# 44.794# 95.598#
2. Standardzed ð"t =h0:5
t Þ
Skewness 0.185 0.033 0.307
Kurtosis 4.627 1.091 3.691
Qð10Þ 11.191 16.283 18.115
Qð20Þ 18.901 22.967 23.221
Qð30Þ 27.586 24.985 30.136
Qð40Þ 40.927 29.613 37.192
ARCH 0.059 0.412 0.079
3. Standardized
squared ð"2t =ht Þ
Skewness 10.882 4.302 9.725
Kurtosis 160.318 25.748 134.340
Qð10Þ 15.731 33.881# 15.734
Qð20Þ 24.217 44.369# 25.283
Qð30Þ 28.695 50.600# 30.824
Qð40Þ 33.513 61.061# 38.899
ARCH 0.000 0.034 0.023

Notes:
1. # implies significance at 1% level.
2. QðkÞ represents the Ljung–Box statistic with k number of lags after adjusting the
number of parameters.
Stability of the day of the week effect 561
GARCH can not be rejected, signifying the adequacy of Janakiraman Committee subsequently appointed by the
the estimated conditional variance specifications. As Reserve Bank of India to investigate the scam noted wide-
regards the values of skewness and kurtosis coefficients, it spread use of many irregularities:
is found that these values have reduced for the series of
. . . The Committee’s terms of reference were largely
residuals of these estimated models as compared to those
restricted to the examination of the securities transac-
based on the original return series for the Entire Period, as
tions of banks. However, the Committee has identified
also for Period 1 and Period 2. For instance, it is found
that there has also been diversion of funds through other
from Table 1 and Table 5 that while for the entire series,
means, for example, call money transactions and the
the skewness and kurtosis coefficients were 0.4308 and
discounting of bills. Thus, large payments such as call
7.9150 respectively, the same based on the standardized
money placed with other banks, have been found to be
residuals of estimated models are 0.307 and 3.691.
credited to individual brokers’ accounts. On due date,
An attempt is now made to provide an interpretation of
these alleged call loans have been repaid by payment
the results. It may be noted that for the Indian market, the
out of brokers’ accounts in the same or other banks.
findings pertaining to the early 1990s appear to be consis-
The Committee has also noticed cases where brokers’
tent with the earlier findings of a significant Friday effect.
funds have been placed in the call money market under
However, the findings highlight that the Friday effect at the
the banks’ names. (Janakiraman, 1993, pp. 292–3)
Indian market used to take place mostly on non-reporting
Fridays. Among different theories, it appears that in the It is possible that as the CRR is maintained by the banking
Indian context, the risk-return tradeoff proposed by Ho sector in India is on average basis for a fortnight, irregular
and Cheung (1994) does not make sense because high availability of funds on specific days within a fortnight used
returns at the Indian market on specific days were not to affect prices of bourses at the Indian capital market on
always associated with high volatility. Similarly, transmis- those days, the direction being dependent on the direction
sion of shocks from other markets (e.g., the US market) of the flow. It is also possible, although difficult to quan-
could not be a major factor because during the early 1990s, tify, that after widespread computerization, currently the
the Indian economy was relatively closed in nature. nature of interaction between the banking sector and capi-
Although during the recent period, capital markets in tal market, especially the irregular flow of funds, has chan-
India have become somewhat integrated to the rest of the ged and the said changes have been manifested in the
world, such integration has not been manifested in the change in the day of the week effect.
seasonalities in returns. For example, if the ‘root cause’ The second factor affecting the day of the week effect
of seasonality is the transmission of shocks from the US could be stock exchange regulations.5 The seasonality in
market, the ‘Monday Effect’ at the US market would have volatility at the Indian capital market could possibly be
been manifested in the returns at the Indian market on explained in terms of these regulations allowing arbitrage
Tuesday, because of the geographical distance and the opportunities across different stock exchanges in India. It
consequent time-difference between the two markets. may be noted that an account period settlement cycle
Although for post-1994 data, Goswami and Angshuman begins in BSE on Monday and ends after eight working
(2000) have observed evidences of a significantly negative days. Thus a trade done during the week generally settles
Tuesday returns at the Indian capital market, the present on the Friday of the next week. However, NSE’s account-
studies do not indicate similar results. ing period settlement cycle starts every Wednesday and
Although it has not been possible to provide a full expla- ends on a Tuesday. All open and outstanding positions
nation of the day of the week effect in returns, two factors on Tuesday at the NSE are required to be settled and
are identified which could have affected the weekday pat- delivered. This is similar to a futures style settlement
terns. The first is the nature and the extent of interaction of wherein shares traded Wednesday onwards are effectively
the banking system with the capital market in India. a forward contract for settlement during the next week.
During early 1990s in India, the interaction could have Thus on the NSE an investor may take speculative position
led to significant positive returns on specific days of a for a week, i.e., he trades on Wednesday and can reverse
week. At that time, the banking sector as well as the capital the trade by the following Tuesday (close of account per-
markets in India were not fully computerized, leading to iod) and book his gains and losses. It is likely that spec-
lack of transparency and efficiency which in turn funnelled ulative activities are at its peak and ebb, respectively at the
movements of funds from the banking sector to the capital beginning and the end of a cycle. Thus, the post-1995 high
market. The securities market scam experienced by India in volatilities on Mondays and Wednesdays could be attrib-
1992 was a direct result of this type of activity. The uted to beginnings of settlement cycles respectively at the

5
Wong et al. (1999), for the Shanghai Stock Exchange (SSE), have arrived at a similar inference.
562 K. Bhattacharya et al.
BSE and the NSE. Similarly, the post-1995 low volatility The day of the week effect in volatility of returns have
on Tuesdays and Fridays could be because of the pressure also been found to be different in the two sub-periods.
of closing one’s position on those days. Based on entire sample, significant positive effect were
The significantly positive volatility on both reporting found on non-reporting as well as reporting Mondays
and non-reporting Thursdays during Period 2 presents an and reporting Thursday. While no day-level seasonality
interesting case and may be explained in terms of kerb has been found in volatility in Period 1, significant positive
trading across stock exchanges situated in different cities day effect has been found to be present on both reporting
in India. Earlier, the kerb market was occasionally used and non-reporting Monday, reporting Tuesday, both
by the big operators and insiders to manipulate share prices reporting and non-reporting Wednesday, both reporting
(Krishnan and Narta, 1997, p. 403). Although academic and non-reporting Thursday and reporting Friday.
research on this topic is rare, a few newspaper reports While testing the hypothesis that day of the week effect
and articles have pointed out the existence of a grey market could in reality be a day of the fortnight effect these find-
in many stock exchanges in India, especially at the Kolkata ings along with consideration to ‘reporting’ and ‘non-
(erstwhile Calcutta) stock exchange on Thursdays.6 It is reporting’ days in a fortnight suggest that stock exchange
possible – although difficult to prove – that the significantly regulations and the nature of interaction between the bank-
high volatility at the Indian capital market on Thursday’s ing sector with the capital market could possibly throw
during Period 2 was because of this type of activities and valuable insights on the origin of the day of the week/fort-
the results of this paper could be interpreted as an indirect night effect in returns. It is possible that the day of the week
evidence in this direction. Incidentally, although Choudhry effect, in reality could be a day of the fortnight effect.
(2000) obtained a significantly positive coefficient pertain- Following Wang et al. (1997), it is also possible that the
ing to volatility of returns on Thursdays during the early periodicity of the seasonality could be even bigger. So far
1990s, a similar result for Period 1 has not been obtained. as the seasonality in volatility is concerned, the analysis has
revealed that interexchange arbitrage opportunities due to
the existence of differences in account period settlement
IV. CONCLUSION cycles could lead to such seasonality. As research on sea-
sonality in volatility of returns is still at a nascent state, the
In this paper the stability of the day of the week effect has above finding could possibly be of help in understanding
been studied in mean and in conditional variance of returns and explaining such seasonality for the capital markets in
at the Indian capital market using a GARCH framework. other economies.
Results based on OLS have also been reported to examine
as to what extent the day of the week effect may be inap-
propriately estimated when volatility aspect of the data is
ignored. Strong evidence is found that the significant posi- ACKNOWLEDGEMENTS
tive returns at the Indian capital market during the early The authors wish to express their sincere thanks to
1990s on Friday were mostly due to non-reporting Fridays. Himanshu Joshi, G. P. Samanta and Surya Choudhury
In fact, the analyses based on the series covering the entire for discussions which led to substantial improvement in
sample period show significant positive effect on non- the exposition of an earlier draft. The views expressed in
reporting Friday as also on non-reporting Thursday. this paper are the authors’ and not of the institutions’ they
OLS-based study on the other hand, has yielded positive belong. The authors bear full responsibility for any error
effect only on non-reporting Monday. Subperiod wise ana- that remains.
lyses have revealed that the day-level seasonality is quite
different in the two subperiods in the sense that while
Monday (non-reporting), Wednesday (reporting) and
REFERENCES
Friday (non-reporting) have been found to have significant
negative, negative and positive effects respectively in Period Agarwal, A. and Tandon, D. (1994) Anomalies or illusions?
Evidence from stock markets in eighteen countries, Journal
1, the same day effects have also been observed in Period 2 of International Money and Finance, 13, 83–106.
but with opposite signs. In Period 1, an additional day Arumugam, S. (1997) Day-of-the-week effects in stock returns: an
effect i.e., significant positive effect on non-reporting empirical evidence from Indian equity market, Mimeo, UTI
Thursday has also been found. Institute of Capital Markets.

6
For example, M. G. Damani, a former president of the Bombay Stock Exchange noted in an article at the Indian Express that ‘For
years, Kolkata has been getting around this inconvenience by entering into financial deals outside the market’. The same article points
out the existence of huge monetary flows between Mumbai and Kolkata market on specific days. In another recent article titled ‘BSE
brokers default on badla funds, divert to CSE (Economic Times, 9 April 2001), Rishi Chopra observed a similar phenomenon.
Stability of the day of the week effect 563
Baillie, R. and Bollerslev, T. (1989) The message in daily exchange Hsieh, D. (1989) Modeling heteroskedasticity in daily foreign
rates: a conditional variance tale, Journal of Business and exchange rates, Journal of Business and Economic Statistics,
Economic Statistics, 7, 297–305. 7, 307–17.
Bera, A. and Higgins, M. (1995) On ARCH models: properties, Jaffe, J. and Westerfield, R. (1985) The week-end effect in com-
estimation and testing in Surveys in Econometrics, (Eds) L. mon stock returns: the international evidence, Journal of
Oxley, D. A. R. George, C. J. Robers and S. Sayer, Basil finance, 40, 433–54.
Blackwell Limited, Cambridge, Massachusetts, pp. 215–72. Janakiraman, R. (Chairman) (1993) Reports of the Committee to
Bollerslev, T. (1986) A generalized autoregressive conditional het- Enquire into Securities Transactions of Banks and Financial
eroskedasticity, Journal of Econometrics, 31, 301–27. Institutions, Reserve Bank of India.
Bollerslev, T. (1987) A conditional heteroscedastic time series Krishnan, B. and Narta, S. S. (1997) Security Markets in India,
model for speculative prices and rates of return, Review of Kanishka Publishers, New Delhi.
Economics and Statistics, 69, 542–7. Lakonishok, J. and Levi, M. (1982) Weekend effects on stock
Campbell, J. Y., Lo, A. W. and MacKinlay, A. C. (1997) The returns: a note, Journal of Finance, 37, 883–9.
Econometrics of Financial Markets, Princeton University Lakonishok, J. and Maberly, E. (1990) The weekend effect: trad-
Press, Princeton, New Jersey. ing patterns of individual and institutional investors, Journal
Chaudhury, S. K. (1991) Seasonality in share returns: preliminary of Finance, 45, 231–43.
evidence on day-of-the-week effect, Chartered Accountant Lakonishok, J. and Smidt, S. (1990) Are seasonal anomalies real?
(India), 40, November, 107–9. a ninety-year perspective, Review of Financial Studies, 1, 403–
Choudhury, T. (2000) Day of the week effect in emerging Asian 25.
stock markets: evidence from the GARCH model, Applied Lumsdaine, R. L. and Ng, S. (1999) Testing for ARCH in the
Financial Economics, 10, 235–42. presence of a possibly misspecified conditional mean, Journal
Endo, T. (1998) The Indian Securities Market: a Guide for Foreign of Econometrics, 93, 253–79.
and Domestic Investors, Vision Books, New Delhi. Miller, E. (1987) Why a weekend effect? Mimeo, University of
Engle, R. F. (1982) Autoregressive conditional heteroskedasticity New Orleans.
with estimates of the variance of UK inflation, Econometrica, Misra, B. M. (1997) Fifty years of the Indian capital market,
Reserve Bank of India Occasional Papers, 18, 351–84.
50, 987–1008.
National Stock Exchange of India Limited (1998) Indian
Fields, M. J. (1931) Stock prices: a problem in verification,
Securities Market: A Review, National Stock Exchange of
Journal of Business, 7, 415–18.
India Limited.
French, K. (1980) Stock returns and the weekend effect, Journal of
Peiró, A. (1994) Daily seasonality in stock returns: further inter-
Financial Economics, 8, 55–70.
national evidence, Economics Letters, 45, 227–32.
French, K. and Roll, R. (1986) Stock returns variances: the arri- Pena, J. I. (1995) Daily seasonalities and stock market reforms in
val of information and the reaction of traders, Journal of Spain, Journal of Financial Economics, 5, 419–23.
Financial Economics, 8, 55–70. Penman, S. H. (1987) the distribution of earning news over time
Gibbons, M. and Hess, P. (1981) Day of the week effects and asset and seasonalities in aggregate stock returns, Journal of
returns, Journal of Business, 54, 579–96. Financial Economics, 18, 199–228.
Goswami, R. and Angshuman, R. (2000) Day of the week effect Poshakwale, S. (1996) Evidence on weak form efficiency and day
on the Bombay stock exchange, ICFAI Journal of Applied of the week effect in the Indian stock market, Finance India,
Finance, 6, 31–46. 10, 605–16.
Gupta, L. C. (1997) India’s Stock Market Crisis, Society for Rogalski, R. J. (1984) New findings regarding day-of-the-week
Capital Market Research and Development, Research returns over trading and non-trading periods: a note,
Monograph No. 5, Vivek Financial Focus Limited. Journal of Finance, 39, 1603–14.
Gupta, L. C. and Choudhury, U. (1999) Return on Indian equi- Solnik, B. (1990) The distribution of daily stock returns and set-
ties, 1980–99, Economic and Political Weekly, 35, 357–64. tlement procedures: the Paris bourse, Journal of Finance, 45,
Hall, A. (1994) Testing for a unit root in time series with pretest 1601–10.
data-based model selection, Journal of Business and Economic Wang, K., Li, Y. and Erickson, J. (1997) A new look at the
Statistics, 48, 201–12. Monday effect, Journal of Finance, 52, 2171–86.
Hansen, E. B. (1992) Testing for parameter instability in linear Wong, K. A., Hui, T. K. and Chang, C. Y. (1992) Day-of-the-
regression models, Journal of Policy Modeling, 14, 517–33. week effect: evidence from the developing stock markets,
Hawawini, G. and Keim, D. B. (1995) On the predictability of Applied Financial Economics, 2, 49–56.
common stock returns: world-wide evidence in Handbooks in Wong, K. E., Chen, R. and Shang, X. (1999) The weekday effect
Operations Research and Management Science, Vol. 9 on on the Shanghai stock exchange, Applied Financial
Finance (Eds) R. A. Jarrow, V. Maksimovic and W. T. Economics, 9, 551–65.
Ziemba, North Holland, Amsterdam. Woolridge, J. M. (1991) On the application of robust, regression-
Ho, R. and Cheung, Y. (1994) Seasonal pattern in volatility in based diagnostics to models of conditional means and con-
Asian stock markets, Applied Financial Economics, 4, 61–7. ditional variances, Journal of Econometrics, 47, 5–46.
Copyright of Applied Financial Economics is the property of Routledge and its content may not be copied or
emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.