Day-of-the-week effect on stock returns in India : effect of economic reforms

DAY-OF-THE-WEEK EFFECT ON STOCK RETURNS IN INDIA:

EFFECT OF ECONOMIC REFORMS

NITYANANDA SARKAR*

It is w ell know n that India initiated m a jo r m a rk e t-o rie n te d and lib eralized eco n o m ic p o lic ie s since 1992 although Indian capital m arkets began to reo rg an ize itse lf a little e a rlie r in the late eighties. In the w ake o f th ese reform s, especially, in the capital m arkets, it seem s w orthw hile to exam ine w h eth e r th e re is any indication o f change sim ila r to those observed in developed capital m arkets. In this p aper we ex am in e this issue with resp e ct to the day-of-the-w eek effect on daily returns on S ensitive Index o f B om bay Stock I :\c h a n g e sp anning the period January 2, 1984 to M ay 14. 1996, by applying ap p ro p riate reg ressio n analysis along with residual-based diagnostic tests. O ur findings suggest that th ere has indeed been a change in the d ay -of-the-w eek effect indicating partial sim ilarity to th o se observed for developed capital m arkets.

INTRODUCTION

D uring the last d ec ad e Indian capital m arkets have show n a sp e cta cu la r grow th in term s o f investor p opulation, num ber o f listed co m p an ies w ith various stock exchanges in India, capital raised from prim ary m arket etc. For instance, the num ber o f listed com panies w as o v er 8000 with a m arket capitalizatio n o f Rs 4.5 3 ,0 0 0 cro res as on N ovem ber 1996; the capital raised from prim ary m arkets increased from Rs 8,298 crores in 1990 to Rs 2 2 .0 9 6 crores in 1996-97. This phenom enal grow th had necessitated the need to control and regulate the capital m arket and to protect interests o f the investors. A ccordingly, bold and far-reaching steps w ere initiated since the late eighties (e.g., the S ecurities and E xchange B oard o f India w as set up in 1988) in o rd er to a c h ie v e ra p id e c o n o m ic g ro w th a n d d e v e lo p m e n t, g lo b a liz a ti o n a n d in te r n a tio n a l c o m p etitiv e n ess o f Indian capital m arkets. It is w idely know n that such reform s w ere u ndertaken not only in the financial sector but also in o th e r sectors o f the eco n o m y w ith the aim tow ards p ursuing m ark et-o rien ted and liberalized eco n o m ic policies. W hile the ch a racter o f the Indian e c o n o m ic system began to change to w ard s o p en n e ss very slow ly since th e late eig h ties, the large-scale reform s w ere initiated from 1992 onw ards.

In the wnke o f these econom ic reform policies in the Indian econom y as a w hole and the Indian capital m arkets in particular, the question that we exam ine in this paper is w hether the behaviour o f the Indian stock prices in the sense o f day-of-the-w eek effect on returns has changed in the period o f econom ic reform s since 1992 as com pared to the p r e -1992 period. The day-of-the-w eek effect refers to the existence o f a pattern on the part o f the stock returns' in the sense that these returns are linked to any particular day(s) o f the week. Naturally, the day-of-the-week effect poses interesting buy and hold strategy issues for the potential investors, and hence studying the presence o f such effects on

* Professor. Economic Research Unit. Indian Statistical Institute. Calcutta. India.

/ It m a y b e n o te d th a t d ire ct sta tis tic a l a n a ly s is o f s to c k p r ic e s is d iffic u lt sin c e c o n s e c u tiv e s to c k p u c e s a re lik e ly to be h ig h ly c o rr e la te d (cf. T aylor 1986), a n d th e v a ria n c e o f th e p r ic e in c r e a s e s w ith lime. In o th e r w ords, th e u n it ro o t n a tu re o f s to c k p r ic e s p o s e s p r o b le m f o r a n a ly s is w ith slo c k p r ic e d a ta h is f o r th is re a so n th ill f i r s t d iffe re n ce s o f p r ic e s lo g a rith m o f p r ic e s o r w h a t is b e tte r k n o w n a s sto ck r e tu r n s are in s te a d u s e d fo r e m p iric a l a n a lysis. M oreover, it is " sto c k r e tu rn s " r a th e r th a n " p ric e s "

w hich " p r ic e s " w h ic h p r im a r ily m o tiva te in v e s to r s ' d e c isio n m aking.

the returns o f the stock m a rk e ts are im portant and useful. A lm ost all the studies reported on th is effect are concerned w ith the returns on the m ajor stock m arkets in the d eveloped e c o n o m ie s, particularly in th e U SA . In th e se studies the researchers ((see for exam ple, C ross (1973), F re n c h (1980), K eim and S tam baugh (1984) and Keim (1 9 8 5 )) have found that the average retu rn on Fridays, the last trad in g day o f the w eek, are significantly positive w hile those on M ondays, th e first trad in g day o f the w eek, are often negative. As stated by L akonishok and Levi ( 1982), th e s e findings on the prev alen ce o f “ day effect” in the stock returns is rather difficu lt to ex p lain in term s o f th e tw o w ell-know n hypotheses - calendar tim e hypothesis and trad in g tim e h y p o th e sis (cf. French (1 9 8 0 ))2. H ow ever, som e researchers like L akonishok and Levi (1 9 8 2 ) and R o g a lsk i (1 9 8 4 ) have explain ed th ese findings in term s o f d ifferences in returns o v er trad in g and n o n ­ trad in g periods, and the delay betw een trading and settlem ents in stocks and in clearing ch e q u e s . S in ce th e se tw o d a y -o f-th e-w e ek effects hav e b een , in g en e ral, o b serv ed in all the s tu d ie s concern in g th e stock p rices o f developed capital m arkets in advanced countries, it may be a s se rte d that sim ilar day-of-the-w eek effects are likely to be observed in em erging stock m arkets w h e re capital m arket reform s and econom ic liberalization p olicies have been effectively initiated. It is with this view in m ind th a t we study in this paper w h eth er the behaviour o f ex p ected retu rn s on Indian stock m arkets has ch an g ed in the period b eginning 1992 as com pared to pre-1992 p e rio d . If the findings are som ew hat sim ilar, i.e., m ostly negative M onday effect and significant p o sitiv e Friday effect, it may then be concluded th at Indian stock prices show signs sim ilar to th o se o f developed capital m arkets in the w ake o f reform s.

Insofar as the d ay-of-the-w eek effect studies with Indian stock m arket returns are c o n c e rn e d , there is only one notable reference in the work o f Poshakw ale (1996), who studied this beh av io u r o f expected returns on the daily N ational Index data o f Bom bay Stock E xchange covering th e period 1984-94, and found evidence based on sam ple-based values o f daily m ean return and sta n d ard deviation o f daily returns that average returns are different on each day o f the w eek and that th e returns achieved on Fridays are higher com pared to other d ays o f the week. But inferences b ased on values o f such descriptive m easures, without actually carrying out appropriate tests, are som ew hat flawed, especially because the return data are m ost often highly autocorrelated. It appears that th e re is no other significant study on this particular behaviour o f expected returns on Indian stock prices from consideration o f either other standard stock prices/indices or application o f m ore appro p riate tim e-series based m ethodologies. In this paper we try to fill in this void to som e extent in the sen se that we study the change, if any, in the tw o periods - pre-1992 and post 1992J in the day-of-the- week effect on daily returns on Bom bay Stock Exchange (B SE), Sensitive Index (SE N SE X ) by using linear regression analysis with dum m y explanatory variables representing the effects o f different days o f week.

T he pap er is organized as follow s. We describe the m ethodology in the next section, and th e data and em pirical findings in Section-111. T he pap er ends with som e concluding rem arks in Section-lV .

2. The c a le n d a r tim e h y p o th e sis m a in ta in s th a t e x p e c te d re tu rn s on M o n d a y s w o u ld b e h ig h e r in o rd e r to c o m p e n s a te f o r the lo n g e r h o ld in g p e rio d . The tra d in g tim e view s, o n the o th er hand, sta te s th a t e x p e c te d r e tu rn s are e q u a l on d iffe r e n t d a y s o f the week.

3. P o st-1 9 9 2 p e r io d i.e., sub-period-11, c o ve rs the p e r io d 2 A p r il 1992 to 14 M ay 19 9 6 w h erea s p r e - 1 9 9 2 p e r i o d i.e., su b -p erio d -1 , s p a n s fr o m 2 J a n u a ry 1984 to 3 0 M a rch 1992.

II. METHODOLOGY:

We have stated in the preceding section that the unit root nature o f stock prices posses’ serious problem , and hence stock returns are used for the purpose o f em pirical analy­

sis. It is, therefore, appropriate that we first exam ine w hether the returns are indeed stationary from consideration o f trend. This is done by using Augm ented Dickey-Fuller test o f Said and Dickey (1984) and Phillips-Perron test (1988). We then consider the following regression model to study the day-of-the-w eek effect4.

R = ct + ccmdM, + a T ° T , + a T h DTh, + a F PF , + M = 1 - n ... ( ! ) W here, R is th e return on t-th day calcu lated as Rt = In P - In P the stock p rice on t-th day, a is the in tercep t term representing the com m on effect as on W ednesday, D vi, D (, D |h and Dk d en o te th e dum m y variables for M onday, Tuesday, T hursday and F riday respectively, a a T a,!,and a pare the co rresp o n d in g coefficients attached to these dum m y variables in dicating the d ifferen tial effects w ith W ednesday and e is the t-th disturbance term , w hich is assum ed to be w hite noise. By d efinition, any dum m y v aria b le representing a particu lar day takes the value 1 for that day and 0 otherw ise. A ssum ing that stock exchanges follow 5-day w eek w orking system , four du m m ies are co n sid ered for the regression m odel. All holiday w eekdays are rem o v ed to elim in ate any kind o f undue effect on the actual seasonality o f non-holidays.

T he m odel is estim ated using the ord in ary least squares (O LS) m ethod o f estim ation. U nder the assu m p tio n o f n o rm ality o f the erro rs, t-ra tio s are obtained and th e significance o f the dum m y v aria b les tested. S ignificant p o sitiv e/n eg ativ e value o f the coefficient for any particu lar day w ould m ean th a t the average return on that day is significantly p ositive/negative as com pared to the com m on day effect (i.e., W ednesday in ou r case), and hence conclusion may be draw n to the effect that the distrib u tio n o f stock returns d epends on the day(s) o f the week.

As w e are using regression analysis for ou r study, it is im perative that diagnostic tests on resid u als be carried out to find w hether erro rs are indeed w hite noise. This is done by com puting, L jung-B ox test statistic (1978), w hich is given by

Q (k) = n(n+2) £ r /(n-j) .M

W here r is the sam ple autocorrelation o f lag j o f O LS residuals and n is the total num ber o f observations. T his statistic follow s a x : distribution asym ptotically with k degrees o f freedom under the null o f w hite noise errors.

In case the null hypothesis is rejected i.e., the errors show significant autocorrelations, the regression m odel is then respecified by incorporating lagged values o f R( i.e.,

q

R , = a + aMDMt + ^cxtD^t| + aTiiDii-,i + ^ockD^f^ S P R^h + s t, l,2 ....n ... (2) j= l *

W here P^s are coefficients associated with the lagged values o f the dependent v ariable and q is ap p ro p riately chosen so that s ’s could indeed be assum ed to be w hite noise. It m ay thus be n o te d th a t in this situ atio n the re g ressio n eq u atio n in (1) w ould be c o n sid e re d to be a m isspecified m odel and the inferences based on (1) m ay be inappropriate. As regards applicatio n o f O LS to (2), it is w ell-know n that O LS w ould in this case produce consistent and asy m ptotically efficient estim ates.

III. EMPIRICAL RESULTS:

In this section we discuss the em pirical findings o f this study. As 4. It m a y be n o te d th a t sto c k returns m a y be r o n -s ta lio n a ry fr o m c o n sid era tio n o f sea sonality.

already stated, we have used the daily closing prices on Bom bay Stock Exchange (B SE) as m easured by BSE Sensitive Index (SEN SEX ). The data cover the period 2nd January 1984 to 14th M ay 1996.

This period has been divided in two parts - p re -1992 and post-1992, more specifically, 2nd January 1984 - 30th M arch 1992 (called sub-period I) and 2nd April 1992 - 14th May, 1996 (called sub- period-U) - keeping in mind the purpose o f this study viz. studying the m ovem ent o f stock prices in the wake o f m ovem ent tow ards econom ic liberalization. The returns have been calculated by using the first differences o f the logarithms o f SENSEX values. The analysed series thus represents the continuously com pounded rate o f return for holding the (aggregate) securities for one day.

In order to find if the return series is indeed stationary or not, we applied, as already stated, Augm ented Dickey Fuller (A D F) test [cf. Said and Dickey (1984)] and Phillips-Perron (P P ) test (1988). For both these tests a linear tim e trend was included to allow trend stationarity in th e data.

Both the tests soundly rejected the existence o f unit root in favour o f stationarity for each o f the two sub-periods viz., 1984-92 and 1992-96 as well as for j;he entire period spanning 1984-96. For instance, ADF and PP test statistic values were found to be - 1 5 . 186 and - 1950.159 respectively for sub-period 1. By com paring these values with the critical values as tabulated originally by Fuller (1976) and then extended by Guilkey and Schm idt (1989) and M acKinnon (1990), we find th at the hypothesis o f unit root is rejected by both the tests. N eedless to mention that the coefficient asso ci­

ated with the time trend t turned out to be insignificant, the resulting t-statistic value for sub-period 1 being 1.059 and 0.181 by AD F and PP tests respectively.

The results o f regression analysis are given in Tables-1 and 2. W hile the estim ated coefficients along with their t-statistic values are given in Table-1, the results o f diagnostic tests b ased on ordinary least squares (OLS) residuals are the contents o f Table-2. We observe from results o f regression (1) in Table-1 that for sub-period-I the coefficient associated with Tuesday is significant (only at five per cent level o f significance) and negative in sign while all other day effects are insignificant. The findings o f regression analysis with data o f sub-period-II is quite different from those o f sub-period-1 in that here the coefficient associated with Tuesday is no longer statistically significant; instead only Friday coefficient is significant and positive in sign at both five p er cent and one per cent levels o f significance (one-sided alternative considered). This clearly indicates how the day-of-the-week effect on SEN SEX returns has changed during p o st-1992 period as co m ­ pared to pre-1992 period. We thus find that the behaviour o f SENSEX after m ajor reform s were initiated in sub-period II (i.e., April 2nd, 1992 till May 14th, 1996 for the data) is sim ilar to that o f the developed capital markets insofar as “positive Friday effect” is concerned. The other day-of-the- week effect observed in the developed markets viz., “negative M onday effect” is still not prevalent in Indian stock prices as exem plified with SEN SEX data. A plausible explanation for this as w ell as the finding o f “negative Tuesday effect” in the pre-1992 period is that during m ost part o f eighties as also earlier the operational and information efficiency in Indian stock m arkets were rather insig­

nificant and m arket behaved m ore as a custom ary response to officially regulated im pulses rather than those based on free market stimuli. It is towards the end o f eighties and thereafter that this character o f Indian capital markets began to change along with some m ajor changes in the Indian econom ic system as well. O ur findings in the two sub-periods thus seem to suggest that in the w ake o f econom ic reforms in 1992 and onwards, the m ovem ent in the tim e series o f stock prices as given by SEN SEX shows som e sign o f behaviour (in respect o f day-of-the-week effect) sim ilar to that observed in developed capital markets..

TAB LE -1 : DA Y- TO -T HE -'W EE K EF FE CT ON DA IL Y RE TU RN S ON SE NS EX DATA

SOO '

&

2

.2 2

v 2 a . -b £—

e ua t

cS

3

_C

ctf

vOO'

"T 2

■g ^O <N

‘C os V O' O-I JZ

3

in

•v© 'uo>

fi-».O

3

<N ON x:0

■£© CO1 TfO0 O'

&

<a c_CO

■oc (N

0/)d>

erf

&D 0)

r_i

4

eo

ofi -8

co tf)a>

ob

(NiC

©C

/5

trta>

OJ)<L>

erf

co a>

So erf

c

e

_<u

UO

ro ^ J °~

<n ^ 2 d O o <N —: o

*-* o o O © © © o o o 6

\0 — '■O cO O '

O ' co

coTf O — CS (N

O ^ £T

m so O ' O ' r t — m r"

• • • uo r-. r r — •

—. — O ' o

^r( \0 m c-i

S rK J® S

o ee ^

o o 9 9

00

<N CO CO vO 00 O rr>

^ ^ •

O O O o

*_ co ^ *o r ? 00_{(N o} tN o ^ 0C[ sC> wo ro ^

^ ^ ^ ^ S

[ r o ^ o ' 3

£ <n ~ 0 -

« P <? o 9

o 't <N n

— r- o 00 vo o rt o ^ r -

^ o (N p © O ' W W w p. r~- o vo

© © *2 2© © q o

© P 9 9

o d

S o . (N

04

O o o o

© o

2 ^ p £ g~

^ , t«s [ S *— © O'1/1 fO_

T o o o n

s i s s g

© © © © © o d d o o

0 0 /—S

VO (N r - 00 On

1—* VO t o 00 ^{0 0} r s

O ' i n

<N r f CO ⁰⁰( 0 <N »n CO

'w ' r f ' ^i n r f 0 0 CO O '

<N O 0 0 00 CO VO

O p 0

d d d 0 0 0 0

CO ( S ‘

f'i °i K

— <N n M

3

•rt- t— (N P o o

9 9

° r—■ »0 5 ; * N r i

^ ^ 'O

^ vO in r , s l / 6

g | S 2 P • O '—

d ^ o 0

CO ro

• 0000 O'

•— o (N

o p d o

v~> ^

^ /-v h ro O' (N 00 (S

r ^

P ^ SO (N

o o CO ^ Tt ro CO t£

CO o o 0 s Os o o o 00 r^;

0 0 0 0 9

/— V ^ '

i n « « ^ ro O' CO in 00 (N ^

(N O 00

v ^

90

^{O ' h}

O' r f <n r^*

«£) 'q ^ ^

P in co ri

_{^ w}

f\J co 00 00 o — <N 2

vc? «n ^

d o

so uo t^r JQ r~ 0 53 ^

CO CO .

o p o 9 9

T * 00

— ^n

^ P

d d

00^ n O' ^ f ' ^ C

^ p rri S

9

^{^ ® S ,}

<-vj sO 0 s w0 r—«

q o p p

d d d d

© — 2J

■5 t_ £ L_ — <N rr? cCL CQ. CQ- CD.

a s 2 a H b d c a . c a c 2 . c Q . a 3 . c a c Q . c 3 . c J .

-Si v>

cx

*c

§ -N C S

_^

3

^{^ ?}

^

5

^,

^ ■ s Cl. o ^

£• “O os:

■§^

5

^.

' .1*a

S

^.§)

3 ^ to ^

§ 2> a . 2 . 8 ^

O - S

TABLE : 2 LJ UN G- BO X TE ST ST AT IS TI C (Q (K )) VA LU ES OF M OD EL S FO R SE NS EX DA TA

* * — ON * * ON

o OO Tf * * r"- o r -

l/~) rn ON

vb Tf

— vb — cs

00

c«^ ro ro

^r

* r f

* On

00 00

vo

vo

* vo

«■ vO

p vO On * *

00

o r -

iT) O vO r-* vb •rf

— — VO On tT CO ro ro

r t — r - r - vo vo

fS CN /«~S

• ON * * ON vo * * r f

Os o ON VO * * On ON O CN

tri © ON ON ON ON vb

OO x i — vb o \ o (N ro ro ro

u >*

r - r -

C

9

^{vo vo}

£

* o

*

m CO

a

£ * * wo S *

o

* (N rr

(N o

n

1

vS O

VO ■**

00

^{ON (N Tf vb}ro ro

fN ro

*

1

^{VO VO ON}

1

vo vo

* * o <N

00

*

00

o CN ON * * ON o <N ro

CN

Os ir! ro

CN

ONwm

00

CN rn CN

ON vb

ro T^' ro

r -r - u

«

TT tT

U Q.

VO vo u

eq

3

^{VO vo ON}

3

C

*

O *

i/i ro < * * t"- C

W

*O •N- r i

ON

“5

^{K r -vq <N ro}d

VO (N

LT) vb

ro ro CNir \

N Tf

* * •o *o

.2

VO

* O

* ro

■o o ON o * * uo O n o

VO E

VO

0 4

-

1

>

J

33

ro d

VO ON

VO OJ

CL vb ro ro

IT)O u

CUi

•'I*

* Tfr

*

d cy

u VO

* vo

3

^{© r - C}

/5

^* *

00

^C O -N-(N

C/>

*0

^ro_CN

00

^fN CN u

vb VO

ro i— «o VO ro ro

^r

* *

CN CN (N d vo

* VO

■«■

d

© CN * * fN O <N

00

wS O vq

vb ON vd

^ro vb

ro ro o o

■ ' d — d vo vo d

OX)

00

oo Vi jo C/3 CO

CO CO C

3 ”3 13

"W

03

J

3

J

2

_c

/5

s_

3

rs

001

_i-,^>

3

*v

5

e

3

• 'oo<L>

u.

T3

3

Vi<U u

3

rsi/5 a)u

<L>

_3

>

c o

"O

co C

c .200

C/3t>

<N C toI7i

oo<u

CO cu.

c .2v_oo

5

0>

(N C .2V) 00u

inC v*

C .200 V54>

(N C

#o i/5Vi 1>

Uh ST Ofi

O'

0)CO cd CD

3

4J 0 *

013

L .

<U Od

ob«

oC

•*-*3

<U u«OX) 1)

u

« ) t *

3 0) C*

s»«

<u

a : Indicates significanceatboth5 per cent and 1 per cent levels * Indicates significanceat5 per cent levelonly.

We have already stated in Section II that d iagnostic tests based on residuals are very im por­

ta n t for p ro p er inferences about a m odel. T ow ards this end we com puted L jung-B ox Q (k) sta tis­

tic values with O L S residuals o f regression (1) for both the su b -p erio d s and these are given in T ab le 2. H ow ever, we first com puted the values o f L jung-B ox statistic w ith the return data in o rd e r to find the extent o f linear dep en d en ce am ongst the o b serv atio n s. All co m p u tatio n s o f Q (k ) statistic w ere done for lags up to 24, but we report for som e lagged values only for reasons o f brevity. We o bserve from this table that auto co rrelatio n s am ongst the o b serv atio n s are highly sig n ific a n t in su b -p erio d -I; the sam e is the conclusion w ith o b se rv a tio n s from su b -p erio d II although Q (k) statistic values are now m uch reduced yet significant. L jung-B ox diagnostic test b ased on O LS residuals o f regression (1) also produces the sam e conclusion viz., the residuals ex h ib it high au to co rrelatio n s - far from being w hite noise process as assum ed in ( I ). O bviously, in feren ces based on this regression m ay then be m isleading o r im proper.

To take account o f the existing au to co rrelatio n s in the errors, we estim ated regression e q u a ­ tio n in (2) in w hich q was chosen to be sufficiently large at 12 i.e., 12 lagged values o f R w ere tak en to be reg resso rs in addition to the four dum m ies representing the day effects. It is evident from the results o f regression (2) in T a b le -1 th at the day-o f-th e-w eek effects viz., “ negative Tuesday effect” in pre-1992 period and “ positive Friday effect” in p o s t- 1992 period rem ain the sa m e as in regression (1). As for the significance o f the lagged coefficients w e note that w hile all th e 12 lagged coefficients are significant for the data from su b-period I, for sub-period-II only e ig h t lagged coefficients are significant. T he results o f the d iagnostic test based on O L S residu­

a ls from regression (2), as given in T able-2, show clearly that ex c ep t for lagged values 12 and 15 fo r sub-period-I at five per cent level o f significance, no significant auto co rrelatio n exists in the resid u a ls o f regression (2) for both the sub -p erio d s. Thus, w e find that the ap propriate regres­

s io n equation for studying the d ay-of-the-w eek effect in the tim e serie s o f S EN S E X is given by reg ressio n (2) since th e residuals o f this regression turn out to be a w hite noise process.

Finally, b efo re concluding, we m ention th at w e have reported in Tables-1 and 2 the results o f regression analy sis based on th e entire d ata set spanning the perio d 2nd January 1984 to 14th M ay 1996. It is clea r that the findings in th is case are exactly the sam e as those for sub-period- I. F or instance, the conclusions reg ard in g the prevalence o f the d ay-of-the-w eek effect rem ain th e sam e viz., significant (with negative sign) T uesday effect and no o th e r significant day effect.

H ence, it m ay be concluded that the analysis with the eiftire data set, as used in this study, show s h o w conclu sio n s on the day-of-the-w eek effect could be m isleading if no a p rio r i consideration to m ovem ent to w ard s econom ic liberalization since 1992 is explicitly introduced in the analysis.

IV. CONCLUSION:

In this paper we have examined whether the behaviour o f Indian stock prices in the sense o f day-of-the-week effect on returns shows any sign o f change during the period o f major economic reform s beginning in 1992 as compared to the preceding period. We were motivated to this study from consideration o f the fact that in the wake o f more emphasis on reorganization o f Indian capital markets since the late eighties and the subsequent thrust since 1992 towards m ajor economic reforms in most sectors o f the economy including financial sector, the behaviour o f potential investors (for whom such effects pose interesting buy and hold strategies) might show some indication o f change sim ilar to those observed in developed capital markets. To study this “day effect” we have used the method o f regression analysis with dummy variables

and also carried out ap p ro p riate diagnostic checks based on least squares residuals. This im portant finding o f this study is th a t the day-of-the-w eek effect has changed from T uesday (with a negative sign) in p re-1992 period to F riday (with a positive sign) in p o s t-1992 period, the later o b servation being the sam e as in d e v e lo p e d capital m arkets. Since the other notable d ay-of-the-w eek effect observed in studies w ith return data on developed capital m arkets viz., “ M onday (with a negative sign) effect,” has not been found to be significant with returns on S EN SEX , we may conclu d e that th ere is som e evid en ce in Indian stock prices tow ards change in the d ay-of-the-w eek effect in th e sense o f it being p artially sim ilar to those observed for developed m arkets.

T h is fin d in g should b e useful to potential investors since it m ight help them in d ecid in g th eir buy and hold strategy issues. Further, this result m ight suggest that further researches may be d one to find if in th e w ake o f econom ic reform s Indian stock prices are show ing signs o f change in term s o f o th e r im portant characteristic features o f developed capital m arkets. In 1998, Basu and M orey have stu d ied how reform s have changed the Indian stock prices in term s o f obeying F am a’s efficient m arket hypothesis, and found signs o f g reater efficiency after the b e­

ginning o f econom ic reform s.

REFEREN CES

Basu, P. ctndM.R. M orey (1998): “S to c k M arket ‘P rices in India after E co n o m ic Liberalisation, "

E conom ic a n d P olitical Weekly, vol. X X X III, No. 7, 353-358.

Cross, F (1973): “The b eh a v io u r o f S to c k P rices on F ridays a n d M ondays, ” F in a n c ia l A n a ly sis Journal, Vol. 29, 67-69.

French, K.R. (1980): “S to c k R eturns a n d the W eekend E ffe c t,” Jo u rn a l o f F in a n cia l E co n o m ­ ics, Vol. 7, 55-69.

Fuller, W.A. (1976): Intro d u ctio n to S ta tistic a l Time Series, N ew York: Wiley

Guilkey, D .K a n d P. S c h m id t (1989): “E xten d ed Tabulations f o r D ickey-F u ller Tests, " E c o ­ nom ic Letters, Vol 31, 355-357.

Keim, D (1985): " S ize -R ela te d A no m a lies a n d S to c k R eturn Seasonality: F urther E m p iric a l E vid en ces," Jo u rn a l o f F in a n c ia l Econom ics, Vol. 12, 13-32.

Keim, D .B a n d F. S ta m b a u g h (1984): “A F urther Investig a tio n o f the W eekend E ffects in S to c k R e tu rn s,” Jo u rn a l o f F inance, Vol. 39, 819-839.

Lakonishok, J a n d M. L ev i (1982): “W eekend E ffects on S to ck R eturns: A Note, " J o u r n a l o f Finance, Vol. 37, 883-889.

Ljimg, G .M a n d G.E.P. B ox (1978): “O n a M easure o f Lack o f Fit in tim e S eries M odels, "

Biom etrika, Vol. 66, 67-72.

M acK innon, J.G . (1990): “C ritica l Values f o r co-in teg ra tio n Tests, ” UC S a n D iego D iscussion Paper, 90-94.

Phillips, P.C.B. a n d

P.

P erro n (1988): “Testing f o r U nit R o o ts in Time S e rie s R egression, "

B iom etrika, Vol. 75, 3 3 5 -3 4 6

Poshakw ale, S (1996): “E v id en ce on Weak fo r m E fficien cy a n d D ay o f the Week Effect in the In d ia n S to c k M arket, " F in a n c e India, Vol. X, No. 3, 605-616.

Rogalski, R (1984): "N ew F in d in g s R egarding D ay o f th e Week R eturns over Trading a n d N on- T rading P e r io d s ,” J o u r n a l o f Finance, Vol. 38, 1603-1614.

Sqid, S.E. a n d D.A. D ic k e y (1984): "Testing f o r Unit R oots in A utoregressive M o vin g A verage m o d els w ith U nknow n O rders, ” Biom etrika, Vol. 71.599-607.

Taylor, S. (1986): M o d e lin g F in a n c ia l Times Series, N ew York: Wiley.