Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Literature Review
There is wide literature available on random walk hypothesis
and stock market efficiency. The random walk model was
first developed by Bachelier (1900) in which he confirmed
that price changes are nearly uncorrelated resulting to
a random-walk like behavior or random nature of price
changes. Samuelson (1965) and Fama (1965, 1970) triggered
keen interest in this area. Samuelson (1965), Fama (1965)
and Jennergeen and Korsvold (1974) examined the behavior
of stock returns by applying serial correlation tests and they
found markets as efficient.
Poshakwale (1996) examined on weak form efficiency in
Bombay Stock Exchange by using serial correlation and runs
test on the selected data. He concluded that market is not
weak form efficient. Alam et al. (1999) tested the random
walk hypothesis for Bangladesh, Sri Lanka, Hong Kong and
Taiwan. They discovered that except Sri Lankan all other
stock indices follow a random walk. Mobarek & Keasey
(2000) in their study on Dhaka Stock exchange applied
Runs test, Smirnov normality test, Auto-correlation, Autoregression, and ARIMA on daily price indices, empirical
results revealed that Dhaka stock exchange is not weak form
efficient.
Worthington & Higgs (2006) studied daily stock returns
of emerging markets like China, India, Korea, Malaysia,
9
2009*
2010
2011
2012
5441.12
6368.80
6754.92
5695.12
5760.84
4198.4
6067.4
3013.2
2366.2
4319.2
6.3
23.9
11.3
14.1
23.0
3388
2285
1317
991
1065
3346
2246
1274
981
1025
42
39
43
10
40
248
246
247
246
249
13.66
9.29
5.33
4.03
4.28
7912
9093
10902
10342
11665
10
Skyline Business Journal, Volume IX-Issue 1-2013-2014
ButBut
in this
serial
correlation
create
in this
serial
correlation
cr
But in this serial correlation creates problem, to counter this Augmented Dickey Fuller
includes
lagsDickey
of the
(ADF)
equation
includes
lags
of t
But
serial
correlation
creates Dickey
problem,
to(ADF)
counter
this Augmented
ut in this
serial correlation
problem,
to counter
this Augmented
Fuller
testequation
extending
till 30th June,creates
2013
for
dailyin&this
weekly
returns.
of aof
unit
(ADF)
equation
includes
lags ofThe
the null
firsthypothesis
differences
Pt. root is rejected in each case if
The return is calculated as the logarithmic difference between the test statistic is more negative than the critical value. A
ADF
unitunit
rootroot
testtest
is conducted
The
is conduc
(ADF)
equation
lags ofwith
theunit
firstroot
differences
P
ADF) equation
includes lags
differences
of Pincludes
two consecutive
pricesofinthe
a first
series,
yielding
continuously
series
is said The
to of
be
non-stationary
and
follow
t.ADF
t.
The
ADF
unit
root
test
is
conducted
using
the
following
regression
equation:
compounded returns.
random walk. But in this serial correlation creates problem,
counter this
Augmented
Dickey Fuller
test (ADF)
equation
The
unit root
test is to
conducted
using
the following
regression
equation:
he ADF unit root test is conducted using
theADF
following
regression
equation:
The MSM Index (MSM-30) was established in 1992 with includes lags of the first differences of Pt.
a base year of June 1990 and base value of 1000. It serves The ADF unit root test is conducted using the following
Pit =Pit =
In equation
PP
the
firstfirs
d
tPit-1
is
the
In iequation
it = and
tPcompanies
it-1iPit-i
it is
it
high capitalized most
are+ it
is the
first difference term i.e. (Pit - Pit-1), it is the white noise or e
In equation
(1), Pit that
listed on MSM. Index includes 30 companies selected on the In equation (1), Pit is the first difference term i.e. (Pit - PitIn 40%),
equation
(1),(P
P
difference
term
i.e.
(P
P
-1the
),constant,
constant,
noi
equation
(1),
Pit capitalization
is the first difference
term
i.e.
),the
it itfirst
isisthe
whitenoise
noise
error
basis
of market
(weight
liquidity
(weight
the
white
or or
error
term,
is
itthe
it is the
term,
it is-is
the
constant,
white
isthe
coe
it - itPit-is
11),
term,
is
is the
term,
15%).
is the The
constant,
is thethe
coefficient
and i are
coefficients
45%) and earnings per share
(weight
components
coefficientfor
for the
the trend
trend and
i are
coefficients to
to be
be estimat
MSM-30 index are revised every
on
the constant,
basis of estimated.
term,year
the
is the
the coefficient
the trend and i are coefficients to be e
constant, is the coefficient
for
trend
and areiscoefficients
to befor
estimated.
rm, isofthe
performance and index components are changed every year i
One
advantage
of the
PP
teststests
ov
One
ofthat
the
in the month of July.
advantage of the PP tests
over
theadvantage
ADF tests
thePP
One advantage of the PP testsOne
over
the ADF tests is that
the
PP tests isare
robust
to s
tests are basis,
robust to
general forms of
3. The free float stocks and capping is revised on a PP
quarterly
byserial
endcorrelations
of March,and
June,
correlations
and
general
of ho
correlations
and
forms
One
advantage
of
the
PP
tests
over
the
ADF
tests
is
that
the
PP general
testsforms
are
robus
The
Muscat
Securities
Market
Index,
MSM
30,
is
a
heteroskedasticity
in
the
error
term
ut.
Another
advantage
is
One advantage
of
the
PP
tests
over
the
ADF
tests
is
that
the
PP
tests
are
robust
to
serial
3. The free float stockscorrelations
and capping
isgeneral
revisedforms
on a of
quarterly
basis, by end
of March,
June,
advantag
and
heteroskedasticity
intothe
errora lag
term
ut. Another
capitalization-weighted
index
of
the
30
most
highly
that
the
user
does
not
have
specify
length
for
the
test
3. The freeand
float
stocks and
is sample
revisedamendment
on a quarterly
endbeginning
of March,ofJune,
September
December.
Butcapping
the index
takesbasis,
place by
in the
July
capitalized,
liquidforms
and profitable
companies listed
on
the
regression.
KPSS testadvantage
also that
uses
autocorrelation
correction
the
user
does
not
have
to spec
that
the
user
does
have
to ad
sp
correlations
and
general
forms
of
heteroskedasticity
in
error
unot
.
Another
isthe
orrelations
and
general
of
heteroskedasticity
in
the
error
term
u
t. Another
t
September
and
December.
But
the
index
sample
amendment
takes
place
in
the
beginning
of
Julyterm
the user
does ofnot
lag PP
length
forinthe
test parametric
regression.
KPSS is
test also
Muscat Securities Market. that
The main
objective
thehave
MSMto specify
approachalike
test but
KPSS
approach
September
and December.
But the inindex
sample amendment
takes
place in
the
beginning
of July
each
year.
30
isdoes
to
represent
all to
thespecify
stocks listed
the
exchange
and
adopted
to overcome
thelength
problem
of the
autocorrelation.
In KPSS
autocorrelation
correction
approach
autocorrelation
correction
approa
that
the
user
does
not
have
to
specify
a
lag
for
test regression.
KPSS
tes
at the user
not
have
a
lag
length
for
the
test
regression.
KPSS
test
also
uses
each
year.
to
work
as a benchmark index
for the investors.
To achieve
the null
hypothesis
different
fromparametric
the earlier two
tests as itis adopte
autocorrelation
correction
approach
like
PP test isbut
in KPSS
approach
eachMSM-30
year.
The
includes
sectors features:
namely banking
and investment,
and
these
objectives,
the MSM three
30 hasmain
the following
assumes
stationarity ofindustry,
the variable
andservices
the
alternative
isoftheautocorre
overcome
theparametric
problem
overcome
the
problem
of autoc
autocorrelation
approach
like root
PPistest
but
in
KPSS
approach
is
utocorrelation
correction
approach
like
PP sectors
test
but
in correction
KPSS
parametric
approach
adopted
to
The
MSM-30
includes
three
main
namely
banking
and
investment,
industry,
and
services
1.
Only
freely floating
shares
for
trading
are
included
in
the
presence
of
unit
or
non-stationarity
of
the
variable
or from
overcome the problem of autocorrelation. In KPSS the null hypothesis is different
The insurance.
MSM-30
includes
three
sectors
namely
banking
investment,
industry,
and Banking
services
construction
of the
Therefore,
shares
like shares
of30
random
To increase
the robustness of the
and
Inindex.
MSM
-30 main
for other
year
2011-2012
out
ofpresence
theand
total
indexwalk.
constituents,
earlier
two
tests
asroot
itasassumes
statio
earlier
two
tests
it assumes
st
overcome
the
problem
of
autocorrelation.
In
KPSS
the
null
hypothesis
is differen
vercomeheld
the
problem
of
autocorrelation.
In
KPSS
the
null
hypothesis
is
different
from
the
by
the
government,
promoters,
strategic
holdings
and
result
all
the
three
tests
are
used
to
detect
unit
the
and insurance. In MSMearlier
-30 fortwo
year
2011-2012
out ofstationarity
the total 30ofindex
constituents,
Banking
tests
as it assumes
the variable
and the
alternative
isin the
presenc
locked-in
shares which
are not
for
least 3 months
Insector
MSM
-30tradable
for year
2011-2012
out ofaseries.
the
constituents,
and insurance.
investment
continues
to at
dominate
with
totaltotal
of 30
13 index
companies,
A totalBanking
of 11
unit
rootroot
or or
non-stationarity
thp
unit
earlier
two
tests
as
it
assumes
stationarity
of
the
variable
and
alternative is of
theof
rlier two
tests
as
it
assumes
stationarity
of
the
variable
and
the
alternative
is
the
presence
of
are
excluded.
and investment sector unit
continues
dominate with of
a total
of 13 companies,
A of
total
of the
11 non-stationarity
root orto non-stationarity
the
variable
or
presence
random
walk.
To
increase
Ratio
Test
and investment
sector
continues
dominate
aVariance
totalin of
13
companies,
total
of 11
companies
from the
Industry
sectortoand
about 6 with
companies
the
Services
sector.AThe
break-up
robustness
of
result
all
the
three
In
this
study
variance
ratio
test
developed
by
Lo
&
Mackinlay
2.
To
ensure
wider
representation
of
smaller
companies
a
robustness
of
the
result
all the
unit
root
or
non-stationarity
of
the
variable
or
presence
of the
random
walk.
To thr
in
nit root companies
or non-stationarity
of
the
variable
or
presence
of
random
walk.
To
increase
the
from the Industry
sector and about 6 companiestests
in the
sector.unit
Theroot
break-up
robustness
areServices
used
detect
in therandom
series.walk
(1988) is used
on thetoMSM-30
index to study
10% capping (CAP) is set in
the index. of the result all the three
companies
from the Industry
sector andforabout
companies
theVariance
Services
sector.
The
break-up
of
market capitalization
for MSM-30
year62011-2012
isin53%
for Banking
and
Investment
ratio
after
test,
robustness
the result
allhypothesis.
the
tests
are
usedistoapplied
detect
unit the
rootunit
in root
the series.
bustnessofofmarket
the result
all the threefor
tests
are usedfor
toofdetect
unit root
inisthree
the
series.
capitalization
MSM-30
year
2011-2012
53%
for
Banking
and
Investment
3. The free float stocks and capping is revised on a quarterly as the variance ratio test is more powerful and robust than
of market
year
2011-2012
isIndustry
53%
Banking
sector,
for Services
& MSM-30
Insurance
sector
and
14.5%
forroot
sector.
unit
tests.for
Variance
ratioand
test Investment
is based on the concept
basis,
by32.5%
endcapitalization
of March,
June,for
September
and for
December.
But
Variance
Ratio
Test
Variance
Ratio
Testa
sector,
32.5%
for
Services
&
Insurance
sector
and
14.5%
for
Industry
sector.
of
random
walk
that
if
a
time
series
of variable
Pt follows
the index sample amendment
takes
place
in
the
beginning
of
Variance Ratio Test
sector,
for Industry
random
walk thensector.
the variance of increments of Pt is linear in
July
each32.5%
year. for Services & Insurance sector and 14.5%
Variance Ratio Test its data interval. This means that the variance
ariance Ratio Test
In this
study
variance
ratio
In this
variance
ra
of study
In this
study
variance
ratio
test
developed
by
Lo
&
Mackinlay
(1988)
is
used
on
The
MSM-30
includes
three
main
sectors
namely
banking
and
(P
P
)
=
(P
P
)/n,
or
the
variance
of
its
n-differences
is
t
t-1
t
t-n
Unit Root Test
n timesratio
the(1988)
variance
its first
difference
for
a to
random
walk
industry,
and
services
and
insurance.
In
MSM
MSM-30
index
study
random
MSM-30
index
to study
random
In
this
study
variance
test
developed
Mackinlay
(1988)
is wu
In investment,
this
study
variance
ratio
test
developed
by
Lo
&
Mackinlay
isofused
onby
theLo &
Unit Root Test
toconstituents,
study random
walk hypothesis. Variance ratio is applied after the unit
-30 for year 2011-2012 out MSM-30
of the total index
30 index
series.
Unit Root Test
Banking and investment sector continues to dominate with
test,
as the
variance
testtest
is mor
test,
as the
variance
ratio
isthm
index
to study
walkafter
hypothesis.
Variance
ratio
is ratio
applied
after
SM-30 index to study random walk MSM-30
hypothesis.
Variance
ratiorandom
is applied
the unit
root
null
and alternative
hypothesis
the root
test are
stated
as
a total of 13 companies, Atest,
totalasofthe
11 variance
companiesratio
fromtest
the is The
more
powerful
and robust
thanofunit
tests.
Variance
ratio
unit
tests whether
a time
series variable is nonstationary using an
Industry A
sector
androot
abouttest
6 companies
in the Services
sector.
is robust
based
on
the
concept
of random
w
istest
based
on an
the
concept
ofVarianc
random
test,
as MSM-30
the
ratio
test
is more
and
than
root
tests.
st, as theThe
variance
test
moretests
powerful
andvariance
robust
than
unit
root
tests.powerful
Variance
ratio
H
The
variance
at series
lag
n, VR
= 1 unit
break-up
of market
for
for
year
A ratio
unit
rootiscapitalization
test
whether
a
time
series
variable
nonstationary
using
0: walk
is based on the concept of random
that is
if aratio
time
of (n)
variable
Pt follows a random w
Thetovariance
at lag n,ofVR
(n)using
1 an
2011-2012
53%model.
for
and
Investment
32.5%
A isunit
rootBanking
test
tests
whether
aroot
time
nonstationary
autoregressive
Three
different
unitsector,
testsseries
areHused
test is
theratio
presence
a unit
root
in
1: variable
then
the the
variance
of increments
Pot
then
variance
of
follows of
a ra
is
based
on
the
concept
of
random
walk
that
if
a
time
series
of
variable
Ptincrements
follows
a
random
walk
based on
the
concept
of
random
walk
that
if
a
time
series
of
variable
P
t
for
Services
&
Insurance
sector
and
14.5%
for
Industry
sector.
autoregressive model. Three
different
unit
root
tests
are
used
to
test
the
presence
of
a
unit
root
in that
then the variance of increments of Pt is linear in its data interval. This means
the variance
VR(n)
=the
Var[rt(n)]
/[n.
is the
model.
Three different
unit root tests
areWhere
usedthe
to
test
presence
ofVar(rt)]
atest,
unitand
root
invariance ratio
aautoregressive
series: namely,
the Augmented
Dickey-Fuller
(ADF)
test,
Phillips-Peron
(PP)
the
then
the
variance
of
increments
of
P
is
linear
in
its
data
interval.
This
at
lag
n
Unit
Root
Test
en the variance
of
increments
of
P
is
linear
in
its
data
interval.
This
means
that
the
variance
of
t
t
a series: namely, the Augmented Dickey-Fuller (ADF) test, the Phillips-Peron (PP) test, and the means that the v
A unit root test tests whether a time series variable is non
a series: namely,
the Augmented
Dickey-Fuller
(ADF)
the Phillips-Peron
(PP) test,
and the
Kwiatkowski,
Phillips,
Schmidt and
Shin (KPSS)
test.test,
Dickey-Fuller
test involves
fitting
stationary using an autoregressive model. Three different Due to the above relationship, Lo & Mackinlay (1988) says
Kwiatkowski,
Schmidt
and Shin
(KPSS)
involves
fittingmust
the be five times
thatDickey-Fuller
the variance of test
weekly
price changes
unit
root tests arePhillips,
used to test
the presence
of a unit
root intest.
Kwiatkowski,
Phillips,
Schmidt
and
Shin
(KPSS)
test.
Dickey-Fuller
test
involves
fitting
the there are
regression
model
as:
a series: namely, the Augmented Dickey-Fuller (ADF) test, the variance of a daily price change. As in general
regression
model(PP)
as: test, and the Kwiatkowski, Phillips, five working days in a week for the stock markets.
the
Phillips-Peron
regression
model
as:
=
P
+
u
, where
denotes the price
at time t
P
Schmidt
and
(KPSS)
test.PtDickey-Fuller
test involves
t
t-1 Shin
t
fitting
regression
model
as: Pt denotes the price at time t Findings
+ ut
, where
P t = thePt-1
the root
unit versus
root tests
wereis applied
once on daily return
,
where
Pt
denotes
the price
at
time
t at time the
P
=
P
+
u
,
where
Pt denotes
pricecontains
t Allunit
t
t-1
t
Test examines the null hypothesis
thatthe
series
series
stationary.
and
then
on
weekly
return
of
MSM-30
Test
examines
the
null
hypothesis
that
series
contains
the
unit
Test examines the null hypothesis that series contains the unit root versus series is stationary. index. Results of
Augmented
Dickey-Fuller
Phillips-Perron (PP)
root
versus
is stationary.
=1 seriesthe
H
Test
null hypothesis that series contains the
unit root versus
series is(ADF),
stationary.
0: examines
tests
and
Kwiatkowski,
Phillips,
Schmidt,
and Shin (KPSS)
H0: =1
test
for
the
daily
observations
are
presented
in Table II,
H :: =1
<1
10
11
H1 : < 1
Skyline Business Journal, Volume IX-Issue 1-2013-2014
The
H1 : null
< 1hypothesis of a unit root is rejected in each case if the test statistic is more negative than
Difference
ADF t- statistics
ADF p-value
ADF t- statistics
ADF p-value
-1.485875
0.5405
-23.53574
0.0000
PP t- statistics
PP p-value
PP t- statistics
PP p-value
-1.568487
0.4983
-23.87755
0.0000
KPSS t- statistics
Asymptotic critical
KPSS t- statistics
Asymptotic critical
values
1.347724
1% level- 0.739000
0.174591
5% level- 0.463000
5% level- 0.463000
0.551229
0.231355
0.125146
0.054104
Daily
-12.34733
-9.652798
-7.129039
-5.236520
12.34733
Daily
Probability
0.0000
0.0000
0.0000
0.0000
0.0000
Data
Statistics
q=2
q=5
q=10
q=20
Joint Test
(at period 2)
Variance
Ratio
Z*(q)-statistic
0.551229
0.231355
0.125146
0.054104
Daily
-3.013568
-2.850846
-2.579530
-2.372500
3.013568
Daily
Probability
0.0000
0.0020
0.0100
0.0210
0.0100
Data
Statistics
q=2
q=5
q=10
q=20
Joint Test
Type
Max |z|
Under Homoskedasticity
Weekly
For thethe
KPSSKPSS
tests of thetests
null hypothesis
unit root,
the LM-statistic
asymptotic
For
of theof no
null
hypothesis
ofexceeds
no the
unit
root,
critical
value at the .01 level
at the levelthe
series,asymptotic
indicating these series
are non-stationary.
the
LM-statistic
exceeds
critical
value atSince
the
.01
level
at
the
level
series,
indicating
these
series
are
nonthe ADF, PP and KPSS tests on the log of prices accepted the presence of unit roots, there is no
stationary.
Since the ADF, PP and KPSS tests on the log of
evidence against weak form efficiency when Muscat securities market was tested on daily data.
prices accepted the presence of unit roots, there is no evidence
against weak form efficiency when Muscat securities market
was tested on daily data.
Max |z|
Under Heteroskedasticity
Daily
1% level- 0.739000
(at period 2)
Variance
Ratio
Z(q)-statistic
Type
values
Max |z|
Under Homoskedasticity
(at period 2)
0.540925
0.208433
0.091159
0.047472
Weekly
Variance
Ratio
Z(q)-statistic
-5.584890
-4.395398
-3.274661
-2.331633
5.584890
Weekly
Probability
0.0000
0.0000
0.0011
0.0197
0.0000
Data
Statistics
q=2
q=5
q=10
q=20
Joint Test
Type
Max |z|
Under Heteroskedasticity
Weekly
(at period 2)
0.540925
0.208433
0.091159
0.047472
Weekly
Variance
Ratio
Z*(q)-statistic
-4.375854
-3.779718
-2.998561
-2.287134
4.375854
Weekly
Probability
0.0000
0.0000
0.0010
0.0030
0.0000
1.
2.
3.
ADF p-value
ADF t- statistics
ADF p-value
-1.573953
0.4934
-11.67795
0.0000
PP t- statistics
PP p-value
PP t- statistics
PP p-value
-1.573953
0.4934
-11.67002
0.0000
KPSS t- statistics
Asymptotic critical
KPSS t- statistics
Asymptotic critical
values
0.552271
1% level- 0.739000
Difference
values
0.245666
1% level- 0.739000
5% level- 0.463000
5% level- 0.463000
12
Concluding Remarks
The rationale of this study is to explore and test the random
walk and weak-form informational stock market efficiency in
the Oman. To satisfy the requirement two different procedures
are employed on daily and weekly returns of MSM-30 index:
(i) ADF, PP and KPSS unit root tests and (ii) Variance ratio
test. The findings show empirical evidence that Muscat stock
index (MSM-30) exhibit unit root for both daily and weekly
returns. All three unit root tests are consistent and are giving
the same result that there is no evidence against weak form
efficiency of Muscat securities market. In variance ratio
test the rejection of null hypothesis of a random walk under
assumptions of both homoskedasticity and heteroskedasticity
is rejected for MSM-30 index concluding that Muscat
Securities market is not weak form efficient. Therefore, it
is expected that investors can generate abnormal or undue
returns using past information and technical analysis which
helps in identifying under-priced shares. The findings of this
study will supplement existing literature on testing of random
walk and weak-form market efficiency in the emerging
markets, especially in GCC countries.
References:
1.Abraham, A., Seyyed, F. J. & Aisakran, S.A. (2002).
Testing the random walk behavior and efficiency of the Gulf
stock markets. The Financial Review, 37:469-480.
16. Pan, M.S., Chiou, J.R., Hocking, R., & Rim, H.K. (1991).
An examination of mean-reverting behavior of stock prices in
Pacific -Basin stock markets. Pacific -Basin Capital Markets
Research, 2: 333-342
21. www.msm.gov.om
Copyright of Skyline Business Journal is the property of Skyline University College 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.