Testing Random Walk Hypothesis: An Empirical Analysis of National Stock Exchange

Indices

Dr. Gurmeet Singh

Unitedworld School of Business, Karnavati University
Gandhinagar, Gujarat, India
Abstract

This paper investigates weak form of efficiency in Indian equity market. For this purpose,
informational efficiency of National Stock Exchange of Indian’s indices i.e. NIFTY, bank
NIFTY and IT NIFTY is examined. The NSE indices returns under the study do not confirm
to normal distribution. The index log returns are found to be nonstationery at levels, whereas
first differences are found stationery. The result of autocorrelation test finds evidence of
statistical dependence in the returns generating process. In order to check the randomness in
returns, ARIMA model is applied. The results of ARIMA model for all the three indices of
NSE are found significant which implies that the returns are predictable and do not follow the
random walk except in case of bank NIFTY, whose autoregression of first order (AR1) was
found to be insignificant, but the serial correlation test of bank NIFTY is found significant.

Keywords: Random walk, ARIMA, Price Discovery, NIFTY, Efficient Market

JEL Code Classification: C13, C22, C32, D81, D82, G14, N15
Testing Random Walk Hypothesis: An Empirical Analysis of National Stock Exchange
Indices

Introduction
Reforms were introduced in the Indian capital market in 1991. Derivative trading was
introduced as one of the reform process and risk management in Indian stock markets. Risk
hedging and price discovery are the important role of the derivatives. Futures are expected to
reflect the prospective price movement in the underlying asset. In this connection futures of
the underlying assets are expected to reflect all current information and prospective events,
which are likely to happen in near time. Investigation of randomness in stock returns is
answer to this question.
Academicians and professional stock analysts have contradictory views on the price behavior
in the speculative markets. According to professional analyst there are trends in the
movement of stock prices, but one should have ability to read these trends timely to earn
super normal profits. Those who are able to read information earlier than others will have an
opportunity to secure super normal profits. Both ‘fundamentalists’ and ‘technicians’ agree on
these basic assumptions. The difference lies in the methodology to read the information early
(Alexander, 1961). The fundamentalists study the macro economic factors that cause the
price movements. The study of industry trends and individual firms within those industry,
future plans of the firms and general business conditions are used for fundamental analysis.
The technical analysts study the price movements of the immediate past to predict the future
price movements. Thus both school of analyst assumes the persistence of trends in the price
movements of stock prices. For this purpose, trends must depend for their existence on the
legged response of the market prices to the underlying factors governing these prices. It
might, seem possible that the trends arise not from a legged response of the market price to
the fundamental circumstances, but form a trend in those underlying circumstances
themselves (Working, 1934). Thus, if there is no lagged response, there should be no trend I
prices. The academic students of speculative markets don’t agree the very existence of trends
in the speculative stock prices. The academic students claim that where trends are observable,
they are merely interpretations, read in after the fact, of a process that really follows a random
walk (Fama, 1965) (1966). A price movement in the stock price can be characterized as a
random walk, if at any time the expected change can be represented by result of tossing a
coin, not necessarily a 50-50 coin. In particular, the random walk would mean that the next
move of the stock price is independent of all the past moves or events (Nordhaus, 1987).
Fama (1970) characterized the speculative market as efficient in weak form, if it satisfies
three conditions i.e. there are no transaction costs, all information is causelessly available to
all market participants and all agree on the implications of current information for the current
price and distributions of future prices of each security. He stated that there are other factors
also responsible for the market inefficiency but the above said conditions are potential
sources of market inefficiency.
Thus, the futures price movements should have lagged correlation coefficient close to zero
and said to be efficient only if the current and futures price reflects all information i.e.
information based on the events that have already occurred and on events, which as of now
the market expects to take place in the future. And the historical price movements do not
have significant impact on the current and future price movements. Thus, the efficient futures
market in the weak form is a primary condition for futures to be an efficient price discovery
vehicle. This impels no arbitrage opportunities persist in the futures market (Singh, 2001).
Base on the above discussion and in order to study the weak form of efficiency of Indian
equity futures market, the current paper studied the spot and future indices of National Stock
Exchange of India i.e. NIFTY, Bank NIFTY and IT NIFTY. The current study is divided in
to five sections. First section covered introduction, second section reviews the literature of
the study, the third section deals with data and methodology, fourth deals with the analysis
and fifth is the conclusion of the study.

Literature Review
We can find the study of random walk hypothesis way back in the beginning of 20th century
Cowels (1933) studied the behavior of stock market prices by evaluating the portfolio of
institutional investors. He has studied twenty fire insurance companies, sixteen financial
services companies and twenty four financial publications and found that the returns
generated were higher than returns generated in normal investments but these returns were
not in the position to beat the returns generated by investing outright investment in respective
stocks for the period of study. In his study he has also found that since patterns were missing
so technical theories are of no use. Further, Kendall and Hill (1953) made an effort to study
the behavior and trends of stock market by calculating the first twenty nine lagged serial
correlations of the first difference of twenty stock prices. They have done this study on
weekly basis. They have studied cash wheat at Chicago on weekly basis and spot cotton price
at New York on monthly basis. They have found that the past history of stock price changes
does not repeat itself.
Similar to the Kendall and Hill (1953) study, Working (1934) and Alexander (1961), finding
were consistent. Working (1934) studied the first difference of the series and not the series
itself, which are random number. The limitation of his study was the lack of force, which was
later provided by Kendall through the extensive study of legged stock prices. Further
Alexander (1961) quoted the study of Kendall & Hill (1953) and Osborne (1959) and stated
that the price movements are random, in an efficient speculative market. He has also
concluded that the price move, once initiated tends to persist until the market reverses the
same proportion. He has also concluded that the more accurate results regarding future price
movements can be achieved, if filters test is applied. Defining the random movements in the
stock markets, he observed unbiased movements in the stock prices. He has observed that if
the stock price has moved x percent, it is likely to move up more than x percent further before
it moves down by x percent. Mandelbrot (1966) in his study presented the Martingale Model,
he implied that price series may not be random but its unbiased property does not allow the
traders to exploit price dependences to secure super normal profits. Smidt (1968) in his study
stated that there is no complete absence of systematic elements. He has stated that the study
of random walk hypothesis would be more fruitful if they were conducted in the spirit of
attempting to determine the size and extent of systematic tendencies that may exists in the
stock price series. Even if the largest part of sequence of price changes may be described as
following a Martingale process, it is the remaining systematic components that may be more
critical. LeRoy (1989) and Roll (1994) studied that it is extremely hard to make profit even
from the most extreme violations of the market efficiency.
Cootner (1962) investigated the efficiency of NYSE and applied the mean square successive
difference test. He concluded that the price movements in the speculative markets are tends to
be significantly confirming leptokurtic distribution and found to be random. He also
concluded that the price movements are systematic and found very weak evidence of
randomness. Fama (1966) in his discussion paper stated that returns in the speculative market
do not confirm to normal distribution, the returns are more skewed and the kurtosis is
relatively higher. Speculative market returns may follow either platykurtic or leptokurtic
distribution. Stevenson and Bear (1970) has studied the corn and soyabean commodity
futures at Chicago Board of Trade. They have conducted these study for the period of 1951-
1967 and found that the futures returns confirms to the leptokurtic probability distribution,
however they have witnessed a very weak evidence of random walk in the price movements.
They found that under the mechanical trading patterns, long term segments profitable
throughout the time period covered both in an absolute sense and with respect to buy and
hold policy. Fama (1965) in his study evidently stated that the stock price movements are
independent and the returns confirms to the leptokurtic distribution and observed strong
evidence of random walk. Evans (1968) applied the filter rule and studied the price
movements of 470 securities listed in S&P Index over a period of 1958-67. He found that
employment of ‘Fixed Investment Proportion Maintenance’ strategy yields significantly
superior returns to those yielded by naïve buy and hold strategy, irrespective of the degree of
randomness or from characterizing the empirical distribution of security price changes. He
has also found that degree of non randomness is not of a magnitude and should not be
considered meaningful to the investor dealing with portfolio of securities. Jensen and
Benington (1970) examined the buy and hold policy and the mechanical trading rules for
1952 securities listed on NYSE for the period of 1926-1966 and concluded that the behavior
of security prices on NYSE is close to that predicted by the efficient market theories of
security price behavior. Fama (1970) categorized the informational efficiency of the
speculative markets. He clarified that if the stock prices show no legged response, it implies
that all past information and information related to the future events has already been
impounded in the current stock prices and the market will be efficient in the weak form. He
has categorized the informational efficiency in the three ways i.e. weak form efficiency,
semi-strong form efficiency and strong form efficiency. He has found that if the stock prices
are sensitive to the public announcements, market will be efficient in semi-strong form.
However, market sensitivity to information will be efficient in the strong form, if they may
affect the future price movements, and are not disclosed by the management.
Rendleman and Carabini (1979) applied autocorrelation test and examined the efficiency of
the Treasury bill future market in CME during 6th May 1976 to 31st March 1978, they found
market inefficiency due to quasi-arbitrage opportunities prevalent in the future market of T-
Bill. Rauser & Carter (1983) applied ARIMA, x2, t, Mean Square Prediction Tests on
soyabean, soyabean meal and soyabean oil and found that the multivariate and ARIMA
models outperform the futures markets for soyabean and soyabean meal but not soyabean oil
for both long term and short term horizons. They have also found the existence of arbitrage
opportunity in soyabean complex. Brannen & Ulveling (1984) observed the spot market of a
variety of commodities market was not a good predictor of prospective spot market prices.
They have compared how well current spot prices predict future spot prices for a variety of
commodities in a non futures market environment. However they have observed strong
evidence that the futures market was a good predictor of prospective spot prices and
recommended futures as an efficient price discovery vehicle. Kumar & Makhija (1986)
studied the efficiency of non stationery stock price series. They found that non stationery
time series confirms to the efficient market hypothesis. Their study contradicted the Shiller
(1981) results. Brown & Easton (1989) studied the daily data of London Stock Exchange and
tried to find out whether in historical periods LSE was efficient. They have used run test,
serial correlation and x2 as a tools and found that the London Stock Market was efficient in
the historical time period as it evidenced in the contemporary markets. The found serial
correlation coefficient close to zero and price movements were independent. Dokko &
Edelstein (1989) found that the Livingstone Stock Market forecasts to be adaptive and stock
prices follow the random walk. They have applied the same set of statistical techniques using
data for a period ranging 1955-1985. Dorfman (1993) studied the future price movements in
Chicago Mercantile Exchange by applying Bayesian Monte Carlo Integration Technique and
found that the future price series was stationery and does not confirm to the efficient market
hypothesis. Chen (1996) studied the price movement in S&P 500 stock price composite
monthly index and S&P common stock composite dividend yield using autocorrelation,
spectral analysis and filter techniques. He found that 70% of fluctuation in S&P Indexes,
detrained by filter, could be explained by deterministic color chaos. They suggested the
existence of persistent chaotic cycles revealing new perspective of market reliance and new
source of economic uncertainties. He concluded the non existence of random walk. Reddy
(1997) used ARCH and GARCH models to investigate the efficient market hypothesis in
Indian stock market and found that the stock market returns were not normally distributed. It
means the price movements are speculative and volatile. He held low spread of information
dissemination, the cause for Indian stock market inefficiency. Most of the studies conducted
in the developed economies confirm the weak form of market efficiency, while in the case of
developing or under developed economies the reverse situation prevails (Mobarek & Keasey,
2000).
Dimison & Mussavian (1998) supported the efficient market hypothesis by demonstrating the
difficulty of beating the market, whether by analyzing publicly available information or by
employing professional investment advisory services. Kamath (1998) examined the 10
industrial indices and SET index of Thailand over a period of 15 years. He has observed
significant day of the week effects in Thailand stock market. He has also concluded that the
index movements confirm to leptokurtic distribution. Mishra (1999) and Anshuman &
Goswami (2000) studied the day of week effect in the Indian stock market and concluded that
weekday anomalies present in the Indian stock markets. Mobarek & Keasey (2000) applied
run test, autocorrelation test, autoregression test and ARIMA model to study the behavior
stock price movement in Dhaka Stock Exchange of Bangladesh for a period 1988-97. They
have found market inefficiency and reported huge transaction cost, weak communication
system long processing time and low volumes as the reasons for inefficiency in the market.
Marrisetty (2003) found that prices take 19 days to adjust to their intrinsic value in Indian
stock market and observed inefficiency in information dissemination. He has also observed
the overreaction in the stock prices before adjusting to their intrinsic values. Mangala &
Mittal (2005) studied the daily closing returns and found the returns to be negative on Friday
and most positive on Wednesday. They have also found inefficiency in the Indian Stock
Market implying that trader can design their strategy and can exploit the situation to secure
extra profits.
In nutshell, on the basis of above-mentioned studies it can be stated that the significant efforts
have been made at the international level to evaluate the weak form of efficiency in stick
prices, whereas in India it has not been well investigated. Therefore, the current paper is one
more attempt to study the weak form efficiency of Indian futures market by studying the
NIFTY Index spot and futures, Bank NIFTY Index spot and futures and IT NIFTY Index spot
and futures prices of National Stock Exchange of India.

Data & Methodology

As the central stock exchange of India, NSE witnesses’ maximum trading volume in F&O
segment, Therefore indices of NSE and its current month futures are considered for the study.
To accomplish the research objective daily data ranging from January-2011 to December-
2015 are obtained which comprises 1233 data points for the analysis. The choice of study
period is based on the availability of data series. The series of return is computed from daily
closing data for the NIFTY spot and future index, Bank NIFTY spot and future and IT
NIFTY spot and future indices of National Stock Exchange of India. The daily returns are
continuous rates of return, computed as log of ratio of present day’s price to previous day’s
price (i.e. Rt = ln (Pt /Pt-1)). Descriptions of variables and data sources are presented in Table
1.
Table 1: Description of Variable
Acronyms Construction of Variable Data Source
LNNFSR Natural Logarithm of the NIFTY Index Spot Returns NSE Website
LNNFFR Natural Logarithm of the NIFTY Index Future Returns NSE Website
LNBNFSR Natural Logarithm of the Bank NIFTY Index Spot Returns NSE Website
LNBNFFR Natural Logarithm of the Bank NIFTY Index Future Returns NSE Website
LNINFSR Natural Logarithm of the IT NIFTY Index Spot Returns NSE Website
 LNINF
FFR Natuural Logarithm
m of the IT NIIFTY Index Future
F Returnss NSE Website

In timee series anaalysis data should be stationery,, otherwisee the resultts might prrovide a
spuriouus results, meaning
m thaat, the meann and varian
nce should be constantt over time and the
value of covariancce between two
t time peeriods depen
nds only onn the distancce between the two
time period and noot the actual time at which the covaariance is coomputed. T
The widely used
u and
popularr test for staationary is thhe unit roott test. The prresence of unit
u root inddicates that the data
series is
i non-statioonery. The standard procedures
p of unit rooot test nameely the Aug
gmented
Dickey Fuller (AD
DF) (1979) (1981)
( and Phillips and
d Perron (19988) is perfformed to ch
heck the
stationaary nature of
o the series. In the Auggmented Diickey Fullerr (ADF) tesst null hypo
othesis is
that datta set being tested has unit
u root. Asssuming thaat the series follows an AR (p) pro
ocess the
Augmennted Dickeey Fuller (A
ADF) test makes
m a paarametric coorrection annd controls for the
higher order
o correlation by addding the laagged diffeerence termss of the deppendent varriable to
the righht hand sidee of the regression equuation. The complete model
m with deterministtic terms
such as intercepts and
a trends is
i shown in equation (1
1).

(1)
Where, is a coonstant, iss the coefficcient on a time trend and is th
he lag ordeer of the
autoreggressive proccess.

Phillipss and Perronn (1988) prropose an allternative (n

nonparamettric) methodd of controllling for
serial coorrelation when
w testingg for a unit root.
r The PP unit root test methodd estimates the
t non-
augmennted DF tesst equationn, and modifies the t-ratio of thee  coefficcient so thaat serial
correlattion does noot affect thee asymptoticc distributio The PP test is based
on of the tesst statistic. T
on the statistic:
s

(2)
Where ^ is the esttimate, and t the t-ratio of ^, se((^) is coeffficient stanndard error, and S is
n, 0 is a coonsistent esstimate of the
the stanndard error of the test regression.. In addition t error
variance in non-auugmented DF
D equationn. The ADF
F unit root test
t and PP
P unit root test
t also
providee the order of integration of the tiime series variables.
v Inn a multivaariate context if the
variablee under connsideration are found to
t be I (1) (i.e. they are
a non-stattionary at level but
stationaary at first difference),
d nation of thhe integratedd variables is I (0),
but the linnear combin
then thee variables are
a said to be
b co-integrrated (Enderrs, 2004).
One of the key asssumptions of
o the ordinaary regressiion model iss that the errrors have the
t same
variance throughouut the sampple. This is also called the homosscedasticity model. If the
t error
variance is not connstant, the data
d are saidd to be heterroscedastic. Findings of heterosced
dasticity
in stockk returns aree well docum
mented (Maandelbrot, 1963) (Famaa, 1965) (Boollerslev, 19
986).

To Tesst whetherr the successive retuurns of NIIFTY indicces are inndependent or not
autocorrrelation teest and AR
RIMA proccess are employed.
e The Breussch–Godfrey
y serial
correlattion LM tesst (1979), (1978) is a test
t for auto
ocorrelationn. Autocorreelation is a reliable
measuree for testinng of indeppendence off random variable
v in return seriees. Kendalll & Hill
(1953), Fama (19665), Stevennson & Beaar (1970), Rendleman
R & Carabinii (1979), Nordhaus
N
(1987), Chen (19996), Reddyy (1997), annd Mobarek
k & Keaseey (2000) have appliied auto
correlattion test inn various speculative
s stock marrkets. The serial corrrelation coefficient
measurees the relatiionship betw
ween the vaalues of a random
r variiable at tim
me t and its value in
usch–Godfreey serial correlation LM
the prevvious periodd. The null hypothesis of the Breu M test is
that the residuals are
a not seriaally correlateed.

The AR
RIMA proccedure analyyzes and foorecasts equ
ually spaceed univariatte time seriies data,
transferr function data,
d and inntervention data using the Auto Regressive
R Integrated Moving
Average (ARIMA
A) or autoregressive moving-averaage (ARMA
A) model. A
An ARIMA
A model
predictss a value in a responsee time seriess as a linearr combinatioon of its ow
wn past valu
ues, past
errors (also called shocks or innnovations)), and curren
nt and past values
v of otther time seeries.

The AR
RIMA approoach was firrst popularized by Box
x and Jenkinns (1976), aand ARIMA
A models
are ofteen referred to as Box-JJenkins models. The general
g transsfer functioon model em
mployed
by the ARIMA prrocedure was
w discusseed by Box and Tiao (1975). Onee subset of ARMA
models are the so-ccalled autorregressive, or
o AR mod
dels. An AR
R model exppresses a tim
me series
as a lineear functionn of its pastt values. Thhe order of the
t AR moddel tells how
w many lagg
ged past
values are
a includedd. An AR (p)
(p (Auto Regressive
R of
o order p) model
m is a discrete tim
me linear
equations with noise, of the foorm

(3)

Here p is the orderr, α1,…,αp are the parrameters or coefficients (real num
mbers), εt is an error

w intensityy σ2. The moving

term, ussually a whhite noise with m averrage (MA) model is a form of
ARMA
A model in which
w the time series iss regarded as
a a movingg average (uunevenly weighted)
w
of a ranndom shockk series εt. A MA (q) (Moving Average
A withh orders p and q) mod
del is an
explicitt formula foor Xt in term
ms of noise of
o the form

(4)

The proocess is giveen by a (weeighted) aveerage of thee noise, but not an averrage from tiime zero
to the present
p timee t; instead, an averagge moving with
w t is takken, using only the last q + 1
times. An
A ARMA (p, q) (Autto Regressivve Moving Average with
w orders p and q) mo
odel is a
discretee time linearr equations with noise, of the form
m

(5)

Empiriical Analysis
The desscriptive staatistics for all
a the variaables are prresented in Table
T 2. Thhe value of kurtosis
and skeewness show
ws the lackk of symmettric in distriibution. The observed distribution
n is said
to be normally
n disstributed, iff the value of skewnesss and kurttosis are 0 and 3 respeectively.
From Table
T 2 it is observed thhat the freqquency distrribution of the
t underlyying variablees is not
normal.. The null hypothesis of the Jarrque-Bera statistics
s is the frequeency distrib
bution is
normallly distributeed. The siggnificant coeefficient off Jarque-Berra statistics also indicaates that
the freqquency distrributions off the considdered series are not norrmal. The pprobability value
v of
less thaan 0.05 of Jaarque-Bera statistics inndicates thatt the frequenncy distribuutions of con
nsidered
series are
a not norm
mally distribbuted, whichh is the preccondition foor any markket to be effficient in
the weaak form ((F
Fama E. , 1965), (Steevenson & Bear, 1970), and Reeddy (1997
7)). The
evidencce for rejecttion of norm
mality in inndex futuress returns is consistent with those fond by
Kendalll (1953), Fama (1965)), Stevenson & Bear (1970),
( Cheen (1996), Reddy (199
97), and
Kamathh (1998) in the
t differennt financial markets.
m

Tabble 2: Descriiptive Statisstics of Variiable

LNNFSR LNNFFR LNBNFSR LNBNNFFR LNIINFSR LN
NINFFR
Meean 0.0176 0.0179 0.0186 0.0198 0.0316 0.0294
Meddian 0.0232 0.0360 0.0255 0.0177 0.0337 0.0227
Maxiimum 1.5704 1.6180 2.3474 2.3261 2.3353 2.3418
Minimum -1.9337 -1.9259 -2.0665 -2.14411 -2..5449 -2.5452
-
Std. Dev. 0.6294 0.6509 0.8035 0.8296 0.7007 0.7030
Skew
wness -0.0394 -0.0447 0.0051 -0.00069 -0..1148 -0.0938
-
Kurttosis 2.3350 2.3152 2.2092 2.1705 2.6522 2.6231
 Jarque-Bera 23.0353 24.5001 32.1361 35.3594 8.9234 9.1053
Probability 0.0000 0.0000 0.0000 0.0000 0.0115 0.0105
Observations 1233 1233 1233 1233 1233 1233
Source: Author's Estimation

Further, to check the stationarity of the underlying data series, ADF test and PP test are
employed and results are presented in Table 3.On the basis of the Augmented Dickey Fuller
test and Phillips and Perron test, all the series are found to be non-stationery at level with
intercept. However, after taking the first difference these series are found to be stationery at
1% significance level. Thus the stationery test indicates that all series are individually
integrated of the order I (1). Thus, in order to give robust interpretations on weak form of
efficiency of Indian stock market, ARIMA model is applied.

Table 3: Result of Augmented Dickey-Fuller & Philips Perron Unit Root Test
ADF Test PP Test

Variable With Drift & With Drift &

Without Drift* Without Drift*
Trend* Trend*

D(LNNFSR) -17.02756 -17.01311 -520.5248 -528.9095

D(LNNFFR) -20.20933 -20.19343 -449.0205 -451.2833
D(LNBNFSR) -16.69332 -16.67929 -405.0239 -410.6576
D(LNBNFFR) -16.01045 -15.99732 -467.3833 -475.536
D(LNINFSR) -15.62296 -15.6123 -440.9522 -439.8993
D(LNINFFR) -15.45261 -15.44197 -562.6176 -561.4593
Base on MacKinnon (1996) one-sided p-values. *Significant at 1% level of significance.
Source: Author's Estimation

The test of serial correlation in the NIFTY indices for weak form of efficiency in the Indian
futures market is presented in the Table 4 reveal that the NIFTY futures returns are serially
dependent and the null hypothesis for all the indices of NIFTY are accepted at 1% level of
significance. But acceptance or rejection of random walk hypothesis in futures returns can’t
inferred by just checking the serial correlation, Thus, in order to check the randomness
further investigation is required. Thus with this purpose ARIMA model is applied.

Table 4: Breusch-Godfrey Serial Correlation LM Test

Indices Null Hypothesis
NIFTY
F-statistic 75.4017 Prob. F(2,1229) 0.0000
Accepted
Obs*R-squared 134.7588 Prob. Chi-Square(2) 0.0000
Bank NIFTY
F-statistic 78.88707 Prob. F(2,1229) 0.0000
Accepted
Obs*R-squared 140.2792 Prob. Chi-Square(2) 0.0000
IT NIFTY
F-statistic 102.8265 Prob. F(2,1229) 0.0000
Accepted
Obs*R-squared 176.7467 Prob. Chi-Square(2) 0.0000
Source: Author's Estimation

After checking the stationarity & serial correlation of the time series data, as discussed above,
ARIMA (1, 1, 1) model is applied for NIFTY spot and future, Bank NIFTY spot and future
and IT NIFTY spot and future indices of NSE of India, which means Autoregression of first
order (AR1), first order difference of time series and Moving Average of first order (MA1).
Table 7, 8 and 9 discusses ARIMA results.

Table 5: Result of ARMA Model of NIFTY

Variable Coefficient Std. Error t-Statistic Prob.
C 0.000156 0.001686 0.092678 0.9262
LNNFFR 0.975675 0.006095 160.0684 0.0000
AR(1) 0.269242 0.050698 5.310736 0.0000
MA(1) -0.707919 0.039703 -17.83034 0.0000
SIGMASQ 0.019741 0.000573 34.48083 0.0000
R-squared 0.950123 Akaike info criterion -1.078764
F-statistic 5848.142 Schwarz criterion -1.058013
Prob(F-statistic) 0.000000 Hannan-Quinn criter. -1.070958
Source: Author's Estimation

ARIMA results of NIFTY index of NSE of India are shown in the Table 5. The t-ratio of
AR1 and MA1 for NIFTY future returns is significant at 5% level of significance, which
implies that NIFTY futures returns are predictable and do not follow the random walk.

Table 6: Result of ARMA Model of Bank NIFTY

Variable Coefficient Std. Error t-Statistic Prob.
C -0.000836 0.002583 -0.323637 0.7463
LNBNFFR 0.981767 0.006506 150.8946 0.0000
AR(1) 0.026585 0.057252 0.464349 0.6425
MA(1) -4.85E-01 0.049953 -9.704104 0.0000
SIGMASQ 0.027393 0.000725 37.76708 0.0000
R-squared 0.957537 Akaike info criterion -0.751275
F-statistic 6922.854 Schwarz criterion -0.730524
Prob(F-statistic) 0.000000 Hannan-Quinn criter. -0.743469
Source: Author's Estimation
ARIMA results of Bank NIFTY index of NSE of India are shown in the Table 6 are quite
interesting. The t-ratio of (MA1) for Bank NIFTY future returns is significant at 5% level,
which implies that NIFTY futures returns are predictable and do not follow the random walk,
but Autoregression of first order (AR1), the first order difference of bank NIFTY index future
returns are insignificant, as the probability value of (AR1) is 0.6425, more than significance
level of 5%.

Table 7: Result of ARMA Model of IT NIFTY

Variable Coefficient Std. Error t-Statistic Prob.
C 2.65E-03 0.002085 1.270351 0.2042
LNINFFR 9.80E-01 0.006819 143.739 0.0000
AR(1) 1.77E-01 0.053338 3.327394 0.0009
MA(1) -6.25E-01 0.04438 -14.07173 0.0000
SIGMASQ 0.024569 0.000801 30.66346 0.0000
R-squared 0.949918 Akaike info criterion -0.86005
F-statistic 5822.945 Schwarz criterion -0.839299
Prob(F-statistic) 0.000000 Hannan-Quinn criter. -0.852243
Source: Author's Estimation

ARIMA results of IT NIFTY index of NSE of India are shown in the Table 7. The t-ratio of
AR1 and MA1 for IT NIFTY future returns is significant at 5% level of significance, which
implies that IT NIFTY futures returns are predictable and do not follow the random walk.

Conclusion
The results of the current study are very important for the investors, traders and regulators of
the Indian stock markets. As observed in the other speculative markets, the NSE indices
returns under the study do not confirm to normal distribution. As the first difference NSE
indices under the study are found to be stationery and the serial correlation LM test signifies
first order autoregression thus, ARIMA (1, 1, 1) was found to be best suited model for the
NSE indices under the study. With one exception, t-ratio of (AR1) and (MA1) was found
significant, which implies that returns are predictable. When the auto correlation of bank
NIFTY indices was studied, whose autoregression of first order (AR1) was found to be
insignificant, it was found that the serial correlation was significantly close to zero. Low
transaction cost, flexibility, easy and flexible margin requirements, huge trade volumes, high
liquidity, easy to arbitrage are the main characteristics of an efficient equity markets. In India,
equity futures market returns are predictable, which signifies that the Indian equity market is
not informational efficient. High volatility, complex derivative products, may be the reasons
for inefficiency. Moreover high transaction cost may be another reason which is responsible
for the non random movement of futures prices.
The traders and investors are advised to use these results cautiously as the Indian equity
futures market are reported as informational inefficient but price discovery and causality
between the cash and futures market returns will give correct results.
Since the small investor of Indian equity market is not well verse with the stock market
variations and don’t know when to enter or quit the market. Information efficiency of equity
futures market will be great help to them and will lead to operational efficiency. An
informational efficient futures market will make the role of fund managers easier and will be
helpful to institutional investors as well, as the interest of the small investors is involved in
these institutions.

Bibliography

Alexander, S. S. (1961). Price Movements in Speculation Markets: Trends or Random Walk. Industrial
Management Review , 2, 7‐26.

Anshuman, R., & Goswami, R. (2000). Day of the Week Effects on the Bombay Stock Exchange. UTI
Indian Capital Market Conference. Mumbai: UTI.

Bollerslev, T. (1986). Generalized Autoregressive Conditional Heteroskedasticity. Journal of

Econometrics , 31 (3), 307‐327.

Box, G., & Jenkins, G. (1976). Time Series Analysis: Forecasting and Control. San Francisco: Holden‐
Day.

Box, G., & Tiao, G. (1975). "Intervention Analysis with Applications to Economic and Environmental
Problems. JASA , 70, 70‐79.

Brannen, P. P., & Ulveling, F. E. (1984). Considering an Informational Role for a Futures Market. The
Review of Economic Studies , 51 (1), 33‐52.

Breusch, T. S. (1979). Testing for Autocorrelation in Dynamic Linear Models. Australian Economic
Papers , 17, 334–355.

Brown, L. R., & Easton, A. S. (1989). Weak‐Form Efficiency in the Nineteenth Century: A Study of
Daily Prices in the London Market for 3 per cent Consols, 1821‐1860. Economica , 56 (2), 61‐70.

Chen, P. (1996). A Random‐Walk or Color‐Chaos on the Stock Market?‐ Time‐Frequency Analysis of S

& P Indexes. Studies in Nonlinear Dynamics & Econometrics , 1 (2), 87‐103.

Cootner, P. H. (1962). Stock Prices: Random vs. Systematic Changes. Industrial Management Review ,
3, 24‐45.
Cowels, A. (1933). Can Stock Market Forecasters Forecast? Econometrica , 1 (3), 309‐324.

Dickey, D. A., & Fuller, W. A. (1979). Distribution of the Estimators for Autoregressive Time Series
with a Unit Root. Journal of American Statistical Association , 74 (366), 427‐431.

Dickey, D. A., & Fuller, W. A. (1981). Likelihood Ratio Statistics for Autoregressive Time Series with a
Unit Root. Econometrica: Journal of the Econometric Society , 49 (4), 1057‐1072.

Dimson, E., & Mussavian, M. (1998). A Brief History of Market Efficiency. European Financial
Management , 4 (1), 91‐104.

Dokko, Y., & Edelstein, H. R. (1989). How well Do Economists Forecast Stock Market Prices? A Study
of the Livingstone Surveys. The American Economic Review , 79 (4), 865‐871.

Dorfman, H. J. (1993). Bayesian Efficiency Tests for commodity Futures Markets. American Journal of
Agriculture Economics , 75 (5), 1206‐1210.

Enders, W. (2004). Applied Econometric Time Series (2nd ed.). Wiley Series in Probability and
Statistics.

Evans, L. J. (1968). The Random Walk Hypothesis, Portfolio Analysis and the Buy‐and‐Hold Criterion.
Journal of Financial and Quantitative Analysis , 3 (3), 327‐342.

Fama, E. F. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. Journal of
Finance , 25 (2), 383‐417.

Fama, E. F. (1966). Random Walk in Stock Market Prices. Selected Paper No. 16, Graduate School of
Business, University of Chicago , 1‐17.

Fama, E. F. (1965). The Behavior of Stock‐Market Prices. The Journal of Business , 38 (1), 34‐105.

Fama, E. (1965). The Behaviour of Stock Market Prices. Journal of Business , 38 (1), 34‐105.

Godfrey, L. G. (1978). Testing Against General Autoregressive and Moving Average Error Models
when the Regressors Include Lagged Dependent Variables. Econometrica , 46, 1293–1302.

Jensen, C. M., & Benington, A. G. (1970). Random Walks and Technical Theories: Some Additional
Evidence. The Journal of Finance , 25 (2), 469‐482.

Kamath, R. R. (1998). Return distribution and the Day‐of‐the‐Week Effects in the Stock Exchange of
Thailand. Journal of Economics and Finance , 22 (2‐3), 97‐106.

Kendall, M. G., & Hill, B. A. (1953). The Analysis of Economic Time Series Part I. Prices. Journal of
Royal Statistical Society , 96, 11‐25.

Kumar, R., & Makhija, K. A. (1986). Volatility of Stock Prices and Market Efficiency. Managerial and
Decision Economics , 7 (2), 119‐122.

LeRoy, F. S. (1989). Efficient Capital Markets and Martingales. Journal of Economic Literature , 27 (4),
1583‐1621.
Mandelbrot, B. (1966). Forecasts of Futures Prices, Unbiased Markets and ‘ Martingale’ Model. The
Journal of Business , 39 (1. Part‐2), 242‐255.

Mandelbrot, B. (1963). The Variation of Certain Speculative Prices. Journal of Business , 36, 394‐419.

Mangala, D., & Mittal, R. K. (2005). Efficiency of Indian Stock Market: An Evidence of Day of the
Week Effect. Gyan – A Bi‐Annual Journal of Management & Technology , 1 (1), 3‐8.

Marisetty, V. B. (2003). Measuring Productive efficiency of Stock Exchanges Using Price Adjustment
Coefficients. International Review of Finance , 4 (1‐2), 79‐99.

Mishra, B. (1999). Presence of Friday Effect in Indian Stock Market. Paradigm , 3 (2), 57‐64.

Mobarek, A., & Keasey, K. (2000). Weak‐Form Market Efficiency of an Emerging Market: Evidence
from Dhaka Stock Market of Bangladesh. ENBS Conference . Oslo.

Nordhaus, D. W. (1987). Forcasting Efficiency: Concepts and Applications. The Reviwe of Economics
and Statistics , 69 (4), 667‐674.

Osborne, M. M. (1959). Brownian Motion in the Stock Market. Operation Research , 7 (2), 145‐173.

Phillips, P. C., & Perron, P. (1988). Testing for a Unit Root in Time Series Regression. Biometrika , 75
(2), 335‐346.

Rauser, C. G., & Carter, C. (1983). Futures Market Efficiency in the Soyabean Complex. The Review of
Economic and Statistics , 65 (3), 469‐478.

Reddy, S. Y. (1997). Efficiency of Indian Stock Markets: An Empirical Analysis of Weak Form EMH of
the BSE. UTI Indian Capital Market Conference (pp. 91‐115). UTI.

Reddy, S. Y. (1997). Efficiency of Indian Stock Markets: An Empirical Analysis of Weak‐Form EMH of
the BSE. UTI Indian Capital Market Conference, (pp. 91‐115).

Rendleman, J. R., & Carabini, E. C. (1979). The Efficiency of the Treasury Bill Futures Market. The
Journal of Finance , 34 (4), 895‐914.

Roll, R. (1994). What Every CEO Should Know About Scientific Progress in Economics: What is Known
and What Remains to be Resolved. Financial Management , 23, 69‐75.

Shiller, R. J. (1981). Do Stock Prices Move too Much to be Justified by Subsequent Changes in
Dividends? American Economic Review , 71, 421‐436.

Singh, J. R. (2001). Weak Form Efficiency of Indian Commodity Futures. UTI Indian Capital Market
Conference (pp. 3‐23). UTI.

Smidt, S. (1968). A New Look at the Random Walk Hypothesis. The Journal of Financial and
Quantitative Analysis , 3 (3), 235‐261.

Stevenson, A. R., & Bear, M. R. (1970). Commodity Futures: Trends or Random Walks? The Journal of
Finance , 25 (1), 65‐81.
Stevenson, A. R., & Bear, M. R. (1970). Commodity Futures: Trends or Random Walks? The Journal of
Finance , 25 (1), 65‐81.

Working, H. (1934). A Random Difference Series for Use in the Analysis of Time Series. Journal of the
American Statistical Association , 29, 11‐24.