MARKET∗
Anand Pandey†
October 2003
∗
The paper was a part of project for the Time Series Course by Dr. Susan Thomas.
†
Anand Pandey is Research Scholars at IGIDR, Mumbai. E-Mail: anand@igidr.ac.in
1
Abstract
risk. On the other hand if markets are not efficient, excess returns can
three popular stock indices is carried out to test the efficiency level in
Indian Stock market and the random walk nature of the stock market
by using the run test and the autocorrelation function ACF (k) for
The study carried out in this paper has presented the evidence of
analyses and runs test we are able to conclude that the series of stock
indices in the India Stock Market are biased random time series. The
does not confirm the applicability of the random walk model in the
India stock market. Thus there are undervalued securities in the mar-
ket and the investors can always excess returns by correctly picking
them.
2
1 INTRODUCTION
Efficient Market Hypothesis (EMH) states that security prices fully reflect
all available information. The weak form of EMH states that the current
prices fully reflect the information implied by the past prices (historical se-
quence of prices). In an efficient market at a given instant of time the prices
3
are assumed to reflect all available information. One would expect the cur-
rent price of security to be good estimate of its intrinsic values. If the ad-
justment to new information is instantaneous successive price changes will
be independent.
2 Literature Survey
Sharma and Kennedy (1977) compared the behaviour of stock indices of the
Bombay, London and New York Stock Exchanges during 1963-73 using run
test and spectral analysis. Both run tests and spectral analysis confirmed the
random movement of stock indices for all the three stock exchanges. They
concluded that stock on the BSE obey a random walk and are equivalent in
the markets of advanced industrialized countries”.
Kulkarni (1978) investigated the weekly RBI stock price indices for Bom-
bay, Calcutta, Delhi, Madras and Ahmedabad stock exchanges and monthly
indices of six different industries of six different industries by using spectral
method. He concluded that there is a repeated cycle of four weeks for weekly
prices and seasonality in monthly prices. This study has thus rejected the
hypothesis that stock price changes were random.
Yalawar (1988) studied the month end closing prices of 122 stocks listed
on the Bombay Stock Exchange during the period 1963-82. He used only the
non-parametric tests Spearman’s rank correlation test and runs test. 21 out
4
of 122 lag 1 correlation coefficients were significant at 5significant difference
for 9 out of 122 stocks.
Thus in the Indian context, except for some of the studies, the available ev-
idence in general indicates that the successive price changes are independent
and the random walk model is appropriate to describe the stock behaviour.
3 METHEDOLOGY
Since the test of weak form of EMH, in general, have come from the random
walk literature, so I am interested in testing whether or not successive price
changes were independent of each other. In this paper, I will use Autocorre-
lation and Runs Test for testing the efficiency of the stock market.
Autocorrelation is one of the statistical tools used for measuring the depen-
dence of successive terms in a given time series. Therefore it is used for mea-
suring the dependence of successive share price changes.It is the basic tool
used to test the weak form of EMH.The autocorrelation function ACF(k) for
the time series Yt and the k-lagged series Yt−k is defined as :-
5
where ȳ is the overall mean of the series with n observations.
The standard error of ACF(k) is given by:
1
SeACF (k) = q (2)
(n − k)
1
SeACF (k) = q (3)
(n)
To test whether ACF (k) is significantly different from zero, the following
distribution of ’t’ is used, i.e. t=ACF (k)/ SeACF (k) .
For both random variable series and series with trends, ACF (k) are very
high and decline slowly as the lag value (k) increases. At the same time the
ACF (k) of the first difference series (price changes or returns) are statis-
tically insignificant when the series is a random walk series.A random walk
series drifts up and down over time. In some situation it may be difficult
to judge whether a trend or drift is occuring. Hence to determine whether
a series has significant trend or whether it is a random walk, the t-test is
applied on the series of first differences.
6
3.2 Run Test
Run test is a non-parametric test. It depends only on the sign of the price
changes but not on the magnitude of the price. It does not require the spec-
ification of the probability distribution. It depends only on the sign of the
price. They are essentially concerned with the direction of changes in the
time series.
Under the hypothesis that the successive price changes are independent
and the sample proportion of positive, negative and zero price changes are
unbiased estimates of the population proportions, the expected number of
runs of all the types is computed as follows, (by Wallis, Robert (1956)).
N (N + 1) − Σ3i=1 n2i
M= (4)
N
7
where M = Expected number of runs,
ni = Number of price changes of each sign (i = 1,2,3) and
N = Total number of price changes = n1 + n2 + n3
The standard error of the expected number of runs of all signs may be
obtained as :-
4 DATA
The sample period is January 1996 to June 2002. The data consist of daily
and weekly closing values of three leading stock indices namely CNXdefty,
CNX Nifty, CNX Nifty Junior.
Autocorrelations of the weekly changes in the three stock indices are given
in table-1:-
8
Table 1: Autocorrelations of Weekly Changes in stock Indices
For CNX Defty, the autocorrelation coefficient for lag 1 is 0.1018, which
is very much larger than twice the standard error ( =2*0.0130). Thus the
autocorrelation differ significantly from zero.From the table, we can see that
out of the 60 autocorrelation computed for the three stock indices, 39 differ
significantly from zero. That is 65% of the autocorrelations differ signifi-
cantly.
9
The t-values of the autocorrelations corresponding to the stock indices
CNX Defty, CNX Nifty and CNX Nifty Junior were computed. It was found
that the autocorrelations are significantly different from zero and the corre-
sponding t-values are greater than 1.96 (at 5% level of significance). Thus
the stock indices are biased random time series and the stock market is not
weakly efficient in pricing its securities.
To verify this, in the next section we perform the runs test. We test the
null hypothesis that price changes are independent. The result of three stock
indices are given in table-2 :-
Index n n1 n2 n3 R σm Z P
CNX Defty 295 147 148 0 137 38.27 -1.34 0.18
CNX Nifty 311 155 156 0 141 44.17 -1.76 0.08
CNX Nifty Jr. 311 155 156 0 139 104.96 -1.99 0.05
The results show that all indices of NSE(except CNX Nifty Junior) show
weak form of market efficiency.The information regarding yesterday’s indices
are effectively absorbed by today’s indices.The stocks in the index absorb the
price information effectively. But the results regarding CNX Nifty Junior is
different, having Z value of -1.99 which is significant at 5% level.
10
6 CONCLUSION
The assumption that the stock prices are random is basic to the Efficient
Market Hypothesis and Capital Asset Pricing Models. The study carried
out in this paper has presented the evidence of the inefficient form of the
Indian Stock Market. From autocorrelaion analyses and runs test we are
able to conclude that the series of stock indices in the India Stock Market
are biased random time series. The auto correlation analysis indicates that
the behaviour of share prices does not confirm the applicability of the random
walk model in the Indian stock market. Thus there are undervalued securities
in the market and the investors can always make excess returns by correctly
picking them.
11
References Yalawar, Y.B. [1988] ” Bombay Stock Exchange: Rates
of Return and Efficiency”, Indian Economic Journal, Vol. 36, No.4 (April-
June), 68-121
12